Analisis Metode Logistik Regresi Ensemble untuk Klasifikasi dengan Pra-Pemrosesan Menggunakan Principal Component Analysis

Abstract

Logistic Regression Ensemble (Lorens) is one of the ensemble concepts of binary logistic regression which is usually used on high dimensional data to develop binary logistic regression in classification analysis. This method is built by dividing the variables into a number of subspaces randomly, each subspace is modeled, predicts a category and combines the prediction results by calculating the average category for each subspace or majority voting. PCA is a statistical analysis that can be used to reduce a number of independent features/variables. PCA is often used for the pre-processing stage. The PCA used to reduce the number of variables in each subspace in this study gave results of increased accuracy for high-dimensional data scenarios of multicollinearity effects generated/simulated as many as 160 continuous variables and 40 discrete variables. The performance results for the total accuracy value for each scenario are: the imbalance ratio and the multicollinearity effect are respectively 94.37% and 71.25% while for real data it is 86.68%