Course Scheduling Optimization Using Genetic Algorithms with Fuzzy Tsukamoto-Based Fitness Adjustment

Abstract

This study develops a course scheduling optimization system that integrates a genetic algorithm with the Tsukamoto fuzzy inference system to dynamically adjust fitness values. The objective of this research is to overcome the limitations of conventional genetic algorithms that rely on static fitness functions, which only evaluate schedule quality based on the number of constraint violations. The Tsukamoto fuzzy inference system is designed with three input variables: constraint violation level, lecturer workload distribution, and classroom utilization efficiency. It employs 27 fuzzy rules based on triangular and trapezoidal membership functions to produce a fitness adjustment factor.The research methodology consists of four stages: requirements analysis and problem modeling, Tsukamoto fuzzy inference system design, hybrid genetic algorithm implementation, and performance testing and evaluation. Experiments were conducted using a synthetic dataset comprising 50 courses, 20 lecturers, 15 classrooms, and 30 weekly time slots. The results show that the proposed hybrid genetic algorithm achieves 42% faster convergence with an average fitness value of 0.89 compared to 0.76 in the conventional algorithm. Constraint satisfaction improved from 82.4% to 94.7%, lecturer workload distribution became more balanced with the coefficient of variation decreasing from 0.34 to 0.19, and classroom utilization efficiency increased from 76.8% to 88.5%. Statistical tests indicate a significant difference (p-value < 0.001) with a substantial effect size (Cohen’s d = 1.23). This research contributes to the development of a hybrid approach that integrates the Tsukamoto fuzzy inference system into genetic algorithms, resulting in more optimal, adaptive, and efficient course schedules compared to conventional methods.