Comparison of Boolean OR, AND, and OR–AND Models for Monthly Rainfall Classification in Bawean Island

Abstract

Rainfall classification plays an important role in climate monitoring, water resource management, agricultural planning, and hydrometeorological disaster mitigation. While machine learning techniques have been widely used for rainfall classification, they often require substantial computational resources and complex training processes. This study proposes a simple, interpretable, and computationally efficient Boolean-based framework for monthly rainfall classification on Bawean Island, East Java, Indonesia. Monthly climatological data from 1972–2023, including rainfall, rainy days, mean temperature, and minimum temperature, were analyzed, yielding 624 observations. Rainfall was classified into three categories: low (<100 mm), moderate (100–299 mm), and high (≥300 mm). Rainy days were converted into ordinal scores, while mean and minimum temperatures were transformed into binary scores. Three Boolean-based rainfall classification models were developed and evaluated using confusion matrices, accuracy, precision, recall, and F1-score. Correlation analysis showed that rainy days had the strongest relationship with rainfall (r = 0.858), followed by minimum temperature (r = −0.592) and mean temperature (r = −0.463). The hybrid OR–AND model achieved the best overall performance, with 66% accuracy, 71% precision, 62% recall, and 61% F1-score, outperforming both the OR and AND models. These results demonstrate that the proposed Boolean-based framework provides an effective, transparent, and computationally efficient approach for monthly rainfall classification.