Pembentukan Rule Otomatis Berbasis Maximum Membership pada Fuzzy Tsukamoto untuk Prediksi Persediaan Pakaian

Abstract

Effective inventory management is essential for clothing retailers to avoid overstock and stockout, both of which can increase operational costs and reduce customer satisfaction. The Fuzzy Tsukamoto method has been widely used for inventory prediction; however, its rule base is generally constructed manually using expert knowledge, making it less adaptable to different datasets. This study proposes an automatic rule generation approach for the Fuzzy Tsukamoto method based on historical inventory data. The proposed method applies the Maximum Membership principle to assign linguistic labels with the highest membership degree for each input variable, thereby generating fuzzy rules automatically without expert intervention. The model uses three input variables, namely initial stock, incoming goods, and outgoing goods, to predict the final stock. The proposed approach was evaluated using 12 historical inventory records, and prediction performance was measured using the Mean Absolute Percentage Error (MAPE). The experimental results show that the proposed method automatically generated 10 unique fuzzy rules and achieved a MAPE of 14.97%, which is lower than the 17.99% reported in a previous study. These findings demonstrate that automatic rule generation can simplify rule construction while providing satisfactory prediction accuracy for clothing inventory management.