Human activity recognition (HAR) is an interesting research area in machine learning. The purpose of human activity recognition study is to automatically detect human activities from the information acquired from different sensors, and to analyze this information using statistical techniques. Big data still a challenging problem in human activity prediction. Several approaches have recently been developed to find practical ways to solve high dimensionality of data problems. The mean objective of this work is to deal with HAR big data modeling in order to maximize the classification accuracy while minimizing the number of factors. The proposed framework has been tested on a publicly HAR available dataset and the results have been interpreted and discussed.