PERBANDINGAN METODE K-NEAREST NEIGHBOR DAN ADAPTIVE BOOSTING PADA KASUS KLASIFIKASI MULTI KELAS

Ade Irma Prianti (Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro)
Rukun Santoso (Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro)
Arief Rachman Hakim (Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro)

Jurnal Gaussian
Vol 9, No 3 (2020): Jurnal Gaussian

The company's financial health provides an indication of company’s performance that is useful for knowing the company's position in industrial area. The company's performance needs to be predicted to knowing the company's progress. K-Nearest Neighbor (KNN) and Adaptive Boosting (AdaBoost) are classification methods that can be used to predict company's performance. KNN classifies data based on the proximity of the data distance while AdaBoost works with the concept of giving more weight to observations that include weak learners. The purpose of this study is to compare the KNN and AdaBoost methods to find out better methods for predicting company’s performance in Indonesia. The dependent variable used in this study is the company's performance which is classified into four classes, namely unhealthy, less healthy, healthy, and very healthy. The independent variables used consist of seven financial ratios namely ROA, ROE, WCTA, TATO, DER, LDAR, and ROI. The data used are financial ratio data from 575 companies listed on the Indonesia Stock Exchange in 2019. The results of this study indicate that the prediction of company’s performance in Indonesia should use the AdaBoost method because it has a classification accuracy of 0,84522 which is greater than the KNN method’s accuracy of 0,82087. Keywords: company’s performance, classification, KNN and AdaBoost, classification accuracy.