A collaborative filtering method based on artificial immune network A. Merve Acilar * , Ahmet Arslan Selcuk University, Eng.-Arch. Fac., Computer Eng., 42003 Konya, Selcuklu, Turkey article info Keywords: Recommender systems Collaborative filtering Artificial immune system Sparsity Scalability aiNet abstract A system is seriously required for helping users to find their path on the shopping and entertainment web sites where the amounts of on-line information vastly increase. Therefore, recommender systems, new type of internet based software tool, appeared, and became an appealing subject for researchers. Collab- orative filtering (CF) technique based on user is the one of the method widely used by recommender sys- tems but they have some problems for waiting to be developed solutions that are more efficient. One of these mainly problems is data sparsity. While the number of products is increase, the ratio of common rated products is decrease so calculating the computations of neighbourhood become difficult. The other one is scalability which is the performance problem of the existing algorithms on the datasets has large amounts of information. In this article, we tackle these two questions: (1) how the data sparsity can be reduced ? (2) How to make recommendation algorithms more scalable? We present an approach to addressing the both of these problems at the same time by using a new CF model, constructed based on the Artificial Immune Network Algorithm (aiNet). It is chosen because aiNet is capable of reducing sparsity and providing the scalability of dataset via describing data structure, including their spatial distribution and cluster inter- relations. The new user-item ratings dataset reduced by applying aiNet (aiNetDS) given more stable results and produced predictions more quickly than the raw user-item ratings dataset (rawDS). Besides, the effects of using clustering for forming the neighbourhoods to the system performance are investi- gated. For this, both of these dataset are clustered by using k-means algorithm and then these cluster par- titions are used as neighbourhoods. As a result, it has been shown that the clustered aiNetDS is given more accurate and quick results than the others are. Ó 2008 Elsevier Ltd. All rights reserved. 1. Introduction The amount of information in the world is growing so quickly with the widespread and easy usage of internet. Therefore, ac- quired data from the sites of e-commerce, entrainment etc. is increasing far more rapidly than our ability to process it. All of them such as Amazon, Yahoo and CDNow suggest a lot of chooses to their potential customers everyday which makes difficult to find the true products that best meet user’s needs and preferences. For overcoming this problem, recommender systems appeared and be- came an appealing subject for researchers. It is a kind of personal software assistant learning the evolving interests of their users by applying the information-processing algorithms to the mass of this information. These recommendations base on the top overall sellers on a site, on the demographics of the consumer, or an analysis of the past buying behaviour of the consumer as a prediction for future buying behaviour. The forms of them include suggesting products to the consumer, providing personalized product information, summariz- ing community opinion, and providing community critiques. Broadly, these recommendation techniques are part of personaliza- tion on a site because they help the site adapt itself to each cus- tomer. Under this broader definition, recommender systems serve to support a customization of the consumer experience in the presentation of the products sold on a Web site. In a sense, rec- ommender systems enable the creation of a new store personally designed for each consumer. Recommender systems enhance E- commerce sales in three ways: helping customers find products they wish to purchase; converting browsers into buyers; improv- ing cross-sell by suggesting additional products for the customer to purchase; improving loyalty by creating a value-added relation- ship between the site and the customer (Schafer, Konstan, & Riedl, 2001). Mainly two distinct techniques are used by today recommenda- tion systems: content-based methods and collaborative methods. Content-based methods analyze the content of information sources (e.g. the HTML source of web pages) that have been rated to create a 0957-4174/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2008.10.029 * Corresponding author. Tel.: +90 332 2233722; fax: +90 332 241 0635. E-mail addresses: msakiroglu@selcuk.edu.tr, msakiroglu@hotmail.com (A. Merve Acilar). Expert Systems with Applications 36 (2009) 8324–8332 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa