Modelling for supply uncertainty of production using adaptive neuro-fuzzy system

The amount of fruit supply is unpredictable because it is seasonal, while the demand for juice constantly increases. Efforts to address uncertainty in both supply and demand necessitate the implementation of accurate production quantity forecasting. Therefore, this study aimed to use the Adaptive Neuro Fuzzy System (ANFIS) approach to model the amount of fruit production based on the uncertainty of the amount of supply. The dataset comprised 48 observation periods, and to develop a robust and reliable model, the data were partitioned into two subsets, including 75% for training and 25% for testing the ANFIS model. The model structure built with ANFIS uses three inputs, including demand quantity, fruit supply quantity, and puree availability. The supply of fruit is seasonal in nature, resulting in substantial availability during the harvest period. During this time, the surplus fruit is processed into puree and juice to ensure continuity of supply beyond the harvest season. The output produced is the amount of juice produced. Based on the most accurate error value based on RMSE (0.063), MAPE (1.55%), MAD (0.027), and R2 (94.4%). The forecasting model for fruit juice production with the ANFIS approach is the Gaussian membership function (Hybrid), with the number of memberships 3 – 4 – 5.