Nitte Education Trust

Nitte Meenakshi Institute of Technology

An Autonomous Institution Approved by UGC/AICTE/Govt. of Karnataka
Department of Electronics and Accredited by NBA (Tier-1) and NAAC A+ Grade
Communication Engineering Affiliated to Visvesvaraya Technological University, Belagavi-590018.
Post Box No. 6429, Yelahanka, Bengaluru-560064, Karnataka, India.

WEATHER PREDICTION
USING MACHINE LEARNING MODEL:

AML LA-1 [Mini-project]

4TH SEM ‘A’ SECTION

By
1.RAKSHITHA H - 1NT22EC128
2.SWAPNA R TADALAGI - 1NT23EC415

Under the Guidance of

Dr. Vishwanath V
Assistant professor, ECE Dept.
NMIT
 Abstract:- Weather prediction is a critical component in
a multitude of sectors, including agriculture, disaster Machine learning (ML) offers a data-driven alternative or
management, and day-to-day planning. Traditional complement to these traditional methods. By learning patterns
weather forecasting relies on numerical weather prediction and relationships from historical weather data, ML models can
(NWP) models, which use complex physical equations to make predictions about future weather conditions. This
simulate atmospheric processes. However, these models can approach has gained significant interest due to advancements
be computationally intensive and sometimes fail to capture in data availability, computational power, and machine learning
local weather patterns accurately. In recent years, machine techniques.
learning (ML) has emerged as a promising alternative or
complement to NWP models, offering the potential for - Time Series Analysis: Models like ARIMA (Auto-
more efficient and accurate weather predictions. Regressive Integrated Moving Average), LSTM (Long Short-
Term Memory networks), and Prophet can handle temporal
This study explores the application of various machine dependencies and trends. Key Concepts and Techniques
learning techniques, including regression models, neural
networks, and ensemble methods, for short-term weather 1. Data Collection and Preprocessing:
forecasting. The focus is on predicting key weather - Historical Weather Data: Collection of past weather data
parameters such as temperature, humidity, precipitation, including temperature, humidity, wind speed, precipitation,
and wind speed. We utilize historical weather data, satellite etc.
imagery, and real-time sensor data to train and validate our - Satellite Imagery: Utilization of satellite images for cloud
models. cover, atmospheric moisture, and other relevant features.
- Data Cleaning and Normalization: Handling missing
The results demonstrate that ML models can achieve values, outliers, and normalizing data for consistency.
competitive accuracy with significantly lower
computational costs compared to traditional NWP models. 2. Feature Engineering:
Neural networks, particularly long short-term memory - Temporal Features: Time-based features such as time of day,
(LSTM) networks, show superior performance in capturing day of the week, seasonality, etc.
temporal dependencies and complex patterns in the data. - Spatial Features: Geographic features such as latitude,
Furthermore, ensemble methods combining multiple ML longitude, altitude, proximity to water bodies, etc.
models enhance prediction reliability and robustness. - Derived Features: Combinations or transformations of raw
features that may reveal more about weather patterns.
This research highlights the potential of machine learning
to revolutionize weather forecasting by providing faster, 3. Machine Learning Algorithms:
more precise, and scalable solutions. Future work will focus - Supervised Learning: Models such as linear regression,
on integrating ML-based forecasts with NWP outputs to decision trees, random forests, and neural networks can be used
leverage the strengths of both approaches, aiming to to predict specific weather variables.
improve the overall accuracy and usability of weather
predictions in various applications. - Ensemble Methods: Combining multiple models to improve
prediction accuracy.

Problem Statement: Weather forecasting relies on 4. Evaluation Metrics:

numerical weather prediction (NWP) models that simulate the - Mean Absolute Error (MAE): Measures the average
atmosphere's behavior based on mathematical equations. magnitude of errors in predictions.
However, these models have limitations in representing small- - Root Mean Squared Error (RMSE): Penalizes larger errors
scale features, such as local wind patterns, precipitation, and more than MAE.
atmospheric turbulence. - Accuracy and Precision: Especially relevant for categorical
predictions like rain/no rain.
Objectives: ML can be incredibly effective in enhancing the
5. Deployment and Integration:
accuracy of weather forecasting models. By identifying
- Real-Time Prediction Systems: Implementing ML models
patterns in historical data, ML models can predict weather
in real-time systems for continuous weather monitoring and
events (like storms, temperature changes, and rainfall) with
forecasting.
remarkable precision – even in highly complex and dynamic
- Integration with Traditional Methods: Combining ML
systems.
predictions with NWP models for enhanced accuracy.

Applications and Benefits

I. INTRODUCTION
- Short-Term Forecasting: ML models can provide accurate
Traditional weather forecasting relies heavily on numerical short-term forecasts, which are crucial for daily activities,
weather prediction (NWP) models, which use mathematical agriculture, and disaster management.
representations of the atmosphere based on physical laws and - Climate Research: Analyzing long-term weather patterns and
principles. These models simulate atmospheric conditions over trends to study climate change.
time by solving complex equations related to fluid dynamics, - Extreme Weather Prediction: Early warning systems for
thermodynamics, and other relevant scientific fields. However, severe weather conditions like hurricanes, floods, and storms.
these models can be computationally intensive and sometimes - Energy Sector: Forecasting weather conditions for optimizing
struggle with accuracy, especially for short-term and hyper- renewable energy production and consumption.
local forecasts.
Challenges and Future Directions
II. METHODOLOGY
- Data Quality and Availability: Ensuring high-quality, high- We have created a Machine Learning model which helps in
resolution data is available for training models. predicting hepatitis C virus using Logistic Regression
- Model Interpretability: Understanding and explaining the Algorithm.
predictions made by complex ML models. To develop this ML model, we used Kaggle, where we
- Computational Resources: Managing the computational gathered the necessary dataset.
demands of training and deploying large-scale models.
- Hybrid Approaches: Combining the strengths of NWP and
ML models to create more robust forecasting systems. Here is how we created our ML model:
Fig 5: Train a Linear Regression

STEP 1: Import Libraries

Fig 2: Import libraries

Step 1: Import Necessary Libraries

We import essential libraries: pandas for data manipulation,

matplotlib.pyplot for plotting, and sklearn modules for model
building and evaluation

Fig 1: Workflow of machine learning

1.Data collection and preprocessing: Compile a dataset with

pertinent elements, such as demographic data, medical history,
and test results from labs. By handling missing values,
encoding categorical variables, and normalizing numerical
features, the data are preprocessed.
2.Feature selection: Identify the characteristics that are most
useful and highly connected with hepatitis C. Stepwise
regression, feature importance, and correlation analysis are a Fig 3: Preapare the Dataset
few examples of approaches that can be used for this.
3.Model training : Split training and testing sets from the Step 2: Prepare the Dataset
dataset. Using the chosen features as input variables and the A dataset is created containing dates and corresponding
Hepatitis C status as the target variable, fit a logistic regression temperature values.
model to the training data. To improve the performance of the This data is organized into a pandas DataFrame for easy
model, adjust hyperparameters like regularization strength. handling and processing.
4.Model assessment: Using appropriate evaluation metrics like
accuracy, precision, assess the logistic regression model's
performance. Examine the model's accuracy
in identifying people as positive or negative for hepatitis C
using the testing set.
5.Results interpretation: The logistic regression model's
coefficients should be examined to determine each
characteristic's impact on the risk of having hepatitis C. Using
the size and direction of the coefficients, determine the main
risk factors for the disease.
6.Model implementation and suggestions:
After it has been developed and tested, think about applying the Fig 4: Split Data into Training
logistic regression model in the healthcare industry. Offer
healthcare providers guidance in the form of early intervention Step 3: Split the Data into Training and Testing Sets
methods or targeted screening programmes based on the The features (X) and target variable (y) are defined. Here, the
model's predictions. index of the DataFrame is used as the feature, which is a
simplistic representation.
The data is split into training and testing sets using an 80-20
split. This helps in evaluating the model's performance on
unseen data.

Fig 8: Visualize the Data

Step 7: Visualize the Data and the Regression Line

A scatter plot is created to visualize the actual temperature data
Fig 5: Train a Linear Regression points.
The regression line, representing the model's predictions over
Step 4: Train a Linear Regression Model the range of data, is plotted.
A linear regression model is instantiated and trained using the The predicted temperature for the new date (index 15) is
training data (X_train and y_train). highlighted on the plot.
The model learns the relationship between the index and the The plot includes labels, a legend, and a grid for better
temperature. readability and interpretation.

Fig 6: Evaluate the Model

Step 5: Evaluate the Model

Predictions are made on the testing set (X_test).
Two metrics are used to evaluate the model's performance:
Mean Squared Error (MSE): Measures the average of the
squares of the errors. Lower values indicate better fit.
R² Score: Represents the proportion of variance in the
dependent variable predictable from the independent variable.
Values closer to 1 indicate a better fit.

1. Mean Squared Error (MSE)

- MSE measures the average of the squares of the errors—that
is, the average squared difference between the actual
temperatures and the temperatures predicted by the model.
- A lower MSE indicates a better fit of the model to the data.

2. R² Score
Fig 7: Predict Using the Model
- R² Score (R-squared) is a statistical measure that represents
the proportion of the variance for the dependent variable
Step 6: Predict Using the Model (temperature) that's explained by the independent variable
The trained model is used to predict the temperature for a new (index).
date (index 15 in this case). This demonstrates the model's - An R² value closer to 1 suggests that the model explains a
ability to make future predictions based on learned data. large portion of the variance in the dependent variable,
indicating a good fit.

3. Predicted Temperature
- This is the temperature value predicted by the model for a new
date index (15 in this case). It demonstrates the model's ability
to extrapolate and make predictions on data points outside the
initial training range.
Graph Explanation

Scatter Plot of Actual Data

- Blue Dots: These represent the actual temperature values for
each date index. This visualizes the original data points used to
train and test the model.

Regression Line
- Red Line: This line represents the linear relationship that the
model has learned from the data. It shows the predicted
temperature values over the range of date indices used in the
dataset.
- The slope and intercept of this line are determined by the
linear regression model to minimize the error between the
actual and predicted values.

Highlighted Predicted Point

- Green Dot: This point represents the temperature predicted by Fig 9: Data Flow Diagram
the model for the new date index (15). It highlights the model's
prediction capability beyond the initially provided dataset.

Plot Elements
- Title: "Linear Regression Model for Temperature" describes
what the graph represents.
- X-axis Label: "Index" indicates the independent variable,
which in this simplified model is just the index of the dates.
- Y-axis Label: "Temperature (Celsius)" shows the dependent
variable, which is the temperature.
- Legend: Helps differentiate between actual data points, the
regression line, and the predicted point.
- Grid: Adds a grid to the background for better readability and
alignment with the data points.

Visual Interpretation Fig 10: Algorithm Linear Regression

1. Trend Line: The red regression line provides a visual STEP 1: Load the dataset into the python and import the
representation of the general trend in temperature over time libraries STEP 2: Data pre-processing
(indexed by days). The slope of the line indicates the direction
and rate of temperature change. Step 2.1: Drop the unnecessary column from the dataset
2. Model Fit: The closeness of the blue data points to the red
line indicates how well the model fits the actual data. If the
points are closely clustered around the line, the model has a Step 2.2: Delete all values from te pressure which has a value -
good fit. 9999
3. Prediction Capability: The green dot (predicted temperature
for index 15) shows the model's ability to predict future Step 2.3: Taking all the features into x variable and y for
temperatures. The position of this dot relative to the regression prediction
line shows where the predicted temperature lies in the context
of the model's learned trend. Step 2.4: Set the dummies value as a level for the weather
classification Step
In summary, the outputs and the graph together provide a
comprehensive understanding of how the linear regression
model performs, its accuracy, and its prediction capabilities.
 CONCLUSION

Weather forecasting using the linear regression algorithm and

the Naïve Bayes algorithm is critical for improving people’s
Step 2.5: Set the dummies value as a level for the weather future results. The linear regression algorithm and the Naïve
classification Bayes algorithm were used to forecast the weather using
weather datasets. Using some selected input variables obtained
Step 2.6: Delete last dummies value which is null from kaggle, GitHub we created a model to predict the weather.
The issue with the current weather situation is that we are
unable to organize ourselves and complete essential tasks. As a
Step 2.7: Concatenate the dummies value with the input feature
result, this model was developed in order to know the weather
X
scenario with high precision while taking into account all of the
factors that influence the weather scenario.
Step 2.8: Created the new dataset after apply the preprocess

STEP 3: Train and test the data

Step 3.1: Splitting Dataset into train set and test set

STEP 4: Fitting the Linear Regression model

STEP 5: Histogram of data (visualization)

STEP 6: Analyze the original with predicted data (using

visualize)

STEP 7: Output

Fig 11 Algorithm Linear Regression