Tribhuvan University

Institute of Science and Technology

A Project Report
On
“FLOOD PREDICTION SYSTEM”
Submitted To:
Patan Multiple Campus
Patandhoka, Lalitpur, Nepal

In Partial fulfilment of the requirements

For the degree of Bachelors In Information Technology

Submitted by:
Binod Timilsaina (BIT 09/076)
Deni Lama (BIT 12/076)
Sanjeev Ghimire (BIT 29/076)

Under the supervision of

Mr. Bishnu Khadka
March 2024
 Acknowledgement
We would like to express our sincerest gratitude to Department of Bachelor In Information
Technology for providing us with the opportunity to undertake this project. Being grateful
to our program coordinator Mr. Jyoti Prakash Chaudhary who gave us the valuable time,
suggestion, help, support to solve problems that occurred during the project selection. We
also like to thank our supervisor Mr. Bishnu Khadka for encouraging to do something
different.
We are grateful for the support and guidance offered by BIT department and the valuable
insights provided by our mentors, who have helped us shape our understanding and skills
in this field. We would also like to thank our classmates and peers who have contributed to
this project through collaboration and constructive criticism. Their support has been
invaluable in helping us bring this project to completion. This project would not have been
possible without the generosity and support of Patan Multiple Campus, and we are deeply
thankful for the resources and opportunities that were provided to us.
We would like to extend our heartfelt gratitude to Department of Hydrology and
Meteorology for providing us with the necessary datasets for this project. Your generosity
and support have been instrumental in making this project a success. The access to high-
quality data has been invaluable in allowing us to conduct meaningful analysis and reach
significant findings. We are grateful for the resources and opportunities provided to us by
the department. Thank you again for your support.

Your Regards,
Binod Timilsaina
Deni Lama
Sanjeev Ghimire

i
 Abstract
Flooding is the most common natural disaster on the planet, affecting thousands of people
per year. In order to reduce casualties and property damage, this study emphasizes the value
of flood prediction models. In order to forecast the occurrence of floods, the Flood
Prediction System presented in this work leverages machine learning, specifically the
Linear Regression method. Over the past two decades, the development of more affordable
and precise flood prediction systems has benefited greatly from the use of machine learning
techniques.
Keywords: Rainfall, Machine Learning, Linear Regression, Python

ii
 Table of Contents
Acknowledgement ................................................................................................................ i

Abstract ................................................................................................................................ii

List of Figures ...................................................................................................................... v

List of Tables ...................................................................................................................... vi

List of Abbreviations .........................................................................................................vii

CHAPTER 1: INTRODUCTION ..................................................................................... 1

1.1 Introduction ................................................................................................................ 1

1.2 Problem Statement ..................................................................................................... 2

1.3 Objectives .................................................................................................................. 2

1.4 Scope and Limitation ................................................................................................. 2

1.5 Development Methodology ....................................................................................... 3

1.6 Report Organization ................................................................................................... 4

CHAPTER 2: BACKGROUND STUDY AND LITERATURE REVIEW .................. 5

2.1 Background Study...................................................................................................... 5

2.2 Literature Review....................................................................................................... 6

CHAPTER 3: SYSTEM ANALYSIS ............................................................................... 9

3.1 System Analysis ......................................................................................................... 9

3.1.1 Requirement Analysis ......................................................................................... 9

3.1.2 Feasibility Study ............................................................................................... 10

3.2 Analysis.................................................................................................................... 11

3.2.1 Proposed System ............................................................................................... 11

3.2.2 Working Mechanism ......................................................................................... 11

3.2.3 Dataset Description ........................................................................................... 12

3.2.4 Process Model ................................................................................................... 13

CHAPTER 4: SYSTEM DESIGN .................................................................................. 14

4.1 System Design ......................................................................................................... 14

iii
 4.1.1 Activity Diagram .............................................................................................. 14

4.1.2 Sequence Diagram ............................................................................................ 15

4.1.3 Flowchart .......................................................................................................... 15

4.1.4 Class Diagram ................................................................................................... 16

4.2 Algorithm Details..................................................................................................... 18

CHAPTER 5: IMPLEMENTATION AND TESTING ................................................ 20

5.1 Implementation ........................................................................................................ 20

5.1.1 Tools Used ........................................................................................................ 20

5.1.2 Implementation detail of modules .................................................................... 21

5.2 Data Collection ........................................................................................................ 22

5.3 Testing And Result Analysis ................................................................................... 23

5.3.1 System Testing .................................................................................................. 23

5.3.2 Result Analysis ................................................................................................. 23

CHAPTER 6: CONCLUSION AND FUTURE RECOMMENDATION ................... 24

6.1 Conclusion ............................................................................................................... 24

6.2 Future Recommendation .......................................................................................... 24

References

Appendix

iv
 List of Figures
Figure 1. 1: Agile Method .................................................................................................... 3
Figure 2. 1: Use Case Diagram of Flood Prediction ............................................................ 9
Figure 3. 1: Gantt Chart ..................................................................................................... 11
Figure 3. 2: Working mechanism of Flood Prediction System .......................................... 12
Figure 3. 3: Class Diagram ……………………………………………………………….13
Figure 4. 1: Activity Diagram of Flood Prediction ............................................................ 14
Figure 4. 2: Sequence Diagram of Flood Prediction.......................................................... 15
Figure 4. 3: Flowchart of Flood Prediction ........................................................................ 16
Figure 4. 4: Class Diagram of Flood Prediction……………………………………….…17
Figure 4. 5: Linear Regression Algorithm Workflow ........................................................ 18

v
 List of Tables

Table 5.1: System Testing Results………………………………………………………..23

vi
 List of Abbreviations
CSS : Cascading Style Sheet
CSV : Comma Separated Values
GDP : Gross Domestic Product
GPU : Graphics Processing Unit
HTML : Hyper Text Markup Language
IDE : Integrated Development Environment
JS : Java Script
LR : Linear Regression.
ML : Machine Learning
PC : Personal Computer
RAM : Random Access Memory
SDLC : Software Development Lifecycle
UML : Unified Modeling Language
VS : Visual Studio

vii
 CHAPTER 1
INTRODUCTION
1.1 Introduction

Floods are very harmful for nature, which are very complex to model. The flood prediction
model will give risk reduction & it minimizes the future loss of human life. Every year
Nepal is affected by flood which hampers a lot. Machine learning is a method which
provides intelligence to predict the result in future. The performance comparison of ML
models is based on the speed, time and accuracy of the result. There exist a lot of machine
algorithms which generate models with more accuracy. Machine learning provides
capabilities to learn from past data. Also based on past data it generates models for future
prediction. This technique will be very useful for flood prediction. Earlier everyone need
to give instruction to system for generating output and result. But now with the help of
machine learning technique it generates models and gives result itself. Most of the work
related to machine learning for flood either predict and helps to make precaution measure
and suggest future flood. Nepal is a prone flood danger place and thousands of life are taken
every year. This motivated us to do study on this pattern of rainfall.
There was much research for prediction of flood ahead, but not many methods give the
estimate with high accuracy. The flood prediction analysis majorly uses Machine Learning
(ML). There are many methods in machine learning to predict the problem with higher
accuracy in this work, This model have proposed to estimate the flash flood to prevent
places that are prone to flood risk. The approach is to the establishment of the ML algorithm
model. It incorporates the flood factor to estimate short term prediction in an urban area
with higher accuracy.
Python is used as the programming language for that. One of the best programming
languages for machine learning is Python. Python offers a variety of features for performing
big data analysis and creating machine learning models. Python comes with a large number
of built-in libraries. These libraries enable users to carry out various operations like
artificial intelligence, data mining, image recognition, machine learning, and more.
In this project, django has also been used for the integration of frontend and backend. In
user interface, there are two input fields for rainfall amount and river amount where user
provides the input and gets the output. Django is a python framework which provides a
feature which is used for web development. User interface is built from HTML, CSS, JS

1
and Bootstrap. Django is used for backend that is used to process the input and provide
output to the user based on input. Python libraries like pandas, numpy and scikit-learn are
used. Numpy is used to process mathematical operations. It calculates partial derivative.
Pandas is used for preprocessing of data. Scikit-learn trains the model using Linear
Regression.
Linear Regression algorithm consists of function like fit, update weight, predict. Fit accepts
the input. Update weight consists of main calculation which calculates partial derivate, dw
and db is also calculated which is gradient descent. Predict function calculates the output
which is stored in a variable.
Joblib is used for serializing the data so that training the model is not repetitive. Once the
serialized data is stored, condition is checked and output is provided according to it.
1.2 Problem Statement
On average, floods cause over 175 deaths each year and average annual economic losses
exceeding USD 140 Million. Nepal is the tenth highest country in the world in terms of
relative physical exposure to fluvial flooding, exposing possible damage to physical assets
as a value equivalent to 1.4% of its GDP. Without comprehensive flood protection or
appropriate risk-sensitive planning, increased ongoing economic and development activity
in the floodplains will likely increase the economic damage associated with floods in the
coming years. The effects of any problem can be reduced if known whether or not the
problem will occur. Similarly, developing of flood prediction system will help us to know
about flood and effective precautions can be applied so that it does not hamper any lives,
resources and so on. .
1.3 Objectives
The primary Objectives of developing the system “Flood Prediction “are as follows
• To develop a system that can predict occurrence of flood.
1.4 Scope and Limitation
Scope and limitations:
The Prediction system is a useful innovation that is implemented. This project is relatively
low-cost and can be successfully deployed in a diverse range within a country. This project
can advise and guide community about flood risk. This system has a scope of being as a
flood forecasting division of Department of Hydrology and Meteorology.
This system still has some limitations. A flood prediction system is not sufficient on its
own to reduce risk people’s reactions to warnings their attitude and the nature of their

2
response has an important bearing upon the effectiveness of a system. Sometimes it may
give inaccuracy that may lead to fear causing unnecessary anxieties.
1.5 Development Methodology
Agile SDLC model is a combination of iterative and incremental process models with focus
on process adaptability and customer satisfaction by rapid delivery of working software
product. Flood Prediction System is based on agile method.
Firstly, requirements were gathered for generating a model. The gathered requirements
were raw rainfall and river data. These datasets were collected from Department of
Hydrology And Meteorology. After requirement gathering phase the raw data were pre-
processed. The features were extracted and the label were differentiated. The training of
model was performed using bought datasets and Linear Regression. Following training of
model the accuracy was calculated.
After that the interface for user was built using Front end tools like HTML, CSS, Bootstrap
and JavaScript. Django was used for backend and also for integrating both front end and
back end. Joblib was used for serializing the data.
The whole analysis of system along with testing, accuracy analyzing and documentation
was performed.

Figure 1. 1: Agile Method

a) Requirement Phase
In this phase, all the necessary requirements were analyzed. All the necessary requirement
for further analysis of project was gathered from end-user, Internet and teachers. And as a
result, final specification of the project was gained.

3
b) Planning and Designing
In this phase, analysis was performed in order to find out the requirements of the system.
The outcome of this phase was System Requirement Specifications. In this phase, the SRS
was translated into the system design. Use case diagram, and component diagram were
developed.
c) Development phase
After the analysis and design the coding was done according to the specifications.
d) Testing
Once a system was developed, series of testing was performed in order to remove bugs and
errors. Testing was done by splitting the datasets into training sets and testing sets and
comparing the difference of actual output and expected output.
e) Evaluation
Evaluation was the last step performed after all the prior steps, where the project was
evaluated to check if it met the specification or not.
1.6 Report Organization
This document is categorized into several chapters and further divided into sub chapters
including all the details of the project.
 The Chapter One is about the introduction of the whole report. It includes short
introduction of the system, scope and limitations and objectives of the system.
 The Chapter Two includes the research methodologies in the project. Background
study and literature review has been covered.
 The Chapter Three is all about system analysis. It also includes feasibility study and
requirement analysis.
 The Chapter Four include the System Design and details of the Algorithm.
 The Chapter Five is about the implementation and testing procedures. It contains the
detail about the tools that are required to design the system. In the testing section,
different testing processes are included.
 The Chapter Six includes conclusion of the whole project. It also provides information
about what further can be achieved from this project.

4
 CHAPTER 2

BACKGROUND STUDY AND LITERATURE REVIEW

2.1 Background Study
The task of predicting flood requires in-depth knowledge of rainfall and river. In this
project, the occurrence of flood is discussed and its prediction using Machine Learning
algorithm. The primary aim of the project is to predict whether or not flood occurs through
the given set of user input. Datasets are extracted from Department of Hydrology and
Metrology. The learning model were trained and parameter tuned to obtain optimal
accuracy.
 Features: The individual observation units or data points for a training sample are
called variables or features. A tabular data contains continuous numerical data.
 Target Variable: It's the variable or feature of a dataset about which you want to gain
a deeper understanding. It is an annotations of a training sample for multiple concepts
like label for an image having flood vs no flood.
 Equations: The simple equation is
Y = β0 + β1.X1 + ϵ (Simple Linear Regression)
But the flood prediction system is based on:
Y = β0 + β1.X1 + β2. X2 +………+ βN. XN + ϵ (Multiple Linear Regression)
X1, X2, XN = Predictor Variables
Β0 = Y-intercept (always a constant)
Β1, β2, βN = regression coefficients
ϵ = Error terms (Residuals)
 Forward Pass: Step in the model training process that deals with calculating ‘y’ (or
predicting output) for each training sample is called Forwards pass.
 Loss/ Cost Function: It simply helps us to understand the difference between the
predicted value & the actual value for each iteration. The Function used to quantify this
loss during the training phase in the form of a single real number is known as “Loss
Function” and it helps us to reach the optimal solutions for the equation. Loss function
refers to the error for a single training example whereas Cost function refers to an
average of the loss functions over an entire training dataset.
 Back Propagation: Technically, back-propagation computes the gradient of the loss
function w.r.t each coefficient which helps in updating each coefficient such that in the

5
 next iteration these are closer to the optimal solution reducing the loss further. It is
performed by computing the partial derivative.
 Gradient Descent: Gradient descent is the optimization algorithm used to update the
coefficients of the equation. It is used to minimize the cost/ loss function by iteratively
moving in the direction of steepest descent as defined by the negative of the gradient.
 Learning Rate: The size of each step at each iteration while moving towards the
minimum of the loss function. With a high learning rate or large steps, there is a risk of
missing the minimum of the loss function while with a very low learning rate there is a
risk of getting stuck at a local minimum, and model training can be time-consuming
too.
 Training Iterations: Each step of forward pass, loss computation, backward pass and
gradient descent leads or updating weights refers to as training iterations.
 Weights and models: Weights are all the coefficients of the equation that are computed
after all the training iterations. The complete algorithmic equation with coefficients
computed while minimizing loss based on the training dataset is called a model.
2.2 Literature Review
This study proposed a neural network for predicting the flow rate of water at Dongola
station in the Nile River, Sudan. The model’s input is upstream flow data, and it tries to
predict the downstream flow rate with one day lead. The proposed model achieved an
accuracy of around 97% in detecting flood hazards. The significance of this study was that
the author could achieve this performance by only using one variable, i.e., upstream flow
data. Mathematical models that simulate the hydrodynamic process of the water’s flow can
also achieve good accuracy, but they often require large datasets. This shows an advantage
of NNs in flood forecasting over conventional models. [1]
This study uses a deep neural network to estimate the total accumulative overflow. Since
the amount of available heavy rainfall and actual flood is limited, they used data
augmentation techniques to train the model with more data. Their study area was the
Samseong-dong district in Seoul, Korea. Due to its low land and insufficient drainage
system, this area experienced an extreme urban flood in 2010 and 2011. Since it is a
metropolitan area, modelling the flood flow requires complex calculations. Statistical
properties of each rainfall event were used as the input. They compared the prediction of
their model with the results of computer simulation. They confirmed that the performance
of their model is improved when they use data augmentations. Lack of historical data is an

6
important challenge in most flood prediction problems. On the other hand, the performance
of machine learning models heavily depends on the amount of available training data. [2]
The significance of this study is that the authors proposed using data augmentation and
confirmed that it could improve the model’s performance. [2]
In this study the authors applied an LSTM model for forecasting floods on the Da River,
which is one of the largest river basins in Vietnam. Model tries to forecast the river’s flow
using rainfall and flowrate data acquired at hydrological stations of the study area Da River
originates from Yunnan province in China and flows through high mountains in Vietnam.
It has an abundant flow which makes a suitable source for hydroelectric power. [3]
Due to the existence of dams upstream of the Da River, forecasting its downstream flow is
a challenging task. The paper attempted to predict the downstream flow with one, two, and
three days of lead time using the stations in the Hoa Binh area. An LSTM is a special kind
of RNN that can learn long-term dependencies of the data. The network consists of memory
blocks called cells. These cells are organized in a way that the network can remember long
term dependencies in the data. To evaluate the performance of their method, they used the
NSE and RMSE criteria and have shown that their model can accurately predict the flow
rate with one, two, and three days lead. As expected, the accuracy of their prediction with
one day lead is better than other scenarios. It is worthy of mentioning that the LSTM model
does not use any expert knowledge but instead learns the physical principles such as
principles of mass conservation or momentum conservation during training from the input
data. This study showed that LSTMs could be reliable tools for predicting the flow rate and
flood risk with fair accuracy. [3]
F. Cruz, M. G. Binag, US, "Flood Prediction Using Multi Layer Artificial Neural Network
in Monitoring System with Rain Gange, Water Level, Soil Moisture Sensors," TENCON
2018-2015 IEEE Region 10 Conference, 2018, pp. 2499-2503, doi:
10.1109/TENCON.2018.8650387. In this study, a system is developed to predict flood
level based on real time monitoring sensors and systems. [4]
F. A. Ruslan, K. Haron, A. M. Samad and R. Adnan, "Multiple Input Single Output (MISO)
ARX and ARMAX model of flood prediction system: Case study Pahang," 2017 pp. 179-
184, doi: 10.1109. This paper compared the prediction performances of a developed flood
prediction models that were designed using Multiple-Input Single-Output (MISO) Auto
regressive with Exogenous Input (ARX) and MISO Auto regressive Moving Average with
Exogenous loput (ARMAX) structure. [5]

7
J.M.A. Opella and A.A. Hernandez, "Developing a Flood Risk Assessment Using Support
Vector Machine and Convolutional Neural Network: A Conceptual Framework," 2019
IEEE 15th International Colloquium on Signal Processing. The proposed study aims to
exploit the data available from the Geographical Information Systems. [6]

8
 CHAPTER 3
SYSTEM ANALYSIS
3.1 System Analysis
All-inclusive analysis of the system starting from problem identification, defining systems
objectives, gathering requirements, feasibility study to requirements analysis was done
using use case diagram.
3.1.1 Requirement Analysis
3.1.1.1 Functional Requirements
Use Case Diagram
User enters an input and get a predictions and system can use dataset, train and predict a
values.

Figure 2. 1: Use Case Diagram of Flood Prediction

 Input Data: The data is provided by the user which is used for prediction.
 UseDataset: Dataset is obtained from Department of Hydrology And Meteorology.
That dataset is used by system for training the data.
 Model Training: Model is trained by Linear Regression algorithm.
 Predicted Values: System provides the predicted value after training the model.
 Predictions: User gets the predictions based on input data.

9
3.1.1.2 Non-Functional Requirement
 Processing of the request should be done in 5 sec from the time of request.
 This system is easy to use, understand and can easily be navigated by a user.
 This system is designed in such a way that it is easy to maintain and upgrade over time.
 This system is able to easily integrate with other similar type of systems.
3.1.2 Feasibility Study
Feasibility analysis, in simple words is an analysis and evaluation of a proposed project to
ensure if it is technically, economically and operationally feasible. As the name suggests,
a feasibility analysis is a study of the viability of an idea. It focuses on answering the
essential question of “should this proposed project idea be proceeded?”
3.1.2.1 Technical
The project is technically feasible; complies with current technology, including both the
hardware and the software. All the technical requirements for this project are listed below:
 A laptop with at least 2GB RAM with GPU
This application is supported by almost all latest personal computers with minimum
hardware and software requirements.
3.1.2.2 Operational
This project can be conducted with a minimum human resource. Three developers are
working in the project which is more than enough manpower required for this project. This
project aims to create a Flood Prediction System that predicts flood.
3.1.2.3 Economic
The project to be developed is very cost effective because the project will be using open-
source software like python etc. which are freely available to download. There are many
websites which host other websites for free until certain storage limit is met which is
currently efficient for the project. Also, the benefits provided by the project easily outweigh
the cost. So, the project can be considered economical feasible for the time being.
3.1.2.4 Schedule
In scheduling feasibility, an organization estimates how much time the project will take to
complete. When these areas have all been examined, the feasibility analysis helps identify
any constraints the proposed project may face, including: Internal Project Constraints:
Technical, Technology, Budget, Resource, etc.

10
Gantt Chart
S.N Task Name Weeks

1 2 3 4 5 6 7 8 9 10 11 12

1. Study and
Analysis
2. Data
Collection
3. Implementati
on
4. Testing

5. Documentatio
n
6. Review

7. Presentation

Figure 3. 1: Gantt Chart

3.2 Analysis
3.2.1 Proposed System
In this paper a model is build based on the Linear Regression. Since this problem is a kind
of regression, implementing algorithm like Backward Propagation, Logistic Regression,
etc. will be best as this is standard for numeric processing. The actual goal is that a machine
is trained with a dataset of rainfall and river. After the model is trained, an input is passed
from user side and then the system predicts whether or not a flood occurs through the
parameters passed by the user.
3.2.2 Working Mechanism
The datasets is provided to the model that is trained using Linear Regression. These datasets
are preprocessed. The dataset is trained. After these steps, a user provides input. A input is
then processed which provides result based on trained data. It sees certain pattern and
predicts the output.

11
 Figure 3. 2: Working mechanism of Flood Prediction System
3.2.3 Dataset Description

Dataset description for a food prediction system containing rainfall, river, and label as
attributes:
 Name: Rainfall-and-River.csv
 Data Source: Department of Hydrology and Meteorology.
 Data Format: Data are stored in CSV format.
 Data Size: Data set had 325 rows and 3 columns. Its size is 5 KB.
 Data Fields: The fields are Rainfall, River and Label. Rainfall and River includes
numeric data.
 Data Cleaning: The dataset was cleaned to remove any duplicate records and missing
values.
 Data Analysis: It shows the proximity to river and amount of rainfall significantly.
 Conclusion: The system uses this dataset to predict the flood in Khokana, Manohara
and Khumaltar based on rainfall and river proximity.
 Limitations: The dataset only considers rainfall and river proximity as factors affecting
flood. Other factors such as soil quality, temperature, and vegetation type were not
considered.

12
3.2.4 Process Model

Figure 3. 3: Class Diagram

In this diagram, the LinearRegression class has eight instance variables: learning_rate,
no_of_iterations, m, n, w, b, X, and Y. The init method initializes the learning_rate and
no_of_iterations variables. The fit method takes in a feature matrix X and a label vector Y
and trains the model by optimizing the weights and biases. The update_weights method
updates the weights and biases using gradient descent. Finally, the predict method takes in
a feature matrix X and returns the predicted label vector.

13
 CHAPTER 4

SYSTEM DESIGN
4.1 System Design
The proposed design of this system was modeled using following diagrams:
4.1.1 Activity Diagram
4.1.2 Sequence Diagram
4.1.3 Flowchart
4.1.4 Class Diagram
4.1.1 Activity Diagram

The left most section is for system where the system use dataset, train the model and predict
the values required. The rightmost section is for user where user enters the input and get
the predictions.

Figure 4. 1: Activity Diagram of Flood Prediction

14
4.1.2 Sequence Diagram

The data of rainfall is validated and is passed into a training model. The trained model
generates a value and determines a flood. A report is then presented to the user.

Figure 4. 2: Sequence Diagram of Flood Prediction

4.1.3 Flowchart

An input of data is passed by a user. The dataset is selected which is pre-processed which
contains data cleaning, removing data redundancy. After pre-processing the data are
extracted from datasets. The data are trained, analyzed, tested and the flood is predicted.

15
 Figure 4. 3: Flowchart of Flood Prediction
4.1.4 Class Diagram

A class diagram is a type of UML diagram that represents the structure and relationships
of classes and their attributes and methods in object-oriented programming. It shows the
classes in a system, their properties and the methods they can perform (behaviors). It is a
graphical representation that helps in understanding and designing the overall structure of
a system by visualizing its classes, their relationships, and their interactions.
The functionality of the linear regression model was divided into various classes in this
enlarged class diagram:
 LinearRegression: The main component of the model is the LinearRegression class,
which handles data fitting, prediction, and gradient descent weight optimization. To

16
 carry out these functions, it communicates with the GradientDescentOptimizer,
LossFunction, and Dataset classes.
 LossFunction: The loss function used to optimize the model weights during training is
represented by the LossFunction class. It has a function called __call__ that accepts
both the actual labels and the predicted labels and outputs the loss value.
 Dataset: The input data must be stored and divided into train and test sets using the
Dataset class.
 GradientDescentOptimizer: Using the gradient descent process, the model weights
are optimized using the GradientDescentOptimizer class.
 Evaluation: The model's performance can be assessed using the methods of the
Evaluation class. It has methods like calculate_mse, which computes the mean squared
error between the predicted and actual values.

Figure 4. 4: Class Diagram of Flood Prediction

17
4.2 Algorithm Details
Linear Regression
Linear Regression is an algorithm that provides a linear relationship between an
independent variable and a dependent variable to predict the outcome of future events. It is
a statistical method used in data science and machine learning for predictive analysis.
The independent variable is also the predictor or explanatory variable that remains
unchanged due to the change in other variables. However, the dependent variable changes
with fluctuations in the independent variable. The regression model predicts the value of
the dependent variable, which is the response or outcome variable being analyzed or
studied.
Thus, linear regression is a supervised learning algorithm that simulates a mathematical
relationship between variables and makes predictions for continuous or numeric variables
such as sales, salary, age, product price, etc.
This analysis method is advantageous when at least two variables are available in the data,
as observed in flood prediction, stock market forecasting, portfolio management, scientific
analysis, etc.

Figure 4. 5: Linear Regression Algorithm Workflow

In the above figure,

X-axis = Independent variable
Y-axis = Output / dependent variable

18
Line of regression = Best fit line for a model
Here, a line is plotted for the given data points that suitably fit all the issues. Hence, it is
called the ‘best fit line.’ The goal of the linear regression algorithm is to find this best fit
line seen in the above figure.
Algorithm
1. Initialize the parameters.
Y = β0 + β1.X1 + ϵ (Simple Linear Regression)
Y = β0 + β1.X1 + β2. X2 +………+ βN. XN + ϵ (Multiple Linear Regression)
X1, X2, XN=Predictor Variables
Β0 = Y-intercept (always a constant)
Β1, β2, βN = regression coefficients
ϵ = Error terms (Residuals)

2. Predict the value of a dependent variable by given an independent variable.

Y = β0 + β1.X1 + ϵ (Simple Linear Regression)
Y = β0 + β1.X1 + β2. X2 +………+ βN. XN + ϵ (Multiple Linear Regression)
Y = Dependent Variable
X1, X2, XN = Independent Variable

3. Calculate the error in prediction for all data points.

S.S.E = Σ(mx+b-y|)2

4. Calculate partial derivative w.r.t a0 and a1.

J = 1/N * Σ (a0 + a1 * xi - yi)2
a0 = a0 – α * 2/N * Σ (predi - yi)

5. Calculate the cost for each number and add them.

J = 1 /N * Σ (Yi (m * xi +b))2

6. The final step involves the update of the forecasts of the previous model

19
 CHAPTER 5
IMPLEMENTATION AND TESTING
5.1 Implementation
After the analysis and design of the system is completed, the implementation of the system
is carried out involving how the system is installed, operated and maintained. The various
system tools that have been used in developing both the front-end, back-end and other tools
of the project are being discussed in this chapter.
5.1.1 Tools Used
Visual Studio Code
Visual Studio Code (VS Code) is an Integrated Development Environment (IDE) providing
a wide range of essential tools for developers, tightly integrated to create a convenient
environment for productive web, and data science development. Software development is
much faster using VS Code. The feature of error spotlighting in the code further enhances
the development process.
Front-end tools
For the purpose of developing user interface of web application, HTML and CSS and
JavaScript is used.
HTML: It is used in our system to make framework as it implements a markup-based
pattern to define behavior of content.
CSS: It is used to describe the presentation of document made in HTML format. It is used
to design presentation and content, including layout, colors, and fonts.
JavaScript: It enables interactive interface of system.
Back-end tools
Advanced Python along with some packages like Pandas, Numpy, Scikit-Learn etc. are
used to perform dataset related operation, along with some mathematical operations and
also in training a model.
Django was used as the language for integrating frontend and backend. It was also used for
handling request made by user and providing user the response for that request.
Pickle
Pickle in Python is primarily used in serializing and de-serializing a Python object structure.
In other words, it's the process of converting a Python object into a byte stream to store it
in a file/database, maintain program state across sessions, or transport data over the
network.

20
Joblib
Joblib provides a better way to avoid recomputing the same function repetitively saving a
lot of time and computational cost. It is also used in parallel processing.
Seaborn
Seaborn was used for generating heatmap and identifying correlation. It was used for
generating pairplot and heatmap.
Heatmap was used for generating a correlation graph. From, the heatmap it concluded that
the dataset was positively correlated. The increase in value of one feature increases the
value of another feature. Hence, Linear Regression was perfect algorithm.
Pairplot was used for extending the information provided by heatmap.
Matplotlib
Matplotlib was used for generating boxplot for identifying the outliers and visualizing the
data for exploratory data analysis.
Boxplot was used for identifying the outliers in dataset.
Hardware Components
• Laptop/PC: Laptop or PC is used as a hardware component to run the system and
visualize the results.
5.1.2 Implementation detail of modules
View Module
The User Interface or View Module contain a simple webpage. This webpage is a front
page that is styled by using bootstrap class.
View Front Page Module
The front page module consist of simple page. It consists of a card which is a bootstrap
class that contains form where user can enter an input and output is also shown in same
page.
Class and Modules Implementation
The project follows object based approach where classes and objects are used to specify
the functionalities. The implemented functions are listed below.
• Linear Regression () Class
Linear Regression is the main algorithm that is used in the system. It specifies overall
working and structure of the system. The attributes and functions used are explained below
• Attributes Used
Learning Rate: It is a hyper parameter that determines the step size at which the optimizer
makes updates to the model parameters during training.
21
W: Slope or Weights
B: Y-Intercept
X: Features or Independent Variable
Y: Label or Dependent Variable
• Functions Used
fit: Takes datasets as input, creates sample dataset from original dataset or parameters for
initialization.
update_weights: Performs numpy calculation of partial derivative and gradient descent for
calculating weights or slopes, intercept everytime.
predict: It also performs numpy calculation. It predicts the output for the given input by
user using the Linear Regression model.
• LossFunction Class():
The loss function used to optimize the model weights during training is represented by the
LossFunction class. It has a function called __call__ that accepts both the actual labels and
the predicted labels and outputs the loss value.
• Dataset Class()
The input data must be stored and divided into train and test sets using the Dataset class.
• GradientDescentOptimizer Class()
Using the gradient descent process, the model weights are optimized using the
GradientDescentOptimizer class.
• Evaluation Class()
The model's performance can be assessed using the methods of the Evaluation class. It has
methods like calculate_mse, which computes the mean squared error between the predicted
and actual values.
5.2 Data Collection
Data were given by the Department of Hydrology and Meteorology. The data sets span the
years 2011 to 2019. The Bagmati River was included in the river dataset. The freely
available rainfall dataset consisted of three stations: Khokana, Manohara, and Khumaltar.
The department's data unit provided access to these datasets.

22
5.3 Testing And Result Analysis
5.3.1 System Testing

System testing were conducted on various rainfall and river amounts and result of the
system was compared with expected output. Accuracy of the system was computed by
changing various parameters.
Table 5 .1: System Testing Results

Rainfall River Expected Observed

S.N
Amount Amount Output Output

1 6.2 3.914839 No Flood No Flood

2 54.9 4.627857 No Flood No Flood

3 452.3 33.98387 Flood Flood

4 13 6.9 Flood Flood

5 0 2.6 No Flood No Flood

6 278.6 15.31333 Flood No Flood

7 428.5 35.73226 Flood Flood

5.3.2 Result Analysis

Total datasets were split into training sets and testing sets in the ratio 4:1. Following results
were found on testing accuracy of algorithm on the given training datasets.
Accuracy on testing data: 86.1%
Accuracy on training data: 90%

23
 CHAPTER 6

CONCLUSION AND FUTURE RECOMMENDATION

6.1 Conclusion
The application has been developed using HTM, CSS, JS and Bootstrap in frontend and
python and django in backend. Hence, all the requirements and the objectives of the project
have been fulfilled by the application.
The current state of ML modeling for flood prediction is quite young and in the early stage
of advancement. The data used in our project is collected from Department Of Hydrology
And Meteorology. The primary aim of this project is to predict the floods as per user data.
The learning models were trained and parameter-tuned to obtain optimal accuracy. Some
models have achieved comparatively higher accuracy than others. Performance is measured
and compared using metrics for each algorithm.
6.2 Future Recommendation
For future work, conducting a survey on spatial flood prediction using machine learning
models is highly encouraged. This important aspect of flood prediction was excluded from
our project due to the nature of modeling methodologies and the datasets used in predicting
the location of floods. Nevertheless, the recent advancements in machine learning models
for spatial flood analysis revolutionized this particular realm of flood forecasting, which
requires separate investigation.

24
References

[1] Pirlo,.C. Amir Mosavi, "Flood Prediction Using Machine Learning Models,"
Research Gate, Trondheim, Norway , 2018.
[2] V. Bellos and G. Tsakiris, "Total Accumulative Overflow," Seoul, 2012.
[3] G. Regan. F. Jack. M. Lyon, "Early warning system," in International Journal of
Computer Applications , 2001.
[4] Morgan.B. Richard. G. Freddy.R. G. Cruz, "Flood Prediction Using MultiLayer
ANN," US, 2018.
[5] Kurt.H. Shawn. Robin. A. Franky. A. Ruslan, "Multiple Input Single Output (MISO)
ARX and ARMAX model of flood prediction system," Pahang, 2017.
[6] A. Hernandez, "SEMANTIC SCHOLAR," 2019. [Online]. Available:
https://www.semanticscholar.org/paper/Developing-a-Flood-Risk-Assessment-
Using-Support-A-Opella-
Hernandez/39651c2a24a734f0977f770957163fd6192a6c41.
Appendix

Home Page
Output No Flood
Output Flood