Comparison of Mamdani Fuzzy Inference System Results Between MATLAB and Python Implementations

Abstract

Planning in determining the amount of production decision-making for the next period depends on the remaining inventory from the previous period and also the estimated amount of demand in the next period. Total demand and supply are uncertain. Fuzzy logic is a technique to analyze uncertainty. The purpose of this study is to compare the results of a Mamdani fuzzy inference system for determining the optimal production quantity of chocolate products on two computing platforms: MATLAB and Python. Both implementations used an identical system design, consisting of the same membership functions, rule base, input-output ranges, aggregation method, and defuzzification method, and the Python implementation was executed with the Scikit-Fuzzy library on Google Colab. The test scenario used input values of production cost of Rp 800.00 per pack and product demand of 45,000 packs. MATLAB produced a crisp output of 64,100 packs, while Python resulted in 80,000 packs. This study shows an output difference of 24.8%, which is attributed to differences in numerical resolution of the universe of discourse and library-specific implementation of the aggregation and centroid defuzzification operations between the two platforms. These findings indicate that Python, using the open-source Scikit-Fuzzy library, offers an accessible and cost-effective alternative for fuzzy modeling, while careful verification of the implementation details remains necessary to ensure numerical consistency between platforms. As this comparison is based on a single test scenario, further validation using additional test cases is recommended.