Multi-criteria Recommender Systems Research Articles

A Neural Networks Approach for Improving the Accuracy of Multi-Criteria Recommender Systems

Accuracy improvement has been one of the most outstanding issues in the recommender systems research community. Recently, multi-criteria recommender systems that use multiple criteria ratings to estimate overall rating have been receiving considerable attention within the recommender systems research domain. This paper proposes a neural network model for improving the prediction accuracy of multi-criteria recommender systems. The neural network was trained using simulated annealing algorithms and integrated with two samples of single-rating recommender systems. The paper presents the experimental results for each of the two single-rating techniques together with their corresponding neural network-based models. To analyze the performance of the approach, we carried out a comparative analysis of the performance of each single rating-based technique and the proposed multi-criteria model. The experimental findings revealed that the proposed models have by far outperformed the existing techniques.

Enhanced multi-criteria recommender system based on fuzzy Bayesian approach

In the area of recommender systems, collaborative filtering is widely used technique for recommending appropriate items to a user based on the available ratings given by similar users. Most recommender systems (RSs) work only on the single criterion rating i.e., overall rating, however overall rating may not be a good representative of a user preference. Single criterion collaborative filtering (CF) does not generate more reliable recommendations because it suffers from correlation based similarity problems. Moreover, representation of uncertain user preferences is another concern of CF. In our work, we develop a novel fuzzy Bayesian approach to multi-criteria CF for handling uncertain user preferences and correlation based similarity problems. Further, incorporation of multi-criteria ratings into CF would be helpful for generating effective recommendations. Through experiments on Yahoo! Movies dataset, we compare our proposed approach to baseline approaches and demonstrate its effectiveness in terms of accuracy, recall and f-measure.

Multicriteria Recommender System for Life Insurance Plans based on Utility Theory

Background/Objectives: This paper proposes a recommender system for life insurance. Insurance is a way of managing risks and has been used as financial instrument for a long time. The remarkable increase in competition within the insurance sector of the India has resulted in an overwhelming number of insurance products being available in the market. Methods: Utility theory is applied to recommend most suitable policy to users. Grey Relation Analysis (GRA) is utilized in intuitionistic fuzzy environment to determine users’ preference over alternatives. Results/Findings: The proposed recommender system has been tested with approximately 600 potential customers of different region of Chhattisgarh (INDIA). The accuracy of the recommender system is 92.6%. Also, our recommendation system has been tested with different parameters by domain experts of different levels (Branch Manager, Insurance Advisor, Development officers). They also found the results significantly accurate. Improvement: Most existing recommender system are based on collaborative filtering technique or content based system, which mainly focuses on finding relations between products and between customers through machine learning techniques. They recommend products without concerning the users’ personalized requirement. Our recommender system takes users’ current need in account and recommends most suitable policy to them. Application: The proposed recommendation technique has worked efficiently with the life insurance products and it can also be successfully applied on the products with specific preference like medical insurance, personal vehicles, and electronic home appliances.

Handling Natural Noise in Multi Criteria Recommender System utilizing effective similarity measure and Particle Swarm Optimization

Multi criteria recommender systems generate quality recommendations to users by incorporating criteria ratings into recommender system using collaborative filtering because ratings over multiple criteria can capture user preferences efficiently. However, aggregation of similarities computed on multiple criteria is still a major concern. Moreover, the concept of natural noise is an emerging trend that is related to inconsistent behaviour of users. Our work in this paper is an attempt towards developing multi criteria recommender systems that deals with inconsistent ratings and uses particle swarm optimization to learn optimal weights for a user over different criteria in the aggregation process.

A multi-criteria recommendation system using dimensionality reduction and Neuro-Fuzzy techniques

Multi-criteria collaborative filtering (MC-CF) presents a possibility to provide accurate recommendations by considering the user preferences in multiple aspects of items. However, scalability and sparsity are two main problems in MC-CF which this paper attempts to solve them using dimensionality reduction and Neuro-Fuzzy techniques. Considering the user behavior about items' features which is frequently vague, imprecise and subjective, we solve the sparsity problem using Neuro-Fuzzy technique. For the scalability problem, higher order singular value decomposition along with supervised learning (classification) methods is used. Thus, the objective of this paper is to propose a new recommendation model to improve the recommendation quality and predictive accuracy of MC-CF and solve the scalability and alleviate the sparsity problems in the MC-CF. The experimental results of applying these approaches on Yahoo!Movies and TripAdvisor datasets with several comparisons are presented to show the enhancement of MC-CF recommendation quality and predictive accuracy. The experimental results demonstrate that SVM dominates the K-NN and FBNN in improving the MC-CF predictive accuracy evaluated by most broadly popular measurement metrics, F1 and mean absolute error. In addition, the experimental results also demonstrate that the combination of Neuro-Fuzzy and dimensionality reduction techniques remarkably improves the recommendation quality and predictive accuracy of MC-CF in relation to the previous recommendation techniques based on multi-criteria ratings.

Exploratory study of multi-criteria recommendation algorithms over technology enhanced learning datasets

Results of previous studies have indicated that the performance of recommendation algorithms depends on the characteristics of the application context. The same algorithms have shown to be performing in totally different ways when a new or evolved data set is considered, thus leading to a need for continuous monitoring of how they operate in a realistic setting. In this paper we investigate such a real life implementation of a multi-criteria recommender system and try to identify the needed adjustments that need to take place in order for it to better match the requirements of its operational environment. More specifically, we examine the case of a multi-attribute collaborative filtering algorithm that has been supporting the recommendation service within a Web portal for organic and sustainable education. Our study particularly explores the experimental performance of the already implemented algorithm, as well as an alternative one, using data from the intended application, a simulated expansion of it, and from similar portals. The results of this study indicate the importance of the frequent experimental investigation of a recommender system's various design options, and the need for the exploration of adaptive implementations in real life recommender systems.

Accuracy and Diversity Improvements for Multi-Criteria Recommender Systems

데이터의 폭발적인 증가와 더불어 개인화 서비스에 대한 수요가 증가함에 따라 다양한 추천 알고리즘이 개발되고 있다. 하지만, 대부분의 알고리즘은 추천의 정확도를 향상시키기 위해 개발되었으며 다양성은 고려되지 않고 있다. 많은 사용자들이 좋아할 만한 인기 제품을 추천하면 어느 정도 높은 정확도를 얻을 수 있지만, 이는 추천 시스템의 사용 이유인 다양하고 개인화된 추천 서비스를 제공하지 못하게 된다. 반대로 개인마다 차별화된 비인기 제품을 추천하면 다양한 제품을 추천할 수 있으나 추천의 정확도는 낮아지게 된다. 따라서 본 연구에서는 이러한 문제점을 극복하기 위해서 기존 연구 결과를 통합하여 -여러 항목별 점수(multi-criteria rating)를 이용한 추천방안과 다양한 순위 결정 방법(ranking-based techniques)을 이용- 추천의 정확도와 다양성을 동시에 증가시키고자 하였다. 본 연구에서 제시한 통합적인 방법을 영화 데이터를 이용하여 분석한 결과, 정확도와 다양성 측면에서 모두 성능이 향상됨을 확인하였다.

HYBRID PROFILING FOR HYBRID MULTICRITERIA RECOMMENDATION BASED ON IMPLICIT MULTICRITERIA INFORMATION

Multicriteria recommender systems typically gather the user preferences by asking a user to rate different aspects of an item on a sliding scale explicitly. However, this approach could possibly cause intrusiveness and conflict on user preferences. For example, an individual's preference on each aspect of an item may conflict with an overall preference. To overcome such limitations, we proposed the hybrid profiling framework to generate a set of useful implicit dataset to support multicriteria recommender systems. We also proposed two hybrid multicriteria recommendation approaches, namely the user-attribute-based (UAB) and the user-item matching (UIM) to improve recommendation accuracy. Finally, we conducted experiments to confirm the efficiency of the proposed approaches. The experiments show that the profiling framework and two hybrid recommendation approaches can alleviate the problem in an intrusive manner and decrease the degree of preference conflict without decreasing the accuracy of the recommendation. They also show that our proposed hybrid multicriteria recommendation approaches can significantly outperform both the traditional collaborative filtering and the simple multicriteria filtering approaches.

Genetic Algorithms for Feature Weighting in Multi-criteria Recommender Systems

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 multicriteria CF.

Analysis and Classification of Multi-Criteria Recommender Systems

Recent studies have indicated that the application of Multi-Criteria Decision Making (MCDM) methods in recommender systems has yet to be systematically explored. This observation partially contradi...

EXPERIMENTAL ANALYSIS OF DESIGN CHOICES IN MULTIATTRIBUTE UTILITY COLLABORATIVE FILTERING

Recommender systems have already been engaging multiple criteria for the production of recommendations. Such systems, referred to as multicriteria recommenders, demonstrated early the potential of applying Multi-Criteria Decision Making (MCDM) methods to facilitate recommendation in numerous application domains. On the other hand, systematic implementation and testing of multicriteria recommender systems in the context of real-life applications still remains rather limited. Previous studies dealing with the evaluation of recommender systems have outlined the importance of carrying out careful testing and parameterization of a recommender system, before it is actually deployed in a real setting. In this paper, the experimental analysis of several design options for three proposed multiattribute utility collaborative filtering algorithms is presented for a particular application context (recommendation of e-markets to online customers), under conditions similar to the ones expected during actual operation. The results of this study indicate that the performance of recommendation algorithms depends on the characteristics of the application context, as these are reflected on the properties of evaluations' data set. Therefore, it is judged important to experimentally analyze various design choices for multicriteria recommender systems, before their actual deployment.