Year-wise Longitudinal Comparative Analysis of Temperature Prediction Models for Hyderabad (2017–2024)

Objective: The study aims to compare and analyze temperature forecasting models for Hyderabad from 2017 to 2024 in the context of modeling long-term accuracy and the influence of critical meteorological conditions on temperature. Methods: Past meteorological data, including temperature, humidity, solar radiation, wind speed, and precipitation, were analyzed. Five models for predicting were applied: Multiple Linear Regression, Decision Tree Regression, Random Forest Regression, Artificial Neural Networks (ANN), and Long Short-Term Memory (LSTM) Neural Networks. A chi-squared test determined the connections between temperature and qualitative meteorological features. The performance of models was evaluated based on R2, Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). Findings: Maximum and minimum temperatures had strong and persistent correlations with mean temperature throughout all years, whereas the effect of humidity decreased significantly in 2024. Between models, Multiple Linear Regression presented the best accuracy (R2 = 0.989, RMSE = 0.4246) in 2024, closely followed by Random Forest. Decision Tree and ANN models had mid-level performance, whereas LSTM had poor performance with negative R2 values for some years. These findings emphasize the strength of linear models and the weakness of deep learning for this data. Novelty: Through its comparative analysis spanning several years, this study presents revealing insights into the resilience of regression models, the drawbacks of deep learning, and the need for adaptive forecasting techniques in urban climate modeling. Keywords: Temperature Prediction, Weather forecasting, Meteorological variables, Machine learning models, Hyderabad Climate