Journal of Engineering Science and Technology Review 10 (4) (2017) 132-153 Review Article Classifications of Recommender Systems: A review Shahab Saquib Sohail 1,* , Jamshed Siddiqui 2 and Rashid Ali 3 1,2 Department Computer Science Aligarh Muslim University, Aligarh, India 3 Department of Computer Engineering, Aligarh Muslim University, Aligarh, India Received 31 July 2017; Accepted 18 September 2017 ___________________________________________________________________________________________ Abstract This paper presents the state of art techniques in recommender systems (RS). The various techniques are diagrammatically illustrated which on one hand helps a naïve researcher in this field to accommodate the on-going researches and establish a strong base, on the other hand it focuses on different categories of the recommender systems with deep technical discussions. The review studies on RS are highlighted which helps in understanding the previous review works and their directions. 8 different main categories of recommender techniques and 19 sub categories have been identified and stated. Further, soft computing approach for recommendation is emphasized which have not been well studied earlier. The major problems of the existing area is reviewed and presented from different perspectives. However, solutions to these issues are rarely discussed in the previous works, in this study future direction for possible solutions are also addressed. Keywords: recommender systems, collaborative filtering, reclusive methods, knowledge based recommender systems, hybrid recommender systems, context aware recommender systems. __________________________________________________________________________________________ 1. Introduction The recommender systems (RS) have grown exponentially in recent few years and its applications have spread over various domain of life including online shopping of books, home appliances, movies, electronic gadgets, recommendation of doctors and hospitals for patients, institute recommendation for students and teachers, hotel recommendations for tourists and so forth. The philosophy behind the success of recommendation technology is the fact that it is human tendency to rely upon experiences of their neighbors and friends prior to making decision of any kind, especially regarding purchase of any items, taking admissions in institutes for higher education, opting an apartment for rent or buying it, spending weekend at some holiday places, etc. The advancement of Internet technologies has caused data overload due to which the buyers face more difficulties in finding the exact destination which meet their needs out of a huge collection of the available options. If a student who wishes to spend his/her vacations at some hill stations and would like to stay in a hotel with peace and calm, there would be thousands of places all around the world which might come to him/her as options. In such a situation recommender systems can provide a better option according to the need and requirement of the user and depending upon his/her prior preferences. Although there are several definitions which researchers have suggested for recommender systems, we define recommender systems as – “Recommender systems try to identify the need and preferences of users, filter the huge collection of data accordingly and present the best suited option before the users by using some well-defined mechanism.” A formal definition for RS can be stated as; Let ‘S’ be the set of users and ‘I’ be set of all items that fall under their preference category. Let R ⊆ I is the ranked list of items which is in some desired order, and r is an item in list R, i.e. r ϵ R. The recommendation problem is to choose an r ϵ R such that it satisfies the users and also meets their need. Let ‘E’ be the evaluation metric to measure user satisfaction for some real number ‘z’, then we can assume user satisfaction is achieved only if E • z. Mathematically, if ‘f’ defines the function of recommending ‘r’ items to ‘s’ users, our problem can be formulated as: f (r,s) • z (1) In this Paper, we have reviewed more than 200 articles related to recommender system including the manuscript in which very first existence of collaborative filtering has reported in mid 90s [1], [2]. 2. Previous Review Studies The first paper on collaborative filtering (CF) was introduced in mid of 90s [2], [3]. The proposed CF technique provided a platform to design recommender system and laid a strong foundation for the development of such recommender systems. The work in the concerned area has been reviewed extensively in the literature. The study of the surveys and reviews of recommender systems helps in establishing a better understanding of the subject and gives a holistic picture of the technology used in the field along with JOURNAL OF Engineering Science and Technology Review www.jestr.org Jestr ______________ *E-mail address: shahabsaquibsohail@gmail.com ISSN: 1791-2377 © 2017 Eastern Macedonia and Thrace Institute of Technology. All rights reserved. doi:10.25103/jestr.104.18