Abstract

Collaborative filtering recommender system is utilized as a significant method to suggest products to users depending on their preferences. It is quite complicated when the user preference and rating data is sparse. Missing value occurs when there are no stored values for the specified dataset. Typical missing data are of three categories such as (i) Missing completely at random, (ii) Missing at random, and (iii) Missing not at random. The missing values in the dataset affect the accuracy and cause deprived prediction outcomes. In order to alleviate this issue, the data imputation method is exploited. Imputation is the process of reinstating the missing value with a substitute to preserve the data in a dataset. It involves multiple approaches to evaluate the missing value. In this paper, we reviewed the progression of various imputation techniques and their limitations. Furthermore, we endeavored k-recursive reliabilitybased imputation (k-RRI) to resolve the boundaries faced in existing approaches. Experimental results evince that the studied methodology appreciably improves the prediction accuracy of the recommendation system.