Attribute Selection Techniques to Clustering the Entrepreneurial Potential of Student based on Academic Behavior

Abstract

A key factor in the process of knowledge discovery in databases is the quality of data that consists of a set of attributes that explain the characteristics of the data. For that, we need the right attribute selection method for optimal data mining performance. In this case, the attributes tested with machine learning are the result of mapping factors is affecting entrepreneurship of students based on behavioral science theory on the attributes of Indonesia Higher Education Database. Testing dataset attributes using four different methods, namely Correlation, Information Gain, OneR, and Relief F. The results of clustering experiments with the Simple K-Means algorithm show that OneR method decrease in the largest drop of Sum of Squared Errors (17%) compared to the other three methods. With the most important attribute differences in each attribute selection method, the instances cluster profile generated is also different.