Predictive Big Data Analytics Using Multiple Linear Regression Model

Abstract

Today fast trending technology era, data is growing very fast to become extremely huge collection of data in all around globe. This so-called “Big Data” and analyzing on big data sets to extract valuable information from them has also become one of the most important and complex challenges in data analytics research. The challenges of limiting memory usage, computational hurdles and slower response time are the main contributing factors to consider traditional data analysis on big data. Then, traditional data analysis methods need to adapt in high-performance analytical systems running on distributed environment which provide scalability and flexibility. Multiple Linear Regression which is an empirical, statistical and mathematically mature method in data analysis is needed to adapt in distributed massive data processing because it may be poorly suited for massive datasets. In this paper, we propose MapReduce based Multiple Linear Regression Model which is suitable for parallel and distributed processing with the purpose of predictive analytics on massive datasets. The proposed model will be based on “QR Decomposition” in decomposing big matrix training data to extract model coefficients from large amounts of matrix data on MapReduce Framework with large scale. Experimental results show that the implementation of our proposed model can efficiently handle massive data with a satisfying good performance in parallel and distributed environment providing scalability and flexibility.