www.ijcrt.

org © 2023 IJCRT | Volume 11, Issue 5 May 2023 | ISSN: 2320-2882

Weather Prediction Using Machine Learning

Mrs. Anjali Kadam, Shraddha Idhate, Gauri Sonawane, Rucha Sathe, Poonam Gundale Department of
Computer Engineering, Bharati Vidyapeeth’s College of Engineering for Women, Savitribai Phule Pune
University
Maharashtra , India

Abstract—To predict the weather of a particular utility of certain meteorological variables for
place at a specific time is known as Weather precision farming is considered when identifying
Forecasting. It is an application of science and them. Additionally, these surface parameters can
technology. Weather forecast is more helpful for be measured using a variety of sensors and a
people as it predicts the possibilities of changes in nearby weather station at a selected site.
weather conditions. The weather forecast can be done
in many ways, collecting data and analysing it, image Applications of weather forecasting:
processing, by using methodologies like
Normalization, Clustering, and cloud mask After the invention of instruments for measuring
algorithms, etc. This paper is based on the process of
atmospheric conditions in the 17th century,
Weather Forecasting. This system will be using
Machine learning algorithms like Support Vector systematic meteorological records were
Machine(SVM), Time series based recurrent neural kept. Undoubtedly, these early records were used
network, Random Forest, Naive Bayes, Artificial primarily by people working in agriculture.
Neural Network, and Decision Tree. This system will Planting and harvesting can be better
accept data on current weather conditions such as
current temperature, wind conditions, and humidity.
planned and executed more efficiently if all long-
It will predict the weather conditions of a certain term weather patterns are predicted in advanced.
location on a given day and time. This system will be A weather alert is a special type of short-term
useful in Air Traffic, Marine, Agriculture, Forestry, forecast. It is necessary to protect human life
Military, and Navy, etc. from abnormal weather. Weather warnings are
Keywords: Weather Forecasting, Weather prediction,
issued by governments and
machine learning, SVM, ANN, Naive Bayes. military organizations around the
world for all types of threatening weather events,
I. INTRODUCTION such as hurricanes, typhoons, or tropical
Weather forecasting is the process of predicting the storms known as tropical cyclones, depending
weather of a certain day and time in a given on locations.
location on the basis of its climate, atmosphere, and
pattern. Weather forecasting is an application of During the 1920s and 1930s, weather forecasting
science and technology which has been growing as developed into a crucial aviation instrument. In
many important factors depending on the weather order to save time loss, potential damage, and fuel
forecast. Ancient weather forecasting methods consumption in rough seas, many oceangoing
usually relied on the observations of the patterns of commerce vessels and military ships employ
events, also termed pattern recognition. Now, the optimum ship routing forecasts to plan their
weather forecast systems predict the weather based
routes. Any observer who is familiar with the
on parameters such as temperature, humidity, and
wind. These days, Machine learning and Data natural indicators in the sky can "foretell the
science algorithms are of great help in predicting weather" by interpreting the sky's appearance, the
things on the basis of old information and patterns. wind, and other local influences
Here in this system, we will be using Machine
A scientist can efficiently make the same
learning algorithms to predict the weather on the
determination using equipment at a single place.
basis of collected parameters and datasets by
accepting the location of the particular region. This The data from numerous such observations
system will be a website with an effective gathered at various locations are used in the
graphical user interface. The user will log in to the contemporary method of weather forecasting.
system using his user ID and password. The Experts at various
system will need the user to enter current
information such as current temperature, humidity,
and wind. The system will take this data and predict
the weather from previous data in the database.
This system will prove to be helpful in air traffic,
marine, agriculture, military, navy, and forestry.

Application:

The analysis of the quantitative temporal weather

data is the main objective of this empirically
based work. In order to predict the weather, this
study uses 10 surface weather parameters. The

IJCRT23A5325 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org l99

www.ijcrt.org © 2023 IJCRT | Volume 11, Issue 5 May 2023 | ISSN: 2320-2882
weather stations swiftly exchange these and input them on humidity, evaporation, from the ground, clouds, rain,
a synoptic weather map. The patterns of pressure, wind, snow, and the interaction of air with the ground and sea. A
temperature, clouds, and precipitation at a certain moment daily weather forecast model is one of them. A mobile
are shown on these synoptic weather maps. Meteorologists phone displays these predictions.
can create the governing mathematical equations to
explain these changes if they have a good understanding of (3) Statistical methods:
how the atmosphere changes over time in response to In addition to numerical weather forecast calculations,
different variables. Numerical models are created using mainly statistical methods are used. These methods often
these equations to determine how the atmosphere is complement numerical methods. Statistical methods use
changing and will look in the future. Forecasters can past records ofweather data with the
utilise the results from these models to help them when assumption that the future is a repeat of
creating short- and long-term forecasts [1]. past weather. The main purpose of studying past weather
Different methods used in modern weather forecasting data is to find aspects of weather that are good indicators
are: of future events. Once these relationships are
established, the correct data can be safely used to
(1) Synoptic weather forecasting: predict future states. Only global weather can be
predicted with this method. It's especially useful when he
This is the traditional and basic approach in weather only projects one side of weather at a time.
forecasting. This method was used until the late 1950s.
“Synoptic” means that observations of various Network:
meteorological elements are related to specific observation
times. A weather map representing the state of the An increasing number of networks of telegraphed
atmosphere at a point in time is therefore a synoptic map weather stations have made synoptic forecasts possible. It
for meteorologists. To get an average view of changing happened towards the end of the 19th century. The
weather patterns, modern forecaster is able to produce synoptic weather maps of
meteorological centres produce a series of overview maps the upper atmosphere twice a day, based on
each day. An overview map like this is the basis for all radiosonde observations. Radar observations of
general weather forecasts. The task of creating the growth, movement, and
overview maps on a regular basis involves extensive features of such storms provide clues to their severity.
collection and analysis of observational data collected Meteorological observation by satellite and aircraft.
from thousands of weather stations. Years of careful Numerous requirements and purposes are served by
study of weather maps have formulated certain rules of weather forecasting. Making informed decisions and
thumb. These rules help forecasters estimate the making plans for the future are both made possible by it.
speed and direction of movement of the weather Understanding atmospheric processes and subsequently
system. making the most accurate weather predictions are the core
goals of meteorologists. The ability to adapt to the climatic
(2) Numerical methods: environment, which includes coping with common
occurrences and withstanding unfavourable extremes, is a
Numerical methods involve a lot of mathematics. crucial component of life's existence. Therefore, using
Numerical weather forecasting methods are currently used weather forecasts.
in modern weather forecasting. This approach is based on
the idea that atmospheric gases are governed by many II. LITERATURE REVIEW
laws of physics. These laws of physics can be used to
predict future weather conditions for current atmospheric From all the papers we have observed and studied that
conditions. there are many factors which are responsible for weather
prediction. Because daily weather data has multiple
A set of mathematical formulas is used to develop a parameters such as temperature, humidity, rainfall
theoretical model of the general circulation of the amount, cloud distance and size, wind speed and direction,
atmosphere. These equations are used to describe etc. All these parameters are non- linear, but they
atmospheric changes over time. These equations consider
required to be processed together to determine
certain meteorological factors such as air movement,
temperature, rainfall, humidity or weather status for the
temperature,
future day. Such type of applications need complex
models and can able to produce the required result by
generating the patterns on its own

IJCRT23A5325 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org l100

www.ijcrt.org © 2023 IJCRT | Volume 11, Issue 5 May 2023 | ISSN: 2320-2882
by performing self-learning using the training data given using real time API driven applications. Many algorithms
to the model. One of the paper was based on ANN were used as a part of model building like temperature with
algorithm which was supposed to give maximum accuracy, sampling error methods lilke RMSE, Mean, Standard
which considers above parameters and factors for weather Deviation, Artificial Neural Network (ANN), Auto-
change[1],[2],[4]-[7]. Regressive Models with eXogenous inputs (ARX), ARX-
NN, Time series modelling, Random Forest Regression
In the paper by Sumit Saha we studied an efficient (RFR). In one of the papers the incorporated regression
temperature forecasting model proposed on hybrid techniques used were Ridge Regression (Ridge), Support
Principal Component Analysis (PCA) empowered Vector (SVR), Multi-layer Perceptron (MLPR), Extra-
machine learning techniques. Datasets comprises of 8760 Tree Regression (ETR),etc [7].
rows with seven attributes, for testing purpose they used
876 numbers of data. Overall operation has been Such type of applications need complex models and can
performed into three phases. In first phase, PCA is applied able to produce the required result by generating the
to eliminate the irrelevant attributes from the datasets to patterns on its own by performing self-learning using the
improve the model accuracy. In second phase five machine training data given to the model. One of the papers talks
learning algorithms: KNN, DT, RF, SVM & AdaBoost are about smart weather forecasting where application
used to predict the test datasets. On the last stage four restricted to only temperature using various cities dataset
statistical performance indicators: MAE, MSE, RMSE & .In another paper 4 separate models were used for
Regression and training time used to identify the best experimentation on single set of data of 48 hours
fitted model [4]. consecutively , which they highly aimed to predict of
A H M Jakaria et al preferred reliability in using AI some other geographical pressure areas. Previously, use
learning models like genetic algorithms, neuro- fuzzy of Python API for data collection from a particular city
logic and neural networks. Among which neural networks based Meteorological Institutes /stations were done
is more preferred. In one of the papers based on smart , their prime try was to use API for multiple factors in
weather prediction it was mentioned that the collected real which they were not successful. [7].
weather data for the city of Nashville from
The researchers have used binary and multiclass
wunderground.com , as well as nine more cities around
Nashville: Knoxville, Chattanooga, Jackson, Bowling classification models to predict weather conditions will
Green, Paducah, Birmingham, Atlanta, Florence, and occur within a certain time limit like 24 hrs, 48 hrs,
Tupelo. For a given place and date, the wunderground API weeks, months etc. The approach in the paper suggests a
returns a list of weather observations data. The target mechanism to identify the pattern that leads to failures. It
variable is always the next day hourly temperature for city is also observed that 80% of prediction is due to tropical
they chose was Nashville. The Training set contains two temperature that varies city to city state wise, process
months of weather data starting from the 1st day of July, issues and 20% is due to technical issues like
2018. In contrast, the test set contains 7 days of data incorporating and choosing learning models for
starting from September 1, 2018 and ending on September application.
7, 2018. Essentially, the trained model predicts hourly
temperature of the 2nd September while inputting 1st III. EXISTING SYSTEM
September as test data. Similarly, temperature of
The existing models have the ability to predict the
September 3rd will be predicted based on data from
weather but they precisely focus on the present weather
September 2nd, andso on [6].
conditions which can be obtained by observations from
Uday Patkar et al compared and applied ANN and the ground, ships, aircrafts, radio sounds, doppler radar
ARXNN (Autoregressive Neural Network with and satellites. Meteorologists generally use the
Exogenous Input) two different models on input data combination of several different methods to come up
learning models and resulted in an application Introducing with the predictions i. e. persistence forecasting, synoptic
precipitation as an input in the ARX model was shown to forecasting, statistical forecasting and computer
slightly improve the prediction performance. We have forecasting. Now there are forecast systems that use the
studied researchers are onto developing learning models dominant technology for climate prediction.
with utmost accuracy in prediction using historical and
some researchers

IJCRT23A5325 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org l101

www.ijcrt.org © 2023 IJCRT | Volume 11, Issue 5 May 2023 | ISSN: 2320-2882
IV. PROPOSED SYSTEM Analysis of Historical Data for the Contiguous United
Our proposed model precisely focuses on data collection States. Journal of Climatology & Weather Forecasting. 2.
by fetching real-time data via APIs of various 10.4172/2332- 2594.1000110.
[3] Afeera Saeed 110, Shamsa Muhammad 153, Muhammad
Meteorological Institutes /stations per city , but fetching SaadButt 153, Mohsin Aamir 134, Zohaib Khalid 16
with better and efficient time complexity including Software Requirements Specification for
multiple factors with more accuracy and easy to understand ACCUWEATHER Version
GUI display of data on applications using weather forecasting 1.0 FURC Date: 1 Jan, 2020
updates on various platforms. The system is supposed to be [4] Susmita Sena , Sumit Sahab , Sudipta Chakic , Payel
Sahad , Pijush Dutta, Analysis of PCA based AdaBoost
as accurate as possible as it uses all the accurate, fast and Machine Learning Model for Predict Mid-Term Weather
proper machine learning algorithms. From all the papers Forecasting,
we concluded using multiple learning models for our October 2021
application as expecting the reliability of the input data to [5] P. Kuppusamy, K. Jayalakshmi, C. Himavathi,
be more complex and multiple in numbers. “Application of Machine Learning for Weather Forecasting
Using Artificial Neural Networks,” International Journal
We hope to cover our learning model with fulfilling of Computer Sciences and Engineering, Vol.07, Issue.06,
pp.24-27, 2019.
essential requirements in user-based applications like
[6] A H M Jakaria, Md Mosharaf Hossain, Mohammad
weather alert adversaries considering present global Ashiqur Rahman, Smart Weather Forecasting Using
warming climate changes, Navy and Marine board of Machine Learning: A Case Study in Tennessee; 25 August
panels, Military applications and the most important the 2020
agriculture sector that highly gets affected by climate [7] Prof, Uday Patkar, Mr. Sanskar Maske and Mr. Saffa
Ahmad , Mr. Rushikesh Mengade, Mr. Gaurav Sadawarti;
change. “Machine Learning for Weather Forecasting Using Freely
Available Weather Data in Python”; August 2022
V. CONCLUSION
[8] Biswas, Munmun & Dhoom, Tanni & Barua, Sayantanu.
(2018). Weather Forecast Prediction: An Integrated
1) In the era of global warming, nobody can guarantee
Approach for Analyzing and Measuring Weather Data.
when the climate will change and how it will affect major International Journal of Computer Applications. 182. 20-24.
factors like agriculture, etc. So, it's necessary to know what 10.5120/ijca2018918265.
is going to happen weather- wise to take the further steps. [9] Swati Pandey, Shruti Sharma, Shubham Kumar, Kanchan
It is extremely important for accuracy and it widely Bhatt, Dr. Rakesh Kumar Arora, "Weather Forecast
through Data Mining", International Journal of Scientific
benefits the economy.
Research in Computer Science, Engineering and
Information Technology(IJSRCSEIT), ISSN : 2456-3307,
2) Weather forecasting and weather monitoring are being
Volume 7 Issue 3, pp.90-95, May- June 2021.
more and more relevant these days, considering the latest [10] Rahul, Govind & Singh, Saumya & Dubey, Saumya.
technologies. The latest technologies used for weather (2020). Weather Forecasting Using Artificial Neural
forecasting include the machine learning algorithms. Data Networks. 21-26. 10.1109/ICRITO48877.2020.9197993.
science algorithms play an important role in the
predictionstoo.

3) This study demonstrates the use and application of the

weather forecasting technology that will be helpful in the
near future to build prediction machinery and websites
that can be used to forecast the climate change of a certain
location just by use of a little information about it. It can be
an importanttool to expand the economy, worldwide.

REFERENCES:
[1] Iseh. A. J., Woma. T. Y., 2013, Weather Forecasting
Models, Methods and Applications,
INTERNATIONAL JOURNAL OF ENGINEERING
RESEARCH &
TECHNOLOGY (IJERT) Volume 02, Issue 12
(December 2013)
[2] Zhu, Audrey & Pi, Halton. (2014). A Method for
Improving the Accuracy of Weather Forecasts Based
on a Comprehensive Statistical

IJCRT23A5325 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org l102