Rainfall Prediction using Wavelet Transform and Transformers

Abstract

Rainfall prediction is crucial for various applications, yet time series data in meteorological datasets poses a significant challenge. This research proposes a novel approach that combines the power of wavelet transform and transformer architecture to develop an accurate rainfall prediction model. By decomposing time series data and leveraging self-attention mechanisms, this model captures complex temporal dependencies and spatial patterns. Through extensive evaluation and comparison using different metrics, the proposed algorithm demonstrates the superiority over existing methods in terms of predictive accuracy. The proposed model has been compared with LSTM model to evaluate its effectiveness in rainfall prediction and has measured a loss of 0.060, mean absolute error of 0.05, mean absolute percentage error of 0.05 and root mean square error of 0.10. The proposed model empowers decision-makers with reliable rainfall predictions, aiding improved planning and preparedness.