International Journal of Scientific Research in Engineering and Management (IJSREM)

Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

Weather Forecasting using Decision Tree Regression

Amanpreet Kaur1,*, Meenakshi Sharma2
1
M.tech Research Scholar, CSE Department, RIEIT, Railmajra, SBS Nagar, India

2
Head of Department, CSE Department, RIEIT, Railmajra SBS Nagar, India

1
[email protected] , [email protected]

ABSTRACT : Weather forecasting [1] is one of the most scientifically and technologically challenging problems around
the world in the last century. To make an accurate prediction is indeed, one of the major challenges that meteorologists are
facing all over the world. This research work is based on weather prediction using machine learning using the DTR which
will help us in getting good accuracy for the weather prediction and prediction of the future weather. The research also
demonstrates the existence of a long term trend in the accuracy of the forecasts.
KEYWORDS : LR(Linear Regression), DTR(Decision Tree Regression).

1. INTRODUCTION

1.1 OVERTURE: Weather forecasting entails predicting how the present state of the atmosphere will change. Present
weather conditions are obtained by ground observations, observations from ships, observation from aircraft, radio sounds,
Doppler radar and satellites. This information is sent to meteorological centers where the data are collected, analyzed and
made into a variety of charts, maps and graphs. Modern high-speed computers transfer the many thousands of observations
onto surface and upper-air maps. Weather forecasts provide critical information about future weather. There are various
techniques involved in weather forecasting, from relatively simple observation of the sky to highly complex computerized
mathematical models. Weather prediction could be one day/one week or a few months ahead. The accuracy of weather
forecasts however, falls significantly beyond a week. Weather forecasting remains a complex business, due to its chaotic
and unpredictable nature. It remains a process that is neither wholly science nor wholly art. The primary aim of the current
study is to provide such an assessment to serve:

1. Improvements in weather forecasting using old and huge dataset and showing the various trends.

2. The good accuracy in order to get the perfect mean temperature.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 1

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

2. BACKGROUND

Stern, H. (2008) reviewed the accuracy of weather forecasts for Melbourne, Australia. He proposed [2] that the analysis
shows that skill is evident in forecasts of temperature, rainfall, and qualitative descriptions of expected weather up to 7 days
in advance.

Mark Holmstrom, Dylan Liu, Christopher Vo (2016) proposed that Two machine learning algorithms were
implemented: linear regression [3] and a variation of functional regression. The input to these algorithms was the weather
data of the past two days, which include the maximum temperature, minimum temperature, mean humidity, mean
atmospheric pressure, and weather classification for each day. The output was then the maximum and minimum
temperatures for each of the next seven days.

Sue Ellen Haupt, Jim Cowie, Seth Linden, Tyler McCandless, Branko Kosovic, Stefano Alessandrini (2018) proposed
that the first big advance was in terms of numerical weather prediction (NWP), i.e. integrating the equations of motion
forward in time with good initial conditions. But the more recent improvements have come from applying artificial
intelligence (AI) techniques to improve forecasting and to enable large quantities of machine-based forecasts.

Tanvi Patil 1, Dr. Kamal Shah2 (2021) proposed LR has been used for forecasting the minimum and maximum
temperature and wind speed. The major objectives of Linear Regression: Linear regression has been used for the following
two objectives.

In order to find the relationship among variables and to estimate the values of some attributes so that new observations are
entertained.

3. PURPOSE

The purpose of the current paper is to show the non-linear trends in order to show the temperature trends since the past
years and also to predict the future weather with a good accuracy using Decision Tree Regression algorithm. In the earlier
researches where different algorithms have used like LR, Bayesian Networks [4], Neural Networks, Functional Regression
[5] etc. The accuracy that we get is quite low than what should we actually expect. We get a very good accuracy with the
approach that is being used by us. Decision trees supports non linearity, where LR supports only linear solutions [6]. When
there are large number of features with less datasets (with low noise), linear regressions may outperform Decision
trees/random forests. For categorical independent variables, decision trees are better than LR. Decision tree builds
regression [7] or classification models in the form of a tree structure. It breaks down a dataset into smaller and smaller
subsets while at the same time an associated decision tree is incrementally developed. The final result is a tree with decision
nodes and leaf nodes. A decision node (e.g., Outlook) has two or more branches (e.g., Sunny, Overcast and Rainy), each
representing values for the attribute tested. Leaf node (e.g., Hours Played) represents a decision on the numerical target.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 2

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

The topmost decision node in a tree which corresponds to the best predictor called root node. Decision trees can handle
both categorical and numerical data.

4. Results and Discussions

4.1 Data cleansing

Weather data cleaning [8] is fundamental to the provision of high quality weather data. Weather forecasters collect the data
for the weather prediction. The dataset which has been used in this model is taken from the government data portal. It
contains month wise mean weather all over India. This dataset contains mean temperature of India from the past years.

First of all, the data has been cleaned by applying numerous ways for instance, keeping the data date for January months
across all the years, converting string to the date time objects.

4.2 SHOWING THE TRENDS

The trends [9] have shown in every way like

a. Warmest, Coldest, median Monthly Temperature

b. Temperature clusters of Months giving it the interactive different colors for every month of the year
c. Frequency chart of temperature readings
d. Yearly mean temperature
e. Seasonal mean temperature throughout years.
f. Month wise temperature have been shown in an animation frame
The trend can be linear and non-linear. Now, according to our work done above, we come to know that the data is definitely
not having he linear trend.

4.3 DECISION TREE REGRESSION:

I am using Decision Tree Regression [10] as the data does not actually have a linear trend that we have proved above. This
algorithm basically breaks down a dataset into smaller and smaller subsets while at the same time an associated decision
tree is incrementally developed. The final result is a tree with decision nodes and leaf nodes. DTR [11] observes features
of an object and trains a model in the structure of a tree to predict data in the future to produce meaningful continuous
output. Continuous output means that the output/result is not discrete, i.e., it is not represented just by a discrete, known set
of numbers or values.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 3

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

4.3.1 ALGORITHM STEPS:

Start

Input the dataset of Weather collected from Govt. website and clean the data

Pre-process the data to make it look better and showing the non-linear
trends

Import libraries, datasets and then Training and testing of

the model.

Apply Decision Tree Regression Algorithm

Get the accuracy predicted with our model

Get the future weather prediction

Fig 4.3.1.1 shows the flowchart of the steps used

a. Importing the libraries: The first step will always consist of importing the libraries that are needed to develop the
ML model. The NumPy, plotly and the Pandas libraries are imported.
b. Importing the dataset: In this step, we shall use pandas to store the dataset
c. Splitting the dataset into Training Set and Testing Set: In the next step, we have to split the dataset as usual into
the training set and the test set. For this we use test size 0.3 from our dataset which means that this will only be used
as test set and the remaining will be used as training set for building the model.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 4

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

d. Training the decision Tree regression on the Training set: We import the DecisionTreeRegressor class from
sklearn.tree and assign it to the variable ‘dtr’. Then we fit the train_x and the train_y to the model by using the dtr.fit
function.
After the above steps, we find the accuracy and the accuracy [12] that we get from this model is 96% which is a way more
better than all the existing weather prediction models till now. The achievement of this accuracy is because of the huge
dataset and the model that we have used for the prediction. The huge the data, the more accurate are the results. With this
good accuracy we have predicted the next year data.

ADVANTAGES OF DECISION TREE REGRESSION

a. The decision tree model can be used for both classification and regression problems, and it is easy to interpret,
understand, and visualize.
b. The output of a decision tree can also be easily understood.
c. Compared with other algorithms, data preparation during pre-processing in a decision tree requires less effort and
does not require normalization of data.
d. The implementation can also be done without scaling the data.
e. A decision tree is one of the quickest ways to identify relationships between variables and the most significant
variable.
f. Decision trees are not largely influenced by outliers or missing values, and it can handle both numerical and
categorical variables.

Fig a) shows warmest. Coldest and median monthly temperature which will help us in showing the trends of the temperature
throughout the year.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 5

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

Fig b and c shows seasonal mean temperature throughout the year and Forecasted temperature respectively.
The data that we input is weather dataset from the past years which is to be taken from the govt. website. By inputting the
dataset to our model, we can have a mean temperature predicted for the next year.

5. CONCLUSION

This paper documents the non-linear trends in the weather forecasting and also the prediction of the next year with the 96%
accuracy using the DTR algorithm of Machine Learning. The mean temperature of the future year is to be predicted. The
seasonal weather trends has also been shown which includes summer, winter, monsoon and autumn. Knowing the weather
prior will help in many ways in each sector. Weather forecasting is the application of science and technology to predict the
state of the atmosphere for a given location. Weather forecasts are made by collecting quantitative data about the current
state of the atmosphere and using scientific understanding of atmospheric processes to project how the atmosphere will
evolve. There are a variety of end users to weather forecasts. Weather warnings are important forecasts because they are
used to protect life and property.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 6

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 06 Issue: 08 | August - 2022 Impact Factor: 7.185 ISSN: 2582-3930

REFERENCES

[1] Mark Holmstrom, Dylan Liu, Christopher Vo, “Machine Learning Applied to Weather Forecasting”, Stanford
University(Dated: December 15, 2016).

[2] Casta˜n´on, J. (10). Machine Learning Methods that Every Data Scientist Should Know. Consultado em Outubro, 16,
2019

[3] Sue Ellen Haupt, Jim Cowie, Seth Linden” Machine Learning for Applied Weather Prediction” IEEE

[4] Abramson, Bruce, et al.” Hailfinder: A Bayesian system for forecasting severe weather.” International Journal of
Forecasting12.1 (1996): 57-71.

[5] W. Myers, G. Wiener, S. Linden, and S. E. Haupt, “A consensus forecasting approach for improved turbine hub height
wind speed predictions,”in Proc. WindPower 2011, Anaheim, CA, May 24, 2011

[6] Tanvi Patil 1, Dr. Kamal Shah2(2021) ”Weather Forecasting Analysis using Linear and Logistic Regression Algorithm”
Volume: 08 Issue: 06 | June 2021

[7] Stern, H. (2008), “The accuracy of weather forecasts for Melbourne, Australia”. Met. Apps, 15: 65?71.
doi:10.1002/met.67

[8] Rahm, Erhard and Hong Hai Do. (2000) “Data Cleaning: Problems and Current Approaches.” IEEE Bulletin of the
Technical Committee on Data Engineering (23): 3-13.

[9] Sushmitha Kothapalli, S. G. Totad, “A Real-Time Weather Forecasting and Analysis”, IEEE International
Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017), pp 1567-1570

[10] L. Breiman, J. H. Friedman, R. A. Olshen, and C. J. Stone. Classification andRegression Trees. CRC Press, 1984

[11] W.-Y. Loh. Regression tree models for designed experiments. Second E.L. Lehmann Symposium, Institute of
Mathematical Statistics Lecture Notes-Monograph Series, 49:210-228, 2006

[12] T. Oates and D. Jensen. The effects of training set size on decision tree complexity. In D. H. Fisher, Jr., editor, Proceedings
of theFourteenthInternationalConferenceonMachineLearning, pages 254–262, San Francisco, CA, 1997. Morgan Kaufmann.

© 2022, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM16084 | Page 7