1 Hybrid Book Recommendation system Miss. Anagha Vaidya (ME Student), Dr. Subhash Shinde (Vice Principal) Department of Computer Engineering, Lokmanya Tilak College of Engineering, Mumbai, India. Abstract: In this age of information, it is very difficult to find the right information from the enormous amount of data present in the online platforms. Recommendation system sorts through massive amounts of data to identify interest of users and makes the information search easier. In this paper, we have presented a model for a web-based personalized hybrid book recommendation system which exploits varied aspects of giving recommendations apart from the regular collaborative and content-based filtering approaches. Temporal aspects for the recommendations are incorporated. Also, for users of different age, country and their interests, personalized recommendations can be made on these demographic parameters. We are taking some information from user while signup which help to get more appropriate recommendations based on individual user interest and thus an attempt to overcome cold start problem. Three types of scenarios are covered in this paper viz. if user is new then recommendations are made depending upon user interests, second is recommendations based on past purchase history and the last is recommendation by using different algorithms namely K Nearest Neighbor (KNN), Singular Value Decomposition (SVD), Restricted Boltzmann Machines (RBM) and cosine similarity. It reduces dependency of rating-based system. Index Terms: Hybrid Recommendation system, Collaborative filtering, content filtering, and Demographic filtering I. INTRODUCTION Recommendation system (RS) is a subclass of information filtering system that seeks to predict the "rating" or "preference" a user would give to an item. Recommender systems typically produce recommendations in one of two ways – through collaborative filtering or through content-based filtering (also known as personality-based approach). Collaborative filtering approaches build a model from a user's past behaviour (items previously purchased or selected and/or numerical ratings given to those items) as well as similar decisions made by other users. This model is then used to predict items (or ratings for items) that the user may have an interest in. Content-based filtering approaches utilize a series of discrete characteristics of an item in order to recommend additional items with similar properties. These approaches are often combined as Hybrid Recommender System [1]. Different techniques have been developed over time to give accurate recommendations. As E- commerce is getting bigger, people are moving from retail shops to online store. On E- commerce, availability of numerous options makes the finding of most suitable item a hefty task, so RS makes this task easy by finding behaviour pattern from past history or user input. There are two major issues in existing recommendation system one is cold start and another one is accuracy [2]. In proposed model, we tried to overcome cold start problem by taking some inputs from user while creating account and accuracy by implementing different algorithms. If recommendations given are varying too much from the user’s likes and tastes, he/she may simply stop using the system. So, in order to build trust, recommendations need to be personalized. Demographic recommendations are a good way of giving personalized predictions [3]. Filtering the results using collaborative approach leads to a better recommendation output. Recommendations suited to the user’s age, region and Interests can be made more personalized. The cold start problem is a major issue in many recommendation systems. In such a scenario, the system is unable to give appropriate predictions until it has a better idea about the user’s preferences. Demographic recommendations could help alleviate this problem to some extent, if not entirely in case of a newly added user. A user always would like to stay abreast of their liked category or liked author’s books. The traditional filtering techniques may not always be able to keep a user updated about the recent trends in books. There are three types of user, firstly, the new users. To deal with the cold start issue, we are beginning by asking users about categories (e.g. Suspense and thriller, romance etc.) and writers they are interested in. Based on these criterions, recommendations are being made. Along with this, a parallel approach is followed where we find users with similar interests and a bigger and more accurate set of recommendation is returned based on the rating profile. The second types of user are the ones who don’t prefer to rate. This case may lead to a failure of rating-based system. To tackle this issue, we are making recommendations based on their past orders. The third set of users consists of those who give feedback (Ratings). We are using 4 algorithms namely SVD, KNN, RBM, Hybrid for more options for user. Figure 1: Types of users To summarize the underlying approach, we are using hybrid model to provide personalized recommendations to individual user. This system is hybrid of content based as well as collaborative approach of recommender system. We are showing