Abstract:

Human activity recognition (HAR) is a very hot research topic in computer vision nowadays. Recently, numerous application (HAR) systems and approaches have been proposed. The prediction in human mobility using big data however still remains a challenge to the classification problem. This is mainly due to the huge variations, such as growing information amount and high dimensionality of data, the difficulty in modeling the precise relationship between the large number of feature variables, and the class variable. In such cases, it is highly desirable to reduce the information to a small number of dimensions in order to improve the accuracy and effectiveness of the classification process. Nevertheless, the performance of (HAR) is not high enough yet. This paper aims at improving the performance of human mobility modeling and mining by employing dimension reduction based on statistical techniques. The developed method has been applied on a new and publicly available human activities dataset. The obtained results are effectively interpreted, and the efficacy of the suggested method over the well-known methods is discussed.