Open AccessMODEL HIBRID GENETIC-XGBOOST DAN PRINCIPAL COMPONENT ANALYSIS PADA SEGMENTASI DAN PERAMALAN PASAR
Author(s) -
Rimbun Siringoringo,
AUTHOR_ID,
Resianta Perangin-angin,
Jamaluddin Jamaluddin
Publication year - 2021
Publication title -
methomika
Language(s) - English
Resource type - Journals
eISSN - 2620-4339
pISSN - 2598-8565
DOI - 10.46880/jmika.vol5no2.pp97-103
Subject(s) - boosting (machine learning) , parameterized complexity , principal component analysis , hyperparameter , artificial intelligence , support vector machine , computer science , logistic regression , gradient boosting , mathematics , statistics , pattern recognition (psychology) , machine learning , algorithm , random forest
Extreme Gradient Boosting(XGBoost) is a popular boosting algorithm based on decision trees. XGBoost is the best in the boosting group. XGBoost has excellent convergence. On the other hand, XGBoost is a Hyper parameterized model. Determining the value of each parameter is classified as difficult, resulting in the results obtained being trapped in the local optimum situation. Determining the value of each parameter manually, of course, takes a lot of time. In this study, a Genetic Algorithm (GA) is applied to find the optimal value of the XGBoost hyperparameter on the market segmentation problem. The evaluation of the model is based on the ROC curve. Test result. The ROC test results for several SVM, Logistic Regression, and Genetic-XGBoost models are 0.89; 0.98; 0.99. The results show that the Genetic-XGBoost model can be applied to market segmentation and forecasting.