Machine Learning as an Effective Technique for Rainfall Forecasting: A Literature Review

Accurate rainfall forecasting is vital for socio-economic planning in climate-sensitive regions. Traditional statistical models often fail to capture the non-linear and stochastic nature of rainfall. This paper conducts a systematic review of machine learning (ML) techniques applied to rainfall forecasting, covering 45 studies published between 2012 and 2024, following PRISMA guidelines. The analysis identifies four high-performing algorithms: Long Short-Term Memory (LSTM), Random Forest (RF), NeuralProphet, and Support Vector Machines (SVM). LSTM models optimized with Modified Particle Swarm Optimization (M-PSO) achieved the lowest Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). RF models demonstrated robustness for short-term forecasts, SVMs performed well with smaller datasets, and NeuralProphet offered explainability through a hybrid statistical–deep learning approach. Model choice depends on data characteristics, forecasting horizon, and the balance between accuracy and interpretability. The findings highlight the comparative strengths of these algorithms across different forecasting horizons.