User Classification in Electronic Devices Using Machine Learning Methods

User classification is a major concern for electronic device providers, because accurate and efficient classification can cut operating cost of company significantly. To deal with this problem, manufacturers try to classify customer into several categories and recognize the characteristic of users, then adopt different promotion strategies to improve sales revenue. This study aims to build models to divide users into several categories, and identify critical and controllable features which dramatically affects classification results. We test proposed models with real data coming from an electronic device producer. Results shows that random forest model performs best. Our main contributions are: 1) we focus on user classification of electronic devices; although many existing studies have discussed similar problem, few of them focus on applications in electronic devices; 2) we consider imbalance sample, and datasets are from real company. This work will be helpful for electronic device producers to improve operation and enhance marketing competitiveness.