PENERAPAN ALGORITMA RELIEF-F UNTUK FEATURE SELECTION PADA PREDIKSI KESESUAIAN TINGKAT PENDIDIKAN DENGAN BIDANG PEKERJAAN PADA ALUMNI UNIVERSITAS SRIWIJAYA
Daftar Isi:
- Sriwijaya University has a talent and career development center for alumni. The Career Development Center (CDC) website provides a questionnaire form for alumni to fill out. The features on the questionnaire form have relevance about the level of education of alumni with their fields of work such as lectures, research projects, internships, practicum, field work, discussions, time to get a job, fields of science, outside the field of science, English, internet, computers, and the relationship of study programs with work. Based on data from the CDC in 2013 to 2015 there were 1143 alumni of Sriwijaya University who filled out the questionnaire form completely, but there were 1019 alumni who had jobs. Feature Selection Relief-f algorithm is a feature selection algorithm based on the amount of data or records. Feature selection is done by calculating the weight difference for each data chosen at random with data selected as near hit (nearest neighbor selected data in the same class) and near miss (nearest neighbor selected data in different classes). This study uses the k-fold cross-validation of the naive bayes and KNN methods to see the success rate of the feature selection function Relief-f. The accuracy results of the data before feature selection process was 73.43% for the naive bayes method and 66.24% for the KNN method, after feature selection process increased to 74.38% for the naive bayes method and 72.22% for the KNN method. The best features were selected as many as 8 features, namely the relationship of study programs with work, science, English, research projects, outside the field of science, field work, internships, and discussions. From the accuracy obtained it can be concluded that the feature selection Relief-f algorithm works well in the feature selection process and improves accuracy.