Human Activity Recognition Using For Smartphone Sensors To Predict The Best Accuracy Based On Machine Learning Algorithms

Abstract

Human activity recognition requires predicting the action of a person based on sensor-generated data. Due to the enormous number of applications possible by modern ubiquitous computing devices, it has sparked a lot of attention in recent years. It categorizes data into actions such as walking, sitting, standing, and lying. The accelerometer and gyroscope were used to generate the sensor data, and the sensor signals were pre-processed with noise filters. The goal of this research is to anticipate the optimum accuracy for machine learning-based approaches for Human Activity Recognition. In this study, supervised machine learning methods such as Logistic Regression, SVM, Decision Tree, and Random Forest were utilized to recognize human behavior using smartphone sensors in a detailed experiment. The proposed machine learning-based technique for accurately predicting human activity involves predicting human actions such as walking, sitting, standing, and lying.  When compared to other supervise classification machine learning techniques, the Random Forest classifier algorithm predicts 97.32 percent accuracy. While other classifiers, such as Logistic Regression, scored 94.45%, Support Vector Machine 94.55%, and Decision Tree 95.75% also show good performance. The dataset containing evaluation classification report and confusion matrix to categorize data from priority and the outcome reveals that the efficacy of the suggested machine learning algorithm technique can be compared to the best accuracy using precision, recall, and F1 Score.