Hybrid Weather Forecasting: Integrating LSTM Neural Networks and Random Forest Models for Enhanced Accuracy

In the context of Delhi's escalating air pollution crisis, characterized by alterations in atmospheric constituents leading to temperature variations and adverse effects on public health, there is an immediate need for effective measures. This paper focuses on developing an advanced predictive model to help implement timely precautions. The proposed model integrates LSTM (Long Short-Term Memory) neural networks and Random Forest algorithms into a powerful hybrid system. This hybrid approach combines LSTM's ability to capture temporal dependencies with Random Forest's ensemble-based learning, enhancing the accuracy of weather predictions. A quantitative comparison is conducted between the hybrid model and individual LSTM and Random Forest models, evaluating metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The results show that the hybrid model outperforms the individual models, with an MAE of 0.0126, MSE of 0.0002, and RMSE of 0.0168, indicating improved prediction accuracy. The study underscores the importance of such integrated models in enabling precise forecasting and empowering decision-makers to take informed actions to address atmospheric conditions in Delhi and similar urban environments.