ABSTRACT – Recommender system is the technique which helps the users to fetch usefull and required data according to their interests and preferences by filtering large amount of information. Over the past decade, a large number of recommendation systems for a variety of domains have been developed and are in use. Entertainment is one of the significant domain as it is necessary for each one of us to refresh our mood and energy. Movie recommender systems help customers to access preferred movies from a huge online multimedia library automatically.
To implement this system, collaborative filtering technique is most widely used as it is more advantageous than other techniques. But this technique also suffers from many issues such as, cold start, data sparsity, shilling attack etc.KEYWORDS – Movie recommendation system, collaborative filtering, KNN clustering, Pearson correlation, User’s psychological profile.1. INTRODUCTIONRSs are continuously being explored and developed because of information overload and popularity of social media. RSs enhance the ratings and revenues for effectiveness of items or products.
There are six classifications of RS in social media based on its wider usability in recommendation approaches which are, domains, data mining techniques, datasets, recommendation types, and performance metrics”. However, the focus of recommender system research is more on entertainment domain (mainly movies). This is due to the change in the taste and preferences of users day by day and also freely and publicly available datasets by research community in this domain.Movie recommender engine can be implemented mainly using content based, collaborative filtering or hybrid methods. Collaborative filtering technique has more advantages over other techniques. The main advantage over content based filtering technique is that it improves the performance of the recommender system. because it also considers other similar user’s interest and past history of that similar users for giving the recommendation to the user. Though hybrid method gives better results, it is quite expensive and complex compared to collaborative approach. Collaborative filtering also suffers from many issues. The problem of recommendations given to the new user and a new product is one of the most challenging issue among them. Recommendations can be predicted for the active user based on the known ratings of movies that have already been given a user to this. For a new user who is registered for the first time in the system has no ratings and known only demographic information such as: age, sex, occupation, country of residence, and education. Hence, there is a huge scope of exploration in movie recommendation systems for improving scalability, accuracy and quality of recommendations given to new users.2. LITERATURE SURVEYThere are mainly two types in collaborative filtering. One is memory based, which recommend items to user based on his past data, another one is model based which implement some model to give recommendation item to user. Memory based method is preferred over other methods due to its easy implementation and it is also less expensive.Memory-based algorithms use the whole data set to compute the recommendation to the users. Generally, they use similarity measures to select users or items that are similar to the active user. Then, the recommendation is calculated from the ratings of these neighbours or groups. Memory-based filtering technique mainly classified into two types such as user-based and item-based filtering technique.To partition the set of movies based on user rating data, Clustering algorithms are most commonly used. Clustering is scalable, simple, and suitable for datasets with compact and highly distributed spherical clusters. In clustering K-Nearest Neighbour(KNN) method is used as it is simple algorithm to implement, moreover it is easy to explain, understand and interpret.The interaction between user item can be implicit or explicit. Implicit interactions include sessions and cookies of browser, whereas explicit interactions are through the ratings and feedback provided by the user.Pearson correlation coefficient method is employed to calculate similarity to optimize results when compared to cosine similarity and Linear Regression is used to make predictions that gave better results. The asymmetric method describes that similarity of user A with B is not the similar as the similarity of B with A . The predictions generated by RS in social media can be evaluated using a variety of performance metrics. The review showed that precision and recall are highly used evaluation metrics. They are important in information retrieval and popular with RS in social media.To address the problem of cold start, a method that includes tags and keywords that provide information about user and item is computed . The psychological profile was used in one of the content-based features and was determined with the help of a yes/no psychological test. The test is carried out by posing different queries regarding individual’s personality.The purpose of this test was to determine to what degree the four main psychological profiles (choleric, sanguine, phlegmatic, melancholic) can be found in the user’s own personality. The main aim of this method is to provide movie recommendations specifically tailored to the user and to observe if there truly is a correlation between a persons’ psychological profile and his/her preferences in movies .The temporary feature of user profile is among the main challenges of capturing user interests. This feature poses the user interests and may be varied over time. At two state of the art studies on temporal recommender systems, the temporary feature is captured using time bins. Each bin represents a group of movies that are rated at the same time slice. Context-aware recommender systems employ contextual information such as time, location, and social data to make recommendations .One more interesting challenge of collaborative filtering technique is shilling attack, which means that some malicious users can create fake profiles and insert into the network in such a way that it affects the recommender system. To reduce this various trust calculation techniques are used. One of them is the Novelty method which can be summarized in it is the ability to show how to add different weights on the social trust relationships among users based on the trustee’s competence and trustworthiness. By using this method, it can be make sure that recommendations always comes from the trusted users.4. CONCLUSIONThe paper analyses the current state of development and application of movie recommendation systems. The optimal way of implementing movie recommender system is studied. Though collaborative filtering is one of the most widely implemented method, it also suffers from many issues. The main issues such as new user and new item problem, shilling attack are discussed. Various methods to reduce these problems are also depicted.5. REFERENCES Lakshmi Tharun Ponnam (Author), Sreenivasa Deepak Punyasamudram, Siva Nagaraju Nallagulla , Srikanth Yellamati Movie Recommender System Using Item Based Collaborative Filtering Technique 2016, IEEE.  Yuri Stekh, Mykhoylo Lobur, Vitalij Artsibasov, Vitalij Chystyak Methods and Tools for building Recommender systems CADSM 2015, 24-27 February, 2015, Polyana-Svalyava(Zakarpattya), UKRAINE.  Rahul Katarya, Om Prakash Verma Effective collaborative movie recommender system using asymmetric user similarity and matrix factorization ,International C onference on Computing, Communication and Automation(ICCCA2016).  Shreya Agrawal, Pooja Jain, “An Improved Approach for Movie Recommendaton System. International Conference on I-SMAC( IOT in Social, Mobile, Analytics and Cloud)(I-SMAC 2017).  Bansari Patel, Palak Desai, Urvi Panchal Methods of Recommender System: A Review” 2017 International Conference on Innovations in Information Embedded and Communication Systems(ICIIECS).  M V V R Murali Krishna Rao ” A Collaborative Filtering Recommender System with Randomized Learning Rate and Regularized Parameter ” 2017. Soanpet .Sree Lakshmi, Dr.T.Adi Lakshmi Recommendation System: Issues and Challenges (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (4) 2014.  Bei-Bei CUI Design and Implementation of Movie Recommendation System Based on Knn Collaborative Filtering Algorithm ITM web of conferences 2017.Ningning Yi, Chunfang Li, Xin Feng, Minyong Shi Design and Implementation of Movie Recommender System Based on Graph Database, 2017 14th Web Information Systems and Applications Conference.Anitha Anandhan, Liyana Shuib, Maizatul akmar ismail and Ghulam mujtaba Social Media Recommender System: Review and Open research issues 2018 The University of Malaya’s Research Grant Programme SBS (Equitable Society) under Grant RP059E-17SBS. Costin-Gabriel Chiru, Vladimir-Nicolae Dinu, Catalina Preda,Matei Macri Movie Recommender System Using the User’s Psychological Profile 2015 IEEE. Gaurav Arora, Ashish Kumar, Gitanjali Sanjay Devre, Prof. Amit Ghumare movie recommendation system based on users’ similarity 2014 International Journal of Computer Science and Mobile Computing.Rupali Hande, Ajinkya Gutti, Kevin Shah, Jeet Gandhi, Vrushal Kamtikar moviemender – a movie recommender system2016 International Journal of Engineering Sciences & Research Technology. Steven Postmus Recommender System technique applied to Netflix movie data” Reasearch Paper Business Analytics,Vrije Universiteit Amsterdam, February 2018. Bagher Rahimpour Cami, Hamid, Hassanpour Hoda Mashayekhi A Content-based Movie Recommender System based on Temporal User Preferences 2017 3rd Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS).