Comparison Analysis of Weight Value Changing in Function Point Analysis Between Fuzzy Inference System Mamdani and Tsukamoto for Software Size Estimation

Abstract

Key of successful software development is planning. To make a good planning is needed estimate the size of the software to be built. The size of the software is usually presented in the form of Lines of Code (LOC). Function Point Analysis (FPA) is the most common method used to estimate the size of the software in units LOC. Accuracy of the FPA can be improved by adjusting the weight values in the Function Point Complexity table. Previous research does the adjustment of the weight values of fuzzy logic Mamdani, and this study was conducted to analyze the accuracy of the result comparison between Mamdani method from previous studies with Tsukamoto method. The evaluation is done by comparing the relative error between the measurement of pure FPA, FPA with Mamdani modification, Tsukamoto modification, and with the actual value of the software LOC. A total of 13 software was used for testing. The result recorded from the smallest error are FPA with Mamdani modification (2.3%), FPA with Tsukamoto modification (2.75%) and the pure FPA (3.51%). The average yield obtained processing time are: the smallest is with pure FPA method 1.1 ms, while the Tsukamoto method obtained processing time 1.9 ms and the largest is the Mamdani method is 3 ms.