In the real time scenario, the volume of data used linearly increases with time. Social networking sites like Facebook, Twitter discovered the growth of data which will be uncontrollable in the future. In order to manage the huge volume of data, the proposed method will process the data in parallel as small chunks in distributed clusters and aggregate all the data across clusters to obtain the final processed data. In Hadoop framework, MapReduce is used to perform the task of filtering, aggregation and to maintain the efficient storage structure. The data are preferably refined using collaborative filtering, under the prediction mechanism of particular data needed by the user. The proposed method is enhanced by using the techniques such as sentiment analysis through natural language processing for parsing the data into tokens and emoticon based clustering. The process of data clustering is based on user emotions to get the data needed by a specific user. The results show that the proposed approach significantly increases the performance of complexity analysis. © 2015 The Authors. Published by Elsevier B.V.