Optimization of Neural Network Based PSO Feature Selection in the Classification of Graduates Working According Their Field
Very Kurnia Bakti (a), Dairoh(b), Muhammad Naufal(c)

(a) DIII computer engineering study program, Parapan Bersama Polytechnic, Tegal, Indonesia.
(b) DIV engineering Information, Harapan Bersama Polytechnic, Tegal, Indonesia.
(c) DIII computer engineering study program, Parapan Bersama Polytechnic, Tegal, Indonesia.

Abstract

The many college graduates who work not in accordance with their field of knowledge. The figures obtained show that the horizontal alignmet<80% in the last three years has not reached the idea value. The question that often arises is why this can happen and what influences can determine the quality result of graduates why they don^t work in their fields.

we need a model that is used in order to see a pattern of graduates in order to work according to their scientific fields. Neural network is an algorithm method that can be used as a reliable classification algorithm but has shortcomings in its selection of features, where with the combination PSO has a good ability to solve problems that have non-linear and non-differentiable characteristics, multiple optima, large dimensions through good adaptations. derived from social psychological theory.

The combination method between the Neural Network and the PSO selection feature obtained an output accuracy of 71.51% greater than the accuracy of the Neural Network method alone, namely with a value of 64.32%.

Keywords: College, neural network , PSOPSO Feature Selection, Graduate

Topic: Engineering