Analysis of Weather Data for Rainfall Prediction using C5.0 Decision Tree Algorithm

Abstract

Rainfall has an essential role in human life, including the agricultural aspect. By knowing the estimated intensity of rainfall that will fall in an area at a particular time, we can determine a good planting period for commodities that require rainfall prediction. This study aims to produce a rainfall prediction model using the C5.0 Algorithm on the Bogor Regency daily weather dataset in the previous five years (2017 - 2021). The dataset is divided into two categories, namely nine explanatory factors (date, month, minimum temperature, maximum temperature, average temperature, average humidity, sun exposure, maximum wind speed, and average wind speed) and one target class rainfall category (low, medium, and high). The best model variation was generated using a 5-fold CV, which resulted in five model partitions with a total accuracy of 86.33% in the training data and 84.22% in the test data. The resultant rules include 72 attributes, two partitions pick the average humidity as root node, and the remaining three choose the average temperature. The model rules produce rainfall prediction information that can assist in determining the best cultivation time for an agricultural commodity to increase yield productivity.