Paper Title
A Comparative Study on Human Activity Recognition using Multi-Layer Perceptron Neural Networks and Multi Probability Logistic Regression

Abstract
The capability of recognizing human physical movements with the help of wearable devices or IoT sensors is known as "Human Activity Recognition." In this pandemic situation, most people, ancient age and the seriously infected people keep in isolation wards. Due to the rapid spread of COVID, it becomes difficult for caretakers or others to monitor them by residing in the same room continuously. So, to monitor them and take necessary precautions, the people are attached to wearable devices, and an IoT-based video capturing device is fixed in the isolation ward. The proposed system is intended to capture 6 general activities like walking, jogging, upstairs, downstairs, sitting, and standing and classify these activities based on the multi-class classification algorithms. The proposed paper implements and compares Logistic Regression, a traditional machine learning algorithm with Multi-Layer Perceptron (MLP), a deep learning algorithm, and MLP performs better than the machine learning algorithm. The data received from the IoT sensors stores in the cloud through the Raspberry Pi wifi module, and the proposed MLP model notifies immediately if any unexpected activity finds in the human. Keywords - Multi-Layer Perceptron, Logistic Regression, Internet of Things, Human Activity Recognition.