Genetic Algorithms for Feature Weighting in Multi-criteria Recommender Systems Genetic Algorithms for Feature Weighting in multi-criteria Recommender Systems Chein-Shung hwang Dept of Information Managemen Chinese culture university aipei, Taiwan hwang@faculty. pccu. edu dor: 10.4156/jcit vol5. issue813 Abstract Recommender systems have been emerging as a powerful technique of e-commerce. The majority of existing recommender systems uses an overall rating value on items for evaluating user's preference opinions. Because users might express their opinions based on some specific features of the item, recommender systems solely based on a single criterion could produce recommendations that do not meet user needs. In this paper, we propose a mechanism for integrating multiple criteria into th Collaborative Filtering (CF) algorithm. Specifically, we present the implementation of Genetic Algorithms (GA) for optimal feature weighting. The proposed system consists of two main parts. First, the weight of each user toward each feature is computed by using GAs. The feature weights are then incorporated into the collaborative filtering process to provide recommendations. Empirical studies have shown that our weighting scheme can be incorporated to improve the performance of multi criteria cF Keywords: Genetic Algorithms, Collaborative Filtering, Multiple Criteria, Recommender Systems 1 Introduction Although the rapid development and popularity of the Internet have brought convenience for people to access a variety of information, products, and services, the internet has also caused a scenario where people have difficulty obtaining relevant information. This scenario is commonly referred to as the problem of"information overload". Recommender systems [1, 2] are emergent to solve the information overload challenges. Recommender systems are personalized information filtering technologies used to suggest items to users that they might like or find interesting. In recent years, recommender systems ave been successfully applied in a broad range of applications, including recommending movies books. news articles. music. etc. Recommender systems can be based on content-based filtering, collaborative filtering, and hybrid Itering [3, 4]. Content-based recommendation suggests items to users that are similar to those items that they were interested in previously. Collaborative filtering( CF)recommends items based on the information about similar items or users. Hybrid recommendation combines both algorithms into one hybrid approach to gain better performance and avoid drawbacks from each individual one p The majority of existing recommender systems uses an overall rating value on items for evaluating s'preferences. The overall rating depends on one single criterion that usually represents the overall preference of users to a particular item. As users might express their opinions based on some specific features of the item, recommender systems merely based on a single criterion could produce recommendations that do not meet user needs. For example, in a movie recommender system, two sers A and B both assign a single-criterion rating of 12(out of 13)for Avatar. The recommender systems will conclude they have the same tastes even if A likes its story and B likes its visuals. This is called a"without distinction of interest problem. Furthermore, even if both users like the same movi features(e.g. actors, visuals, etc. ) they might select different movies. This is because people usually select a movie based on different movie features This situation is referred to as an"unsuitable weight feature" problem [5] Many recommender systems [6-8 have been developed to tackle the above-mentioned problems by using multiple criteria. Roux et al. [6] described a course recommender systemGenetic Algorithms for Feature Weighting in Multi-criteria Recommender Systems Chein-Shung Hwang Genetic Algorithms for Feature Weighting in Multi-criteria Recommender Systems Chein-Shung Hwang Dept. of Information Management Chinese Culture University Taipei, Taiwan cshwang@faculty.pccu.edu.tw doi: 10.4156/jcit.vol5.issue8.13 Abstract Recommender systems have been emerging as a powerful technique of e-commerce. The majority of existing recommender systems uses an overall rating value on items for evaluating user’s preference opinions. Because users might express their opinions based on some specific features of the item, recommender systems solely based on a single criterion could produce recommendations that do not meet user needs. In this paper, we propose a mechanism for integrating multiple criteria into the Collaborative Filtering (CF) algorithm. Specifically, we present the implementation of Genetic Algorithms (GA) for optimal feature weighting. The proposed system consists of two main parts. First, the weight of each user toward each feature is computed by using GAs. The feature weights are then incorporated into the collaborative filtering process to provide recommendations. Empirical studies have shown that our weighting scheme can be incorporated to improve the performance of multi￾criteria CF. Keywords: Genetic Algorithms, Collaborative Filtering, Multiple Criteria, Recommender Systems 1. Introduction Although the rapid development and popularity of the Internet have brought convenience for people to access a variety of information, products, and services, the internet has also caused a scenario where people have difficulty obtaining relevant information. This scenario is commonly referred to as the problem of “information overload”. Recommender systems [1, 2] are emergent to solve the information overload challenges. Recommender systems are personalized information filtering technologies used to suggest items to users that they might like or find interesting. In recent years, recommender systems have been successfully applied in a broad range of applications, including recommending movies, books, news articles, music, etc. Recommender systems can be based on content-based filtering, collaborative filtering, and hybrid filtering [3, 4]. Content-based recommendation suggests items to users that are similar to those items that they were interested in previously. Collaborative filtering (CF) recommends items based on the information about similar items or users. Hybrid recommendation combines both algorithms into one hybrid approach to gain better performance and avoid drawbacks from each individual one. The majority of existing recommender systems uses an overall rating value on items for evaluating users’ preferences. The overall rating depends on one single criterion that usually represents the overall preference of users to a particular item. As users might express their opinions based on some specific features of the item, recommender systems merely based on a single criterion could produce recommendations that do not meet user needs. For example, in a movie recommender system, two users A and B both assign a single-criterion rating of 12 (out of 13) for Avatar. The recommender systems will conclude they have the same tastes even if A likes its story and B likes its visuals. This is called a “without distinction of interest” problem. Furthermore, even if both users like the same movie features (e.g. actors, visuals, etc.), they might select different movies. This is because people usually select a movie based on different movie features. This situation is referred to as an “unsuitable weight feature” problem [5]. Many recommender systems [6-8] have been developed to tackle the above-mentioned problems by using multiple criteria. Roux et al. [6] described a course recommender system 126