Model-based Collaborative Filtering Research Articles

Personal Value-Based User Modeling Without Attribute Evaluation and its Application to Collaborative Filtering

This paper proposes a personal values modeling method that does not require attribute ratings. The proposed method is applied to memory-based and model-based collaborative filtering (CF) to demonstrate its effectiveness. A recent trend in CF is to introduce additional factors than interaction history. A rate matching rate (RMRate) has been proposed for modeling user’s personal values, and it has been shown to be effective in increasing diversity and recommending niche (long-tail or unpopular) items. However, RMRate needs an attribute-level evaluations in addition to rating (total evaluation) to items, which limits its applicability. To obtain users’ personal values model only from a rating matrix, this paper defines users’ personal values as their tendency to select popular/unpopular items and reputable/unreputable items. Ten attributes are proposed to model user’s personal values, all of which can be calculated from a rating matrix without additional information. Experimental results on four datasets show that the proposed attributes have different characteristics from the RMRate, and can improve precision, recall, and normalized discounted cumulative gain of memory-based CF and factorization machines. It is also shown that the proposed modeling method is useful for mitigating a cold-start problem.

Harnessing Item Features to Enhance Recommendation Quality of Collaborative Filtering

Recommendation systems provide ways of directing users to items that may be relevant to them by guiding them to relevant items that will be suitable to the users according to their profiles. Collaborative filtering is one of the most successful and mature techniques of recommender system because of its domain independent ability. Bayesian Personalized Ranking Smart Linear Model (BPRSLIM) is model-based collaborative filtering (CF) recommendation algorithm that usually reconstructs a scanty user-item matrix directly; also, using only user-rating matrix usually prevents the algorithm from accessing relevant information that could enhance its recommendation accuracy. Therefore, this work reconstructs BPRSLIM user-item rating matrix via item feature information in order to improve its performance accuracy. Comprehensive experiments were carried out on a real-world dataset using different evaluation metrics. The performance of the model showed significant improvement in recommendation accuracy when compared with other top-N collaborative filtering-based recommendation algorithms, especially in precision and nDCG with 30.6% and 22.1% respectively.

Collaborative Movie Recommendation System using Enhanced Fuzzy C-Means Clustering with Dove Swarm Optimization Algorithm

Recommender Systems (RSs) aid in filtering information seeking to envisage user and item ratings, primarily from huge data to recommend the likes. Movie RSs offer a scheme to help users categorize them based on comparable interests. This enables RSs to be a dominant part of websites and e-commerce applications. This paper proposes an optimized RS for movies, primarily aiming to suggest an RS by clustering data and Computational Intelligence (CI). Unsupervised clustering, a model-based Collaborative Filtering (CF) category, is preferred as it offers simple and practical recommendations. Nevertheless, it involves an increased error rate and consumes more iterations for converging. Enhanced Fuzzy C-Means (EFCM) clustering is proposed to handle these issues. Dove Swarm Optimisation Algorithm (DSOA)-based RS is proposed for optimising Data Points (DPs) in every cluster, providing effcient recommendations. The performance of the proposed EFCM-DSOA-based RS is analysed by performing an experimental study on benchmarked MovieLens Dataset. To ensure the effciency of the proposed EFCM-DSOA-based RS, the outcomes are compared with EFCM-Particle Swarm Optimization (EFCM-PSO) and EFCM-Cuckoo Search (EFCM-CS) based on standard optimization functions. The proposed EFCM-DSOA-based RS offers improved F-measure, Accuracy, and Fitness convergence.

Collaborative filtering method based on graph neural network

An essential component of contemporary computer application technology is the recommender system. The collaborative filtering is one of RS's most crucial elements. acquiring knowledge about vector representations or, the model benefits from the combination of the graph neural network and model-based collaborative filtering since it can calculate the high-order connectivity in the item-user graph and perform better overall. This connectivity successfully and explicitly introduces the collaboration signal into the embedding process. Therefore, better embeddings also imply greater performance compared to more established collaborative filtering techniques, such as matrix factorization. The neural graph collaborative filtering (NGCF) algorithm will be primarily introduced in this article. In this paper, the performance of the NGCF algorithm is verified on several data sets, and the experimental results show that there is still room for improvement in the process of practical application. For instance, the NGCF algorithm is not appropriate for processing complicated data, and user cold start is an issue. This study offers a remedy for the difficulties the NGCF algorithm ran into in real-world use. Research on how to enhance the NGCF algorithm considering the issues will continue.

Recommender systems: a novel approach based on singular value decomposition

<span lang="EN-US">Due to modern information and communication technologies (ICT), it is increasingly easier to exchange data and have new services available through the internet. However, the amount of data and services available increases the difficulty of finding what one needs. In this context, recommender systems represent the most promising solutions to overcome the problem of the so-called information overload, analyzing users' needs and preferences. Recommender systems (RS) are applied in different sectors with the same goal: to help people make choices based on an analysis of their behavior or users' similar characteristics or interests. This work presents a different approach for predicting ratings within the model-based collaborative filtering, which exploits singular value factorization. In particular, rating forecasts were generated through the characteristics related to users and items without the support of available ratings. The proposed method is evaluated through the MovieLens100K dataset performing an accuracy of 0.766 and 0.951 in terms of mean absolute error and root-mean-square error.</span>

An enhanced matrix completion method based on non-negative latent factors for recommendation system

Among the model-based collaborative filtering (CF) recommendation algorithms, matrix factorization (MF) technology is quite efficient. An ever-increasing focus has witnessed that the non-negative latent factor (NLF)-based MF model is superior to other state-of-art models, due to its great ability to guarantee a desirable prediction accuracy while grasping the non-negativity of the LF matrix. However, most existing NLF models have not adequately considered various LFs in different conditions, which might lead to negative impacts on the model performance. In order to address this issue, a novel NLF model, i.e., efficient NLF model (ENLF), is put forward to adequately reflect the various influences of LFs to the target matrix, thereby more freedom is introduced to find the solution of the established minimization problem. Furthermore, to alleviate the computational burden caused by the introduction of more latent factors, a so-called momentum-based additive gradient descent (MAGD) algorithm is employed to learn the model, where the truncating strategy is utilized during the update so as to guarantee the non-negativity of LFs. Finally, the empirical results on six real industrial data sets show that the ENLF model based on MAGD can achieve a desirable performance with relatively low time consumption.

Parallel Algorithm of Improved FunkSVD Based on GPU

The increasing amount of network data and the development of big data technology have brought development opportunities and challenges to recommendation systems. Model-based collaborative filtering has become a mainstream algorithm in recommendation systems. For example, the Funk singular- value decomposition (FunkSVD) algorithm. However, in big data calculations, data sparseness and iterative oscillations often affect the accuracy of the FunkSVD algorithm. Moreover, when the data volume is in units of GB or more, the FunkSVD algorithm runs slowly and is ineffective. Therefore, we propose an improved FunkSVD algorithm (IFABG) based on the root-mean-square prop (RMSProp) and graphics processing unit (GPU) to solve this problem. First, we used RMSProp algorithm to improve the traditional FunkSVD algorithm, alleviate data sparseness and iterative shock, and improve the prediction accuracy of the algorithm. Next, we implemented parallelization of the improved FunkSVD algorithm in the GPU, which increased the calculation speed of the algorithm. Finally, we verify the IFABG algorithm using the Movielens dataset. The experimental results show that the IFABG algorithm is very suitable for processing sparse data, and it alleviates the iterative shock. The prediction accuracy rate is approximately 30% higher than that of the traditional FunkSVD algorithm. The experimental results also demonstrate that the IFABG algorithm has a good acceleration effect. For the same size dataset, the IFABG algorithm is faster than the traditional FunkSVD, and the acceleration ratio can be as high as 19.27.

A Location-Based Factorization Machine Model for Web Service QoS Prediction

With the prevalence of web services, a large number of similar web services are provided by different providers. To select the optimal service among these service candidates, Quality of Service (QoS), representing the non-functional characteristics, plays an important role. To obtain the QoS values of web services, a number of web service QoS prediction methods have been proposed. Collaborative web service QoS prediction is one of the most popular approaches. Based on the historical QoS data, collaborative QoS prediction methods employ memory-based collaborative filtering (CF), model-based CF, or their hybrids to predict QoS values. However, these methods usually only consider the QoS information of similar users and services, neglecting the correlation between them. To enhance the prediction accuracy, we propose a novel method to predict QoS values based on factorization machine, which leverages not only QoS information of users and services but also the user and service neighbor’s information. To evaluate our approach, we conduct experiments on a large-scale real-world dataset with 1,974,675 web service invocations. The experiment results show that our approach achieves higher prediction accuracy than other QoS prediction methods.

Hybrid high-order semantic graph representation learning for recommendations

The amount of Internet data is increasing day by day with the rapid development of information technology. To process massive amounts of data and solve information overload, researchers proposed recommender systems. Traditional recommendation methods are mainly based on collaborative filtering algorithms, which have data sparsity problems. At present, most model-based collaborative filtering recommendation algorithms can only capture first-order semantic information and cannot process high-order semantic information. To solve the above issues, in this paper, we propose a hybrid graph neural network model based on heterogeneous graphs with high-order semantic information extraction, which models users and items respectively by learning low-dimensional representations for them. We introduced an attention mechanism to allow the model to understand the corresponding edge weights adaptively. Simultaneously, the model also integrates social information in the data to learn more abundant information. We performed some experiments on related datasets. Our method achieved better results than some current advanced models, which verified the proposed model’s effectiveness.

Simulating real profiles for shilling attacks: A generative approach

Collaborative Filtering (CF) approaches are vulnerable to Shilling Attacks, in which malicious users or companies inject a large number of fake profiles in a system in order to manipulate its recommendations. One problem of current Shilling Attack models is that they commonly use straightforward statistical templates, producing profiles with different rating patterns than actual system data, which facilitates its detection, requiring a larger amount of profiles to achieve its goals. To address this problem and create profiles closer to reality, we propose using a generative model, Variational Autoencoder (VAE) to map original data distribution. With VAE, it is possible to generate new profiles based on real data, without explicit copying their actual ratings. Its generated profiles are converted to malicious profiles by adding target item rating value. We test our attack model on MovieLens 100k data set and compare to literature attack models. Our results indicate that our model outperforms all other models in model-based CF system, especially using low attack sizes (from 3% to 5%). Also, analysis comparing profiles generated from it and other approaches shows that our model ratings pattern are very similar to real profiles, which may indicate that attacks mounted using our approach may be less likely to be detected by detection approaches. Thus, we show that our attack model represents an advance on Shilling Attack models, since its superior results in model-based CF and possible indistinction from real profiles may be useful as a baseline to test detection techniques and other tasks among Shilling Attack area.

Evaluation of Collaborative Filtering for Recommender Systems

Recently, due to the increasing amount of data on the Internet along with the increase in products’ purchasing via e-commerce websites, Recommender Systems (RS) play an important role in guiding customers to buy products they may prefer. Furthermore, these systems help the companies to advertise their products to the most potential customers, and therefore raise their revenues. Collaborative Filtering (CF) is the most popular RS approach. It is classified into memory-based and model-based filtering. Memory-based filtering is in turn classified into user-based and item-based. Several algorithms have been proposed for CF. In this paper, a comparison has been performed between different CF algorithms to assess their performance. Specifically, we evaluated K-Nearest Neighbor (KNN), Slope One, co-clustering and Non-negative Matrix Factorization (NMF) algorithms. KNN algorithm is representative of the memory-based CF approach (both user-based and item-based). The other three algorithms, on the other hand, are under the model-based CF approach. In our experiments, we used a popular MovieLens dataset based on six evaluation metrics. Our results reveal that the KNN algorithm for item-based CF outperformed all other algorithms examined in this paper.

Providing reliability in recommender systems through Bernoulli Matrix Factorization

Beyond accuracy, quality measures are gaining importance in modern recommender systems, with reliability being one of the most important indicators in the context of collaborative filtering. This paper proposes Bernoulli Matrix Factorization (BeMF), which is a matrix factorization model, to provide both prediction values and reliability values. BeMF is a very innovative approach from several perspectives: a) it acts on model-based collaborative filtering rather than on memory-based filtering, b) it does not use external methods or extended architectures, such as existing solutions, to provide reliability, c) it is based on a classification-based model instead of traditional regression-based models, and d) matrix factorization formalism is supported by the Bernoulli distribution to exploit the binary nature of the designed classification model. The experimental results show that the more reliable a prediction is, the less liable it is to be wrong: recommendation quality improves after the most reliable predictions are selected. State-of-the-art quality measures for reliability have been tested, which shows that BeMF outperforms previous baseline methods and models.

Improving neighbor-based collaborative filtering by using a hybrid similarity measurement

Memory-based collaborative filtering is one of the recommendation system methods used to predict a user’s rating or preference by exploring historic ratings, but without incorporating any content information about users or items. It can be either item-based or user-based. Taking item-based Collaborative Filtering (CF) as an example, the way it makes predictions is accomplished in 2 steps: first, it selects based on pair-wise similarities a number of most similar items to the predicting item from those that the user has already rated on. Second, it aggregates the user’s opinions on those most similar items to predict a rating on the predicting item. Thus, similarity measurement determines which items are similar, and plays an important role on how accurate the predictions are. Many studies have been conducted on memory-based CFs to improve prediction accuracy, but none of them have achieved better prediction accuracy than state-of-the-art model-based CFs. In this paper, we proposed a new approach that combines both structural and rating-based similarity measurement. We found that memory-based CF using combined similarity measurement can achieve better prediction accuracy than model-based CFs in terms of lower MAE and reduce memory and time by using less neighbors than traditional memory-based CFs on MovieLens and Netflix datasets.

A Hybrid Model- and Memory-Based Collaborative Filtering Algorithm for Baseline Data Prediction of Friedreich's Ataxia Patients

Friedreich's ataxia (FRDA) is the most common inherited ataxia that causes progressive damage of nervous systems and performance deterioration of physical movements. FRDA baseline data analysis plays a crucial role in advancing the disease research, where the main obstacle comes from the baseline data collection primarily due to the degenerative symptoms of the FRDA patients. Inspired by the nowadays popular collaborative filtering (CF) method, a new FRDA baseline data collection algorithm is proposed in this article, with which the patients (or their families) are only required to provide certain reliable baseline data acquired from home and the uncertain/missing parts of the data can then be predicted with acceptable accuracy by utilizing existing patient information. The framework of the proposed algorithm is constructed based on a novel hybrid model combining the merits of model- and memory-based CF methods, thereby facilitating the baseline data collection with improved prediction accuracy. The proposed hybrid algorithm exhibits the following two main features: when a patient does not have neighbors sharing similar baseline data, the model-based CF component is activated to employ certain clustering method to find similar neighbors based on their attributes; and in the case that a patient does have neighbors, a novel similarity measure, which accounts for more statistical characteristics by integrating rating habits and degree of co-rated items, is developed in the memory-based component of the algorithm in order to adjust initial similarities between the patients. To evaluate the advantages of the proposed algorithm, the Scale for the Assessment and Rating of Ataxia is selected from the European FRDA Consortium for Translational Studies database. Experimental results demonstrate that our proposed hybrid CF approach is superior to other conventional approaches.

Model-based Collaborative Filtering with Transparency using Linear Regression

Model-based Collaborative Filtering with Transparency using Linear Regression

Applying landmarks to enhance memory-based collaborative filtering

Memory-based Collaborative Filtering (CF) has been a widely used approach for personalised recommendation with considerable success in many applications. An important issue regarding memory-based CF lies in similarity computation: the sparsity of the rating matrix leads to similarity computations based on few co-rated items between users, resulting in high sensitive predictions. Additionally, the ‘sparse’ similarity computation has high computational cost, due to the dimensionality of the item space. In this paper, we pursue both these issues. We propose a new model to compute similarity by representing users (or items) through their distances to preselected users, named landmarks. Such user modelling allows the introduction of more ratings into similarity computations through transitive relations created by the landmarks. Unlike conventional memory-based CF, the proposal builds a new user space defined by distances to landmarks, avoiding sensitivity in similarity computations. Findings from our experiments show that the proposed modelling achieves better accuracy than the ‘sparse’ similarity representation in all tested datasets, and has also yielded competitive accuracy results against the compared model-based CF algorithms. Furthermore, the proposed implementation has beaten all compared methods in terms of computational performance, becoming a promising alternative to memory-based CF algorithms for large datasets.

Parallel naïve Bayes regression model-based collaborative filtering recommendation algorithm and its realisation on Hadoop for big data

Collaborative filtering (CF) algorithms are widely used in a lot of recommender systems. However, space-time overhead and high computational complexity hinder their use in large-scale systems. This paper implements the parallel naive Bayes regression model based collaborative filtering recommendation algorithm on Hadoop computing platform to scalability problem of CF. Firstly, this paper analysis the inherent parallelism of the naive Bayesian regression model and constructs the theoretical model of naive Bayesian parallelisation. Secondly, the parallel naive Bayes regression model-based collaborative filtering recommendation algorithm is realised on Hadoop platform with distributed Hadoop distributed file system (HDFS) and MapReduce as the transparent distributed infrastructure. And its temporal-spatial overhead, speedup is discussed. Finally, applying parallel the naive Bayes regression model-based collaborative filtering recommendation algorithm to a large dataset. The experiment results on Netflix dataset show that this method has high scalability and less space-time overhead, which is suitable for real-time recommendation on large dataset.

QoS Prediction for Mobile Edge Service Recommendation With Auto-Encoder

In the mobile edge computing environment, there are a large number of mobile edge services which are the carriers of various mobile intelligent applications. So how to recommend the most suitable candidate from such a huge number of available services is an urgent task, especially the recommendation task based on quality-of-service (QoS). In traditional service recommendation, collaborative filtering (CF) has been studied in academia and industry. However, due to the mobility of users and services, there exist several defects that limit the application of the CF-based methods, especially in an edge computing environment. The most important problem is the cold-start. In this paper, we propose an ensemble model which combines the model-based CF and neighborhood-based CF. Our approach has two phases, i.e., global features learning and local features learning. In the first phase, to alleviate the cold-start problem, we propose an improved auto-encoder which deals with sparse inputs by pre-computing an estimate of the missing QoS values and can obtain the effective hidden features by capturing the complex structure of the QoS records. In the second phase, to further improve prediction accuracy, a novel computation method is proposed based on Euclidean distance that aims to address the overestimation problem. We introduce two new concepts, common invocation factor and invocation frequency factor, in similarity computation. Then we propose three prediction models, containing two individual models and one hybrid model. The two individual models are proposed to utilize user similar neighbors and service similar neighbors, and the hybrid model is to utilize all neighbors. The experiments conducted in a real-world dataset show that our models can produce superior prediction results and are not sensitive to parameter settings.

Parallel naïve Bayes regression model-based collaborative filtering recommendation algorithm and its realisation on Hadoop for big data

Parallel naïve Bayes regression model-based collaborative filtering recommendation algorithm and its realisation on Hadoop for big data

A novel Adaptive Genetic Neural Network (AGNN) model for recommender systems using modified k-means clustering approach

The Recommender System (RS) plays an important role in information retrieval techniques in a bid to handle massive online data effectively. It gives suggestions on items/services to the target online user to ensure correct decisions quickly and easily. Collaborative Filtering (CF) is a key approach in RS providing a recommendation to the target online user, based on a rating similarity among users. Unsupervised clustering approach is a model-based CF, which is preferred as it ensures simple and effective recommendation. Such CFs suffer from a high error rate and needs additional iterations for convergence. This paper proposes a Modified k-means clustering approach to eliminate the above mentioned issues to provide well-framed clusters. The novel supervised Adaptive Genetic Neural Network (AGNN) method is proposed to locate the most favored data points in a cluster to deliver effective recommendations. The performance of the proposed RS is measured by conducting an experimental analysis on benchmark MovieLens and Netflix datasets. Results are compared with state-of-the-art methods namely Artificial Neural Network (ANN) and Fuzzy based RS models to show the effectiveness of the proposed AGNN method.