Article Preview
Top1. Background Study
As the Internet continues to grow, it becomes increasingly difficult to sift through all the information that is available. With the overwhelming amount of data and choices that can be made, people need a filter to increase the Internet's usability. Recommender Systems have emerged as important response to the so-called information overload problem and they have been successfully deployed in a variety of e-commerce domains (O’Donovan & Smyth, 2005).
Typically, a Recommender System (RS) analyzes data about items or about interactions between users and items to find associations among items and users (Omisore et. al., 2013). It provides advice to users about items they might wish to purchase or examine. The results obtained are presented as recommendations and such can help users navigate through large information space of product descriptions, news articles or other items (Huang et. al., 2008 and Burke, 2000). The products can be recommended based on the top overall sellers on a site, on the demographics of the consumer, or on an analysis of the past buying behavior of the consumer as a prediction for future buying behavior (Omisore & Samuel, 2014). The forms of recommendation can be to suggest products to consumers, to summarize opinions of a community or provide their critiques (Schafer et. al., 1999), and to provide personalized product information which helps Internet users do away from problem of information overload (Yang et. al., 2004).
In order to meet the need of a specific customer, it is essential to identify his interests, requirements and preferences. Some existing RSs focus on intelligently discovering the requirements of the customer by analyzing his personal information, products he purchased in the past, or logs of browsing history, among others in order to make recommendations. However, these RSs could be useful for frequently purchased products like books, movies and so on. For less frequently purchased products like laptops, digital cameras and mobile phones, it is difficult for the business firm or enterprise to actually get enough information to identify the specific requirements of each customer for a particular product. After identifying the possible products for a specific customer in terms of his requirements, this paper is proposing combining the opinions of previous users of the product in order to obtain the optimal products.
Some existing studies in this line have proposed personalization (Cao & Li, 2007 and Ojokoh & Kayode, 2012) in terms of correlating the users’ preferences with product features, without including other users’ opinions while some like (Ojokoh et. al., 2013) used Collaborative Filtering (CF) technique to recommend items that are not frequently purchased. In the phase of buoyancy in rich data which often overwhelms Internet business, dearth of competent RS makes it difficult to promote online products that are specifically appropriate to consumers (Castellano et. al., 2009).
In order to leverage the strengths of Content Based (CB) and CF recommender techniques, there have been several hybrid methods that were proposed. One approach is to produce separate ranked lists of recommendations from CB and CF methods before merging the results to produce a final list (Smyth & Cotter, 2012), while another is to combine either of the popular methods in a manner that one dominates over the other. Claypool et al. (1999) combines CB and CF techniques by using an adaptive weighted average, where the weight of the Collaborative component increases as the number of users accessing an item increases.
Melville et. al. (1997) proposes a general framework for Content-boosted CF, where CB predictions are applied to convert a sparse user ratings matrix into a full ratings matrix, and then a CF method is used to provide recommendations. This study use a Naïve Bayes classifier trained with documents to describe rated items of each user, and replace the unrated items by predictions from this classifier. Several other hybrid approaches are based on traditional CF, but also maintain a CB profile for each user. These CB profiles, rather than co-rated items, are used to find similar users.