Comparison of The Classification Data Mining Methods to Identify Civil Servants in Indonesian Social Insurance Company

Abstract

Indonesian civil servants already have social security; however, the benefits' value has not sufficed life necessities in retirement. Indonesian social insurance company provides additional insurance products for civil servants, yet only 7 percent of civil servants are interested. Improved marketing by identifying civil servants through data mining will help boost product sales. Data mining uses the CRISP-DM approach, starting from understanding business processes, civil servant data, data preparation, and modeling to evaluation. Data mining techniques use classification with three algorithms: Decision Tree, Naive Bayes, and Neural Network. Data mining results show six influential attributes of civil servants, including sex, the number of children, age, remaining working period, marital status, and years of service. The neural network algorithm has better performance with an accuracy value of 71.7%, a F1-score value of 73.4%, a precision value of 69.7%, a recall value of 77.6%, and an AUC value of 79.1%.