CHAPTER ONE

INTRODUCTION

1.1 Background to the Study

The use of soft computing techniques such as Artificial Neural Network (ANN), support vectors

regression (SVR), and neuro-fuzzy system (NFS) in hydrology and water resources modeling has

been popular in recent years. A neuro-fuzzy system (NFS) is a system that takes advantage of the

learning ability of neural networks and the reasoning ability of fuzzy systems (Cho et al. 2009).

Fuzzy models that assume local model presentations with local function dynamics at the

consequent or rule-layer of the models are known as Takagi–Sugeno–Kang (TSK) (Takagi and

Sugeno 1985) models. The adaptive network-based fuzzy inference system (ANFIS) is an

example of TSK-models where global parameter tuning is implemented by minimizing the

global error of the model (Jang 1993). To date, ANFIS has been employed in a wide range of

hydrological modeling including rainfall-runoff modeling (Nayak et al. 2004; Tayfur and Guldal

2006; Talei et al. 2010; Talei and Chua 2012; Tsai et al. 2014), flood forecasting (Nayak et al.

2005; Mukerji et al. 2009; Talei et al. 2013), water resources management (Abolpour et al. 2007;

Chang and Chang 2006), water quality modeling (Yeon et al. 2008), and rainfall forecasting (El-

Shafie et al. 2011).

The application used to forecast the state of the atmosphere for a specific region at any particular

moment is known as weather forecasting in science and technology. Humans have been

attempting weather prediction for millennia, and in a formal capacity since the nineteenth

century. Weather forecasts are produced by gathering numerous quantitative data regarding the

current state of the atmosphere and applying scientific knowledge of the process to predict how it

will evolve.

1
Once a purely human undertaking, weather forecasting today relies on computer-based

techniques that take into consideration many atmospheric parameters. It was formerly primarily

relied upon changes in barometric pressure, present weather conditions, and sky condition. In

order to choose the optimal prediction model and provide the basis for the forecast, human input

is still needed. This entails a variety of abilities related to pattern identification, teleconnections,

model performance, and model biases. Forecasts become less accurate as the difference between

the current time and the time for which the forecast is being made (the range of the forecast)

increases due to the chaotic nature of the atmosphere, the massive computational power required

to solve the equations that describe the atmosphere, error involved in measuring the initial

conditions, and an incomplete understanding of atmospheric processes.

Weather forecasting is made possible by gathering as much data as possible about the current

state of the atmosphere, such as temperature, humidity, and wind, and then using meteorology to

understand atmospheric processes to predict how the atmosphere will change in the future.

Conventionally, trained observers, automated weather stations, or buoys gather data on air

pressure, temperature, wind direction, speed, humidity, and precipitation that are observed at the

surface. The meteorological analysis is then produced by combining the most recent forecast

from a numerical model for the time that observations were made with information gathered

from the observations during the data assimilation process.

Numerical weather prediction models, or computer simulations of the atmosphere, start with

analysis and use knowledge of physics and fluid dynamics to modify the atmosphere's condition

over time. Supercomputers are needed to solve the complex equations that govern how a fluid's

condition changes over time. The model's outputs then enable the creation of weather forecasts.

1.2 Statement of the Problem

2
Accurate weather forecasting presently has been a major challenge in Nigeria. The primitive

methods have been met with many unresolved challenges which have posed a major concern for

Meteorology department in Nigeria. In the field of meteorology this task is complicated because

all decisions are made within a visage of uncertainty associated with the weather system (Hasan

et al, 2008). Also, in Nigeria most instruments such as rain gauges, evaporation pans, windvanes,

anemometers, barometers, thermometers, hygrometers, and sunshine recorders etc are poorly

maintained in meteorological stations in Nigeria. This amounts to inadequate gathering of

weather information or data which in turn poses a serious challenge on accurate weather forecast.

The challenges are mainly related to inadequate infrastructure for weather observation and

Ineffective dissemination of weather information to vulnerable communities.

1.3 Objectives of the Study

1.3.1 Aim

This project primarily aims at developing a weather forecasting system that will predict accurate

weather conditions and allow authorized users to access weather information remotely.

1.3.2 Objectives

The objectives are as follows:

i. To develop an accurate weather forecasting model which incorporatee fuzzy logic

principles:

ii. To develop a model which adapt to changing weather patterns:

iii. To develop algorithm for weather forecasting using ANFIS.

iv. To implement the algorithm in (iii)

3
1.4 Justification of the Study

Weather forecasting system has been selected and approved for research in order to solve the

different challenges faced by the old methods of weather forecasting in Nigeria as the need for

accurate weather prediction increases. This system will therefore improve infrastructure in the

meteorological department.

1.5 Significance of the study

The significance of this project is to offer improved accuracy in predicting weather patterns by

utilizing a dynamic and adaptable fuzzy logic approach.

1.6 Scope and Limitation of the study

The project entails the creation of a platform through which the Yandev area of the Gboko local

government in Benue State will have its regional data parameters collected and a sophisticated

method of weather prediction will be applied. Only local weather data from Benue State's

Apkeran Orshi College of Agriculture Agro-Meteorological Station in Yandev, Gboko local

government, is used in the study. In addition, a lot of variables affect how a storm forms and how

the atmosphere interacts with the planet. On the other hand, computerized weather predicting has

a far higher accuracy rate, but it also has limitations because the researcher is unable to see every

possible component that affects the weather.

4
 CHAPTER TWO

REVIEW OF RELATED LITERATURE

2.1 Historical Background of Weather Forecasting

There has been a surge in climate change-induced extreme weather events such as floods,

droughts, and storms in the past decade (2010—2020), particularly in tropical regions such as

sub-Saharan Africa (Codjoe and Atiglo, 2020; Dube et al., 2021). The effects of such extreme

events are more profound in Africa due to limited capacity to manage risks, weaker revenue

capability, and lower institutional ability to upgrade infrastructure (Kayaga et al., 2013; Horne et

al., 2018). Extreme events can result in the loss of human lives and livelihoods, decreased crop

productivity, destruction of electricity distribution infrastructure, and interruption to water

supplies (Connolly-Boutin and Smit, 2016; Curtis et al., 2017; Ife-Adediran and Aboyewa, 2020;

Intergovernmental Panel on Climate Change, 2021).

One way to minimize the impact of extreme weather is by improving weather forecasting and

better communication of weather warnings. The ability to provide timely and precise weather

forecasts offers the potential to reduce the vulnerability of people to the impacts of extreme

weather (Singh et al., 2018; Alley et al., 2019; Nost, 2019; Antwi-Agyei et al., 2021a). However,

accurate forecasting of weather variables remains a major challenge for the scientific community

(Mani and Mukherjee, 2016), especially in tropical regions. The need to safeguard lives and

livelihoods from extreme weather events has made the provision of timely weather forecasts an

essential adaptation initiative (Oyekale, 2015). Improving the weather forecast’s accuracy is

crucial for people’s safety and protecting key economic sectors, including agriculture, aviation,

water, energy, and emergency response (Parker et al., 2021). In addition, availability,

accessibility, and usability of accurate weather information are key for climate-informed

decision-making (Dinku et al., 2014; Mabe et al., 2014; Nkiaka et al., 2019). As weather

5
forecasts become progressively more accurate and timelier, their use can inform climate change

risk mitigation strategies. However, the benefits linked to these improvements will only be

appreciated if the forecasts are used directly in management decisions (Blum and Miller, 2019).

There is an urgent need to assess the forecasting of high-impact weather in sub-Saharan Africa to

minimize the devastating effects of extreme events on the economy and livelihoods (Woodhams

et al., 2018; Carter et al., 2019). With improved warning systems, it is possible to reduce the

extent of damage by improving the resilience of at-risk populations and reinforcing the

preparedness of the population to cope with extreme events (Anaman et al., 2017). The

production and use of improved weather information to deliver early warnings and adequately

manage these extreme events are essential in societal preparedness against extreme weather

events (Anaman et al., 2017). Studies have highlighted the importance of weather and climate

services in mitigating risks such as drought, flooding, and loss of lives and livelihoods (Mabe et

al., 2014; Sena et al 2014; Oyekale, 2015; Mittal, 2016; Guido et al., 2021; Kayaga et al., 2021)

These studies have considered the effects of extreme weather events on agriculture, health,

water, and energy sectors in South America, Asia, and Africa. Despite the plethora of

information on climate services, the effect of observations of climate change on using weather

forecasts has not been investigated in depth in East and West Africa regions (Seo, 2014; Nkiaka

et al., 2019). Studies have not synthesized all of the evidence on the contribution of weather and

climate information in Africa into a single study to enhance decision-making. The choice of East

and West Africa was influenced by the geographic and socioeconomic characteristics, which

make these regions most vulnerable to climate change. In addition, gaps remain between the

production of climate services, its use in decision-making, and the societal benefits derived from

climate services (Webber, 2019; Harvey et al., 2021).

6
2.2 Review of Previous Works on Weather Forecast

Timely and precise weather information could reduce the exposure of people to climate change

and their risks. In this project, we conduct a systematic review to harmonize the existing

evidence on weather information services in Benue and Nigeria, and we point out the most

important areas for future research. The review identified relevant peer publications using

Scopus database and Science.

This revealed that weather information studies focus more on weather information, mainly

information related to rainfall and temperature, and the information was accessed through radio,

mobile phones, and television. Most of the information provided focused on generic forecasts

instead of tailored impact-based forecasts. Only very few users can access, or benefit from the

information produced due to poor communication and technical understanding of weather

information. In addition, a lack of downscaled information, logistics, and trust hinders the uptake

and use of climate information. Consequently, mainstreaming capacity-building of key

stakeholders is required to promote effective adoption and strengthening of climate information

services across Benue and Nigeria. (Thomas Peprah Agyekum, Philip Antwi-Agyei and Andrew

J. Dougill)

A model for analyzing meteorological data using data mining techniques was presented by

(Kohail and El-Halees 2011). The authors tried to extract useful knowledge from weather daily

historical data collected at Gaza strip city during the duration of 9 years period (1977-1985).

Daily average relative humidity, average temperature, wind speed, wind direction, the time of the

highest wind speed and rainfall are the parameters considered. Time series analysis is used to

identifying temperature and linear interpolation method is used to fill missing values. Outliers

are removed by using outlier analysis. The authors use cluster analysis to partitioning data by

7
using K-means algorithm at K=4. ANN and least median squares linear regression method are

the prediction methods. Classification and association rules are also used.

A methodology for weather forecasting using incremental K-means clustering was also presented

by (Chakraborty et al 2011). The weather forecasting is done based on the incremental air

pollution data base of west Bengal in the years of 2009 and 2010. In this approach, the authors

proposed a generic methodology for time series forecasting through clustering. The air pollution

data is collected from “West Bengal Air Pollution Control Board.” This data base consists of 4

air pollution elements or attributes. They are carbon dioxide (CO 2), Respirable Particulate Matter

(RPM), Sulphur dioxide (SO2) and Oxides of Nitrogen (NOx). K-means is applied for initial data

and incremental K-means is applied to the new coming data by using Manhattan metric measure.

presented A data mining technique for wind speed analysis was presented (Al-Roby and El-

Halees 2011). The data recorded between 2004 to November 2006 daily historical data by

meteorological station of Gaza is considered. After preprocessing the data, data mining

techniques, association rules, classification, cluster, outlier analysis were applied and used. The

observed data of wind speed contain 4 years from January 2003 to November 2006 daily

historical data. To normalize the data, Normalization is applied.

A methodology proposed for measuring weather prediction accuracy using sugeno based

adaptive Neuro Fuzzy Inference System, Grid partitioning and Guassmf was presented (Answer

et al 2011). These authors solved weather event puzzle for known industrial city Pakistan,

Sialkot by implementing a fuzzy rule based system using sugeno Fuzzy Inference during the

period of 7 years from January 2003 to December 2010. Two separate experimental settings

were used to developing a Fuzzy Inference System, the first experimental dataset consisted of

2100 instances with 14 inputs and 5 weather events while the second dataset also consisted of

8
2100 instances but 6 input parameters. Adaptive Neuro Fuzzy Inference System (ANFIS) model

was used for fuzzy inference system.

(Ghosh et al. 2011) presented a methodology for weather Data mining using artificial neural

network. The main focused was based on predictive data mining which they have extracted

interesting patterns or knowledge from huge amount of meteorological data. They used back

propagation neural (BPN) network for initial modeling. The results obtained by BPN model are

fed to a Hopfield Network. The performance of this approach is tested on three years weather

dataset comprises 15000 records containing attributes like temperature, humidity and wind

speed. Data cleaning, relevance analysis, data transformation and reduction are the preprocessing

steps applied to the data to improve accuracy and efficiency. (Torabi and Hashemi 2012)

presented an approach for short-term energy consumption to forecast weather by using data

mining. This approach is used to predict ranging from hourly electric energy usage to weather

data in a predefined period of time. The authors used Neural Networks and Support Vector

Machine (SVM) methods to extract energy consumption patterns. The authors used three

different kinds of data. Those are Hourly electric energy usage for each day, Zero power

consumption data and weather data for each hour. In data cleaning process missing values

missing values of electric energy consumption is calculated as the hourly average value and for

everyday Load patterns are described through charts. ANN has capability of identifying patterns

in unsupervised clustering. SVM is based on statistical learning theory and it improves

performance. Data mining techniques such as decision trees, fuzzy logic, artificial neural

networks, rule based techniques and some other techniques for the climate prediction. The result

also showed neural network based algorithms are performed well when compared to other data

mining techniques. In order to improve the presentation of neural network algorithms, other

statistical based feature selection techniques and fuzzy techniques can be incorporated to achieve

better predictability (Kaur 2012). The procedure for estimating minimum, maximum and average

9
daily temperature values using data mining techniques was conducted with experiments with

regression algorithms using temperature data from the city of Patras in Greece. Data was

collected from meteorological station of the Laboratory of Energy and Environmental Physics of

the Department of Physics of University of Patras during the period 2002-2005. Classical linear

least-squares regression, model trees and MS rules algorithms were used to predict the data. The

K-nearest neighbor learning algorithm for grouping the data and to estimate the values of the

weights of a neural network was used. They presented the regression method in predicting

minimum daily temperature values as input (Kotsiantis et al 2012).The use of data mining

techniques in forecasting temperature, rainfall, evaporation, wind speed and the data was

collected between 2000 and 2009 from the city of Ibadan, Nigeria where artificial neural

networks and decision trees were used to meteorological data and to fill the missing values, data

cleaning techniques were also used. The evaluation metrics such as co-relation co-efficient mean

squared error, the mean-squared error to select the algorithms, parameters and to predict the

values of weather forecasting were put to use (Olaiya, 2012). A different model for interpreting

the weather using k-means clustering attempt is made to analyze the temperature at atmospheric

pressures with 400hpa, 500hpa, 700hpa. The experimentations are also followed by varying the

humidity and for presenting the model the weather data set of Bangladesh is considered (Kavita

2012). A time series forecasting methodology through clustering was implemented and used

where subset of the data set to build up the system model by compressing the information

through clustering and coming up with inherent patterns for data. The pattern was represented as

the curves such that any time series from the given set is expected to follow. The authors used

linear regression method by matching to the closest pattern to each time series that has to be

predicted (Vipul et al 2012).(Nazeri and Zhang 2012) also presented a methodology to analyze

severe weather impacts on National Airspace System (NAS) performance by using data mining

applications. The framework of this approach consists of three phases: data preparation, feature

extraction and data mining. For this experiment, three sources were used: Airline Service Quality

10
Performance (ASQP), Enhanced Traffic Management System (ETMS) and National Convective

Weather Forecast (NCWF) supplied by National Center for Atmosphere Research from April to

September in 2000 to represent the severe weather season. Complete weather data were used

which was captured within 152 days . The feature extraction phase consists of four steps in this

work: image segmentation, weather feature extraction, air traffic feature extraction and

representation conversion. Segmentation algorithm was used to identify the severe weather areas

and considered correlation analysis to calculate the eight weather and air traffic features with the

performance. They applied clustering to group the days with similar weather impacts on NAS

performance. K-means clustering algorithm was also applied. The classification method to

partition the performance measure into three classes: bad, medium and good was used. An

approach to analyze and predict data of temperature, humidity values for future using clustering

technique was introduced. Here, the outlier analysis is considered to detect the outliers with

respect to the data and clustering analysis used in partitioning the data based on the similarity of

the objects. K-nearest neighbor (K-NN) algorithm is used to predict the values of temperature

and humidity parameters of climate. To find the distances between data sets in the K-NN

algorithm absolute distance and Euclidean distance measuring are used (Badhiye et al 2012). An

efficient clustering method that classifies cities based on atmospheric conditions like

temperature, pressure, wind speed and humidity was also presented using algorithms to predict.

The data represents the month wise atmospheric conditions under two parameters namely

temperature and pressure; and collected from the ten different cities namely Delhi, Kolkata,

Bhopal, Mumbai, Jaipur, Amritsar, Cochin, Luck now, Bhubaneshwar, Guwahati (Singh et al

2012).

In 2013, (Kalyankar and Alaspurkar) presented a data mining technique which is used to analyze

the meteorological data. Here the weather data is extracted to find hidden patterns inside a huge

dataset by using clustering technique i.e., K-means partitioning algorithm. 10 channel midi-data

11
logger systems that provide weather data in form of excel sheets and can automatically collect

data on a 24-hour basis was used. A proposed model for predicting the weather data based on

classification technique and considered several attributes such as wind pressure, humidity, vapor,

wind speed and cynical results obtained. The results showcased good accuracy by correlating the

above parameters (Valmik and Meshram 2013). Works on k-nearest neighbor classification are

utilized to estimate the wind speed by using the Manhattan and miniskowski distance, the speed

of wind is estimated. The authors (Mehmet et al 2013), have also analyzed these works using

three parameters wind direction, air temperature and atmospheric pressure.(Mayilvahanan and

Sabitha 2013) approach for estimating the availability of sunshine using data mining techniques

deals with estimating the temperature values of four cities in Tamil Nadu, South India namely

Chennai, Coimbatore, Madurai and Kanyakumari. The dataset was collected from the India

Meteorological Department portal. The dataset consists of monthly mean maximum, minimum

temperature and total rainfall based upon 1901-2000 data of various cities in India. The attributes

present in the data are station name, month, period(years), number of years, maximum mean

temperature in degrees Celsius, minimum mean temperature Celsius and mean rainfall in

millimeters. Data preprocessing was applied to the data to clean and improve the quality of data.

K-Means clustering algorithm and Expectation Maximization algorithms were used to compare

the availability of sunshine in the cities. The authors claimed that the maximum amount of

sunshine is recorded in the city of Chennai when compared to the other cities.The use of

Artificial Neural Network in data mining to forecast weather was also used by (Sawale and

Gupta 2013). A neural-network based algorithm was used to predict the atmosphere for a future

time. They used Back Propagation Neural (BPN) Network for initial modeling, afterwards the

results obtained by BPN model are fed to a hop field network. The performance of the proposed

ANN-based method tested on 8 years weather dataset comprising 15000 records containing

attributes like temperature, humidity and wind speed. The focus is based on predictive data

mining by which they can extract interesting patterns or knowledge from huge amount of

12
meteorological data. The use Data cleaning to fill the missing values and remove noisy data was

employed. Relevance analysis can be used to detect attributes that do not contribute to the

classification or prediction task. Normalization is applied to the data scaling all values the

performance of the Back Propagation Network and Hope field Network Model was satisfactory

as there were not sustain number of errors.

In 2014 (A1-Matarneh et al) developed a temperature-based weather forecasting model using

neural networks and fuzzy logic, a proposed computer based models for weather forecasting

based on the temperature to predict the daily temperature using two techniques, artificial neural

networks and fuzzy logic was employed. The main from aim from study was to develop different

weather forecasting models based on two techniques over different regions. The models had

established on two different regions, Amman airport and Taipeil China. The Back propagation

algorithm was used in artificial neural networks technique. They evaluate the efficiency of

weather forecasting models using two measures, Variance Accounted for (VAF), and Mean

Absolute Error (MAE). The application of cluster analyzes technique to the verification of

quantitative precipitation forecasts. Clustering analyzes is adopted as an objective method to

create groups of rain gauges displaying interrelated measurements. Most of the events in this

work are considered with an upper-level throw and a surface level low approaching northern

Italy from the west. The authors detected the cluster by using the average linkage algorithm. The

results of the verification of the precipitation forecasts are highly affected by distribution of rain

gauges (Gutierrez et al 2014).

Weather forecasting from satellite image sequences using picture fuzzy clustering and

spatiotemporal comprises of detailed description of the current weather along with forecasts

obtained by the extrapolation for every short-range period of zero to six hours ahead (Hoa et al

2014). A semi supervised technique for weather condition prediction using DBSCAN and K-

Nearest Neighbor rule (KNN) was presented and in this work, certain atmospheric parameters

13
are taken for four years on a day wise basis in a certain city where weather attributes such as

snow, rain, fog, are considered. They proposed a new approach that combines data mining

technologies like classification and clustering for the semi supervised clustering algorithm to

identify clusters in large spatial data sets Sharma et al (2014).

Rajashekar and Rajinikanth (2014) Weather analyzers of Guntur district of Andhra region using

SVM (Support Vector Machine) data mining techniques developed a machine learning algorithm

that is hybrid SVM (Support Vector Machine) model. The model is used for effective weather

prediction by analyzing the given weather data and to recognize the patterns exiting it. An

application was presented using clustering methods for statistical down scaling in short range

weather forecasts. Here, the clustering technique compared with the standard nearest neighbors

analogue methods. Some validation results daily precipitation and maximum wind speed

operative downscaling (lead time 1 to 5 days) on a network of hundred stations in the Iberian

peninsula for the period 1998 to 1999 are collected as data with self-Organizing maps (SOMS) to

define weather classes and study the associated distributions of local precipitation (Gutierrez et

al 2014).The fluctuating wind series are modeled by (Trombe and Pinson2014) using the time

series analysis together with Markovian switching model and thereby a methodology is carried

out for estimating the weather changes in the sea using radars. The classification is used to find

the relationship between precipitation and wind attributes. A renewable energy source

optimization model is proposed by (Iqbal et al 2014) for classifying the wind sources using

optimization classification algorithm. The works are further extended (Vibav 2015) by

integrating clustering and classification by which wind speed evaporation is calculated and

minimum and maximum temperatures can be obtained with ease. K-means algorithm is utilized

for clustering the data.

An adaptive neuro fuzzy approach on the estimation of wind turbine effect in adaptive neuro

fuzzy inference system (ANFIS) is designed and adopted to estimate the wind form wake effect

14
in wind form according to the position of and distance of wind turbine between turbines in the

wind form and router radius A1-Shammari et al (2015).

Gokila et al (2015) presented another approach based on clustering and classification in support

of climatology to mine weather data using clustering techniques applied on the climate data helps

to produce similar patterns of climate with the consideration of spatial nature. The classification

techniques are used to relate the attributes of weather data to predict the future climate.

2.3 Conclusion of Literature Review

From the above review, it is clear that there is a need for much better approach to weather

forecasting which can handle weather parameters more intelligently rather that a crisp theory.

Therefore, it is concluded that soft computing techniques especially Neural Network and Fuzzy

System has become interesting preference for researchers to solve weather forecasting problems.

15
 CHAPTER THREE

RESEARCH METHODOLOGY

3.1 Introduction

The main goal is to increase the accuracy of weather forecasting by utilizing the capabilities of

the Adaptive Network-Based Fuzzy Inference System (ANFIS). As input for this process,

meteorological data including pressure, humidity, wind direction, speed, and temperature must

be gathered.

The ANFIS framework is primarily used to build fuzzy logic models that can capture the

numerous interactions between various aspects in metrology because of its capacity to adapt and

learn from data. The model will be trained using historical weather data and related observed

outcomes. This will allow the ANFIS system to dynamically modify its fuzzy rules and

membership functions, reflecting variations and trends in the underlying atmospheric processes.

3.2 Description and Analysis of Existing System

In order to handle imprecise and uncertain information and produce more accurate and flexible

weather predictions, a fuzzy system can handle complex and non-linear data relationships, such

as those involving local terrain and microclimates, which the existing weather forecasting system

frequently finds difficult to incorporate and analyze.

3.3 Description of the Existing System

The current weather forecasting system creates predictions of future weather conditions using

real-time observations from weather stations and satellites, historical weather data, and computer

models. This aid in helping people and organizations plan ahead and get ready for a variety of

atmospheric occurrences.

16
 3.4 Analysis of the Existing System

1. Uncertainty Handling: The system struggles with effectively managing uncertainties and

complex interactions in the atmosphere, which can lead to inaccuracies, especially in long-

term predictions.

2. Local Variability: It may overlook local terrain and microclimate effects, resulting in less

accurate forecasts for specific locations.

3. Extreme Events: Predicting extreme weather events like hurricanes and tornadoes remains

challenging, and there is room for improvement in accuracy and lead-time.

4. Data Gaps: In some regions, the availability of weather data can be limited or sparse,

impacting forecast quality.

5. Continuous Improvement: The existing system requires ongoing calibration and

improvement to keep up with evolving meteorological science and technology.

3.5 Limitations of the Existing System

The present weather forecasting system, though functional, is not without its limitations. The

observations which call for modification in the way this whole system will be changed are as

follows:

i. Lack of Real-time Updates:

ii. Handling Uncertainty

iii. Climate Change Adaptation

3.6 Analysis of the Proposed System

A weather forecasting system which is proposed using Adaptive Network-Based Fuzzy Inference

System (ANFIS) offers several advantages:

17
 i. Improved Accuracy

ii. Early Warning for Extreme Events:

iii. Uncertainty Handling

iv. User-Friendly Outputs

3.7 Design goals for the Proposed System

The architecture, components and modules from now, along with different interfaces of those

components and the data that goes through the proposed system will be considered. The design

goals are listed below:

i. Ensure that weather forecasts are easily accessible through various platforms and devices,

including websites, mobile apps, and automated alert systems.

ii. Ensure seamless integration of the ANFIS-based forecasting system with existing

meteorological infrastructure, data sources, and information dissemination channels.

iii. Design an intuitive and user-friendly interface for accessing and interpreting weather

forecasts, catering to a broad range of users, including the general public, researchers,

and professionals.

iv. Design the system to be adaptable to evolving technologies, data sources, and climate

change scenarios.

v. Provide probabilistic forecasts that convey the likelihood of different weather scenarios,

enabling users to assess risks and uncertainties associated with each forecast.

3.8 System Requirements

The implemented system is intended to support algorithm of accurate weather forecast. The

implementation attribute include:

18
 i. Interface Enhancements: The forecasting interface will provide an enhancement thus

will help the end-user to operate the system easier.

ii. Error Management: Managing and handling of errors will be carried out directly by the

system and will be invisible to the user to make it bug free.

3.9 System Specification

Specification gives the actual requirement for an optimal use of the system. For optimal

performance, the new system will require such specifications.

3.10 Software Requirement

The system will perform well on Windows, Mac OS X, and Linux operating systems. Since the

system was designed using MATLAB, differences may occur among the operating system

platforms for the following reasons:

 MathWorks builds its products with a different compiler on each platform, and each has

its own performance characteristics.

 MathWorks incorporates third-party libraries into its products that may perform

differently on each platform.

 The operating systems carryout functions differently, especially in the situation of disk or

graphics intensive operations.

3.1.1 Hardware Requirements

The Standardization of client hardware is very important. The designed system will work well on

the x86-platform. The advantages of using the x86-platform include:

 The working frequency of more than 1 GHz are met by x86 processors,

19
  Graphics adapters for the x86-platform supports major graphics standards in meteorology

(e.g., OpenGL).

Also, being implemented with MATLAB, every component in a typical computer configuration

has an impact on the new system as stated below:

i. Central Processing Unit (CPU): The dependency of MATLAB performance is the

presence of floating-point hardware. On many CPUs, the number of Floating-Point Units

(FPUs) equals the number of CPU cores. However, some processors, a single FPU may

be shared between multiple CPU cores, potentially creating a performance bottleneck.

ii. Hard disk: The hard disk speed is a vital factor in the start-up time of the system, to

improve the start-up time of system, it is required to take advantage of technologies such

as solid-state drives or Redundant Array of Inexpensive (or Independent) Disks (RAID).

iii. Graphics Processing Unit (GPU) for display: Graphics are offered using OpenGL

technology. A graphics card with superior OpenGL support can perform faster than a

lesser card. The best visual appearance and robustness are recommended to be used with

Up-to-date drivers.

iv. Graphics Processing Unit (GPU) for computation: To speed up computation, Parallel

Computing Toolbox leverages NVIDIA GPUs with compute capability 2.0 or higher.

3.1.2 Design Techniques (Choice of Programming Language)

Using MATLAB, the Adaptive Network Based fuzzy Inference System (ANFIS) technique is

used to design the new system. A high-level language and interactive environment for numerical

computing, programming, and visualization is called MATLAB (MATrix LABoratory). You can

build models and applications, analyze data, and construct algorithms with MATLAB. Unlike

with spreadsheets or conventional programming languages like C/C++ or Java, you can explore

20
several strategies and arrive at a solution more quickly using the language, tools, and built-in

math functions offered.

3.1.3 Use Case Diagram

A use case diagram is a visual representation of how a user and a system interact. It describes

how a system functions from the perspective of an outside observer. The function of a system is

prioritized over its method of operation. They have a tight relationship to scenarios. A scenario is

an illustration of how a user interacts with the system.

Login

Select Location

User
Check Weather Report

Figure 3.7 The Use Case diagram of the Software

21
 CHAPTER FOUR

RESULTS AND DISCUSSION

4.1 INTRODUCTION

This chapter presents the implementation and testing overview of the application: it starts with

the program implementation details of the system and then proceeds to give a more top –level

view of the application and subsequently wraps up with steps and requirements for deploying the

application.

4.2 System Requirements

The diagrams below depict the system and representations, or images of the ANFISM in

weather forecasting system that displays the outcome of the developed system.

4.3 Adaptive Neuro-Fuzzy Inference System (ANFIS)

Fuzzy logic (FL) and artificial neural network (ANN) principles are integrated in the

development of ANFIS. Usually, the goal of this is to enhance the effectiveness of the two

distinct methods that were previously discussed. Because of its capacity for learning, it tends to

solve nonlinear functions and correlates to the fuzzy logic principles of IF-THEN. It is referred

to as the generic algorithms for assessing various inputs. As seen in Figure 4.1 below, just two

inputs will be taken into account for this work.

22
 Figure 4.1 Block diagram of the ANFIS structure.

The input is Multiplexed and first sent into the ANFIS with rules developed by the expert.

Secondly, rules in the ANFIS act on the input and produce an output (forecasted).

4.4 The Study Area

Yandev the study area is a community in Gboko town, Gboko local government area which has a

land area of 4,493 sq. Km located between Latitudes 07 0 08’ 16” and 070 31’ 58”, and

Longitudes 080 37’ 46” and 090 10’ 31” is located at an elevation of 315 meters above sea level.

The area is located within a sub-humid tropical region.

4.5 Method of Data Collection and Presentation

Numerous factors that influence the weather either directly or indirectly can be used to

characterize weather forecasting. Maximum temperature (MAX TEMP°C), minimum

temperature (MIN TEMP°C), actual vapour pressure (AVP mm M), actual vapour pressure

(AVP mm E), relative humidity (RH %) M, relative humidity (%) E, wind speed (WS Km/h),

bright sunshine hours, pan evaporation (mm), and local monsoonal precipitation (mm) are the

variables used in weather forecasting.

23
Figure 4.2 below shows the three sets of data that were gathered from the meteorological station

of Akperan Orshi College of Agriculture Yandev, Gboko: temperature (T), relative humidity (H),

and rainfall (R);

Figure 4.2 Rainfall, temperature and related humidity.

The relationship between input (Temperature and Relative humidity) and output data (Rainfall)

on ANFIS interface is shown in Fig. 3.3 below:

Figure 4.3 ANFIS input/output interface

24
4.6 Fuzzification

As was previously noted, ANFIS is the result of combining FL and ANN. The input/output data

fuzziness must be taken into consideration first. In order for the prediction to come true, the input

and output are often fuzzified depending on the data range (maximum and minimum).

4.7 Classification of Input/Output Data and Membership Function

T and H can be classified as input and R as the output. There are several types of membership

functions (MFs) but in this work generic MF (Bell) was adopted because of its suitability to

handle all kinds of data. It consists of a vector say x and other nonlinear parameters a, b, c,

referred to as antecedents of the ANFIS rules given by

(1)

a, b, c change the generic bell shape if altered accordingly and x can be T, H or R.

Figure 4.4 ANFIS network model

Table 4.1 below compares the components of an adaptive network based fuzzy inference system

(ANFIS) in the block diagram (in Figure 1) with a network model of ANFIS (in Figure 4)

25
Table 4.1 Comparison of block diagram and equivalent network model of ANFIS

Figure 3.1 Figure 3.4

Inputs L1

Multiplexing L2-L3

ANFIS/Rules L4

Output L5

Table 4.2 Linguistic terms and intervals of Temperature

Linguistic terms Intervals a Intervals Intervals c

T1 0.936 2.001 19.00

T2 0.910 2.011 20.87

T3 0.948 2.007 22.82

T4 0.975 1.978 24.62

T5 0.955 1.998 26.49

Input Variables of Temperature

26
 Table 4.3 Linguistic terms and intervals of Relative humidity

Linguistic terms Intervals a Intervals Intervals c

H1 3.125 2.001 32.00

H2 3.105 1.999 38.20

H3 3.130 2.011 44.49

H4 3.126 1.999 50.75

H5 3.125 2.000 57.00

Input variables of Relative Humidity

Table 4.4 Linguistic terms and intervals of Rainfall

Linguistic Terms Intervals

R1 508.10

R2 23.10

R3 100.20

R4 2.64

R5 3.94

Output variables Rainfall

Tables 4.2, 4.3 and 4.4 above present the linguistic terms and the intervals of the membership

function of T, H and R.

27
4.8 Formulation of Rules and Implementation of ANFIS Simulation Model

Since there are two inputs and one output, the ANFIS network model is developed using tables

4.2, 4.3, 4.4 above.

Figure 4.5 and 4.6 present the membership function of the input parameters that correspond to

membership function intervals.

Figure 4.5 Membership of the temperature.

28
 Figure 4.6 Membership function of Related Humidity.

ANFIS equivalent two input networks a 1, a2, b2 and b2 represent the fuzzy sets. The membership

function for a and b is defined by the generalized Bell membership function given in Eq. (1). In

correspondence with Figure 4.1, L1 represents the inputs, L2-L3 represent multiplexer, L4

represents ANFIS/rules and L5 represents the output as shown earlier in Table 4.1.

Based on this network model shown in Figure 3.2, rules are developed.

First rule: IF T is a1 and H is b1, THEN

(2)

Second rule: IF T is a2 and H is b2, THEN

(3)

The average weight is obtained as given in below

(4)

29
L1: Every node in L1 auto-generates a member grade of linguistic value a1, a2, b1, and b2 and

these are used to define the MF as given in equation 5 and 6 called nonlinear functions.

(5)

(6)

Where ai or bi-2 is a linguistic term sometimes referred to as HIGH, LOW or MODERATE and T

or H is the input at the ith node.

L2: All the nodes in this layer are labeled as π, which multiply the two incoming signals. For

example, a1b1 or a1b2 multiplexed, and the firing strength of a rule may be expressed as in the

equation below,

(7)

L3: Each node in this layer is labeled N. ith node determines the ith rule firing strength as given

below

(8)

L4: Here each node is an adaptive node as given in the equation below;

(9)

Where wi is the output of L3 while, p1, q1 and r1 are the antecedents.

L5: The final consequence is obtained as

(10)

4.9 Error analysis

30
Equations 11 and 12, which provide the Mean Absolute Percentage Error (MAPE) and

coefficient of correlation, respectively, are used to calculate the prediction accuracy.

(11)

(12)

(13)

where n is the number of days in the month, R is the amount of rainfall, T or H is the temperature

or humidity, and A and F are the actual and predicted rainfall, respectively.

4.10 Result

4.1.1 Simulated result

To replicate the network shown in Figure 3.6, one randomly chosen sample of data is used. T =

25.5 °C and H = 46%, for example. It yielded R = 3.8 mm, as seen in Figure 4.2, and figure 4.1

below illustrates how to derive the remaining values of R.

31
Figure 4.5 Simulated ANFIS network model

32
 Figure 4.6 Rule View

With a matching efficiency of 95.36%, the MAPE is determined to be 4.64%. Temperature and

rainfall have a positive correlation with correlation coefficient (r) = 0.1277, whereas humidity

and rainfall have a positive correlation with (r) = 0.075. R often lies in the range of -1 to +1. The

link between the two variables under investigation is positive as r gets closer to +1. When r is

negative, it implies a negative correlation between the variables, and when it is 0, it indicates no

correlation at all.

33
 CHAPTER FIVE

SUMMARY, CONCLUSION AND RECOMMENDATION

5.1 Summary

The primary goal of weather forecasting is to predict the weather and deliver important

information for a specific future date. There are many different methods for predicting the

weather, ranging from simple sky observation to more intricate computer and mathematical

models. Predicting the weather is also crucial for a number of other purposes. Because the

atmosphere is dynamic, precisely predicting weather conditions is a challenging undertaking.

Several variables can be used to depict the weather at any given time. The most important

variables are those that are chosen and used in the prediction process. The variables that have

been chosen are location-dependent, meaning that the ranges of each variable are constantly

subject to change.

Nigeria's standard meteorological stations are insufficient to deliver reliable climate data to the

general population. Furthermore, the lack of professional meteorologists delays the development

of accurate meteorological forecasts. To overcome these difficulties, a novel strategy is put out

and developed that makes use of the fuzzy logic notion in meteorological systems. An adaptive

neural network system, which learns patterns from historical data, is one way to comprehend this

kind of inconsistent situation until it reaches the necessary degree of training error. The Adaptive

Neuro Fuzzy Inference System (ANFIS), a mix of a neural network and a fuzzy inference system

technology, makes this possible. This combines fuzzy logic techniques with artificial neural

networks (ANN). By efficiently utilizing the learning capacity of ANN for automatic fuzzy rule

generation and parameter optimization, the fundamental issue in fuzzy system design—defining

the membership function parameters and designing fuzzy if–then rules—is resolved (Bacanli et

al. 2009).

34
Numerical weather models are used in computer-aided atmospheric simulations. These models

take analytical data as a starting point and apply physics and fluid dynamics knowledge to evolve

the atmosphere's current condition over time. Supercomputers are needed to solve the equation

governing how a fluid's condition changes over time; the model's output serves as the foundation

for weather forecasts.

5.2 Conclusion

Researchers are finding that using soft computing techniques to handle weather forecasting

difficulties is an exciting trend in weather forecasting. For weather forecasting, an attempt has

been made to create a hybrid algorithm based on the fusion of the powers of two algorithms, the

fuzzy system and the neural network. In order to predict rainfall, an adaptive network based

fuzzy inference system (ANFIS) model is created utilizing temperature and relative humidity.

Other weather factors can also be predicted using the same approach. Forecasting that is

accurate, timely, and well-structured is made possible by the efficiency of the ANFIS model. In

summary, the forecasting technique performs exceptionally well overall in terms of accuracy.

5.3 Recommendation

One of the greatest options for weather forecasting may be to use soft computing. Creating

weather prediction systems using soft computing techniques is an expensive undertaking. In light

of this, the researcher recommends that government funding be allocated to this industry in order

to guarantee enhanced system performance.

5.3.1 Future Works

In the near future, more work can be done to improve algorithm learning (i.e., to employ

improved hybrid duos of the ANFIS). Additionally, more work may be done to modify or, in

35
fact, take multiple types of membership functions along with altered values rather than the

membership functions' default values.

36
 REFERENCES:

A1-Matarnch, L., Sheta, A., Bani-Ahmad, S., Alshaer, J. & Oquily, L. (2014). Development of
Temperature Based Forecasting Models Using Neural Networks And Fuzzy Logic.
International Journal of Multimedia and Ubiquitous Engineering, Vol: 9, No:12,
PP:343-366,2014.

A1-Shammari, E., Amirmojahedi, M., Shamshirband, S., Petkovic, D., Pavlovic, N. T., &
Bonakdari, H. (2015). Estimation of Wind Turbine Wake Effect by Adaptive Neuro Fuzzy
Approach. Flow Measurement and Instrumentation, Volume: 45, PP: 1-6, 2015

Acqua, F. D. & Gamba, P. (2000). A simple Model Approach to the problem of meteorological
object Tracking. Volume: 1, Issue No: 7803-6359, PP: 2152-2154, Italy.

AI-Roby, M. A., & El-Halees, A. M. (2011). Data Mining Techniques for Wind Speed Analysis.
Journal of Computer Engineering, ISSN:2010-1619, Vol-2 ,No:1, PP:1-5, 2011.

Answer, N., Abbas, A., Mazhar A., & Hassan, S. (2011). Measuring Weather Prediction
Accuracy using Sugeno based Adaptive Neuro Fuzzy Inference System, Grid Portioning
and Guassmf. International Journal of Computer Science and Applications, Vol-9,
No.17-19, PP:214-219.

Bacanli, U. G., Firat, M. & Dikbas, F. (2009). Adaptive Neuro-Fuzzy Inference System for
Drought Forecasting. Stochastic Environmental Research and Risk Assessment, Volume
23: pp1143-1154.

Badhiye S. S., Chatur P. N., & Wakode, B. V. (2012). Temperature and Humidity Data Analysis
for Future Value Prediction using clustering Technique: An Approach. International
Journal of Emerging Technology and Advanced Engineering, ISSN:2250-2459, volume-
2, Issue-1, PP-88-91, January 2012.

Benzon, H., & Bovith, T. (2008). “Simulation And Prediction Of Weather Radar Clutter Using a
Wave Propagator on High Resolution NWP Data. IEEE Transactions an Antennas and
Propagation, volume: 56, No: 12, PP: 3885-3890, December 2008

Chakraborty, S., Nagwani, N. K., & Lopamudra, D. (2011). Weather Forecasting using
Incremental K-means Clustering. International Conference in High Performance
Architechture and Grid Computing, Vol.169, Part-2, PP:338-341, 2011.

Dawid, A. P. (1984). Statistical theory: the prequential approach (with discussion). J. R. Statist.
Soc.A,147, 278–292.

Diyvankov, O. V., lykov, V. A., & Terekhoff, S. A. (1992). Artificial Neural Networks in Weather
Forecasting. PP: 829-835, February 1st, 1992

Evangelos T., & Evangelidis, G. (2010). The effect of Training Set Selection in Meteorological
Datamining, 14th Panhellenic Conference on Informatics, Volume: 3, Issue No: 7695-
4172, PP: 61-65, 2010.

37
Fallah, G., Gholam, A., Mousavi, B., Mohammad, N & Majid H., (2009). Annual Rainfall
Forecasting by Using Mamdani Fuzzy Inference System. Research Journal of
Environmental Sciences, Vol. 3, No. 4, pp 400-407.

Frisinger, H. H. (1977). The History of Meteorology to 1800, Science History Publications, New
York, 148 pp.

Fuller, R. (2000). Introduction to Neuro-Fuzzy Systems. Advances in Soft Computing Series,

Springer-Verlag, Berlin/Heildelberg.

Ghosh, S., Nag, A., Biswas, D., Prakash, J., & Singh, J. P. (2011). Weather Data Mining using
Artificial Neural Network. International Journal of Computer Science and Applications,
Vol-5, No.6, PP: 192-195, 2011.

Gokila, S., Kumar K. A., & Bharathi, A. (2015). Clustering and Classification in Support of
Climatology to Mine Weather Data-A Review. International Conference Computing and
Intelligence Systems, ISSN: 2278-2397, volume: 04, pages: 1336-1340, March 2015.

Gutierrez, J. M., Cofino, A. S., Cano, A., & Rodriguez, M. A. (2004). Clustering methods for
statistical downscaling in short range weather forecasts. American meteorological
society, volume: 132, PP: 2169-2183, September 2004

Hasan, M., Tsegaye, T., Shi, X., Schaefer, G. & Taylor, G. (2008). Model for predicting rainfall
by fuzzy set theory using USDA scan data, An online Journal. Pp 3-5, 2008

Hoa, N. D., Thong, P. H., & Son, L. H. (2014). Weather Now Casting From Satellite Image
Sequences Using Picture Fuzzy Clustering and Spatio-Temporal Regression.
International Symposium on Geo Informatics for Spatial Infrastructure Development for
Earth and Allied Sciences, PP: 1-6, 2014.

https://en.wikipedia.org/wiki/Weather_forecasting. Accessed on 15th October, 2017, at 4:37pm.

Iloeje, N.P. (1982). A New Geography of Nigeria. London: Longman.

Iqbal, M., Azam, M., Naeem, M., Khwaja, A. S. & Anpalagan. (2014). Optimization classification
algorithms and tools for renewable energy: a review. Renewable and sustainable energy
reviews, volume: 39, PP: 640-654, 2014

Kalyankar, M. A., & Alaspurkar, S. J. (2013). Data Mining Technique to Analyse the
Mateorological Data. International Journal of Advanced Research in Computer science
and Software Engineering, ISSN: 2277-128X, Volume-3, Issue: 2, PP: 114-118, February
2013.

Kanth T. V., Balaram V. S., & Rajashekar, N. (2014). Analysis of Data Sets using Data
Mining Techniques. Computer science and technology, DOI: 10.5121/esit.2014.4510,
PP-No 89-94.

Kaur, G. (2012). Meteorological Data Mining Techniques: A Survey. International Journal of

Emerging Technology and Advanced Engineering, Volume 2, ISSN-2250-2459, Issue-8,
PP 325-327, Auguest 2012.

38
Kavita, P. (2012). Clustering Technique to Interpret Numerical Weather Prediction Output
Products for Forecast of Cloudburst. International Journal of Computer Science and
Information Technologies (IJCSIT), vol-3(1), ISSN:0975-9646, PP. 2996-2999.

Kohail, S. N., & El-Halees, A. M. (2011). Implementation of Data Mining Techniques for
Mateorological Data Analysis, International Journal of Information and Communication
Technology Research, ISSN: 2223-4985, Volume-1, No:3, PP:96-100, July 2011.

Kotsiants, S., Kostoulas, A., Lykoudis, S., Argirou, A., & Menagias, K. (2012). Using Data
Mining Techniques for Estimating Minimum, Maximum and Average Daily Temperature
values. International Journal of Computer, Information, Systems and Control
Engineering, Vol-1, PP: 12-19

Kumar, A., Sinha, R., Verma, D. S., Bhattacherjee, V., Singh, S. (2012). Modeling using K-Means
Clustering Algorithm. First International Conference on Recent Advances in Information
Technology, Vol: 4, Issue-1, Issue No: 4577-0697, PP:1-5.

Liu, J. N., Li, B. N. & Dillon, T. S. (2001). An Improved Naïve Bayesian Classifier Technique
Coupled with a Novel Input Solution Method. IEEE Transactions on Systems, Man and
Cybernetics-Part C: Applications and Reviews, Vol.31, No.2, PP:249-256, May 2001.

Lorenz, E. N. (1969). Three approaches to atmospheric predictability, Bulletin of the American

Meteorological Society,vol. 50, pp. 345–349

Lutgens, F K., & Edward J. T. (1998). Weather analysis and forecasting. In The Atmosphere: An
Introduction to Meteorology, seventh edition, Prentice Hall, Upper Saddle River, New
Jersey, pp. 278-302.

Mayilvahanan, M., & Sabitha, M. (2013). Estimating the Availability of Sunshine using Data
Mining Techniques. 2013 International Conference on Computer Communication and
Informatics (ICCCI-2013), PP:1-4,2013

Mehmet, Y., Sagiroglu, S. & Colak, I. (2013). A new approach to very short term wind speed
prediction using K-nearest neighbor classification. Energy conversion and management,
volume-69, PP: 77-86.

Mohander, M. A., Rehman, S., & Halawani, T. O. (1998). A neural networks approach for wind
speed prediction. Renewable Energy, Volume: 13, PP: 345-354, 1998

Mohandes, M. A., Halawani, T. O., Rehman, S., & Hussaine, A. A. (2004). Support Vector
Machines for Wind Speed Prediction. Renewable Energy, VOL: 29, PP: 939-947, 2004

Naik, A. R., & Pathan, S. K. (2007). Weather Classification and Forecasting using Back
Propagation Feed Forward Neural Network. International Journal of Scientific and
Research Publications, ISSN: 2250-3153, volume-2, Issue-12, PP:1-3, December
2012.No:2, PP-382-386, 2007.

Nazeri, Z., & Zhang, J. (2012). Mining Aviation Data to Understand Impacts of Severe Weather
on Airspace Syatem Performance. Proceedings of the International Conference on
Information Technology: Coding and Computing (ITCC’OL), Volume-1, ISSN: 7695-
1506, PP: 1-6, 2002.

39
Okhakhu, P.A. (2010). The Significance of Climatic Elements in Planning the Urban
Environment of Benin City Nigeria. Unpulished Ph.D Thesis, Department of Geography
and Regional Planning, Ambrose Alli University Ekpoma, Nigeria.

Olaiya, F., & Adeyemo, A. B. (2012). Application of Data Mining Techniques in Weather
Prediction and Climate Change Stuies. International Journal of Information Engineering
and Electronic Business, DOI:10.5815/ijieeb.2012.01.07, pp-51-59, February 2012.

Onwubolu, G. C., Buryan, P., Garimella, S., Ramachandran, V., Buadromo, V., & Abraham, A.
(2007). Self-Organizing Data Mining for Weather Forecasting. IADIS European
Conference Data Mining 2007, ISBN: 978-972-8924-40-9,pp.81-88, 2007.

Ozelkan, E. C., Ni, F., & L. Duckstein, L. (1996): Relationship between monthly atmospheric
circulation patterns and precipitation: Fuzzy logic and regression approaches. Water
Resour. Res., 32, p.p.2097–2103.

Pasero, E., & Moniaci, W. (2004). Artificial Neural Networks for Meteorological Newscast.
“ClMSA 2004 - IEEE lntemanonal Conference on Compumional Intelligence for
Measurement System and Applications”,Boston. MA, USA, pp 14-16

Rajashekar, N., & Rajinikanth, T.V. (2014). Weather Analysis of Guntur District of Andhra
Region Hybrid SVM Data Mining Techniques. International journal of Engineering and
Advanced Technology (IJEAT), ISSN: 2449-8958, Volume-3, issue-4, PP: 133-136, April
2014

Sawale, G. J., & Gupta, S. R. (2013). Use of Artificial Neural Network in Data Mining for
Weather Forecasting. International Journal of Computer Science and Applications,
ISSN: 0974-1011, vol.6, No.2, PP: 383-387, April 2013.

Sharma, A., Chaturvedi, S., & Gour, B. (2014). A Semi-Supervised Technique for Weather
Condition Prediction Using DBSCAN and KNN. International Journal of Computer
Applications (0975-8887), Volume 95, No: 10, PP-21-26, June 2014

Singh, A., Bhupesh Gour, B., Khandelwal, A. & Lackhwani, H. (2012). An Efficient Clustering
Method for Atmospheric Conditions Prediction Using Art Algorithm. International
Journal of Advanced Research in Computer Engineering and Technology, Volume-1,
Issue-1, PP: 12-17, March 2012.

Solomatine, D. P., & Dulal, K. N. (2003). Model Trees as an Alternative to Neural Networks in
Rainfall Runoff Modelling. Hydrological Sciences Journal, Volume-48, No-3, PP: 399-
411, 2003

Tektas M. (2010). Weather forecasting using ANFIS and ARIMA, “A case study for Istanbul.
Environmental Research, Engi-neering and Management”, Vol 1(51); pp 5–10

Torabi, M. & Hashemi, S. (2012). A DataMining Paradigm to Forecast Weather Sensitive Short
Term Energy Consumption. The 16th CSI International Symposium on Artificial
Intelligence and Signal Processing, Volume: 4, Issue No: 4673-1479, PP: 579-584, 2012.

40
Trombe, P. J & Pinson, P. (2014). Automatic classification of offshore wind regimes with
weather radar observations. IEEE journal of selected topics in applied earth
observations and remote sensing, volume: 07, No: 1, PP: 116-125, January 2014

Udo, R.K. (1987). A Comprehensive Geography of West Africa. Ibadan: Heinemann Educational
Books Ltd.

Valmik, B. N., & Meshram, B. B. (2013). Modeling Rainfall Prediction using Data Mining
Method. Fifth International Conference on Computational Intelligence, Modeling and
Simulation, Issue No: 2166-8531, PP:132-136.

Villa, D. A., Augusto, L., Machado, T., Laurent, H & Velaseo, I. (2005). Forecast and
Tracking the Evolution of Cloud Clusters (For TRACC) Using Satellite Infrared Imagery:
Methodology and Validation, Weather and Forecasting, Volume: 23, PP: 233-245.

Vipul K., Vamsidhar T., & Kamalakar K. (2012). Time Series Forecasting Through Clustering-A
Case Study. An online journal. Pp 1-9.

Zahoor, J., Abrar, M., Shariq, B., Anwar M. M. (2008). Seasonal to Inter-Annual Climate
Prediction using Data Mining KNN Technique. IMTIC 2008, CCIS 20, PP: 40-51, 2008.

Zhang, G. P. (2003). Time Series Forecasting Using a Hybrid Arima and Neural Network
Model. Neuro Computing, VOL: 50, PP: 159-175, 2003.

41
 APPENDIX

CODE LISTINGS
import numpy as np
import skfuzzy as fuzz
from skfuzzy import control as ctrl
# Step 1: Data Preprocessing
# Load and preprocess historical weather data (input variables) and observed outcomes (output
variable)
# Step 2: Define Input and Output Variables
input_variable_1 = ctrl.Antecedent(np.arange(min_value, max_value, resolution),
'input_variable_1')
# Define input_variable_2, input_variable_3, ...
output_variable = ctrl.Consequent(np.arange(min_value, max_value, resolution),
'output_variable')

# Step 3: Define Membership Functions

input_variable_1['low'] = fuzz.trimf(input_variable_1.universe, [low_a, low_b, low_c])
# Define membership functions for input_variable_1, input_variable_2, input_variable_3, ...

output_variable['low'] = fuzz.trimf(output_variable.universe, [low_a, low_b, low_c])

# Define membership functions for the output variable

# Step 4: Define Fuzzy Rules

rule1 = ctrl.Rule(input_variable_1['low'] & input_variable_2['medium'],
output_variable['low'])
# Define more rules based on your understanding of weather patterns

# Step 5: Create Fuzzy Control System

fuzzy_ctrl = ctrl.ControlSystem([rule1, rule2, ...])

# Step 6: Simulate ANFIS Model

weather_prediction = ctrl.ControlSystemSimulation(fuzzy_ctrl)
weather_prediction.input['input_variable_1'] = input_value_1
weather_prediction.input['input_variable_2'] = input_value_2
# Set input values for all input variables

# Step 7: Compute Forecast

weather_prediction.compute()

# Step 8: Display Forecast

print("Weather Forecast:", weather_prediction.output['output_variable'])

# Step 9: Evaluate and Compare Results

# Evaluate the accuracy of the ANFIS model using validation and test datasets
# Compare ANFIS results with traditional weather forecasting methods

# Step 10: Draw Conclusions and Further Refinements

# Analyze the performance of the ANFIS model, draw conclusions, and consider refinements

42