Search results

In this “Info-plosion” era, recommendation systems (or recommenders) play a significant role in finding interesting items in the surge of online digital activity and e-commerce. The purpose of this paper is to model users' preference evolution to recommend potential items which users may be interested in.

A novel recommendation system, namely evolution-learning recommendation (ELR), is developed to precisely predict user interest for making recommendations. Differing from prior related methods, the authors integrate the matrix factorization (MF) and recurrent neural network (RNN) to effectively describe the variation of user preferences over time.

A novel cumulative factorization technique is proposed to efficiently decompose a rating matrix for discovering latent user preferences. Compared to traditional MF-based methods, the cumulative MF could reduce the utilization of computation resources. Furthermore, the authors depict the significance of long- and short-term effects in the memory cell of RNN for evolution patterns. With the context awareness, a learning model, V-LSTM, is developed to dynamically capture the evolution pattern of user interests. By using a well-trained learning model, the authors predict future user preferences and recommend related items.

Based on the relations among users and items for recommendation, the authors introduce a novel concept, virtual communication, to effectively learn and estimate the correlation among users and items. By incorporating the discovered latent features of users and items in an evolved manner, the proposed ELR model could promote “right” things to “right” users at the “right” time. In addition, several extensive experiments are performed on real datasets and are discussed. Empirical results show that ELR significantly outperforms the prior recommendation models. The proposed ELR exhibits great generalization and robustness in real datasets, including e-commerce, industrial retail and streaming service, with all discussed metrics.

The purpose of this paper is to explore the potential of enriching the library subject headings with folksonomy for enhancing the visibility and usability of the library subject headings.

The WorldCat-million data set and SocialBM0311 are preprocessing and over 210,000 library catalog records and 124,482 non-repeating tags were adopted to construct the matrix to observe the semantic relation between library subject headings and folksonomy. The proposed system is compared with the state-of-the-art methods and the parameters are fixed to obtain effective performance.

The results demonstrate that by integrating different semantic relations from library subject headings and folksonomy, the system’s performance can be improved compared to the benchmark methods. The evaluation results also show that the folksonomy can enrich library subject headings through the semantic relationship.

The proposed method simultaneous weighted matrix factorization can integrate the semantic relation from the library subject headings and folksonomy into one semantic space. The observation of the semantic relation between library subject headings and social tags from folksonomy can help enriching the library subject headings and improving the visibility of the library subject headings.

The purpose of this work is that of showing some efficient techniques to perform PV-PQ node type switching in multiple power flow computations.

Reactive generation limits of generation buses must be taken into account to obtain realistic power flow solutions. This may result computationally demanding when many power flow computations are required as in contingency screening or Monte Carlo simulations. In the present paper, the implementation of efficient PV-PQ node type switching is examined with particular emphasis on the efficiency of computation. Some different methods are proposed and compared on the basis of computation speed and accuracy.

Tests show the efficiency of the proposed methods with reference to actual networks with up to 800 buses.

The classical method of (partial) re-factorisation is not very efficient when many power flow solutions are to be evaluated. In the present work, a different approach is proposed; it is based on grounding each PV node by a fictitious short-circuit branch which is removed when the node type is changed to PQ. This operation is carried out by compensation of the solution and combined with the modifications required for contingency simulation.

The space discretization of eddy‐current problems in the magnetic vector potential formulation leads to a system of differential‐algebraic equations. They are typically time discretized by an implicit method. This requires the solution of large linear systems in the Newton iterations. The authors seek to speed up this procedure. In most relevant applications, several materials are non‐conducting and behave linearly, e.g. air and insulation materials. The corresponding matrix system parts remain constant but are repeatedly solved during Newton iterations and time‐stepping routines. The paper aims to exploit invariant matrix parts to accelerate the system solution.

Following the principle “reduce, reuse, recycle”, the paper proposes a Schur complement method to precompute a factorization of the linear parts. In 3D models this decomposition requires a regularization in non‐conductive regions. Therefore, the grad‐div regularization is revisited and tailored such that it takes anisotropies into account.

The reduced problem exhibits a decreased effective condition number. Thus, fewer preconditioned conjugate gradient iterations are necessary. Numerical examples show a decrease of the overall simulation time, if the step size is small enough. 3D simulations with large time step sizes might not benefit from this approach, because the better condition does not compensate for the computational costs of the direct solvers used for the Schur complement. The combination of the Schur approach with other more sophisticated preconditioners or multigrid solvers is subject to current research.

The Schur complement method is adapted for the eddy‐current problem. Therefore, a new partitioning approach into linear/non‐linear and static/dynamic domains is proposed. Furthermore, a new variant of the grad‐div gauging is introduced that allows for anisotropies and enables the Schur complement method in 3D.

The purpose of this study is to solve the cold start problem caused by the lack of evaluation information about the products.

A recommendation system has been developed by using the image data of the clothing products, assuming that the user considers the visual characteristics importantly when purchasing fashion products. In order to evaluate the performance of the model developed in this study, it was compared with Random, Itempop, Matrix Factorization and Generalized Matrix Factorization models.

The newly developed model was able to cope with the cold start problem better than other models.

A hybrid recommendation system has been developed that combines the existing recommendation system with deep learning to effectively recommend fashion products considering the user's taste.

This is the first research to improve the performance of fashion recommendation system using the deep learning model trained by the images of fashion products.

The purpose of this paper is to apply link prediction to community mining and to clarify the role of link prediction in improving the performance of social network analysis.

In this study, the 2009 version of Enron e-mail data set provided by Carnegie Mellon University was selected as the research object first, and bibliometric analysis method and citation analysis method were adopted to compare the differences between various studies. Second, based on the impact of various interpersonal relationships, the link model was adopted to analyze the relationship among people. Finally, the factorization of the matrix was further adopted to obtain the characteristics of the research object, so as to predict the unknown relationship.

The experimental results show that the prediction results obtained by considering multiple relationships are more accurate than those obtained by considering only one relationship.

Due to the limited number of objects in the data set, the link prediction method has not been tested on the large-scale data set, and the validity and correctness of the method need to be further verified with larger data. In addition, the research on algorithm complexity and algorithm optimization, including the storage of sparse matrix, also need to be further studied. At the same time, in the case of extremely sparse data, the accuracy of the link prediction method will decline a lot, and further research and discussion should be carried out on the sparse data.

The focus of this research is on link prediction in social network analysis. The traditional prediction model is based on a certain relationship between the objects to predict and analyze, but in real life, the relationship between people is diverse, and different relationships are interactive. Therefore, in this study, the graph model is used to express different kinds of relations, and the influence between different kinds of relations is considered in the actual prediction process. Finally, experiments on real data sets prove the effectiveness and accuracy of this method. In addition, link prediction, as an important part of social network analysis, is also of great significance for other applications of social network analysis. This study attempts to prove that link prediction is helpful to the improvement of performance analysis of social network by applying link prediction to community mining.

This study adopts a variety of methods, such as link prediction, data mining, literature analysis and citation analysis. The research direction is relatively new, and the experimental results obtained have a certain degree of credibility, which is of certain reference value for the following related research.

The explosion of data due to the sophistication of information and communication technology makes it simple for prospective tourists to learn about previous hotel guests' experiences. They prioritize the rating score when selecting a hotel. However, rating scores are less reliable for suggesting a personalized preference for each aspect, especially when they are in a limited number. This study aims to recommend ratings and personalized preference hotels using cross-domain and aspect-based features.

We propose an aspect-based cross-domain personalized recommendation (AsCDPR), a novel framework for rating prediction and personalized customer preference recommendations. We incorporate a cross-domain personalized approach and aspect-based features of items from the review text. We extracted aspect-based feature vectors from two domains using bidirectional long short-term memory and then mapped them by a multilayer perceptron (MLP). The cross-domain recommendation module trains MLP to analyze sentiment and predict item ratings and the polarities of the aspect based on user preferences.

Expanded by its synonyms, aspect-based features significantly improve the performance of sentiment analysis on accuracy and the F1-score matrix. With relatively low mean absolute error and root mean square error values, AsCDPR outperforms matrix factorization, collaborative matrix factorization, EMCDPR and Personalized transfer of user preferences for cross-domain recommendation. These values are 1.3657 and 1.6682, respectively.

This study assists users in recommending hotels based on their priority preferences. Users do not need to read other people's reviews to capture the key aspects of items. This model could enhance system reliability in the hospitality industry by providing personalized recommendations.

This study introduces a new approach that embeds aspect-based features of items in a cross-domain personalized recommendation. AsCDPR predicts ratings and provides recommendations based on priority aspects of each user's preferences.

The aim of this paper is to present two independent ways in which a simple approximation to a Green's function for a differential equation can be used to improve the performance of well‐known iterative methods for linear equations.

The purpose of this paper is to reduce issues arising when computing steady‐state solutions for AC machine models using the harmonic balance method.

Generally, currents at steady‐states of AC machines are described by periodic or quasi‐periodic time functions, which Fourier spectra are determined by an infinite set of algebraic equations obtained from a harmonic balance method. To solve them, after reducing to finite dimensions, an iterative algorithm is developed in this paper. It bases on the LU decomposition of an infinite matrix representing the inductance matrix of an AC machine. Since that decomposition is done separately, due to a band type form of this matrix, the equation set determining the Fourier spectra of currents is solved recurrently.

An algorithm for the LU decomposition of an infinite matrix representing the inductance matrix of an AC machine and an iterative algorithm for determining AC machine steady‐state currents in a recursive manner.

The approach is limited to solving of so‐called “circuital” models of AC voltage supplied machines. The approach breaks the large dimension barrier when solving steady‐state equations for AC machines.

Reducing computer requirements in terms of computer memory, workload and computing time to determine a steady‐state solution for AC machines.

A separation of the LU decomposition of an infinite matrix representing the inductance matrix in AC machine steady‐state model from the solution method.

Identifying the frontiers of a specific research field is one of the most basic tasks in bibliometrics and research published in leading conferences is crucial to the data mining research community, whereas few research studies have focused on it. The purpose of this study is to detect the intellectual structure of data mining based on conference papers.

This study takes the authoritative conference papers of the ranking 9 in the data mining field provided by Google Scholar Metrics as a sample. According to paper amount, this paper first detects the annual situation of the published documents and the distribution of the published conferences. Furthermore, from the research perspective of keywords, CiteSpace was used to dig into the conference papers to identify the frontiers of data mining, which focus on keywords term frequency, keywords betweenness centrality, keywords clustering and burst keywords.

Research showed that the research heat of data mining had experienced a linear upward trend during 2007 and 2016. The frontier identification based on the conference papers showed that there were five research hotspots in data mining, including clustering, classification, recommendation, social network analysis and community detection. The research contents embodied in the conference papers were also very rich.

This study detected the research frontier from leading data mining conference papers. Based on the keyword co-occurrence network, from four dimensions of keyword term frequency, betweeness centrality, clustering analysis and burst analysis, this paper identified and analyzed the research frontiers of data mining discipline from 2007 to 2016.