Review on Analysis of User Behavior Based on Prediction Algorithm

Various service provider systems or recommender systems have to manage the large amount of data regarding customer, service and information. This leads to Big Data analysis problem and its handling is also a challenging task. There are various organizations which serves user through service provider systems or recommender systems that provide efficient services in terms of set of good quality products or services to use by giving appropriate data as per their need. Hence the Decision Support System is used to determine the decision making activities in various organizations of recommender system which allows making certain planning, management and operations. This also helps to make decisions for structured and unstructured problem. It can be done in number of ways like descriptive, decision and prediction based. This paper will focus on Predictive Analysis of user behavior towards recommender system. The predictive analysis deals with the identification of relationship between specific performance of unit with respect to set of features or attributes defined for the system. Efficient service provider system have to determine preferences of user towards service through ratings, ranking or free-text given by them like reviews posted by user over internet. This is useful to get by understanding about users behavior for particular system. This can be done through opinion mining, review analysis and sentiment analysis. The text mining allows mining of features and characteristics of product or service efficiently and further processing of features can be done by defining efficient predictive model for it. Hence, this data will be available in plenty of amount which must be processed by using new innovative framework called Hadoop. The Hadoop technology provides facility of Map Reduce mechanism to process large amount of data. The mined data will be processed through various techniques using prediction method. There are various existing methodologies that involve analysis based on probability model, product aspect ranking method, cross-domain sentiment classification, estimating helpfulness from set of data, and dynamic interaction using mashup tools. These methods can be implemented with Hadoop for efficient result.