A new clustering method for collaborative filtering

Abstract. Searching in huge amount of information available on the internet is undoubtedly a challenging task. A lot of new web sites are created every day, containing not only text, but other types of resources: e.g. songs, movies or images. As a consequence, a simple search result list from search engines becomes insufficient. Recommender systems are the solution supporting users in finding items, which are interesting for them. These items may be information as well as products, in general. The main distinctive feature of recommender systems is taking into account personal needs and taste of users. Collaborative filtering approach is based on users' interactions with the electronic system. Its main challenge is generating on-line recommendations in reasonable time coping with large size of data. Appropriate tool to support recommender systems in increasing time efficiency are clustering algorithms, which find similarities in off-line mode. Commonly, it involves decreasing of prediction accuracy of final recommendations. This article presents an approach based on clustered data, which prevents the negative consequences, keeping high time efficiency. An input data are clusters of similar items and searching the items for recommendation is limited to the elements from one cluster.

show abstract

“…Another example is a modified method proposed in [15]. Modification concerned similarity computation in a clustering procedure.…”

Section: Description Of the Clustering Algorithm Used In The Systemmentioning

confidence: 99%

“…To increase the number of clustered input data a procedure of complement clustering was performed [15]. For every not clustered point its distance to the formed previously clusters was calculated according to Equation 5.…”

Section: Fig 1 Components Of the Recommender System Used In The Expmentioning

confidence: 99%

A Modified Clustering Algorithm DBSCAN Used in a Collaborative Filtering Recommender System for Music Recommendation

Kużelewska

Wichowski

2015

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

show abstract

“…But at the same time it deletes some useful hidden data. Clustering [4,5]is mainly according to a certain method to divide the users or items into a plurality of clusters, and having a similar preference of the user, or similar properties as the type of project is assigned to the same cluster. This method can alleviate to some extent the problems of data sparsity, but in the case of the original data is extremely sparse the recommended accuracy of the experimental results is unsatisfactory.…”

Section: Related Workmentioning

confidence: 99%

An Improved Collaborative Filtering Algorithm Based on Sparse Dataset's Optimization with User's Browser Information

Sun

Huang

2013

2013 10th Web Information System and Application Conference

View full text Add to dashboard Cite

Collaborative filtering technology is the mainstream recommendation technology in personalized recommendation system, the sparsity of the dataset plays a leading role in the prediction accuracy of the collaborative filtering algorithm. Virtual data filling and neighbors' calculation etc. are adopted to solve the sparsity problem in traditional methods, which lacked of dynamic changes of rating data and objectivity. For the deficiencies of the traditional methods, making use of the data redundancy and dynamic changes in Big Data environment, to improve the sparse dataset, this paper proposes an improved collaborative filtering algorithm based on optimizing sparse dataset through user's browser information. This approach gets data related with user objective score from various fields through user's IP address to fill the dataset and reduce the sparsity of the dataset of candidate neighbors. The algorithm is compared with other classic algorithms on the performance and analyzing the result in the case of sparse dataset. The experiments results show that the algorithm can effectively reduce the sparsity of the data set, and improve the quality of recommendation system.

show abstract

“…Uma outra técnica amplamente utilizada na área de mineração de dados, conhecida por agrupamento de dados (clustering), tem sido utilizada também em sistemas de recomendação. FOSTER, 1998;PHAM et al, 2011;RONGFEI;MAOZHONG;CHAO, 2010;YANG, 2008). A subseção 2.5.2 introduz o conceito de biclusterização -biclustering ou coclustering (MADEIRA; OLIVEIRA, 2004;CHO, 2008), que se refere a uma classe distinta de algoritmos de agrupamento que executam simultaneamente o agrupamento de linhas e colunas.…”

Section: Recomendações On-lineunclassified

Agrupamento de dados baseado em predições de modelos de regressão: desenvolvimentos e aplicações em sistemas de recomendação

Pereira¹

View full text Add to dashboard Cite

Recommender Systems (RS) are powerful and popular tools for e-commerce. To build its recommendations, RS make use of multiple data sources, capture the characteristics of items, users and their transactions, and take advantage of prediction models. Given the large amount of data involved in the predictions made by RS, is unlikely that all predictions can be well represented by a single global model. Another important aspect to note is the problem known as cold-start that, despite that recent advances in the RS area, it is still a relevant issue that deserves further attention. The problem arises due to the lack of prior information about new users and new items. This thesis presents a hybrid recommendation approach that addresses the (pure) cold start problem, where no collaborative information (ratings) is available for new users. The approach is based on an existing algorithm, named SCOAL (Simultaneous Co-Clustering and Learning). In its original version, based on multiple linear prediction models, the SCOAL algorithm has shown to be efficient and versatile. In addition, it can be used in a wide range of problems of classification and / or regression. The SCOAL algorithm showed impressive results with the use of linear prediction models, but there is still room for improvements with nonlinear models. From this perspective, this thesis presents a variant of the SCOAL based on Extreme Learning Machines. Besides improving the accuracy, another important issue related to the development of RS is system scalability. In this sense, a parallel version of the SCOAL, based on OpenMP, was developed, aimed at minimizing the computational cost involved as prediction models are learned. Experiments using real-world datasets has shown that all proposed developments make SCOAL algorithm even more attractive for a variety of practical applications.

show abstract

A new clustering method for collaborative filtering

Cited by 16 publications

References 7 publications

A Modified Clustering Algorithm DBSCAN Used in a Collaborative Filtering Recommender System for Music Recommendation

A Modified Clustering Algorithm DBSCAN Used in a Collaborative Filtering Recommender System for Music Recommendation

An Improved Collaborative Filtering Algorithm Based on Sparse Dataset's Optimization with User's Browser Information

Agrupamento de dados baseado em predições de modelos de regressão: desenvolvimentos e aplicações em sistemas de recomendação

Contact Info

Product

Resources

About