An Experimental Comparison of Classification Algorithms for Premium Beef Customer Buying Intention

Abstract

This study aimed to explore and compare machine learning performance to predict the customer purchasing decision within premium beef shops. The sampling locations were Thailand. The population used in the study consisted of 436 valid responses from 5 premium beef shops. The data was obtained by using questionnaires consisting of gender, age, three questions of product, one question for the price, one question for the place, and three questions for appearances. The study was used four classifier's algorithms: k- nearest neighbors, decision tree, random forest, and xgboost model. The models were compared to find the highest accuracy for premium beef customer behavior data set. Random forest algorithms were evaluated to have the best performance in predicting premium beef purchasing decisions in Thailand. The model has an accuracy of 88.62 percent, precision of 88.46 percent, recall of 85.19 percent, f1 of 86.79 percent, and AUC of 95 percent. The two important elements that influence purchasing decisions are price and product age. The most accurate algorithms can be used to forecast consumer product purchases and comprehend the principles of elements that influence buying decisions.