Item-based Collaborative Filtering Research Articles

Efficient Optimization of Sparse User Encoder Recommenders

Embedding representations are a popular approach for modeling users and items in recommender systems, e.g., matrix factorization, two-tower models or autoencoders, where items and users are embedded in a small dimensional, dense embedding space. On the other hand, there are methods that model high-dimensional relationships between items, most notably item-based collaborative filtering (CF), which is based on an item-to-item similarity matrix. Item-based CF has been proposed over two decades ago and gained new interest through new learning methods in the form of SLIM [18] and most recently EASE [25]. In this work, we rephrase traditional item-based collaborative filtering as sparse user encoders where the user encoder is an (arbitrary) function and the item representation is learned. Item-based CF is a special case where the sparse user encoding is the one-hot encoding of a user’s history. Different from typical dense user/item encoder models, this work targets high-dimensional and sparse user encoders. The core contribution is an efficient closed form learning algorithm that can solve arbitrary sparse user encoders. Several applications of this algorithm including higher order encoders, hashed encoders, and feature based encoders are presented.

Research on the application of bilateral collaborative algorithm in hotel tourism consumption preference recommendation

SummaryThe continuous expansion of the domestic tourism market has driven the development of the hotel related industry. It is a key focus of the hotel industry to provide accurate supporting services to tourists based on their preferences. However, the current recommendation model mainly considers tourist factors and lacks consideration for hotel factors. In order to address the concerns of recommendation services, a bilateral collaborative recommendation model is proposed, which first analyzes tourist feature data through comprehensive similarity calculation. Firstly, the comprehensive similarity calculation is used to analyze tourist feature data, and the K‐means Clustering Algorithm (K‐means) is introduced to improve the clustering effect between data. And to better distinguish tourist preferences, the Gale Shapley (G‐S) bilateral matching model is introduced to achieve preference matching of consumer features. In the Mean Absolute Error (MAE) analysis of the model, the User‐based Collaborative Filtering (UBCF) model performs the worst. When the k value is 90, the minimum MAE value is 0.7896. The Item‐based Collaborative Filtering (IBCF) model is 0.7776, with the best performance being the improved G‐S model of 0.7325. In the performance testing of three models' preference recommendations, the improved G‐S model achieves an accuracy of over 90% in all three types of feature data recommendations, with the lowest accuracy of 62.5% for UBCF and 71.3% for IBCF. It can be seen that the proposed improved G‐S model can accurately identify tourist and hotel preference data and accurately recommend corresponding projects. The research content provides important technical references for the digital transformation and development of the tourism industry.

A crowd-sourcing recommendation algorithm OPCA-CF using outer-product co-attention mechanism

ABSTRACT With the rapid development of information technology, crowd-sourcing technology is increasingly used in non-invasive monitoring in smart cities. Applying recommendation algorithms in crowd-sourcing can optimise resource allocation, improve task-matching accuracy and enhance participant satisfaction, whereas existing recommendation algorithms cannot be directly applied in crowd-sourcing, as such scenarios have unique features, such as task timeliness and multi-role users. Designed explicitly for crowd-sourcing scenarios, our OPCA-CF (Outer-product Co-attention Collaborative Filtering) algorithm is formed by an upgraded ItemCF (Item-based Collaborative Filtering) algorithm as main-network and OPCA (Outer-product Co-attention) mechanism as a sub-network. Firstly, ItemCF is improved through attribute-level task feature learning, new-role feature and weighted cross-entropy in the loss function. Most importantly, we propose OPCA using outer-product, while the existing co-attention mechanism only uses inner-product. Compared with the best existing algorithm using real-world datasets, OPCA-CF’s performance is proved to be superior by 1.24%, 4.25% and 5.35%, with binary classification indicators AUC (Area under Curve), recommended Lists related indicators HR (Hit Ratio) and MRR (Mean Reciprocal Rank), respectively. All the performance indicators verified the effectiveness of the OPCA-CF algorithm.

How Do Product Recommendations Help Consumers Search? Evidence from a Field Experiment

Product recommendations can benefit consumers’ online product search via multiple underlying mechanisms, such as showing products that offer them high value, facilitating navigation on the website, or exposing more product information. However, it is unclear ex ante which is the primary underlying mechanism that drives the benefits of product recommendations to consumers. We conducted a randomized field experiment to estimate the benefits of an item-based collaborative filtering (CF) recommendation system to consumers. We collect unique data on the affinity scores computed by an item-based CF algorithm to develop measures of a product’s net value and horizontal (taste) fit for consumers. Our results indicate that product recommendations help consumers search for higher-value products that are lower priced, fit their tastes better, or both. Besides that, we find that the ability to find higher-value products (rather than easy navigation or exposure to more product information) is the primary driver for consumers’ higher purchase probabilities under recommendations. We further find a higher benefit of recommendations in product categories with higher price dispersion and heterogeneity in consumers’ tastes, providing additional evidence for the lower price and better horizontal fit mechanisms. Finally, we find that when made available, consumers substitute their usage of other search tools on the website with product recommendations. Our findings have important implications for online retailers, policymakers, regulators, and item-based CF recommendation system design. This paper was accepted by D. J. Wu, information systems. Funding: This work was supported by the Public Utility Research Center of the University of Florida, the Hi! PARIS Fellowship, the HEC Foundation, and the Leavey School of Business at Santa Clara University [Grant 102720]. Supplemental Material: Data and the online appendix are available at https://doi.org/10.1287/mnsc.2023.4951 .

A Hybrid Deep Ranking Weighted Multi-Hashing Recommender System

In countries where there is a low availability of resources for language, businesses face the challenge of overcoming language barriers to reach their customers. One possible solution is to use collaborative filtering-based recommendation systems in their native languages. These systems employ algorithms that understand the customers’ preferences and suggest products or services in their native language. Collaborative filtering (CF) is a popular recommendation technique that simulates word-of-mouth phenomena. However, the accuracy of a CF recommendation can be affected by sparse data. In this research paper, we present a novel hybrid weighted multi-deep ranking supervised hashing (HWMDRH) approach. Our method leverages both user-based and item-based CF by merging the item-based deep ranking weighted multi-hash recommender system prediction with the user-based deep ranking weighted multi-hash recommender system prediction to generate Top-N prediction. We conducted extensive experiments on the MovieLens 1M dataset, and our results show that the proposed HWMDRH model outperforms existing models and achieves state-of-the-art performance across recall, precision, RMSE, and F1-score metrics.

Fashion Recommendation System using Collaborative Filtering

Collaborative Filtering is an method used to build a recommendation system with the concept that conclusions from different clients are used to anticipate things that may be of interest to users. This research uses data from Rent the Runway and the method used is Item-based Collaborative filtering, where the system will look for similarities in products that have been purchased by customers and then look for predictive values. Fashion requires recommendations because it plays a crucial role in helping individuals express their identity, personal style, and personality through clothing choices, accessories, and dressing styles.The recommendation system uses the item method based on analyzing the number of purchases or sales and grouping according to each product category so that it can help consumers in choosing fashion products. It was found that the use of Adjusted Cosine Similarity produces better recommendations with an average MAE value of 0.2750, while Cosine Similarity with an average MAE difference of 0.3989. This proves that the use of adjusted cosine similarity can produce better recommendations because the adjustment algorithm not only considers user behavior, but also produces lower performance errors.

A Novel Algorithm for Professor Recommendation in Higher Education

This paper introduces a novel professor-recommendation system designed specifically for community college and university courses. Building upon an existing algorithm for one-on-one teacher recommendations, we leveraged insights from the literature on Massive Open Online Course (MOOC) recommender algorithms. By analysing various approaches, we combined and refined ideas to develop an optimised system. Our approach utilises a tri-module framework that incorporates supervised and unsupervised learning techniques. The first module employs a Gradient-boosted Decision Tree algorithm, augmented with multiple factors and student dropout rates as ground truth, to generate a ranking score. The second module applies Apriori Association and Density-based Spatial Clustering of Applications with Noise (DBSCAN) algorithms to analyse these factors and identify professors with similar characteristics. In the third module, item-based collaborative filtering is employed, incorporating user ratings and the cosine similarity algorithm. The outputs from these three modules are subsequently integrated through a weighted average. This addition enables the system to prioritise opportunities for new professors, thereby ensuring a balanced recommendation approach. The resulting combined ranking score provides accurate recommendations for course instructors. This approach can be integrated into university course selection software for the benefit of both students and educational institutions.

A movie recommendation method based on knowledge graph and time series

Traditional collaborative filtering algorithms use user history rating information to predict movie ratings Other information, such as plot and director, which could provide potential connections are not fully mined. To address this issue, a collaborative filtering recommendation algorithm named a movie recommendation method based on knowledge graph and time series is proposed, in which the knowledge graph and time series features are effectively integrated. Firstly, the knowledge graph gains a deep relationship between users and movies. Secondly, the time series could extract user features and then calculates user similarity. Finally, collaborative filtering of ratings can calculate the user similarity and predicts ratings more precisely by utilizing the first two phases’ outcomes. The experiment results show that the A Movie Recommendation Method Fusing Knowledge Graph and Time Series can reduce the MAE and RMSE of user-based collaborative filtering and Item-based collaborative filtering by 0.06,0.1 and 0.07,0.09 respectively, and also enhance the interpretability of the model.

A distributed real-time recommender system for big data streams

Recommender Systems (RS) play a crucial role in our lives. As users become continuously connected to the internet, they are less tolerant of obsolete recommendations made by an RS. Online RS has to address three requirements: continuous training and recommendation, handling concept drifts, and the ability to scale. Streaming RS proposed in the literature address the first two requirements only. That is because they run the training process on a single machine. To tackle the third challenge, we propose a Splitting and Replication mechanism for distributed streaming RS. Our mechanism is inspired by the shared-nothing architecture that underpins contemporary big data processing systems. We have applied our mechanism to two well-known approaches for online RS, namely, matrix factorization and item-based collaborative filtering. We conducted experiments comparing the performance with the baseline (single machine). Evaluating different data sets, experiments show online recall improvement by 40% with more than 50% less memory consumption.

A Comparative Analysis of Collaborative Filtering Similarity Measurements for Recommendation Systems

Collaborative Filtering (CF) is a widely used technique in recommendation systems to suggest items to users based on their previous interactions with the system. CF involves finding correlations between the preferences of different users and using those correlations to provide recommendations. This technique can be divided into user-based and item-based CF, both of which utilize similarity metrics to generate recommendations. Content-based filtering is another commonly used recommendation technique that analyzes the attributes of items to suggest similar items. To enhance the accuracy of recommendation systems, hybrid algorithms that combine CF and content-based filtering techniques have been developed. These hybrid systems leverage the strengths of both approaches to provide more accurate and personalized recommendations. In conclusion, collaborative filtering is an essential technique in recommendation systems, and the use of various similarity metrics and hybrid techniques can enhance the quality of recommendations.

Movie Recommendation System Based on Traditional Recommendation Algorithm and CNN Model

As streaming services have expanded in recent years, an excellent recommendation algorithm, as one of the core technologies in movie service, brings huge benefits. Although the current application research of recommendation algorithms is mature, most systems or products usually rely on only one main algorithm. This makes it difficult for the system to overcome the shortcomings of various algorithms and cannot benefit from the combination of multiple recommendation algorithms. In addition, the widely used recommendation models are found to be unable to extract the finer features of users and movies, and the calculation time is very long. Meanwhile, the recommendation results are inaccurate, and they are not user-friendly. In this research, we design and construct a system to recommend movies which is consisted by a model based on a convolutional neural network consisting. By extracting the features of users and movies, we can calculate the direct similarity of different users' movies and then predict movie ratings to give recommendations. When extracting features of users and movies, we refer to traditional algorithms based on content and content. Through the "cosine similarity" and "user movie score matrix", the Item-Based Collaborative Filtering and User-Based Collaborative Filtering can be well implemented. To sum up, our movie recommendation system is based on Convolutional Neural Network (CNN) model and the performance of the system is improved by adopting multiple recommendation algorithms such as content-based recommendation and system filtering. We efficiently trained the neural network and finally built a movie recommendation system with a faster operation speed, although some parts are not perfect.

SoMLier: a South African wine recommender system

ABSTRACT Recommender Systems (RS) are used to generate recommendations of items that a user may be interested in. Several commercial wine recommender systems exist but are largely tailored to consumers outside of South Africa (SA). Consequently, these systems are of limited use to novice wine consumers in SA. In this research, a system soMLier (a combination of the terms ‘sommelier’ and ‘Machine Learning’) is developed for SA consumers that yields high-quality wine recommendations, maximises the accuracy of predicted ratings for those recommendations and provides insights into why those suggestions were made. This system is developed using two datasets – a database containing several attributes of SA wines and the corresponding numeric 5-star ratings made by users on Vivino.com. Using these datasets, several recommendation methodologies are investigated and it is found that collaborative filtering succeeds at generating lists of relevant wine recommendations, matrix factorisation techniques accurately predict ratings and content-based methods are most appropriate for explaining wine recommendations. These methods are optimally combined in the soMLier system. Though it would benefit from more explicit user data to establish a richer model of user preferences, soMLier can assist consumers in discovering wines they will likely enjoy and understanding their preferences of SA wine. Abbreviations: SA: South Africa(n); RS: Recommender System(s); IBCF: Item-basedCollaborative Filtering; CB: Content-Based; MF: Matrix Factorisation; RMSE: RootMean Square Error; COV: Coverage; PER: Personalistion; ARHR: Average ReciporcalHit Rate

A collaborative filtering recommendation algorithm based on DeepWalk and self-attention

Graph embedding is one of the vital technologies in solving the problem of information overload in recommendation systems. It can simplify the vector representations of items and accelerates the calculation process. Unfortunately, the recommendation system using graph embedding technology does not consider the deep relationships between items when it learns embedding vectors. In order to solve this problem, we propose a collaborative filtering recommendation algorithm based on DeepWalk and self-attention in this paper. This algorithm can enhance the accuracy in measuring the similarity between items and obtain more accurate embedding vectors. Chronological order and mutual information are used to construct a weighted directed relationship graph. Self-attention and DeepWalk are utilised to generate embedding vectors. Then item-based collaborative filtering is utilised to obtain recommended lists. The result of the relative experiments and evaluations on three public datasets shows that our algorithm is better than the existing ones.

A comparative analysis of different recommender systems for university major and career domain guidance

Academic advising is inhibited at most of the high schools to help students identify appropriate academic pathways. The choice of a career domain is significantly influenced by the complexity of life and the volatility of the labor market. Thus, high school students feel confused during the shift period from high school to university, especially with the enormous amounts of data available on the Web. In this paper, an extensive comparative study is conducted to investigate five approaches of recommender systems for university study field and career domain guidance. A novel ontology is constructed to include all the needed information for this purpose. The developed approaches considered user-based and item-based collaborative filtering, demographic-based recommendation, knowledge base supported by case-based reasoning, ontology, as well as different hybridizations of them. A case study on Lebanese high school students is analyzed to evaluate the effectiveness and efficiency of the implemented approaches. The experimental results indicate that the knowledge-based hybrid recommender system, combined with the user-based collaborative filtering and braced with case-based reasoning as well as ontology, generated 98% of similar cases, 95% of them are personalized based on the interests of the high school students. The average usefulness feedback and satisfaction level of the students concerning this proposed hybrid approach reached 95% and 92.5% respectively, which could be a solution to similar problems, regardless of the application domain. Besides, the constructed ontology could be reused in other systems in the educational domain.

Recommendation System with Biclustering

The massive growth of online commercial data has raised the request for an automatic recommender system to benefit both users and merchants. One of the most frequently used recommendation methods is collaborative filtering, but its accuracy is limited by the sparsity of the rating dataset. Most existing collaborative filtering methods consider all features when calculating user/item similarity and ignore much local information. In collaborative filtering, selecting neighbors and determining users' similarities are the most important parts. For the selection of better neighbors, this study proposes a novel biclustering method based on modified fuzzy adaptive resonance theory. To reflect the similarity between users, a new measure that considers the effect of the number of users' common items is proposed. Specifically, the proposed novel biclustering method is first adopted to obtain local similarity and local prediction. Second, item-based collaborative filtering is used to generate global predictions. Finally, the two resultant predictions are fused to obtain a final one. Experiment results demonstrate that the proposed method outperforms state-of-the-art models in terms of several aspects on three benchmark datasets.

Recommender System Based on Collaborative Filtering for Personalized Dietary Advice: A Cross-Sectional Analysis of the ELSA-Brasil Study.

This study aimed to predict dietary recommendations and compare the performance of algorithms based on collaborative filtering for making predictions of personalized dietary recommendations. We analyzed the baseline cross-sectional data (2008-2010) of 12,667 participants of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). The participants were public employees of teaching and research institutions, aged 35-74 years, and 59% female. A semiquantitative Food Frequency Questionnaire (FFQ) was used for dietary assessment. The predictions of dietary recommendations were based on two machine learning (ML) algorithms-user-based collaborative filtering (UBCF) and item-based collaborative filtering (IBCF). The ML algorithms had similar precision (88-91%). The error metrics were lower for UBCF than for IBCF: with a root mean square error (RMSE) of 1.49 vs. 1.67 and a mean square error (MSE) of 2.21 vs. 2.78. Although all food groups were used as input in the system, the items eligible as recommendations included whole cereals, tubers and roots, beans and other legumes, oilseeds, fruits, vegetables, white meats and fish, and low-fat dairy products and milk. The algorithms' performances were similar in making predictions for dietary recommendations. The models presented can provide support for health professionals in interventions that promote healthier habits and improve adherence to this personalized dietary advice.

Recommender System with User-Based and Item-Based Collaborative Filtering on Twitter using K-Nearest Neighbors Classification

Netflix is one of the most widely used applications for watching movies online. There are various movie titles that can be watched by users, so a recommendation system is needed to help users who feel confused in choosing movie titles. Twitter is a social media used to express ideas, thoughts, and feelings. Not a few Twitter users who conduct movie discussions, with the movie discussion can be converted into a rating that can be used in the recommendation system. Collaborative Filtering is one of the methods of the recommendation system, by recommending based on the similarity between users (user-based) and based on items that have similarities with user-selected items (item-based). In this research, the Collaborative Filtering method is combined with K-Nearest Neighbors classification which obtains an RMSE value for user-based 1.8244 and item-based 0.5449. K-Nearest Neighbors gets 91.22% precision and 91.07% recall for user-based, while item-based gets 89.44% precision and 91.22% recall with the optimal K as a parameter is 3.

Item Selection With Collaborative Filtering in On-The-Fly Multistage Adaptive Testing

An important design feature in the implementation of both computerized adaptive testing and multistage adaptive testing is the use of an appropriate method for item selection. The item selection method is expected to select the most optimal items depending on the examinees’ ability level while considering other design features (e.g., item exposure and item bank utilization). This study introduced collaborative filtering (CF) as a new method for item selection in the on-the-fly assembled multistage adaptive testing framework. The user-based CF (UBCF) and item-based CF (IBCF) methods were compared to the maximum Fisher information method based on the accuracy of ability estimation, item exposure rates, and item bank utilization under different test conditions (e.g., item bank size, test length, and the sparseness of training data). The simulation results indicated that the UBCF method outperformed the traditional item selection methods regarding measurement accuracy. Also, the IBCF method showed the most superior performance in terms of item bank utilization. Limitations of the current study and the directions for future research are discussed.

Exploiting Long- and Short-Term Preferences for Deep Context-Aware Recommendations

With the tremendous growth in online information, capturing dynamic users’ preferences based on their historical interactions and providing a few desirable items to users have become an urgent service for all businesses. Recurrent neural networks (RNNs) and item-based collaborative filtering (CF) models are commonly used in industries due to their simplicity and efficiency. However, they fail to different contexts that could differently influence current users’ decision-making. Also, they are not sufficient to capture multiple users’ interests based on features of the interacting items. Besides, they have a limited modeling capability for the evolution of diversity and dynamic user preferences. In this article, we exploit long- and short-term preferences for deep context-aware recommendations (LSCAR) to enhance the next item recommendation’s performance by introducing three novel components as follows: 1) the user–contextual interaction module is proposed to capture and differentiate the interaction between contexts and users; 2) the encoded multi-interest module is introduced to capture various types of user interests; and 3) the integrator fusion gate module is used to effectively fuse the related long-term interests to the current short-term part, and the module returns the final user interest representation. Extensive experiments and results for two public datasets demonstrate that the proposed LSCAR outperforms the state-of-the-art models in almost all metrics and could provide interpretable recommendation results.

Water Ecotourism Route Recommendation Model Based on an Improved Cockroach Optimization Algorithm

Aiming to address the problems of the current research on water ecotourism routes, a water ecotourism route recommendation model based on an improved cockroach optimization algorithm is proposed. The aim is to recommend the tour routes with the lowest exhaust emissions. Firstly, depending on tourists’ once-visited water scenic spots, a scenic spot recommendation model based on the improved item-based collaborative filtering algorithm is set up. Then, by combining the recommended scenic spots and integrating the random transportation modes selected by tourists, a tour route recommendation model based on an improved cockroach optimization algorithm is constructed, which can output the tour route that produces the lowest exhaust emissions. Finally, The sample experiment shows that, on the basis of combining with the multivariate random transportation modes, the proposed algorithm has greater advantages than the tour routes planned by the traditional electronic maps, as it can output the tour routes with the lowest exhaust emissions, reduce the damage exhaust emissions cause in the urban water environments and to water resources, and effectively protect the urban water ecological environments.