Towards Estimating Rainfall Using Cellular Phone Signal

A crowd-sourced method in rainfall estimation from mobile phones is attempted. The result of the study indicates high correlation between cellular signal levels and rainfall, suggesting that rainfall could be predicted using cellular signal levels from mobile phones. Custom data preprocessing methods have been employed to ensure significant results. Regression models using machine learning built upon the collected data show a borderline
$\mathbf{R}^{\mathbf{2}}$
score at only 0.39, while classification models show high performance with an average macro F1-score of 0.81 in predicting rain events instead of predicting rainfall levels. The result of this study paves the way for crowd-sourcing cellular signal data from mobile phones to better understand rainfall patterns. Further extensive data collection will need to be carried out to clarify the effectiveness of the method. This study is still limited in terms of data size.