Abstract: Weather forecast is essential for agriculture, transportation, disaster management, and environmental planning. Traditional numerical and statistical models are helpful, but they do not capture the complex, nonlinear, and dynamic relationships among numerous atmospheric variables. To address this problem, the study combines and evaluates several approaches for multivariate weather forecast using Artificial Intelligence (AI) and Deep Learning (DL). By applying meteorological datasets with different parameters, including temperature, humidity, wind speed, and precipitation levels, the study evaluates more sophisticated architectures for deep learning (e.g., Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit based (GRU), and a hybrid Convolutional Neural Networks-Recurrent Neural Networks (CNN-RNN) model), as well as machine learning regressors (e.g., Random Forest, Gradient Boosted Regression). The contribution of the hybrid CNN-RNN model shows that it better captured the underlying spatial-temporal dependencies in the dataset, yielding the best forecasting accuracy and precision compared to all other models. In general, results showed a greater decrease in prediction error (RMSE, MAE) and improved consistency across various climatic conditions. Overall, work serves as a unified framework showing the potential of AI-based forecasting systems to enhance the accuracy, reliability, and representation of weather prediction systems.

Keywords: Weather forecasting, Artificial Intelligence, Deep Learning, Machine Learning, CNN, RNN, LSTM, GRU, Hybrid Models, Multivariate Prediction, Atmospheric Parameters, Time-series Analysis, Meteorology, Predictive Modeling.

Downloads: | DOI: 10.17148/IARJSET.2025.121022