ALGORITMA GENETIKA: STUDI KASUS MASALAH MULTI-CRITERIA DECISION ANALYSIS (MCDA) DALAM HAL ADA DATA KOSONG

Septian Rahardiantoro, Totong Martono, Bagus Sartono

Abstract


Many methods on Multi-Criteria Decision Analysis (MCDA) are used to rank the m alternatives A1, A2, ..., Am based on the n criteria C1, C2, ..., Cn. MCDA data can be presented in a decision matrix [Amn] containing aij , the value in the i-th alternative and j-th criterion. The solution on MCDA methods is obtained by giving wj , weighted value on the j-th criterion which is suitable with its role. The optimization concept of Spearman’s correlation in every pairs of solution candidate with all of criterias as a measure of goodness of the solution using genetic algorithm seems to be an alternative solution method for MCDA, even though, it is assumed that each criterion vector on matrix A should be positively correlated. It is indicated from the results of the simulation against 30 alternatives with 15 criterias, genetic algorithm provides a solution that is high correlated with a result using the AHP method, the correlation of 0.94. Besides the treatment of missing value will be much simpler to use genetic algorithms and the result will be a high correlation between the ranking of alternative simulation from the complete data with alternative rankings contained missing value as much as 10% to 40%; all correlations were worth more than 0.85. A case study of 29 automobile brands with 11 criteria and contains 20% of missing value resulting Model-Y, Model-1, and Model-3 as the best sequence of three consumer preferred brands. 

Keywords-correlation; genetic algorithm; MCDA; missing

value;


Full Text:

PDF


DOI: https://doi.org/10.29244/xplore.v1i1.12402

Redaksi: Departemen Statistika - Jl. Meranti Wing 22 Level 4, Kampus IPB Darmaga Bogor 16680 (email : xplore.stkipb@yahoo.com)

hit
counter pages