FLOOD PREDICTION USING

MACHINE LEARNING

M O H A M E D I R FA N
ALI(22112084)
ASHWIN(22112075)
SABEER
 INTRODUCTION
• Floods are among the most catastrophic natural disasters,
causing extensive damage to property, infrastructure, and
lives worldwide. Accurate and timely flood prediction is
crucial in disaster management, providing early warnings
that help mitigate impacts. In recent years, machine
learning (ML) has emerged as a powerful tool in predicting
floods due to its ability to process large volumes of
historical and real-time data, identifying patterns that might
be challenging for traditional methods
• This project focuses on predicting flood events based on
rainfall data, specifically targeting Kerala, a region prone to
monsoon-induced flooding. By leveraging machine learning
algorithms, the model processes monthly and annual
rainfall data to predict the likelihood of flooding. Various
11/09/2024 2
models, such as Random Forest, are trained to improve
 • Additionally, the project includes deploying this predictive
model through a Flask web application, making it accessible
to users who can input rainfall data and receive real-time
flood predictions. This web-based tool offers a practical, user-
friendly interface that can assist local authorities and
communities in preparing for potential flood events.
Ultimately, this project aims to support proactive flood risk
management, reducing both economic and social impacts by
enhancing flood forecasting capabilities.

11/09/2024 3
LITERATURE
REVIEW

11/09/2024 4
PAPER NAME AUTHORS ALGORITHM AND LIMITATION AND
METHODLOGIES REMARKS

Predicting Flood B. S. Shubham, M. S. This paper discusses The study primarily

Occurrences Using Prabha, S. R. Aditya the application of focuses on a single
Machine Learning Random Forest and geographic region,
Techniques Published on: 2020 Support Vector which may limit the
Machines (SVM) for generalizability of
flood prediction the results to other
based on rainfall areas. Future work
data. The authors could explore
implement feature incorporating more
selection techniques diverse datasets.
to identify the most
significant rainfall
parameters affecting
flood events.
Machine Learning Anjali R. Patil, Vinay The review The paper notes that
Approaches for Flood D. Sakhare, A. S. synthesizes various while machine
Prediction: A Review Gupta machine learning learning models
11/09/2024 algorithms used in show promise, they 5
Published on: 2021 flood prediction, often lack
PAPER NAME AUTHORS ALGORITHM AND LIMITATION AND
METHODLOGIES REMARKS

A Data-Driven R. K. Sharma, P. K. This study utilizes The reliance on

Approach for Flood Mehta, A. Sharma Logistic Regression historical
Prediction Using and Artificial Neural meteorological data
Machine Learning Published on: 2019 Networks (ANN) to can introduce biases
predict floods based if the data is not
on meteorological representative of
data. The authors future conditions.
emphasize the The authors suggest
importance of data exploring real-time
preprocessing and data integration for
outlier detection in improving prediction
achieving high timeliness
prediction accuracy.
Flood Risk K. M. Karthik, S. L. The review covers a The authors highlight
Prediction Using Vijaya Kumar, T. A. wide range of the challenge of data
Machine Learning Ramesh machine learning scarcity in certain
Techniques: A algorithms, including regions, which limits
Systematic Review
11/09/2024 Published on: 2022 Random Forest, SVM, the ability to train 6
and Gradient robust models. They
PAPER NAME AUTHORS ALGORITHM AND LIMITATION AND
METHODLOGIES REMARKS

Evaluation of R. A. Ahmed, M. S. This paper evaluates The authors note

Machine Learning Al-Shammari, Y. I. Al- the performance of that while some
Models for Flood An various machine models perform well,
Forecasting learning models, they can be sensitive
Published on: 2023 including Xgboost to hyperparameter
and Random Forest, settings. They
in forecasting floods. suggest further
The study exploration into
emphasizes cross- automated
validation techniques hyperparameter
and the use of tuning to enhance
performance metrics model performance.
for model evaluation.
A Comprehensive P. K. Ranjan, V. K. The framework The framework's
Framework for Flood Bhat, M. S. Tiwari integrates various complexity can make
Prediction Using data analytics implementation
Data Analytics Published on: 2020 techniques, including challenging in
11/09/2024 time series analysis resource-constrained 7
and machine environments. The
 PROBLEM STATEMENT
The project aims to develop a machine
learning-based flood prediction model using
rainfall data, addressing the need for accurate
forecasting to mitigate the impact of floods. By
leveraging historical rainfall patterns, the
model seeks to provide timely predictions to
aid disaster management and enhance
community preparedness against flooding
events.
11/09/2024 8
 OBJECTIVE
• A machine learning model
capable of predicting flood
occurrences based on monthly
and annual rainfall data.
• Deploy the model using Flask to
create a user-friendly web
interface.
• Enable prediction on user-
provided data for proactive
11/09/2024 9
flood warnings.
System Architecture
Diagram

11/09/2024 10
 Machine Learning
Preprocessing &
Data Collection Model (Random
Outlier Handling
Forest)

Flask Web App Prediction Results

11/09/2024 11
 Modules and
Pseudo code

11/09/2024 12
 Data Collection and Preprocessing

Load and clean the dataset, handle outliers, and prepare

the data for modelling.
Code:

def preprocess_data(data):
for column in monthly_columns:
cap_outliers(data,
column)
return data

11/09/2024 13
 Model Training

Train a Random Forest model on pre-

processed data to predict flood
occurrences.
Code:
rf = RandomForestClassifier()
rf.fit(X_train, y_train)

11/09/2024 14
 Flask Web
Application
Create a web interface for user interaction, where users
input rainfall data and receive predictions
Code:

def predict():
user_data = [float(request.form[field]) for field
in fields]
prediction = rf.predict([user_data])
return render_template('predict.html',
Prediction=prediction)

11/09/2024 15
System Workflow

Data is collected A Random Forest

and pre-processed,
model is trained on
with outliers
handled to avoid the cleaned data
skewing. and saved.

The model returns a Flask is used to

prediction, deploy the model,
displayed on a web accepting user input
page
for prediction.

11/09/2024 16
Experimental
Results
• Training Accuracy and Testing Accuracy:

11/09/2024 17
• Classification Report:

11/09/2024 18
11/09/2024 19
11/09/2024 20
 Evaluation
• The evaluation of the flood prediction model
focused on assessing its accuracy, reliability, and
practical utility for early warning systems. The
model was tested using historical rainfall data
from flood-prone areas, and various performance
metrics, including accuracy, precision, recall, and
F1-score, were used to gauge its effectiveness.
Among the models evaluated, Random Forest
yielded the highest accuracy at 85%, proving
robust in classifying flood events based on
rainfall data patterns. Additionally, cross-
validation ensured the model’s stability across
different data subsets, reducing the risk of
overfitting.
• The model was further evaluated through a
11/09/2024 21
Flask-based web application, simulating real-
 Case Study
This case study explores a machine learning-based
flood prediction system designed to analyses
monthly rainfall data for accurate flood
forecasting. Using a dataset from Kerala, India,
the model applies Random Forest and other
classifiers to predict flood events based on
historical rainfall records. After preprocessing
the data to handle outliers and optimize features,
the model achieved an accuracy of 85%. A Flask-
based web application allows users to input
rainfall data and receive real-time flood
predictions, providing a valuable tool for
authorities and communities in flood-prone areas.
11/09/2024 This project showcases the effectiveness of data-
22
 Conclusion
This project demonstrates the effectiveness of
machine learning in flood prediction by utilizing
historical rainfall data to forecast flood
occurrences. The Random Forest model, which
showed an accuracy of 85%, proved reliable in
identifying patterns associated with flood
events, providing an essential tool for disaster
management. The developed model, integrated
into a user-friendly web application, offers a
valuable resource for authorities and
11/09/2024 23
 Future Work
• Improve model accuracy by incorporating additional
environmental factors (e.g., temperature, humidity).
• Use more advanced machine learning models or deep
learning for enhanced accuracy.
• Deploy the model on cloud services for scalability and
accessibility.

11/09/2024 24
 References
• Jain, S., & Kumar, R. (2018). "Flood prediction using machine learning techniques: A
survey." Journal of Hydrology, 564, 121-133.

• Rath, M., & Mishra, S. (2020). "Predictive modeling of flood risk using machine
learning." Applied Water Science, 10(7), 1-12.

• Ali, I. A., & Mehmood, A. (2021). "A comprehensive review of machine learning
algorithms for flood prediction." Natural Hazards, 106(2), 1335-1357.

• Elshafie, A., & Shokry, A. (2019). "Application of machine learning in flood prediction:
A case study." Water, 11(5), 999.

• Sahu, S. K., & Tripathy, R. (2020). "Flood prediction using random forest algorithm: A
case study of the Mahanadi river basin." Journal of Water and Climate Change, 11(4),
1023-1035.

• Sharma, R. A., & Kumar, S. (2022). "Machine learning techniques for rainfall
prediction and flood forecasting: A review." Environmental Science and Pollution
Research, 29(24), 35851-35865.
11/09/2024 25
 THANK YOU

11/09/2024 26