Hybrid Genetic Algorithm & Learning Vector Quantization for Classification of Social Assistance Recipients

Abstract

The social assistance program called “Rastra” is a government program that aims to ease the burden on poor families by providing food. However, the distribution of the food to prospective beneficiaries is still not accurate. So, a classification method is needed that can help to estimate the right target. In this study, the classification of social assistance recipients by using the Learning Vector Quantization (LVQ) method. LVQ weight vector is very important in the classification process because it affects the classification results. This study applies the Genetic Algorithm to optimize the LVQ weight vector to improve accuracy. The results obtained from this study indicate an LVQ accuracy of 84.16% and GA-LVQ gives a higher accuracy of 87.08%. Produces the best parameters: population size (popSize) 100, crossover rate (cr) 0.5, mutation rate (mr) 0.5, max generation 80, learning rate (a) 0.1 and reduce learning rate (dec a) 0, 1. The use of the LVQ method that is optimized using GA has been shown to provide better results, with higher accuracy values compared to the LVQ method without being optimized.