Advanced Deep Learning Techniques for Multivariate Weather Prediction Using RNN and LSTM Models

This paper investigated advanced deep learning and traditional machine learning techniques for multivariate weather forecasting using time series data. A dataset with weather features like temperature, humidity, wind speed, and precipitation over extended periods of time was collected. Experiments were conducted with each feature of temperature, humidity, wind speed, and precipitation as the target variable to evaluate predictive accuracy in multiple scenarios. Initially, traditional machine learning regression models like Random Forest Regression (RFR), Decision Tree Regression (DTR), Gradient Boosted Regression (GBR), and Multiple Linear Regression (MLR) were applied as base models for predictive evaluation. Later, advanced deep learning models like RNN, LSTMs and GRU were applied and observed to enhance performance with respect to measures MSE, MAE, RMSE, and r-squared values. The results reveal the strengths of advanced methods over traditional ML models. This work contributes to valuable insights into weather forecast modeling.