Application of the Relief-f Algorithm for Feature Selection in the Prediction of the Relevance Education Background with the Graduate Employment of the Universitas Sriwijaya
Career Development Center (CDC) at Universitas Sriwijaya provided a tracer study dataset for graduates. The data contained feature questions about the relevance of background education and graduate employment, namely about lectures, research projects experience, internships experience, English skill, internet knowledge, computer skill and others. the data was filled in by graduates in 2014, 2015, and 2016. Applying the Relief-f algorithm was to select the pattern features that most influence the relevance of education background and graduate employment. This study used Naive Bayes and KNN methods to measure the success rate of the Relief-f algorithm. The results of the accuracy of the data before the feature selection process for the naïve Bayes method were 73.43% and the KNN method was 66.24%, after the feature selection process the accuracy obtained in both methods increased to 74.38% for the Naive Bayes method and 72.22% for the KNN method. The best pattern features selected were 8 features: department relationship with work, the competence of education background, English skill, research projects experience, extracurricular activities, the competence of education background, internships experience, and communication skills. Based on the accuracy obtained, it was concluded that the Relief-f algorithm worked well in the feature selection and improved the accuracy.