Trend Analysis And Machine Learning Based Short – Term Forecasting Of Heat Index And Temperature – A Case Study In Urban Chennai

Abstract

Analyzing the weather particularly heat waves in a urban city like Chennai has become essential since the temperature is high leading to high heat index particularly in hot summer April – June. Forecasting the weather parameters such as temperature and heat index will aid the humans to take necessary precautions for themselves, their pets etc. This necessitates a ease model that would predict next day or next week parameters. The proposed model is developed for short term prediction of temperature and heat index based on machine learning techniques. The proposed model is trained using historical weather data from January 2017 till May 2025 and immediate next day metrics are predicted using Random Forest regression. Since the dataset contains time information like year, days and months Long Short-Term Memory technique best fits the training and upon implementation the metrics for one week is predicted. The model is evaluated using error metrics Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). The training efficiency is validated using regression R². An average of 0.85 is the R² for Random Forest and 0.87 for LSTM. Similarly, the RSME value is less for LSTM. Hence LSTM network outperforms than Random Forest leading to prediction of increased number of days.