Analysis of Crossover Probability (Pc) on Genetic Algorithm Performance in Optimizing Course Scheduling in the Unimed Electrical Engineering Study Program
Rudi Salman (a*),Irfandi(b),Suprapto(c),Sayuti Rahman(d),Herdianto(e)

(a) Electrical Engineering (Universitas Negeri Medan)
*rudisalman[at]unimed.ac.id
(b) Education of Physic (Universitas Negeri Medan)
(c) Mechanical Engineering (Universitas Negeri Medan)
(d) Information system (Universitas Harapan Medan)
(e) Informatics Engineering (Universitas Panca Budi Medan)

Abstract

Genetic Algorithm (GA) performance speed is determined by computation time. Computing time in AG for finding the optimum value is strongly influenced by the following parameters: population size (pop size), Crossover Probability (Pc), Mutation Probability (Pm), and the selected selection method. Pc is one of the essential parameters in GA. A chromosome that will reach the best solution can be obtained from the crossover process of the two parent chromosomes. The Pc value strongly influences the crossover process. Determining the appropriate and correct Pc value indicates how large the parent chromosome will experience crossover.
The method used to analyze the effect of Pc on GA performance is changing the Pc value between 0.80-0.95. The simulation used MATLAB R2012b to obtain the best computational time for each Pc value. Meanwhile, the other GA parameters remained the same: Pm=0.05 and population size=100 for each change of Pc value.
The test results using MATLABR2012 show that the fastest computing time is in the range of Pc values between 0.85-0.95 with an average computation time of 0.14564s. This indicates that for the case of optimizing the scheduling of courses in the Unimed Electrical Engineering study program, the Pc value between 0.85-0.95 will provide the fastest computation time.

Keywords: Crossover probability, Genetic Algorithm, Computation time, Optimization, Course scheduling

Topic: Applied Sciences and Information Technology