ClimateCast: ML-Powered Temperature Forecasting for North 24 Parganas

Global warming is increasing at a rapid rate. Several efforts are made to understand its influence worldwide and in particular regions as well. Therefore, temperature prediction has become one of the most important and influential research topics because of its application in agricultural, ecological, industrial, and environmental sectors. This study reviews dif-ferent machine learning algorithms for temperature prediction and observes their superiorities. The area of study was done in North 24 Parganas, West Bengal, India. Parameters used in the dataset of the North 24 Parganas include wind speed, different types of UV radiance, cloud amount, soil moisture, precipitation rate, surface pressure, and specific and relative humidity. The relevant features are chosen from all the parameters using feature selection. The feature selection technique used is SelectKBest, using Mutual Info Regression. The different ML models used are KNN Regression, Linear Regression (LR), Decision Tree Regression, Support Vector Regression, Random Forest Regression, LightGBM Regression, XGBoost Regression, AdaBoost Regression and Ridge Regression. The experimental results were evaluated using Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE) and R Squared (coefficient of determination). The result of this experiment revealed that XGBoost Regression has MAE, MSE, RMSE, and R Squared respectively 0.361, 0.273, 0.523, 0.989 followed by LightGBM Regression with MAE, MSE, RMSE and R squared followed by 0.366, 0.276, 0.526 and 0.988. These results show that XGBoost Regression possesses a high capability for temperature prediction