Imputing Missing Precipitation Data at Benin Synoptic Stations (West Africa) by Using Machine Learning Methods

Abstract Missing precipitation data is frequently an unavoidable issue in many water-related studies and can negatively affect the study’s results. To overcome this issue, missing data imputation techniques are used. The most suitable model for imputing missing data at six synoptic stations in Benin is identified by using daily rainfall records from 1951 to 2010. Five Machine Learning models were tested: (a) decision tree, (b) random forest, (c) support vector machine (SVM), (d) multi-layer perceptron and (e) multiple linear regression. Among them, the support vector machine model yielded the best performance, although it varied unpredictably across stations and months. In contrast, the multilayer perceptron and decision tree models showed the poorest performance. Moreover, all models generally performed better during dry months than wet ones. No correlation was observed between station latitude and model performance.