IMPLEMENTASI METODE C4.5 DAN NAIVE BAYES BERBASIS ADABOOST UNTUK MEMPREDIKSI KELAYAKAN PEMBERIAN KREDIT

Putu Gede Surya Cipta Nugraha, Gede Rasben Dantes, Kadek Yota Ernanda Aryanto

Abstract


At PT. BPR XYZ credit problems is a very vital issue, where if many debtors are delinquent in payment it will increase the NPL value of the bank itself. Increasing the NPL value above 5% indicates that the bank is not healthy. From the above problems, then in this study aims to perform the implementation process of data mining methods to determine the accuracy level of prediction of creditworthiness at PT. BPR XYZ, so that the future of credit problems can be overcome. Data mining methods used in the prediction process are C4.5 and Naïve Bayes methods, where both methods are implemented and the accuracy level comparison process is used to see which method is more accurate in predicting creditworthiness. Both methods are also embedded AdaBoost method with the aim of increasing the accuracy in the process of prediction of creditworthiness feasibility. The result obtained from the comparison of method accuracy level, stated that the better accuracy is C4.5 method that is 90.00% with the precision level of 86.67%. As for the accuracy of Naïve Bayes method that is equal to 70.00% with the precision level of 79.71%. Then with the addition of AdaBoost method in predicting creditworthiness proved to increase the higher accuracy value of 91.54% in method C4.5 and by 78.13% in Naïve Bayes method. From the description above, with the implementation of AdaBoost method on the method of C4.5 and Naïve Bayes can improve the accuracy of the prediction of creditworthiness of PT. BPR XYZ. In addition, the implementation of the AdaBoost-based C4.5 method can be a recommendation for PT. BPR XYZ in conducting predictive process of credit worthiness in the future.

Full Text:

PDF


DOI: http://dx.doi.org/10.23887/ijnse.v1i2.12470

Refbacks

  • There are currently no refbacks.



International Journal of Natural Science and Engineering indexed by:

  Akreditasi SINTA 5 Crossref    




Creative Commons License

International Journal of Natural Science and Engineering is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.