Decoding Human Behavior: Human Activities Recognition using Machine Learning Approach

The study of Human Activity Detection (HAD) has become a challenging topic, especially when technology such as Internet of Things (IoT) is introduced, allowing for a wide range of applications in the field of healthcare and elderly care. This study looks at new methods developed in HAD, such as sensor-based, mobile devices and visual data processing. The paper points out the potential uses of HAD, from simple to complex behaviors, powered by video frames, wearable sensors and accelerometers. HAD can be integrated with promise in medical diagnostics, surveillance and real-time activity detection. The research goal is to identify and classify human activities based on the NTU-RGB-D dataset, which is used to collect 3D skeletal data. More detailed analysis can be done if a subset of acts is considered, e.g., drinking, picking up, applauding, and saluting. The proposed system emphasizes to develop a machine learning model for activity prediction with high accuracy by implementing Logistic Regression Algorithm. This paper highlights the role, subset selection, algorithm integration and experimental setting of the skeletal-joints dataset from the NTU-RGB-D dataset. The Logistic Regression Algorithm is effective and with decision tree majority voting amazing accuracy is achieved. Logistic Regression is shown to be better through comparative evaluations. The effectiveness of the study verifies its feasibility to understand and predict human behavior.