2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

Flood Prediction using Logistic Regression

Hanshika. V. G Preethi. S Sreemathy. S. P
Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Sri Ramakrishna Engineering College Sri Ramakrishna Engineering College Sri Ramakrishna Engineering College
Coimbatore, India Coimbatore, India Coimbatore, India
hanshika.2001047@srec.ac.in preethi.2001150@srec.ac.in sreemathy.2001228@srec.ac.in

Ezhillin Freeda. S
Department of Computer Science and
Engineering
Sri Ramakrishna Engineering College
Coimbatore, India
ezhilinfreeda@srec.ac.in
2023 International Conference on Circuit Power and Computing Technologies (ICCPCT) | 979-8-3503-3324-4/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICCPCT58313.2023.10245832

Abstract— Flood Prediction is crucial in mitigating its The foundation of machine learning is the notion that
impact on human life, property, and the environment. This computers are capable of learning from data, recognizing
paper proposes a machine learning approach using Logistic patterns, and making decisions without human
Regression for flood prediction, which involves the usage of intervention. The process involves three key components:
rainfall data to determine the probability of flood occurrence. data, algorithms, and models. It is a rapidly growing field
By analyzing the average rainfall and training the model with transforming how we use technology in our daily lives [4].
the vast dataset, our model can generate more accurate It is a branch of artificial intelligence that gives computers
predictions on flood occurrence. Our mobile application the capacity to learn from their experiences and advance
provides an early warning system through the flood risk map
without explicit programming. Finance, healthcare, and
that is accessible to users. Our dataset comprises 37 regions in
Tamil Nadu, which are the meteorological stations and also
marketing [3] are just a few of the industries and
the base stations of dams. The latter allows for notification in applications that use machine learning algorithms.
the event of flood and dam overflow respectively. Supervised learning and unsupervised learning are two
Furthermore, our mobile app enables users to easily identify subcategories of machine learning. The following flow
camps and shelters established during flood periods and even diagram in figure 1 explains the workflow of Machine
add information about them. In addition, our mobile app also Learning algorithms. First, data are collected and which is
provides users with the necessary helpline numbers to reach then analyzed to identify patterns and relationships among
out for assistance during flood times. them. Then, algorithms are developed to process the data
and create models that can make predictions or decisions.
Keywords—logistic regression, rainfall, flood, mobile Finally, the models are tested and refined to improve their
application. accuracy and performance.
I. INTRODUCTION
One of the most common natural catastrophes is
flooding, which is brought on by heavy downpours,
overflowing rivers, breached dams, cyclones, and storm
surges. It is a phenomenon when water runs into normally
dry land, inflicting harm to infrastructure and property,
disrupting travel, and even claiming human lives. In India,
floods are a recurring problem and Tamil Nadu is one of
the states most prone to flooding in India. Moreover,
Machine learning models have become increasingly
popular in flood prediction because of their ability to
analyze large datasets and identify patterns in flood data
[1]. By using these models, we can examine a range of
environmental as well as socioeconomic parameters,
including rainfall and weather patterns, soil moisture,
vegetation cover, land use, and terrain. These models can
aid in identifying regions with a high risk of flooding and
aid in creating an early warning system [2]. Our developed
Logistic regression model to predict the occurrence of the
flood is trained to identify patterns and predict the
likelihood of flooding in a particular area of Tamil Nadu. Fig 1 Flow diagram of Machine Learning model.
Using logistic regression, flood risk maps are created that
can be used more effectively to allocate resources, evacuate The use of machine learning models for flood
residents from high-risk locations, and plan disaster prediction has grown significantly in recent years due to
response activities. Mobile application has been created their effectiveness and potential. Recent statistics show that
that makes the above-mentioned flood prediction model these models have achieved high accuracy rates in
accessible to everyone. predicting floods, surpassing traditional methods. This has
led to a rapid expansion in the adoption of machine learning
for flood prediction, opening up new possibilities for

979-8-3503-3324-4/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1174
 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

managing and mitigating the impacts of floods. With The primary focus of [9] is to create an optimal flood
ongoing advancements, machines are poised to accomplish detection model. In this study, a decision tree model is
tasks that were once considered impossible, promising a constructed, and machine learning algorithms such as
bright future for the application of machine learning in Random Forest and Gradient Boosting are employed. The
flood prediction and beyond. developed model performs multiple computations on
datasets, incorporating an AI algorithm specifically
A statistical technique called logistic regression [5] can designed for flood prediction [9]. By utilizing these
be used to examine a dataset in which one or more techniques, the model aims to accurately forecast and
independent factors affect the outcome. There are only two detect flood events with improved precision.
possible possibilities for the result, hence it is measured by
a bipolar variable. It is an efficient tool that can aid in In the research paper [10], a Flood Prediction Model
understanding and foretelling systems or process behavior. (FPM) is introduced to forecast river floods by employing
One benefit of logistic regression is that it can handle a high the Artificial Neural Network (ANN) approach, chosen for
number of predictor variables and is simple to execute and its ability to address nonlinear problems. The FPM utilizes
analyze. It can be used to handle non-linear interactions rainfall data to predict river water levels and provides
between variables as well. Logistic regression uses a binary corresponding data on river water levels. While various
dependent variable. The objective is to identify the link factors contribute to water level fluctuations, this model
between the independent and dependent variables and to focuses on considering two specific factors in the
generate predictions about the future course of events using prediction process. By emphasizing the ANN's nonlinear
this relationship. A useful technique for assessing and capabilities and incorporating relevant variables, the FPM
forecasting binary outcomes [6] is logistic regression. It aims to enhance flood prediction accuracy and contribute
can handle a lot of predictor variables and has a lot of to effective river management strategies. Table 1 is a
benefits, including being simple to use and analyze. summary of additional strategies [9] now in use and their
drawbacks.
Table.1 Comparison of existing ML models for flood prediction

Sl. Models Description Disadvantages

No

1. K-Nearest The supervised machine Costly

Neighbours learning method K- computationally for
nearest neighbour (KNN) huge datasets. It is
is used for classification necessary to decide on
and regression tasks. The the ideal number of
majority class of its K neighbours. sensitive
nearest neighbours in the to the size of the data
Fig 2 Logistic Regression model
feature space are given a and irrelevant
new data point. attributes.
2. Decision Tree A decision tree is a Overfitting or
II. LITERATURE SURVEY supervised machine underfitting may occur
Understanding the dynamics of flood occurrences and learning system that if the tree is not pruned
bases predictions on properly. Unstable
accurately predicting their occurrence has gained input information on a because even a minor
significant attention in recent years, with advancements in hierarchical structure of change in the input can
data-driven approaches and computational models. The nodes. It recursively have a significant
research paper explores the use of Logistic regression for splits the data based on impact on tree's
flood prediction, analyzing its effectiveness in accurately feature values. structure.
forecasting flood events. Various flood detection models 3. Support Support Vector Machines Expensive in terms of
have been studied and applied in this research. The system Vector (SVM), a supervised computation for huge
aims to advance Machine Learning techniques in flood Machines machine learning datasets. Results are
technique, maximizes the challenging to
prediction and improve living conditions during such margin between classes understand. Sensitive
calamities. By leveraging advanced computational models to choose the best to the hyperparameters
and data-driven approaches, the study strives to enhance hyperplane to classify and kernel function
the accuracy of flood prediction systems for proactive data into distinct groups. selection.
disaster management and effective mitigation strategies 4. Random Random Forest is an Computationally
[7]. There is also a model that utilizes support vector Forest approach for ensemble expensive for large
regression (SVR) for the purpose of forecasting rainfall machine learning that datasets. Difficult to
values [8]. SVR algorithms are known for their superior blends various decision interpret the results.
trees to produce Can overfit if the
generalization capability compared to conventional predictions.It aggregates number of trees is too
artificial neural networks (ANN). The specific objective of the results of individual high. Sensitive to the
this study is to assess the feasibility and practicality of SVR trees to provide more choice of hyper
in accurately predicting the volume of rainfall during accurate and robust parameters.
typhoon seasons. By investigating the performance of SVR predictions, particularly
in complex and high-
in this context, the research aims to contribute valuable dimensional datasets.
insights into improving rainfall forecasting methodologies
during critical weather events [8].

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1175
 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

III. PROPOSED WORK awareness and encourage proactive measures to prevent or

The suggested method is a supervised machine learning mitigate flood-related disasters, equipping users with the
model that enhances flood forecast accuracy by using the information and tools necessary to stay informed and take
logistic regression algorithm. This model overcomes the appropriate action in emergencies.
challenges in existing systems [8] by handling missing
values and outliers, modeling non-linear relationships
between features, and providing interpretable coefficients.
Additionally, it is capable of handling large datasets and is
computationally efficient. Furthermore, this model can
handle both continuous and categorical variables and
provides interpretable coefficients, which help to
understand how each parameter contributes to the
prediction.

Fig 3 Flood Prediction using Logistic Regression workflow Fig 4 System architecture

The dataset includes a range of parameters that are Firstly, the user has to login to the application by giving
relevant to flood prediction, including the name of each your mobile number (figure 5). After logging into the
station, its latitude and longitude, the average rainfall over application, users are directed to the home screen. The
the preceding 10 days, the annual rainfall, and a binary home screen presents a variety of options for users to
indicator [6] of whether or not a flood occurred. Series of choose from as shown in figure 6. The home screen offers
preprocessing steps are conducted to clean and prepare the users a centralized platform to access crucial information
data for analysis. This included removing any incomplete and services related to meteorological conditions and
or erroneous entries, normalizing the data to ensure emergency response. It aims to provide convenience,
consistency across the different parameters, and splitting timely updates, and essential resources for users to navigate
the data into training and test datasets at a ratio of 4:1. The through weather-related challenges effectively.
training data [9] is used to refine and validate the model
and the test data to evaluate its accuracy and effectiveness
in predicting future floods in Tamil Nadu. The Logistic
Regression model is trained to predict using data collected
from 37 meteorological stations across the region for the
previous 10 days.

The development of a user interface focused on

enhancing people's understanding of flood-prone areas.
Information is made available to end-users via a mobile
application [11] . The main objective of the application is
to enable users to identify hazardous zones on a map and
stay informed about them. It also provides supplementary
data such as meteorological stations, camps, shelters, and a
list of dams, including notifications of potential flooding
resulting from dam overflow. The user interface is
designed to be user-friendly, presenting clear and concise
Fig 5 User Login Page
information in a visually appealing manner. It incorporates
interactive features for easy navigation and exploration of
flood-prone regions, aiding users in comprehending
potential risks and taking necessary precautions. The
camps and shelter interface is crucial, providing
information on nearby emergency locations and helpline
numbers for quick access to assistance. For dams, real-time
data updates trigger notifications about potential overflow
if water levels exceed safety thresholds at the base station.
Overall, the user interface aims to increase public

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1176
 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

Fig 6 Home page

Fig 8 Camps and shelters
The map view displays the locations of all 37
meteorological stations in Tamil Nadu as shown in figure The application includes a "Dams" menu that provides
7. Users can view the map and see the precise locations of a comprehensive list of all the dams in the region as shown
these stations. It also provides an option to navigate to a in figure 9. Users can access this menu to obtain
separate screen that specifically highlights the danger information about each dam. One important aspect of the
zones. These danger zones are areas that are prone to dams menu is that it offers administrators the ability to
extreme weather conditions or are at higher risk during adjust the water levels of the dams accordingly.
certain weather events. By accessing this separate screen,
users can focus on these specific areas and gather
information about potential risks or warnings associated
with them.

Fig 7 Meteorological Stations and Danger Zone

Additionally, an admin can add locations of organized

camps that are set up as temporary shelters to provide relief
and support to affected individuals. They are also displayed
on the map after validation from the respective admin.
Users can utilize this feature to find the nearest organized
camps for assistance and support during emergencies. By
submitting the latitude, longitude, and contact details of
these camps to the admin for validation, users can ensure
the accuracy and reliability of the information displayed on
the map.

Fig 9 Base stations and Dams

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1177
 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

In the event of a potential flood caused by water

overflow from a dam, users will receive a notification
through the notifications menu, ensuring timely awareness
and response. These features were carefully designed to
provide users with critical information and aid in effective
decision-making during flood events, emphasizing the
importance of preparedness and proactive measures. The
ultimate goal is to help people stay safe and minimize the
impact of flooding in Tamil Nadu, fostering resilience and
protecting lives and properties. By leveraging the latest
technology and data, a significant difference in protecting
people and communities from the devastating effects of
flooding.
IV. RESULT
After testing Logistic regression with an existing model
[12], it was found that the LR model performed better, as
illustrated by a bar chart in figure 10. The graph shows that
the LR model outperformed the current model in terms of
accuracy rate.

Fig 11 Confusion Matrix of labeled vs predicted data

Fig 10 Comparison of classifier Models

This comparison was important for validating the

effectiveness of our model and providing evidence for its Fig 12 ROC curve
potential application in real-world scenarios. Through our
experimentation with data, we were able to develop a V. CONCLUSION
model with an accuracy rate of approximately 95 percent
when tested on real-world data. To evaluate the The lack of awareness about the possibility and severity
performance of our model, Two important metrics were of floods often leads people to face numerous difficulties
used, ROC [13] and recall. The ROC curve graphically when confronted with such a disaster [14]. Thus, the
displays the true positive rate and the false positive rate at implementation of an effective Flood Prediction System
various categorization criteria. The percentage of actual could prove to be incredibly beneficial in many ways. One
positive cases that the model properly detected is known as of the primary advantages of this system would be its
recall, on the other hand. A higher recall score indicates a ability to raise awareness among the general public about
better performance of the model in correctly identifying the potential risk of floods. By providing accurate and
positive cases. By calculating both ROC and recall values timely predictions of flood occurrences, people can be
in figure 11 for our model, a comprehensive understanding better informed and prepared to face the situation [15].
of its performance is gained and determines its They can take necessary precautions and reach out to
effectiveness in predicting flood occurrences. relevant authorities, camps, and helplines for assistance.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1178
 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT)

REFERENCES

[1] A. Mosavi, P. Ozturk, and K. Chau, "Flood Prediction

Using Machine Learning Models: Literature Review,"
Water, vol. 10, no. 11, p. 1536, Oct. 2018, doi:
10.3390/w10111536.
[2] Mekanik, F.; Imteaz, M.; Gato-Trinidad, S.; Elmahdi,
A. Multiple regression and artificial neural network for
long-term rainfall forecasting using large scale climate
modes. J. Hydrol. 2013, 503, 11–21.
[3] Vapnik, V. N. (1995). The nature of statistical learning
theory. Springer.
[4] Hastie, T., Tibshirani, R., & Friedman, J. (2009). The
elements of statistical learning: data mining, inference,
and prediction. Springer.
[5] J. L. Alzen, L. S. Langdon, and V. K. Otero, "A
logistic regression investigation of the relationship
between the Learning Assistant model and failure rates
in introductory STEM courses," Int. J. STEM Educ.,
vol. 5, Art. no. 56, Dec. 2018.
[6] K. Lahiri and L. Yang, "Forecasting Binary
Outcomes," in Handbook of Economic Forecasting, G.
Elliott, C. Granger, and A. Timmermann (eds.), 1st
ed., vol. 2, ch. 0, pp. 1025-1106, Elsevier, 2013. doi:
10.1016/B978-0-444-62731-5.00019-1.
[7] Saiesh Naik, Srushti Ashok Patil, Anang Verma, Anil
Hingmire “Flood Prediction using Logistic Regression
for Kerala State”, IJERT,2021.
[8] Ping-Feng Pai, Wei-Chiang Hong : Potential
assessment of the support vector regression technique
in rainfall forecasting , Water resource
management,2007.
[9] Y. Roh, G. Heo, and S. E. Whang, "A Survey on Data
Collection for Machine Learning: A Big Data - AI
Integration Perspective," arXiv:1811.03402v2
[cs.LG], Aug. 2019.
[10] Kruti Kunverji, Krupa Shah, Nasim Shah : A Flood
Prediction System Developed Using Various Machine
Learning Algorithms, ICAST2021 , 2021
[11] V. Sharma and A. K. Tiwari, "A Study on User
Interface and User Experience Designs and its Tools,"
World J. Res. Rev. (WJRR), vol. 12, no. 6, pp. 41-44,
Jun. 2021. ISSN: 2455-3956.
[12] Prodipto Das, Abhijit Paul : Flood Prediction Model
using Artificial Neural Network , International Journal
of Computer Applications Technology and Research,
2014
[13] T. Fawcett, "A Gentle Introduction to ROC Analysis,"
Machine Learning, vol. 45, no. 1, pp. 171-196, 2001.
[14] U. Nkwunonwo, M. Whitworth, and B. Baily, "A
review of the current status of flood modelling for
urban flood risk management in developing countries"
Scientific African, vol. 7, p. e00269, March 2020.
[15] D. Perera, O. Seidou, J. Agnihotri, and A.W. M.
Rasmy, "Flood Early Warning Systems: A Review of
Benefits, Challenges, and Prospects," in Proceedings
of the United Nations University - Institute for Water,
Environment and Health (UNU-INWEH), August
2019, DOI: 10.13140/RG.2.2.28339.78880.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:35:01 UTC from IEEE Xplore. Restrictions apply.
1179