Collaborative user modeling with user-generated tags for social recommender systems

With the popularity of social media services, the sheer amount of content is increasing exponentially on the Social Web that leads to attract considerable attention to recommender systems. Recommender systems provide users with recommendations of items suited to their needs. To provide proper recommendations to users, recommender systems require an accurate user model that can reflect a user’s characteristics, preferences and needs. In this study, by leveraging user-generated tags as preference indicators, we propose a new collaborative approach to user modeling that can be exploited to recommender systems. Our approach first discovers relevant and irrelevant topics for users, and then enriches an individual user model with collaboration from other similar users. In order to evaluate the performance of our model, we compare experimental results with a user model based on collaborative filtering approaches and a vector space model. The experimental results have shown the proposed model provides a better representation in user interests and achieves better recommendation results in terms of accuracy and ranking.     

A hybrid collaborative filtering method for multiple-interests and multiple-content recommendation in E-Commerce

Recommender systems apply knowledge discovery techniques to the problem of making personalized recommendations for products or services during a live interaction. These systems, especially collaborative filtering based on user, are achieving widespread success on the Web. The tremendous growth in the amount of available information and the kinds of commodity to Web sites in recent years poses some key challenges for recommender systems. One of these challenges is ability of recommender systems to be adaptive to environment where users have many completely different interests or items have completely different content (We called it as Multiple interests and Multiple-content problem). Unfortunately, the traditional collaborative filtering systems can not make accurate recommendation for the two cases because the predicted item for active user is not consist with the common interests of his neighbor users. To address this issue we have explored a hybrid collaborative filtering method, collaborative filtering based on item and user techniques, by combining collaborative filtering based on item and collaborative filtering based on user together. Collaborative filtering based on item and user analyze the user-item matrix to identify similarity of target item to other items, generate similar items of target item, and determine neighbor users of active user for target item according to similarity of other users to active user based on similar items of target item.In this paper we firstly analyze limitation of collaborative filtering based on user and collaborative filtering based on item algorithms respectively and emphatically make explain why collaborative filtering based on user is not adaptive to Multiple-interests and Multiple-content recommendation. Based on analysis, we present collaborative filtering based on item and user for Multiple-interests and Multiple-content recommendation. Finally, we experimentally evaluate the results and compare them with collaborative filtering based on user and collaborative filtering based on item, respectively. The experiments suggest that collaborative filtering based on item and user provide better recommendation quality than collaborative filtering based on user and collaborative filtering based on item dramatically.     

一种检测兴趣漂移的图结构推荐系统

协同过滤是构造推荐系统最有效的方法之一.其中,基于图结构推荐方法成为近来协同过滤的研究热点.基于图结构的方法视用户和项为图的结点,并利用图理论去计算用户和项之间的相似度.尽管人们对图结构推荐系统开展了很多的研究和应用,然而这些研究都认为用户的兴趣是保持不变的,所以不能够根据用户兴趣的相关变化做出合理推荐.本文提出一种新的可以检测用户兴趣漂移的图结构推荐系统.首先,设计了一个新的兴趣漂移检测方法,它可以有效地检测出用户兴趣在何时发生了哪种变化.其次,根据用户的兴趣序列,对评分项进行加权并构造用户特征向量.最后,整合二部投影与随机游走进行项推荐.在标准数据集MovieLens上的测试表明算法优于两个图结构推荐方法和一个评分时间加权的协同过滤方法.     

The state-of-the-art in personalized recommender systems for social networking

With the explosion of Web 2.0 application such as blogs, social and professional networks, and various other types of social media, the rich online information and various new sources of knowledge flood users and hence pose a great challenge in terms of information overload. It is critical to use intelligent agent software systems to assist users in finding the right information from an abundance of Web data. Recommender systems can help users deal with information overload problem efficiently by suggesting items (e.g., information and products) that match users’ personal interests. The recommender technology has been successfully employed in many applications such as recommending films, music, books, etc. The purpose of this report is to give an overview of existing technologies for building personalized recommender systems in social networking environment, to propose a research direction for addressing user profiling and cold start problems by exploiting user-generated content newly available in Web 2.0.     

An efficient hybrid similarity measure based on user interests for recommender systems

Recommender systems are used to suggest items to users based on their interests. They have been used widely in various domains, including online stores, web advertisements, and social networks. As part of their process, recommender systems use a set of similarity measurements that would assist in finding interesting items. Although many similarity measurements have been proposed in the literature, they have not concentrated on actual user interests. This paper proposes a new efficient hybrid similarity measure for recommender systems based on user interests. This similarity measure is a combination of two novel base similarity measurements: the user interest–user interest similarity measure and the user interest–item similarity measure. This hybrid similarity measure improves the existing work in three aspects. First, it improves the current recommender systems by using actual user interests. Second, it provides a comprehensive evaluation of an efficient solution to the cold start problem. Third, this similarity measure works well even when no corated items exist between two users. Our experiments show that our proposed similarity measure is efficient in terms of accuracy, execution time, and applicability. Specifically, our proposed similarity measure achieves a mean absolute error (MAE) as low as 0.42, with 64% applicability and an execution time as low as 0.03 s, whereas the existing similarity measures from the literature achieve an MAE of 0.88 at their best; these results demonstrate the superiority of our proposed similarity measure in terms of accuracy, as well as having a high applicability percentage and a very short execution time.     

基于用户行为及语义相关实时更新用户兴趣的推荐系统

在现有的推荐系统中,其用户兴趣模型都能够有效地表达出用户的兴趣,但在用户兴趣发生变化时却不能很好地调整用户兴趣模型,不能及时适应用户兴趣的动态变化。本文提出一种基于语义相关实时更新用户兴趣模型的推荐系统。该系统能够及时响应用户兴趣变化,从而改善了以往推荐系统对用户兴趣更新不及时所导致的推荐结果不够全面、准确的问题。实验表明该系统能够准确表达用户兴趣,特别是在用户兴趣发生变化时比以往系统具有更高的准确性。     

How scales influence user rating behaviour in recommender systems

Many websites allow users to rate items and share their ratings with others, for social or personalisation purposes. In recommender systems in particular, personalised suggestions are generated by predicting ratings for items that users are unaware of, based on the ratings users provided for other items. Explicit user ratings are collected by means of graphical widgets referred to as ‘rating scales’. Each system or website normally uses a specific rating scale, in many cases differing from scales used by other systems in their granularity, visual metaphor, numbering or availability of a neutral position. While many works in the field of survey design reported on the effects of rating scales on user ratings, these, however, are normally regarded as neutral tools when it comes to recommender systems. In this paper, we challenge this view and provide new empirical information about the impact of rating scales on user ratings, presenting the results of three new studies carried out in different domains. Based on these results, we demonstrate that a static mathematical mapping is not the best method to compare ratings coming from scales with different features, and suggest when it is possible to use linear functions instead.     

基于极端评分行为的相似度计算

随着互联网技术的迅猛发展,互联网信息急剧增长,信息过载问题愈发凸显。面对海量的互联网信息,用户往往需要耗费大量的时间来搜索所需的信息或产品,而搜索的解往往受到制约。为解决信息过载问题,推荐系统应运而生。推荐系统根据用户的历史行为推测其需求、兴趣等,将用户感兴趣的信息、产品等推荐给用户。作为推荐领域中一类重要的推荐方法,基于记忆的协同过滤方法通常依据用户或产品的近邻信息来构造评分预测函数,其核心在于准确度量用户或产品之间的相似度。传统的相似度量,如皮尔逊、余弦及秩相关系数等,通常只考虑了用户之间的线性关系;而启发式相似度如基于3个特殊因子的PIP相似度及其改进方法,则只刻画了用户之间的非线性关系。事实上,在推荐系统中,就用户之间的相似关系而言,仅用线性或是非线性函数来度量均是不准确的。为了更为精细地刻画用户之间的相似程度,文中提出了基于非线性函数的用户极端评分行为的相似程度度量指数,通过将该指数融入传统的线性相关系数,构造了一个考虑极端评分行为的新的相似度。为验证该方法的有效性,基于Ml(100k)和Ml-latest-small两个数据集,将其与传统相似度以及启发式相似度进行比较,结果显示基于极端评分行为相似度的协同过滤方法在MAE和RMSE指标上能够获得更好的表现。     

面向用户偏好建模的个性化对话推荐算法

推荐系统的目标是从物品数据库中，选择出与用户兴趣偏好相匹配的子集，缓解用户面临的“信息过载”问题。因而近年来推荐系统越来越多地应用到电商、社交等领域，展现出巨大的商业潜力。传统推荐系统中，系统对用户的认知往往来源于历史交互记录，例如点击率或者购买记录，这是一种隐式用户反馈。对话推荐系统能够通过自然语言与用户进行多轮对话，逐步深入挖掘其兴趣偏好，从而向对方提供高质量的推荐结果。相比于传统推荐系统，对话推荐系统主要有两方面的不同。其一，对话推荐系统能够利用自然语言与用户进行语义上连贯的多轮对话，提升了人机交互中的用户体验；其二，系统能够询问特定的问题直接获取用户的显式反馈，从而更深入地理解用户兴趣偏好，提供更可靠的推荐结果。目前已经有不少工作在不同的问题设定下对该领域进行了探索，然而尽管如此，这些工作仍仅局限于关注当前正在进行的对话，忽视了过去交互记录中蕴涵的丰富信息，导致对用户偏好建模的不充分。为了解决这个问题，本文提出了一个面向用户偏好建模的个性化对话推荐算法框架，通过双线性模型注意力机制与自注意力层次化编码结构进行用户偏好建模，从而完成对候选物品的排序与推荐。本文设计的模型结构能够在充分利用用户历史对话信息的同时，权衡历史对话与当前对话两类数据的重要性。丰富的用户相关信息来源使得推荐结果在契合用户个性化偏好的同时，更具备多样性，从而缓解“信息茧房”等现象带来的不良影响。基于公开数据集的实验表明了本文方法在个性化对话推荐任务上的有效性。     

Noisy but non-malicious user detection in social recommender systems

Social recommender systems largely rely on user-contributed data to infer users’ preference. While this feature has enabled many interesting applications in social networking services, it also introduces unreliability to recommenders as users are allowed to insert data freely. Although detecting malicious attacks from social spammers has been studied for years, little work was done for detecting Noisy but Non-Malicious Users (NNMUs), which refers to those genuine users who may provide some untruthful data due to their imperfect behaviors. Unlike colluded malicious attacks that can be detected by finding similarly-behaved user profiles, NNMUs are more difficult to identify since their profiles are neither similar nor correlated from one another. In this article, we study how to detect NNMUs in social recommender systems. Based on the assumption that the ratings provided by a same user on closely correlated items should have similar scores, we propose an effective method for NNMU detection by capturing and accumulating user’s “self-contradictions”, i.e., the cases that a user provides very different rating scores on closely correlated items. We show that self-contradiction capturing can be formulated as a constrained quadratic optimization problem w.r.t. a set of slack variables, which can be further used to quantify the underlying noise in each test user profile. We adopt three real-world data sets to empirically test the proposed method. The experimental results show that our method (i) is effective in real-world NNMU detection scenarios, (ii) can significantly outperform other noisy-user detection methods, and (iii) can improve recommendation performance for other users after removing detected NNMUs from the recommender system.     

Improving sparsity and new user problems in collaborative filtering by clustering the personality factors

In collaborative filtering recommender systems, items recommended to an active user are selected based on the interests of users similar to him/her. Collaborative filtering systems suffer from the ‘sparsity’ and ‘new user’ problems. The former refers to the insufficiency of data about users’ preferences and the latter addresses the lack of enough information about the new-coming user. Clustering users is an effective way to improve the performance of collaborative filtering systems in facing the aforementioned problems. In previous studies, users were clustered based on characteristics such as ratings given by them as well as their age, gender, occupation, and geographical location. On the other hand, studies show that there is a significant relationship between users’ personality traits and their interests. To alleviate the sparsity and new user problems, this paper presents a new collaborative filtering system in which users are clustered based on their ‘personality traits’. In the proposed method, the personality of each user is described according to the big-5 personality model and users with similar personality are placed in the same cluster using K-means algorithm. The unknown ratings of the sparse user-item matrix are then estimated based on the clustered users, and recommendations are found for a new user according to a user-based approach which relays on the interests of the users with similar personality to him/her. In addition, for an existing user in the system, recommendations are offered in an item-based approach in which the similarity of items is estimated based on the ratings of users similar to him/her in personality. The proposed method is compared to some former collaborative filtering systems. The results demonstrate that in facing the data sparsity and new user problems, this method reduces the mean absolute error and improves the precision of the recommendations.     

A Movie Recommendation System – An Application of Voting Theory in User Modeling

Our research agenda focuses on building software agents that can employ user modeling techniques to facilitate information access and management tasks. Personal assistant agents embody a clearly beneficial application of intelligent agent technology. A particular kind of assistant agents, recommender systems, can be used to recommend items of interest to users. To be successful, such systems should be able to model and reason with user preferences for items in the application domain. Our primary concern is to develop a reasoning procedure that can meaningfully and systematically tradeoff between user preferences. We have adapted mechanisms from voting theory that have desirable guarantees regarding the recommendations generated from stored preferences. To demonstrate the applicability of our technique, we have developed a movie recommender system that caters to the interests of users. We present issues and initial results based on experimental data of our research that employs voting theory for user modeling, focusing on issues that are especially important in the context of user modeling. We provide multiple query modalities by which the user can pose unconstrained, constrained, or instance-based queries. Our interactive agent learns a user model by gaining feedback aboutits recommended movies from the user. We also provide pro-active information gathering to make user interaction more rewarding. In the paper, we outline the current status of our implementation with particular emphasis on the mechanisms used to provide robust and effective recommendations.     

Implicit feedback techniques on recommender systems applied to electronic books

The goal of this research is to define and capture a series of parameters that allowed us to perform a comparative analysis and find correlations between explicit and implicit feedback on recommender systems. Most of these systems require explicit actions from the users, such as rating, and commenting. In the context of electronic books this interaction may alter the patterns of reading and understanding of the users, as they are asked to stop reading and rate the content. By simulating the behavior of an electronic book reader we have improved the feedback process, by implicitly capturing, measuring, and classifying the information needed to discover user interests. In these times of information overload, we can now develop recommender systems that are mostly based on the user’s behavior, by relying on the obtained results.     

Modeling and broadening temporal user interest in personalized news recommendation

User profiling is an important step for solving the problem of personalized news recommendation. Traditional user profiling techniques often construct profiles of users based on static historical data accessed by users. However, due to the frequent updating of news repository, it is possible that a user’s fine-grained reading preference would evolve over time while his/her long-term interest remains stable. Therefore, it is imperative to reason on such preference evaluation for user profiling in news recommenders. Besides, in content-based news recommenders, a user’s preference tends to be stable due to the mechanism of selecting similar content-wise news articles with respect to the user’s profile. To activate users’ reading motivations, a successful recommender needs to introduce “somewhat novel” articles to users.In this paper, we initially provide an experimental study on the evolution of user interests in real-world news recommender systems, and then propose a novel recommendation approach, in which the long-term and short-term reading preferences of users are seamlessly integrated when recommending news items. Given a hierarchy of newly-published news articles, news groups that a user might prefer are differentiated using the long-term profile, and then in each selected news group, a list of news items are chosen as the recommended candidates based on the short-term user profile. We further propose to select news items from the user–item affinity graph using absorbing random walk model to increase the diversity of the recommended news list. Extensive empirical experiments on a collection of news data obtained from various popular news websites demonstrate the effectiveness of our method.     

A survey of autoencoder-based recommender systems

In the past decade,recommender systems have been widely used to provide users with personalized products and services.However,most traditional recommender systems are still facing a challenge in dealing with the huge volume,complexity,and dynamics of information.To tackle this challenge,many studies have been conducted to improve recommender system by integrating deep learning techniques.As an unsupervised deep learning method,autoencoder has been widely used for its excellent performance in data dimensionality reduction,feature extraction,and data reconstruction.Meanwhile,recent researches have shown the high efficiency of autoencoder in information retrieval and recommendation tasks.Applying autoencoder on recommender systems would improve the quality of recommendations due to its better understanding of users,demands and characteristics of items.This paper reviews the recent researches on autoencoder-based recommender systems.The differences between autoencoder-based recommender systems and traditional recommender systems are presented in this paper.At last,some potential research directions of autoencoder-based recommender systems are discussed.     

CinemaScreen recommender agent: combining collaborative and content-based filtering

A film recommender agent expands and fine-tunes collaborative-filtering results according to filtered content elements - namely, actors, directors, and genres. This approach supports recommendations for newly released, previously unrated titles. Directing users to relevant content is increasingly important in today's society with its ever-growing information mass. To this end, recommender systems have become a significant component of e-commerce systems and an interesting application domain for intelligent agent technology.     

基于时序行为的协同过滤推荐算法

协同过滤直接根据用户的行为记录去预测其可能喜欢的产品,是现今最为成功、应用最广泛的推荐方法.概率矩阵分解算法是一类重要的协同过滤方式.它通过学习低维的近似矩阵进行推荐,能够有效处理海量数据.然而,传统的概率矩阵分解方法往往忽略了用户(产品)之间的结构关系,影响推荐算法的效果.通过衡量用户(产品)之间的关系寻找相似的邻居用户(产品),可以更准确地识别用户的个人兴趣,从而有效提高协同过滤推荐精度.为此,提出一种对用户(产品)间的时序行为建模的方法.基于该方法,可以发现对当前用户(产品)影响最大的邻居集合.进一步地,将该邻居集合成功融合到基于概率矩阵分解的协同过滤推荐算法中.在两个真实数据集上的验证结果表明,所提出的SequentialMF 推荐算法与传统的使用社交网络信息与标签信息的推荐算法相比,能够更有效地预测用户实际评分,提升推荐精度.     

Enriching one-class collaborative filtering with content information from social media

In recent years, recommender systems have become popular to handle the information overload problem of social media websites. The most widely used Collaborative Filtering methods make recommendations by mining users’ rating history. However, users’ behaviors in social media are usually implicit, where no ratings are available. This is a One-Class Collaborative Filtering (OCCF) problem with only positive examples. How to distinguish the negative examples from missing data is important for OCCF. Existing OCCF methods tackle this by the statistical properties of users’ historical behavior; however, they ignored the rich content information in social media websites, which provide additional evidence for profiling users and items. In this paper, we propose to improve OCCF accuracy by exploiting the social media content information to find the potential negative examples from the missing user-item pairs. Specifically, we get a content topic feature for each user and item by probabilistic topic modeling and embed them into the Matrix Factorization model. Extensive experiments show that our algorithm can achieve better performance than the state-of-art methods.     

Sequence-based trust in collaborative filtering for document recommendation

Collaborative filtering (CF) recommender systems have emerged in various applications to support item recommendation, which solve the information-overload problem by suggesting items of interest to users. Recently, trust-based recommender systems have incorporated the trustworthiness of users into CF techniques to improve the quality of recommendation. They propose trust computation models to derive the trust values based on users' past ratings on items. A user is more trustworthy if s/he has contributed more accurate predictions than other users. Nevertheless, conventional trust-based CF methods do not address the issue of deriving the trust values based on users' various information needs on items over time. In knowledge-intensive environments, users usually have various information needs in accessing required documents over time, which forms a sequence of documents ordered according to their access time. We propose a sequence-based trust model to derive the trust values based on users' sequences of ratings on documents. The model considers two factors – time factor and document similarity – in computing the trustworthiness of users. The proposed model enhanced with the similarity of user profiles is incorporated into a standard collaborative filtering method to discover trustworthy neighbors for making predictions. The experiment result shows that the proposed model can improve the prediction accuracy of CF method in comparison with other trust-based recommender systems.