JAC : A Journal Of Composition Theory ISSN : 0731-6755

RAINFALL PREDICTION USING MACHINE LEARNING

1
Sk Arshiya Julma, 2Kotha Pravalika, 3Bodla Sai Supriya, 4K.Hari Chandhana, 5Challa Harshini
1
Assistant Professor, 2,3,4,5UG Students, 1,2,3,4,5Dept. of CSE (AI & ML), Malla Reddy Engineering College for
Women (Autonomous), Hyderabad, India. E-Mail: arshiyashaik06@gmail. com

ABSTRACT
The prediction of rainfall is one of among the most significant methods of forecasting the weather in any
country. This paper presents a rainfall forecast model using MLR (multiple linear regression) for the Indian
dataset. As a result of the input data's multiple meteorological variables, forecasting rainfall is more accurate.
Accuracy and correlation are the measures used to assess the suggested approach. The results show that the
proposed model for machine learning outperforms the methods currently used in the literature.

INTRODUCTION
The Rainfall Prediction project is aimed at developing machine learning that can predict rainfall based on
historical weather data. This documentation provides an overview of the project, including its objectives, dataset
used, model architecture.
Rainfall prediction is significant to Indian culture and has a significant impact on people's daily lives. The
meteorological department has the difficult task of making uncertain predictions about the frequency of rainfall.
With the meteorological conditions changing, it is challenging to anticipate the rainfall with accuracy. Predicting
the amount of rain during the summer and the rainy season is difficult. Researchers from around the world have
created a variety of models to forecast rain fall, most of which use random numbers and are related to climatic
data.

Data Pre-processing: Handle missing values will fill or remove missing values in the dataset. Normalized
features will perform scaling and it will scale the numerical features to a common range. And it will preprocess
the encode categorical features for convert categorical variables into numerical representations, if applicable
Feature Selection: Analyze the correlation between features and select relevant features for the model.
Model Training: Split the dataset into training and testing sets. Select an appropriate machine learning
algorithm (e.g., Random Forest, Gradient Boosting, etc.). Train the model using the training set.
Model Evaluation: Use suitable metrics, for instance the mean absolute error (MAE) or room mean squared
error (RMSE), to assess the performance of the model. Comparing the algorithm's predictions to the actual
rainfall values for the testing set.
Model Deployment: Once the model achieves satisfactory performance, it can be deployed to make rainfall
predictions against the actual rainfall values from the testing set.
There right now are no reliable ways to forecast whether it will rain today or not. Sometimes even the forecasts
of the meteorological autonomy are wrong.
This problem is related to Rainfall Prediction using Machine Learning because machine learning models tend to
perform better on the previously known task which needed highly skilled individuals to do so.

Volume XVI, Issue VII, JULY 2023 Page No: 333

JAC : A Journal Of Composition Theory ISSN : 0731-6755

LITERATURE SURVEY
There are right now no accurate ways to predict whether or not it will rain today. Even the predictions made by
the meteorological authority can occasionally be inaccurate.

The following keywords ("machine learning" OR "deep learning") AND ("precipitation prediction" OR "rainfall
prediction" OR "precipitation nowcasting") were employed to gather articles from 2016 to 2020. Out of the
almost 1240 findings, only supervised rainfall prediction papers using meteorological information from
platforms such as radar systems, spacecraft, and observatories were chosen., while publications that used data
from regular cameras, for example, pictures, were not included. Although the review's main focus is on rainfall
forecasting, similar techniques can be used to predict other geophysical variables like temp and wind. As a
result, the chapter's conclusions and debates are adaptable to other conditions. Except for one paper that was
issued in 2015 and is an important contribution in this subject, there are a total of 66 reviewed publications,
which are composed of conference and journal articles published from 2016 to 2020.

Figure 1 Pie Chart

EXISTING SYSTEM
Rainfall is triggered on by abrupt changes in the weather. The meteorological department was unable to provide
precise forecasts. The two most popular techniques were statistical analysis and numerical analysis. Data on
rainfall are not linear in nature. The three important features are distribution, precipitation, and amount of
rainfall. These standards can differ from one location on earth to another. Traditional techniques have some
drawbacks. These drawbacks affect predicting the precise amount of rainfall. Only linear procedures will be
compatible with statistical methods.
DISADVANTAGES
Using statistical methods to predict the rainfall we will face many difficulties.
By using numerical analysis and precipitation the accurate results are not found.

PROPOSED SYSTEM
By using Random Forest and SVM classifier this predictor will be giving the accurate values. The rainfall
predictor will analyze the dataset and give the accurate values according to the inputs. In this the data

Volume XVI, Issue VII, JULY 2023 Page No: 334

JAC : A Journal Of Composition Theory ISSN : 0731-6755

visualization is also done together with the help of graphs.

SYSTEM ARCHITECTURE
•

•
• Figure 2 System Architecture
ADVANTAGES
● Accurate prediction of rainfall.
● The ability to recognise trends and patterns in past meteorological data is what machine learning for
forecasting rainfall does.
● Such data can be used to create more precise climate models and gain a better understanding of the
factors that influence weather patterns.
● Limit threats to lives and property as significantly as achievable, and enhance the management of farms.
farms..

DEFINE THE PROBLEM

A more accurate prediction system needs to be used over a timely notify to help minimise dangers to people and
property and also help manage agricultural farms easier. Forecasting rainfall is crucial because heavy and
inconsistent precipitation can have many effects such as the destruction of crops and farms and damage to
property. Each year, people experience disasters like droughts and floods all across the world as a result of
excessive rainfall. To assess the rainfall and forecast the likelihood of rain, many models have been devised.
Both unsupervised and supervised algorithms for machine learning constitute the foundation of these models.
Considering total rainfall won't enable us to determine if it rains under particular circumstances. The main issue

Volume XVI, Issue VII, JULY 2023 Page No: 335

JAC : A Journal Of Composition Theory ISSN : 0731-6755

with machine learning is accuracy. To forecast whether it will rain given certain conditions, we must first
comprehend the data and train the algorithm accordingly.

MODULES AND FUNCTIONALITIES

Three regional SVR and/or ANN made up the modular models. Two data-preprocessing methods, MA and SSA,
were examined using a three-layer feed-forward ANN. The MA was superior to the SSA, according to the
results. As test examples, four rainfall records from India and China—India, Zhongxian, Wuxi, and Zhenwan—
were employed.
Decision trees, k-nearest neighbours, linear regression, and methods based on rules are examples of algorithms
for machine learning used for rainfall prediction.

IMPLEMENTATION
LIBRARIES
Pandas:
Pandas is a sophisticated Python library that is freely available and provides outstanding performance.
Applications for processing information and analysis. Python was mostly used for data munging and cleaning.
On data analysis, it had little of an effect. Pandas found the answer. Whatever source where the data came from,
we may use Pandas to carry out the five typical phases of data processing and analysis: prepare, modify, model,
and analyse. Various professional and academic areas, such finance, economics, research, analytics, etc., use
Python with Pandas.
Numpy:
NumPy is an universal library that manages arrays. It gives you a very fast multidimensional array object in
addition to a way to interact with these arrays. This Python package is key to scientific computing..
It has a number of features, including these important ones:
● A strong object in an N-dimensional array.
● Complex (broadcasting) operations.
● C/C++ and Fortran code integration tools.
● Comprehensive knowledge of the Fourier transform, arbitrary numbers, and linear algebra.
An efficient multi-dimensional data store with numerous uses beyond study is NumPy. NumPy is able to quickly
and simply connect with a variety of databases thanks to its ability to define any data-types..
Flask:
The Flask web-based application framework was developed in Python. Armin Ronacher, the team leader of
Poocco, a global community of Python followers, developed it. Flask is built on the Jinja2 template engine and
the Werkzeug WSGI framework.Both of them are Pocco initiatives. Web framework Flask is a Python tool that
renders it simple to develop web apps. It's a microframework without an object relationship manager or any
other comparable features, but its core is compact and simple to customise. Yes, it includes multiple outstanding
characteristics such as a template engine and url routing. It is a WSGI web-based application framework.
Datatime:
Python comes with a built-in module called DateTime that can handle times and dates in a variety of
ways. We'll look at some fundamental Python DateTime operations in this tutorial. The datetime
module contains the following six primary object classes and their corresponding components:
1. datetime.timezone
2. datetime.datetime
3. datetime.tzinfo
4. datetime.time

Volume XVI, Issue VII, JULY 2023 Page No: 336

JAC : A Journal Of Composition Theory ISSN : 0731-6755

5. datetime.timedelta
6. datetime. date

Pickle:
The pickle module implements binary protocols for serialising and deserializing Python object structures.
Pickling is the process of converting a Python object structure onto a byte stream. Unpickling is the reversal
process of Pickling in which a byte stream is turned into an object hierarchy..

ALGORITHMS
The Flask application uses a machine learning model to predict whether it will be a sunny or rainy day
based on user input. However, the specific machine learning algorithm used for prediction.
The steps involved in the algorithm are as follows:
1. The Flask application is initialized with the necessary dependencies and template folder.
2. The pre-trained model is loaded from the "cat.pkl" file using the pickle.load() function and stored in the
model variable.
3. The home route ("/") is defined, which renders the "index.html" template when accessed via a GET request.
4. The prediction route ("/predict") is defined, which handles both GET and POST requests. If a POST request
is received, the code retrieves the input data from the form submitted by the user.
5. The input data is processed, including extracting the day and month from the provided date, and converting
various form inputs to appropriate data types (float in this case).
6. The processed input features are stored in a list named input_lst.
7. The machine learning models predict () method is called with input_lst as the input to make the prediction.
The predicted output is stored in the output variable.
8. Based on the predicted output, the application renders different templates ("after_sunny.html" or
"after_rainy.html") to display the appropriate message to the user.
9. If the request method is not POST (e.g., it is a GET request), the application renders the "predictor.html"

TESTING THE SYSTEM

The goal of evaluation is to find flaws. Testing is the practise of attempting to find every possible flaw or
vulnerability in a product of work. It additionally makes it feasible to test various components, subassemblies,
assemblies, and/or complete products. Software testing is the method for guaranteeing that it complies with
requirements, meets customer expectations, and does not malfunction in a way that is unacceptable. There are
different types of tests. Every test type was developed to satisfy a particular testing requirement.
TEST CASE
TEST ID TEST INPUT EXPECTED ACTUAL RESULT
DESPRICTION OUTPUT OUTPUT

1 Collecting Temperature Rainfall Rainfall success

information information

2 Collecting Temperature Rainfall Rainfall success

information information

Volume XVI, Issue VII, JULY 2023 Page No: 337

JAC : A Journal Of Composition Theory ISSN : 0731-6755

3 Collecting Temperature Rainfall Windy success

information information

4 Collecting Temperature NoRainfall Sunny success

information information

5 Collecting Temperature NoRainfall Windy success

information information

Table.1. Test Case Report

Test Results: All of the abovementioned test cases were effective. There were however no errors detected..

RESULTS

Volume XVI, Issue VII, JULY 2023 Page No: 338

JAC : A Journal Of Composition Theory ISSN : 0731-6755

Figure.3. Univariate Visualization

Volume XVI, Issue VII, JULY 2023 Page No: 339

JAC : A Journal Of Composition Theory ISSN : 0731-6755

Figure.4. Heat Map

Volume XVI, Issue VII, JULY 2023 Page No: 340

JAC : A Journal Of Composition Theory ISSN : 0731-6755

Figure 5 Weather

Figure 6 Rainfall

CONCLUSION
The general goal is to describe various machine learning approaches that can be used to predict rainfall. The
purpose of this study is to create precise and effective models with fewer features and ttests. The data is first pre-
processed before being used in the modelling process. The most efficient categorization algorithms are K K-
Nearest Neighbour (87% efficiency) and Random Forest (88% efficiency). The Decision Tree classifier, on the
other hand, has the lowest accuracy (73%). This research can be expanded to include other ML approaches such
as time-series
series data, clustering and association rules, and other ensemble techniques. Given the limits of this
research, there will be a need to develop more complicated
complicated and combined models in order to improve the
accuracy of rainfall forecast systems. A study may also be made utilising more detailed monitoring for a specific

Volume XVI, Issue VII, JULY 2023 Page No: 341

JAC : A Journal Of Composition Theory ISSN : 0731-6755

area and creating this type of model for a large dataset so that the computation rate can be enhanced with greater
precision and accuracy.
FUTURE SCOPE
Rainfall prediction is critical due to high and unpredictable precipitation can have many consequences such as
crop destruction and property damage, so an improved forecasting model is required for warning in advance that
can minimise dangers to life and property while also better managing agricultural farms.As a result of early
projections, we can even take preventative measures..

REFERENCES
[1] Manojit Chattopadhyay, Surajit Chattopadhyay, “Elucidating the role of topological pattern discovery and
support vector machine in generating predictive models for Indian summer monsoon rainfall”, Theoretical
and Applied Climatology, pp. 1-12, July 2015, DOI: 10.1007/s00704-015-1544-5.
[2] Kumar Abhishek, Abhay Kumar, Rajeev Ranjan, Sarthak Kumar,” A Rainfall Prediction Model using
Artificial Neural Network”, 2012 IEEE Control and System Graduate Research Colloquium (ICSGRC
2012), pp. 82-87, 2012.
[3] Minghui Qiu, Peilin Zhao, Ke Zhang, Jun Huang, Xing Shi, Xiaoguang Wang, Wei Chu, “A Short-Term
Rainfall Prediction Model using Multi-Task Convolutional Neural Networks”, IEEE International
Conference on Data Mining, pp. 395-400, 2017, DOI 10.1109/ICDM.2017.49.
[4] Aswin S, Geetha P and Vinayakumar R, “Deep Learning Models for the Prediction of Rainfall”,
International Conference on Communication and Signal Processing, April 3-5, 2018, India, pp. 0657-0661.
[5] Xianggen Gan, Lihong Chen, Dongbao Yang, Guang Liu, “The Research Of Rainfall Prediction Models
Based On Matlab Neural Network”, Proceedings of IEEE CCIS2011, pp. 45- 48.
[6] Sam Cramer, Michael Kampouridis, Alex A. Freitas and Antonis Alexandridis, “Predicting Rainfall in the
Context of Rainfall Derivatives Using Genetic Programming”, 2015 IEEE Symposium Series on
Computational Intelligence, pp. 711 – 718.
[7] Mohini P. Darji, Vipul K. Dabhi, Harshadkumar B.Prajapati, “Rainfall Forecasting Using Neural Network:
A Survey”, 2015 International Conference on Advances in Computer Engineering and Applications
(ICACEA) IMS Engineering College, Ghaziabad, India, pp.706 – 713
[8] Sandeep Kumar Mohapatra, Anamika Upadhyay, Channabasava Gola, “Rainfall Prediction based on 100
years of Meteorological Data”,2017 International Conference on Computing and Communication
Technologies for smart Nation, pp.162 – 166.
[9] Sankhadeep Chatterjee, Bimal Datta, Soumya Sen, Nilanjan Dey, “Rainfall Prediction using Hybrid Neural
Network Approach”, 2018 2nd International Conference on Recent Advances in Signal Processing,
Telecommunications & Computing (SigTelCom), pp. 67 – 72.
[10] Mr. Sunil Navadia, Mr. Pintukumar Yadav, Mr. Jobin Thomas, Ms. Shakila Shaikh, “Weather
Prediction: A novel approach for measuring and analyzing weather data”, International conference on I-
SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC 2017), pp. 414.

Volume XVI, Issue VII, JULY 2023 Page No: 342