Cold Start Issues Research Articles

A hybrid personalized scholarly venue recommender system integrating social network analysis and contextual similarity

Rapidly developing academic venues throw a challenge to researchers in identifying the most appropriate ones that are in-line with their scholarly interests and of high relevance. Even a high-quality paper is sometimes rejected due to a mismatch between the area of the paper, and the scope of the journal attempted to. Recommending appropriate academic venues can, therefore, enable researchers to identify and take part in relevant conferences and to publish in impactful journals. Although a researcher may know a few leading high-profile venues for her specific field of interest, a venue recommender system becomes particularly helpful when one explores a new field or when more options are needed. We propose DISCOVER: A Diversified yet Integrated Social network analysis and COntextual similarity-based scholarly VEnue Recommender system. Our work provides an integrated framework incorporating social network analysis, including centrality measure calculation, citation and co-citation analysis, topic modeling based contextual similarity, and key-route identification based main path analysis of a bibliographic citation network. The paper also addresses cold start issues for a new researcher and a new venue along with a considerable reduction in data sparsity, computational costs, diversity, and stability problems. Experiments based on the Microsoft Academic Graph (MAG) dataset show that the proposed DISCOVER outperforms state-of-the-art recommendation techniques using standard metrics of precision@k, nDCG@k, accuracy, MRR, F−measuremacro, diversity, stability, and average venue quality.

CNAVER: A Content and Network-based Academic VEnue Recommender system

The phenomenon of rapidly developing academic venues poses a significant challenge for researchers: how to recognize the ones that are not only in accordance with one’s scholarly interests but also of high significance? Often, even a high-quality paper is rejected because of a mismatch between the research area of the paper and the scope of the journal. Recommending appropriate scholarly venues to researchers empowers them to recognize and partake in important academic conferences and assists them in getting published in impactful journals. A venue recommendation system becomes helpful in this scenario, particularly when exploring a new field or when further choices are required. We propose CNAVER: A Content and Network-based Academic VEnue Recommender system. It provides an integrated framework employing a rank-based fusion of paper-paper peer network (PPPN) model and venue-venue peer network (VVPN) model. It only requires the title and abstract of a paper to provide venue recommendations, thus assisting researchers even at the earliest stage of paper writing. It also addresses cold start issues such as the involvement of an inexperienced researcher and a novel venue along with the problems of data sparsity, diversity, and stability. Experiments on the DBLP dataset exhibit that our proposed approach outperforms several state-of-the-art methods in terms of precision, nDCG, MRR, accuracy, F−measuremacro, average venue quality, diversity, and stability.

Deeply Fusing Reviews and Contents for Cold Start Users in Cross-Domain Recommendation Systems

As one promising way to solve the challenging issues of data sparsity and cold start in recommender systems, crossdomain recommendation has gained increasing research interest recently. Cross-domain recommendation aims to improve the recommendation performance by means of transferring explicit or implicit feedback from the auxiliary domain to the target domain. Although the side information of review texts and item contents has been proven to be useful in recommendation, most existing works only use one kind of side information and cannot deeply fuse this side information with ratings. In this paper, we propose a Review and Content based Deep Fusion Model named RC-DFM for crossdomain recommendation. We first extend Stacked Denoising Autoencoders (SDAE) to effectively fuse review texts and item contents with the rating matrix in both auxiliary and target domains. Through this way, the learned latent factors of users and items in both domains preserve more semantic information for recommendation. Then we utilize a multi-layer perceptron to transfer user latent factors between the two domains to address the data sparsity and cold start issues. Experimental results on real datasets demonstrate the superior performance of RC-DFM compared with state-of-the-art recommendation methods.Deeply Fusing Reviews and Contents for Cold Start Users in Cross-Domain Recommendation Systems

Optical diagnostics of low-temperature ignition and combustion characteristics of diesel/kerosene blends under cold-start conditions

Diesel engine plays a vital role in various fields, but it does suffer from the cold-start issue when operating in low-temperature environment. As an alternative fuel or emergency back-up fuel for diesel engine, kerosene was investigated to reveal the influences of its difference on physical and chemical properties from diesel on low-temperature ignition characteristics. In this work, the spray and ignition characteristics of diesel/kerosene blends were investigated at ambient temperatures of 653–733 K and ambient densities of 10–22 kg/m3. Experiments were conducted in an optical constant volume combustion chamber using Mie-scattering and direct photography methods. The results show that under cold-start conditions, although kerosene has lower cetane number than diesel, its strong volatility significantly promotes the atomization and evaporation of the blends. The addition of 15% volumetric kerosene to diesel effectively shortens the ignition delay. In addition, the low-temperature ignition starts with a blue flame and further formaldehyde signal identification indicates that the blue flame is a cool flame. Kerosene addition enhances the cool flame intensity, which is conducive to the transformation to a high-temperature hot flame. Thus, adding a small proportion of kerosene to the diesel fuel contributes to a rapid cold-start at low temperatures.

Privacy-preserving Cross-domain Location Recommendation

Cross-domain recommendation is a typical solution for data sparsity and cold start issue in the field of location recommendation. Specifically, data of an auxiliary domain is leveraged to improve the recommendation of the target domain. There is a typical scenario that two interaction domains (location based check-in service, for example) combine data to perform the cross-domain location recommendation task. Existing approaches are based on the assumption that the interaction data from the auxiliary domain can be directly shared across domains. However, such an assumption is not reasonable, since in the real world those domains may be operated by different companies. Therefore, directly sharing raw data may violate business privacy policy and increase the risk of privacy leakage since the user-location interaction records are very sensitive. In this paper, we propose a framework named privacy-preserving cross-domain location recommendation which works in two stages. First, for the interaction data from the auxiliary domain, we adopt a differential privacy based protection mechanism to hide the real locations of each user to meet the criterion of differential privacy. Then we share the protected user-location interaction to the target domain. Second, we develop a new method of Confidence-aware Collective Matrix Factorization (CCMF) to effectively exploit the transferred interaction data. To verify its efficacy, we collect two real-world datasets suitable for the task. Extensive experiments demonstrate that our proposed framework achieves the best performance compared with the state-of-the-art baseline methods. We further demonstrate that our method can alleviate the data sparsity issue significantly while protecting users' location privacy.

Spatiotemporal Representation Learning for Translation-Based POI Recommendation

The increasing proliferation of location-based social networks brings about a huge volume of user check-in data, which facilitates the recommendation of points of interest (POIs). Time and location are the two most important contextual factors in the user’s decision-making for choosing a POI to visit. In this article, we focus on the spatiotemporal context-aware POI recommendation, which considers the joint effect of time and location for POI recommendation. Inspired by the recent advances in knowledge graph embedding, we propose a spatiotemporal context-aware and translation-based recommender framework (STA) to model the third-order relationship among users, POIs, and spatiotemporal contexts for large-scale POI recommendation. Specifically, we embed both users and POIs into a “transition space” where spatiotemporal contexts (i.e., a < time, location > pair) are modeled as translation vectors operating on users and POIs. We further develop a series of strategies to exploit various correlation information to address the data sparsity and cold-start issues for new spatiotemporal contexts, new users, and new POIs. We conduct extensive experiments on two real-world datasets. The experimental results demonstrate that our STA framework achieves the superior performance in terms of high recommendation accuracy, robustness to data sparsity, and effectiveness in handling the cold-start problem.

Differentiating Regularization Weights -- A Simple Mechanism to Alleviate Cold Start in Recommender Systems

Matrix factorization (MF) and its extended methodologies have been studied extensively in the community of recommender systems in the last decade. Essentially, MF attempts to search for low-ranked matrices that can (1) best approximate the known rating scores, and (2) maintain low Frobenius norm for the low-ranked matrices to prevent overfitting. Since the two objectives conflict with each other, the common practice is to assign the relative importance weights as the hyper-parameters to these objectives. The two low-ranked matrices returned by MF are often interpreted as the latent factors of a user and the latent factors of an item that would affect the rating of the user on the item. As a result, it is typical that, in the loss function, we assign a regularization weight λ p on the norms of the latent factors for all users, and another regularization weight λ q on the norms of the latent factors for all the items. We argue that such a methodology probably over-simplifies the scenario. Alternatively, we probably should assign lower constraints to the latent factors associated with the items or users that reveal more information, and set higher constraints to the others. In this article, we systematically study this topic. We found that such a simple technique can improve the prediction results of the MF-based approaches based on several public datasets. Specifically, we applied the proposed methodology on three baseline models -- SVD, SVD++, and the NMF models. We found that this technique improves the prediction accuracy for all these baseline models. Perhaps more importantly, this technique better predicts the ratings on the long-tail items, i.e., the items that were rated/viewed/purchased by few users. This suggests that this approach may partially remedy the cold-start issue. The proposed method is very general and can be easily applied on various recommendation models, such as Factorization Machines, Field-aware Factorization Machines, Factorizing Personalized Markov Chains, Prod2Vec, Behavior2Vec, and so on. We release the code for reproducibility. We implemented a Python package that integrates the proposed regularization technique with the SVD, SVD++, and the NMF model. The package can be accessed at https://github.com/ncu-dart/rdf.

A Survey of User Profiling: State-of-the-Art, Challenges, and Solutions

Advancements in information and communication technology, and online web users have given attention to the virtual representation of each user, which is crucial for effective service personalization. Meeting users need and preferences is an ongoing challenge in service personalization. This issue can be addressed through the building of a comprehensive user profile. A user profile is the summary of the user's interests, characteristics, behaviours, and preferences, while user profiling is the system of collecting, organizing and inferring the user profile information. Many reviews on user profiling have been conducted but none focused on the effective profile modeling process. Hence, this article aims to provide a review of the recent state-of-the-art approach to user profiling. These include methods, description, characteristics, and taxonomy of the user profile. The study of the existing user profiling modeling in the aspect of data acquisition, feature extraction, profiling techniques, and profiling approaches (with the identification of their strengths and weaknesses) and the performance measures are also provided. In addition, the research challenges were also discussed with a focus on privacy, datasets, cold start issues, trust issues, and computational complexity. Moreover, the article identified an open research direction that serves as solutions to the identified challenges and motivation for further researchers in advancing user profiling. The findings showed that an effective modeling process enhances the construction of accurate user profile for service personalization.

A scalable and robust trust-based nonnegative matrix factorization recommender using the alternating direction method

Matrix Factorization (MF) has been proven to be an effective approach for the generation of a successful recommender system. However, most current MF-based recommenders cannot obtain high prediction accuracy due to the sparseness of the user–item matrix in collaborative filtering models. Moreover, they suffer from scalability issues when applied to large-scale real-world tasks. To tackle these issues, a social regularization method, called TrustANLF, is proposed, which incorporates users’ social trust information in a nonnegative matrix factorization framework. The proposed method integrates trust statements as an additional information source along with rating values into the recommendation model to deal with the data sparsity and cold-start issues. Moreover, the alternating direction optimization method is used for solving the trust-based nonnegative MF model in order to improve convergence speed as well as reduce computational and memory costs. To evaluate the effectiveness of the proposed method, several experiments are performed on three real-world datasets. The obtained results demonstrate the significant improvements of the proposed method over several state-of-the-art methods.

An explanatory item response theory method for alleviating the cold-start problem in adaptive learning environments.

Electronic learning systems have received increasing attention because they are easily accessible to many students and are capable of personalizing the learning environment in response to students' learning needs. To that end, using fast and flexible algorithms that keep track of the students' ability change in real time is desirable. Recently, the Elo rating system (ERS) has been applied and studied in both research and practical settings (Brinkhuis & Maris, 2009; Klinkenberg, Straatemeier, & van der Maas in Computers & Education, 57, 1813-1824, 2011). However, such adaptive algorithms face the cold-start problem, defined as the problem that the system does not know a new student's ability level at the beginning of the learning stage. The cold-start problem may also occur when a student leaves the e-learning system for a while and returns (i.e., a between-session period). Because external effects could influence the student's ability level during the period, there is again much uncertainty about ability level. To address these practical concerns, in this study we propose alternative approaches to cold-start issues in the context of the e-learning environment. Particularly, we propose making the ERS more efficient by using an explanatory item response theory modeling to estimate students' ability levels on the basis of their background information and past trajectories of learning. A simulation study was conducted under various conditions, and the results showed that the proposed approach substantially reduces ability estimation errors. We illustrate the approach using real data from a popular learning platform.

The media-oriented cross domain recommendation method

With the rapid development of modern high-tech, such as big data and artificial intelligence, the demand for cross-media services is also greatly improved. Since different modes of cross-media data apply different dimensions and different attributes of the underlying features to present data, many tasks need to work collaboratively to handle multiform of information (including text, audio, video, image, etc.), so as to build cross-media analysis and reasoning. Through the cross-media, the method could express the same semantic information from their own side and could reflect the specific information more fully than a single media object and its specific modal. The same information is cross spread and integrated across different kinds of media objects. Only conducting fusion analysis to these multi-modal media, can ones fully and correctly understand the content information contained in the cross-media complex, which also adds difficulty in cross-media information recommendation process. At the same time, data sparsity, cold start and scalability issues of traditional recommendation system have long been unsolved, and as such it cannot adapt to the personalized service needs in cross-media applications. Focusing on the field of media information recommendation and taking the media data, user behavior data and project attribute information as the information source, this paper aims at researching the cross-domain recommendation algorithm. With the help of the label data, matrix decomposition, the author constructs a media-oriented cross-domain recommendation model in order to improve the recommendation accuracy and solve the data sparsity, cold start problems of media information recommendation technology, exploring a high-accurate media-oriented cross-domain recommendation method.

Exploiting Past Users’ Interests and Predictions in an Active Learning Method for Dealing with Cold Start in Recommender Systems

This paper focuses on the new users cold-start issue in the context of recommender systems. New users who do not receive pertinent recommendations may abandon the system. In order to cope with this issue, we use active learning techniques. These methods engage the new users to interact with the system by presenting them with a questionnaire that aims to understand their preferences to the related items. In this paper, we propose an active learning technique that exploits past users’ interests and past users’ predictions in order to identify the best questions to ask. Our technique achieves a better performance in terms of precision (RMSE), which leads to learn the users’ preferences in less questions. The experimentations were carried out in a small and public dataset to prove the applicability for handling cold start issues.

Tensor factorization method based on review text semantic similarity for rating prediction

Recommendation methods have been proved to be successful in eliminating the information overload problem. However, they are still insufficient as far as data sparsity and cold start issues are concerned. Some approaches have attempted to resolve these issues by utilizing relevant information contained in user review text, although, the type of information extracted from the review text and the way such recommendation methods utilize the information affect recommendation accuracy of the models. In this paper, we address such challenges by considering linguistic similarity between review texts and employ it as additional factor in rating prediction, and we propose a recommendation method based on tensor factorization which involves review text semantic similarity. The proposed tensor factorization model supplements the central task of factorization methods of finding similar users, uncovering underlying characteristics of the data and predicting user preferences by introducing text semantic similarity. The proposed method is carried out in two main phases; first phase, by computing semantic similarity between review texts and assigning a similarity score, and second phase by introducing the similarity score as an additional factor in the probabilistic matrix factorization (PMF) model. To evaluate the performance of the proposed approach, several Amazon datasets were experimented and the results verify that the semantic similarity of review texts not only successfully portray user preferences, and extend PMF to include review texts similarities but also increases prediction influence which results in improved performance.

Improving Existing Collaborative Filtering Recommendations via Serendipity-Based Algorithm

In this paper, we study how to address the sparsity, accuracy and serendipity issues of top-N recommendation with collaborative filtering (CF). Existing studies commonly use rated items (which form only a small section in a rating matrix) or import some additional information (e.g., details about the items and users) to improve the performance of CF. Unlike these methods, we propose a novel notion towards a huge amount of unrated items: serendipity item. By utilizing serendipity items, we propose concise satisfaction and interest injection (CSII), a method that can effectively find interesting, satisfying, and serendipitous items in unrated items. By preventing uninteresting and unsatisfying items to be recommended as top-N items, this concise-but-novel method improves accuracy and recommendation quality (especially serendipity) substantially. Meanwhile, it can address the sparsity and cold-start issues by enriching the rating matrix in CF without additional information. As our method tackles rating matrix before recommendation procedure, it can be applied to most existing CF methods, such as item-based CF, user-based CF and matrix factorization-based CF. Through comprehensive experiments using abundant real-world datasets with LensKit implementation, we successfully demonstrate that our solution improves the performance of existing CF methods consistently and universally. Moreover, comparing with baseline methods, CSII can extract uninteresting items more carefully and cautiously, avoiding potential items inferred by mistake.

Collaborative Filtering Recommender System: Comparative Survey on Cold-Start Issue

Objectives: To analyze the issue of cold-start (user cold-start and item cold-start) in Collaborative Filtering Recommender System (CFRS) and to compare its solution with various approaches are summarized in this paper. Methods/Statistical Analysis: The manuscript discussed about the cold-start issue in which the recommender system cannot recommend items to the new user because no ratings made by the new user (user cold-start) as well as for the newly added items, the system cannot be able to provide recommendations to the user because the system has no ratings for the newly added item (item cold-start). The solutions for cold-start issue are analyzed based on the model based approach, demographic data, ask-to-rate technique, and Social Network Analysis (SNA). Findings: The comparative review of the aforementioned approaches provides the detail about how to implement the model based approach, how to collect the demographic data from the new user, how to apply the ask-to-rate technique and how to make use of the SNA concept to solve the cold-start issue in CF recommender system. Application: The recommender system on Amazon helps the user to purchase books, Compact Disks (CDs), Netflix helps the user to choose CDs to purchase/rent and Epinions, helps the users to decide to purchase based on user reviews. Keywords: Ask-To-Rate Technique, Cold-Start, Collaborative Filtering Recommender System, Demographic Data, Model Based Approach, Social Network Analysis

Collaborative Filtering Recommender System: Comparative Survey on Cold-Start Issue

Collaborative Filtering Recommender System: Comparative Survey on Cold-Start Issue

A Multi-Modality Deep Network for Cold-Start Recommendation

Collaborative filtering (CF) approaches, which provide recommendations based on ratings or purchase history, perform well for users and items with sufficient interactions. However, CF approaches suffer from the cold-start problem for users and items with few ratings. Hybrid recommender systems that combine collaborative filtering and content-based approaches have been proved as an effective way to alleviate the cold-start issue. Integrating contents from multiple heterogeneous data sources such as reviews and product images is challenging for two reasons. Firstly, mapping contents in different modalities from the original feature space to a joint lower-dimensional space is difficult since they have intrinsically different characteristics and statistical properties, such as sparse texts and dense images. Secondly, most algorithms only use content features as the prior knowledge to improve the estimation of user and item profiles but the ratings do not directly provide feedback to guide feature extraction. To tackle these challenges, we propose a tightly-coupled deep network model for fusing heterogeneous modalities, to avoid tedious feature extraction in specific domains, and to enable two-way information propagation from both content and rating information. Experiments on large-scale Amazon product data in book and movie domains demonstrate the effectiveness of the proposed model for cold-start recommendation.

Real-time social recommendation based on graph embedding and temporal context

With the rapid proliferation of online social networks, personalized social recommendation has become an important means to help people discover their potential friends or interested items in real-time. However, the cold-start issue and the special properties of social networks, such as rich temporal dynamics, heterogeneous and complex structures, render the most commonly used recommendation approaches (e.g. Collaborative Filtering) inefficient. In this paper, we propose a novel dynamic graph-based embedding (DGE) model for social recommendation which is capable of recommending relevant users and interested items. In order to support real-time recommendation, we construct a heterogeneous user-item (HUI) network and incrementally maintain it as the social network evolves. DGE jointly captures the temporal semantic effects, social relationships and user behavior sequential patterns in a unified way by embedding the HUI network into a shared low dimensional space. Then, with simple search methods or similarity calculations, we can use the encoded representation of temporal contexts to generate recommendations. We conduct extensive experiments to evaluate the performance of our model on two real large-scale datasets, and the experimental results show its advantages over other state-of-the-art methods.

Knowledge Driven Paper Recommendation Using Heterogeneous Network Embedding Method

We search a variety of things over the Internet in our daily lives, and numerous search engines are available to get us more relevant results. With the rapid technological advancement, the internet has become a major source of obtaining information. Further, the advent of the Web2.0 era has led to an increased interaction between the user and the website. It has become challenging to provide information to users as per their interests. Because of copyright restrictions, most of existing research studies are confronting the lack of availability of the content of candidates recommending articles. The content of such articles is not always available freely and hence leads to inadequate recommendation results. Moreover, various research studies base recommendation on user profiles. Therefore, their recommendation needs a significant number of registered users in the system. In recent years, research work proves that Knowledge graphs have yielded better in generating quality recommendation results and alleviating sparsity and cold start issues. Network embedding techniques try to learn high quality feature vectors automatically from network structures, enabling vector-based measurers of node relatedness. Keeping the strength of Network embedding techniques, the proposed citation-based recommendation approach makes use of heterogeneous network embedding in generating recommendation results. The novelty of this paper is in exploiting the performance of a network embedding approach i.e., matapath2vec to generate paper recommendations. Unlike existing approaches, the proposed method has the capability of learning low-dimensional latent representation of nodes (i.e., research papers) in a network. We apply metapath2vec on a knowledge network built by the ACL Anthology Network (all about NLP) and use the node relatedness to generate item (research article) recommendations.

Probabilistic Matrix Factorization Recommendation Algorithm with User Trust Similarity

In this paper, we describe the formatting guidelines for Conference Proceedings. Whether the user similarity calculation is reasonable in the traditional collaborative filtering recommendation algorithm directly affects the result of the collaborative filtering recommendation algorithm. This paper proposes a probabilistic matrix factorization recommendation algorithm with user trust similarity which combines improved similarity of users’ trust and probability matrix factorization recommendation method. The results show that proposed algorithm could relieve user cold start issues and effectively reduce the error of recommendation.