MODELING OF HUMAN BEHAVIOR FOR SMARTPHONE WITH USING MACHINE LEARNING ALGORITHM

The article focuses on exploring human behavior recognition as an alternative means of identifying and authenticating smartphone users. The process involves obtaining raw data, extracting features, and making classifications. In this study, a single accelerometer-equipped smartphone is utilized to sense users' walking patterns for experimental data. Unlike traditional machine learning algorithms, a deep learning approach is employed. The paper introduces a novel Convolutional Neural Network (CNN) model for user identification based on activity patterns. The experiment uses a publicly available walking activity dataset for user identification. The CNN model achieves an impressive 99.88% accuracy in recognizing users from their walking patterns. Additionally, the article conducts a comparative analysis with classical machine learning algorithms such as Ada-Boost, Decision Tree, GaussianNB, Linear Discriminant, Logistic Regression, Quadratic Discriminant, and Random Forest. While Random Forest reaches a commendable accuracy of 95.78%, the CNN model surpasses it in terms of both recognition time and accuracy.