社交网络用户在在线媒体中点播行为预测

在线媒体快速发展,为用户带来丰富多彩信息的同时,用户的参与也给在线媒体本身带来巨大的经济利益。因此,如何通过精确预测用户的偏好以增加在线媒体点击,成为一个学术界和工业界均关注的问题。现有的预测方法主要是借助用户个人信息和历史行为来预测用户行为,然而此类方法没有考虑媒体本身缺乏用户信息造成无法预测的问题。随着社交网络的发展,在线媒体与服务运营商间的兼并或合作的增多,支持用户通过单一账户使用多个媒体网络服务的情况越来越常见,这就为基于用户在社交网络中的资料预测用户在其他媒体中的喜好提供海量可信的基础数据。该文基于社交网络Google+和视频媒体YouTube的数据,首先证明用户在YouTube偏好具有高度的集聚性,并提出用户在社交网络中偏好与其在线媒体点击行为具有关联性,基于这种关联性,该文使用社交网络用户信息预测用户在在线媒体中的点播行为。实验结果显示,使用社交网络用户信息可以有效预测用户偏好,预测准确率比仅使用媒体本身信息提高了17%,而且能满足用户个性化需求。     

利用准私密社交网络文本数据检测抑郁用户的可行性分析

社交媒体的发展为抑郁用户的检测提供了一条新的途径。已有的相关研究通常是利用用户在Twitter、微博等社交网络平台上的用户行为数据或公开发表的文本内容,较少有利用微信朋友圈、QQ空间这种相对比较私密的社交网络数据。直观地,这类准私密社交网络数据更能反映用户的心理健康状况。该文主要讨论利用准私密社交网络文本数据检测抑郁用户的可行性,包括训练样本的选择、特征量化方法、检测模型选择和不同文本特征下的模型分类效果等。实验表明,采用平衡高低分组的方法选择样本比非平衡高低分组样本和离散化的高低分组样本训练的分类器要好;利用Z-score标准化的特征量化方法比直接使用频次或归一化频率要好;随机梯度下降模型SGD较支持向量机SVM等其他用于对比的分类模型要好。实验还发现,相对于词袋、词向量等文本特征,主题特征有较好的效果,可以使社交网络用户抑郁检测模型的F值达到0.753,而对抑郁用户的检测精度达到0.813。     

基于信任关系和词相关关系的冷启动用户词特征重建

文本是社交媒体用户的重要信息之一,从文本中获取用户的词特征是实现用户主题建模、兴趣挖掘及个性化推荐等任务的基础。然而社交媒体中存在许多用户(冷启动用户)只含有少量甚至缺乏文本信息,为此该文提出一种融合用户信任关系及词相关关系的词特征重建方法。该方法通过对用户信任关系矩阵、词相关关系矩阵和用户词特征矩阵进行联合概率矩阵分解来实现对冷启动用户的词特征重建。在新浪微博和Twitter的四组数据集上的实验结果表明,该文所提出的冷启动用户词特征重建算法能够取得较好的词特征重建结果。     

基于文本和用户信息的在线评论质量检测

随着互联网的迅速发展,越来越多的用户评论出现在社交网站上。面对迅速增长的评论数据,如何为阅读评论的消费者提供准确、真实的高质量评论就显得尤为重要。评论质量检测旨在判断在线评论的质量,在传统的研究中,文本信息通常独立地被用于预测评论质量。但是在社交媒体上,每个文本之间不是独立的,而是可以通过发表文本的作者与其他文本相关联,即同一个用户或相近的用户发表的评论质量具有一定的相似性。因此,为了更好的构建文本的表示和研究文本之间基于用户的关联,该文基于神经网络模型分别构建用户和文本的表示,同时,为了放大用户信息的作用,我们进一步将基于注意力机制的用户信息融合到文本中,从而提高文本评论质量检测的效果。在Yelp 2013数据集上进行实验的结果表明,该模型能有效地提高在线评论质量检测的性能。     

基于卷积神经网络的评论文本兴趣点推荐算法

针对基于位置社交网络中的兴趣点推荐存在用户签到数据稀疏、评论文本信息利用不充分、推荐准确度不高等问题, 提出一种基于卷积神经网络的评论文本兴趣点推荐模型(RT-CNN). 首先采用高斯函数利用邻近地理位置加权方法填补矩阵分解模型中缺少的位置信息, 预测用户对未签到位置的潜在兴趣. 然后通过卷积神经网络处理评论文本信息挖掘潜在特征, 深度提取用户情感倾向, 使用Softmax逻辑回归函数获得评论文本与用户和位置兴趣点潜在特征相关的概率, 通过对目标函数的求解提取用户和位置潜在特征向量. 最后融合签到行为、地理位置影响、用户情感倾向、用户潜在特征和位置兴趣点潜在特征进行兴趣点推荐. 在公开的Foursquare网站纽约(NYC)和洛杉矶(LA)两个真实签到数据集进行实验, 结果表明RT-CNN模型相比其他先进的兴趣点推荐模型提高了精确率和召回率, 具有更好的推荐性能.     

基于社交圈层和注意力机制的信息热度预测

社交网络现已成为现实世界中信息传播与扩散的主要媒介,对其中的热点信息进行建模和预测有着广泛的应用场景和商业价值,比如进行信息传播挖掘、广告推荐和用户行为分析等.目前的相关研究主要利用特征和时间序列进行建模,但是并没有考虑到社交网络中用户的社交圈层对于信息传播的作用.本文提出了一种基于社交圈层和注意力机制的热度预测模型SCAP(Social Circle and Attention based Popularity Prediction),首先对社交圈层进行定义,通过自动编码器提取用户历史文本序列的特征,对不同用户的社交圈层进行聚类划分,得到社交圈层特征.进而对于一条新发布的文本信息,通过长短期记忆网络与嵌入层提取其文本特征、用户特征和时序特征,并基于注意力机制,捕获到不同社交圈层对于该文本信息的影响程度,得到社交圈层注意力特征.最后将文本特征、用户特征、时序特征和社交圈层注意力特征进行特征融合,并通过两个全连接层进行建模学习,对社交信息的热度进行预测.在推特、微博和豆瓣等四个数据集上的实验结果表明,SCAP模型的预测表现相比于多个对比模型总体呈优,在不同数据集上均方误差(MSE)分别降低了0.017,0.022,0.021和0.031,F1分数分别提升0.034,0.021,0.034和0.025,能够较为准确地预测社交信息的热度.本文同时探究了不同实验参数对于模型的影响效果,如用户历史文本序列的数量、社交圈层的数量和时间序列的长度,最后验证了模型输入的各个特征和注意力机制的引入对于模型预测性能提升的有效性,在推特数据集中,引入社交圈层和注意力机制,模型的MSE指标分别降低了0.065和0.019.     

基于知识库的客户网购意向预测系统

电子商务网站所产生的海量客户网购行为数据中蕴含着丰富的反映客户网购行为规律的知识。这些知识是进行客户网购意向预测以及在此基础上进一步进行商品推荐的重要依据。基于Hadoop框架构建了“客户购买意向预测系统”,实现了利用隐含在交易数据中的知识对正在浏览中的电商客户的购买意向预测。在基于知识图谱技术构建网购客户购买行为知识库的基础上,利用FP-Growth算法和数据统计方法实现了以增量方式从购买者的交易数据中获取反映用户购买规律的知识,并将取得的知识融合到知识库中。系统能够利用存储于知识库中的知识在客户进行网上购物的过程中对其购买意向进行分析和预测,从而达到提高电子商务平台的运营效率,改善客户购物体验的目的。     

梯度渐进回归树算法在电子商务品牌推荐中的应用

针对电子商务推荐系统中,互联网“信息过载”所造成的难以准确定位用户兴趣并提供准确品牌推荐的问题,通过深入挖掘电子商务网中的用户行为日志,抽取出能辨别出用户对商品品牌购买行为的多个特征,然后将这些特征融入到梯度渐进回归树算法中,建立用户兴趣偏好模型来提高推荐精度。实验结果表明,在数据稀疏的情况下,该算法仍能较好的识别出用户对品牌的偏好,并在推荐准确度方面较其他传统推荐和分类算法有明显的提高。     

一种预测群体用户访问行为的算法

在电子商务发展中,商家需要理解用户访问网站的行为,为用户提供个性化服务,从而吸引用户购买商品。挖掘用户访问网站的行为是商家一个急需解决的问题,通过对Web日志进行挖掘是解决该问题的重要研究方法。提出了网页兴趣信息素的新概念,它是由页面相对浏览时间和点击率构建而成,利用兴趣信息素设计了基于蚁群算法的群体用户访问路径挖掘算法,根据挖掘结果预测用户访问行为。实验结果表明,兴趣信息索可以有效地预测用户的兴趣变化,能准确地反映用户访问模式,提高了预测群体用户访问行为的准确率。     

社交网络用户标签预测研究

随着社交网站的流行以及用户的大规模增加,社交网络用户行为分析已经成为社交网站进行网站维护、性能优化和系统升级的重要基础,也是网络知识挖掘和信息检索的重要研究领域。为了更好地理解社交网络用户添加个人标签的行为特征,该文基于大约263万个微博用户的真实数据,对用户标签的分布进行了研究和分析。我们主要考察了用户标签的宏观分布特征,以及用户标签与关注对象的标签分布之间的联系,发现微博用户给自己添加标签时,在开始阶段倾向于使用反映个性的标签,之后会出于从众心理而选用大众化标签。我们将研究发现运用到基于关注关系的标签预测算法中,结果证实相关分析对于社交网站的标签推荐等课题具有一定的参考意义。     

The effect of different social media marketing channels and events on movie box office: An elaboration likelihood model perspective

Social media has become an important marketing media to attract and retain consumers. This article focuses on the movie industry and aims to investigate how different channels and events in social media marketing achieve box office success through different consumers’ responses. In reference to elaboration likelihood model (ELM), we propose our research model and test hypotheses using the data of 304 movies with the information of box office, attributes, and associated social media posts. The results show that: (1) official microblog activity increases the purchase intent through changing audience's attitude while third-party mention increases purchase intent through catching audience's attention and promoting positive attitude, and there is an interaction effect between official microblog and third-party media; (2) social media marketing events related to contents introduction promote purchase intent through central route while those related to surrounding information promote purchase intent through peripheral route; and (3) movie attributes interact with marketing intensity in affecting purchase intent. Under the same marketing intensity, the marketing effect of domestic sequel movies released on popular holidays is better. Our findings provide both theoretical and practical implications.     

An Experimental Study of Text Representation Methods for Cross-Site Purchase Preference Prediction Using the Social Text Data

Nowadays, many e-commerce websites allow users to login with their existing social networking accounts. When a new user comes to an e-commerce website, it is interesting to study whether the information from external social media platforms can be utilized to alleviate the cold-start problem. In this paper, we focus on a specific task on cross-site information sharing, i.e., leveraging the text posted by a user on the social media platform (termed as social text) to infer his/her purchase preference of product categories on an e-commerce platform. To solve the task, a key problem is how to effectively represent the social text in a way that its information can be utilized on the e-commerce platform. We study two major kinds of text representation methods for predicting cross-site purchase preference, including shallow textual features and deep textual features learned by deep neural network models. We conduct extensive experiments on a large linked dataset, and our experimental results indicate that it is promising to utilize the social text for predicting purchase preference. Specially, the deep neural network approach has shown a more powerful predictive ability when the number of categories becomes large.     

Social Media Visual Analytics

With the development of social media (e.g. Twitter, Flickr, Foursquare, Sina Weibo, etc.), a large number of people are now using them and post microblogs, messages and multi‐media information. The everyday usage of social media results in big open social media data. The data offer fruitful information and reflect social behaviors of people. There is much visualization and visual analytics research on such data. We collect state‐of‐the‐art research and put it into three main categories: social network, spatial temporal information and text analysis. We further summarize the visual analytics pipeline for the social media, combining the above categories and supporting complex tasks. With these techniques, social media analytics can apply to multiple disciplines. We summarize the applications and public tools to further investigate the challenges and trends.     

The influence of eWOM in social media on consumers’ purchase intentions: An extended approach to information adoption

Social media websites have created valuable opportunities for electronic word of mouth (eWOM) conversations. People are now able to discuss products and services of brands with their friends and acquaintances. The aim of this study is to examine the influence of these conversations in social media on consumers' purchase intentions. For this purpose, a conceptual model was developed based on the integration of Information Adoption Model (IAM) and related components of Theory of Reasoned Action (TRA). The new model, which is named as Information Acceptance Model (IACM), was validated through structural equation modelling (SEM) based on surveys of 384 university students who use social media websites. The results confirm that quality, credibility, usefulness and adoption of information, needs of information and attitude towards information are the key factors of eWOM in social media that influence consumers’ purchase intentions. Theoretical and practical implications are discussed as well as recommendations for further research.     

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.     

Distinguishing re-sharing behaviors from re-creating behaviors in information diffusion

Social media plays a fundamental role in the diffusion of information. There are two different ways of information diffusion in social media platforms such as Twitter and Weibo. Users can either re-share messages posted by their friends or re-create messages based on the information acquired from other non-local information sources such as the mass media. By analyzing around 60 million messages from a large micro-blog site, we find that about 69 % of the diffusion volume can be attributed to users’ re-sharing behaviors, and the remaining 31 % are caused by user re-creating behaviors. The information diffusions caused by the two kinds of behaviors have different characteristics and variation trends, but most existing models of information diffusion do not distinguish them. The recent availability of massive online social streams allows us to study the process of information diffusion in much finer detail. In this paper, we introduce a novel model to capture and simulate the process of information diffusion in the micro-blog platforms, which distinguishes users’ re-sharing behaviors from re-creating behaviors by introducing two different components. Thus, our model not only considers the effect of the underlying network structure, but also the influence of other non-local information sources. The empirical results show the superiority of our proposed model in the fitting and prediction tasks of information diffusion.     

Understanding online consumers’ purchase intentions: a contribution from social network theory

There is a dynamic and interconnected international setting shaped by the power of the Internet and social media. To gain more consumers, understand their behaviours and needs, and maintain closest relationships with them, businesses should understand how consumers behave in social media and how they vary in their purchase intentions. In the scope of the study, we integrate the social network theory and the theory of planned behaviour to analyse online consumers’ purchase intentions and to investigate their structural positions by analysing their friendships in social networks. We target Twitter users to conduct analysis due to Twitter's popularity in use, market penetration, and opportunity to work with open-source data. This study contributes to a better theoretical understanding of online consumers’ purchase intentions by integrating multiple theoretical perspectives. It expands the literature by considering both online consumers’ friendship network in Twitter and their individual online purchasing intentions. The study also guides e-marketers to design proper strategies for potential and current consumers and target the right sets of people in the social networks.     

Attribute selection-based recommendation framework for short-head user group: An empirical study by MovieLens and IMDB

Most of recommender systems have serious difficulties on providing relevant services to the “short-head” users who have shown intermixed preferential patterns. In this paper, we assume that such users (which are referred to as long-tail users) can play an important role of information sources for improving the performance of recommendation. Attribute reduction-based mining method has been proposed to efficiently select the long-tail user groups. More importantly, the long-tail user groups as domain experts are employed to provide more trustworthy information. To evaluate the proposed framework, we have integrated MovieLens dataset with IMDB, and empirically shown that the long-tail user groups are useful for the recommendation process.     

Social presence,trust, and social commerce purchase intention: An empirical research

Lacking the presence of human and social elements is claimed one major weakness that is hindering the growth of e-commerce. The emergence of social commerce might help ameliorate this situation. Social commerce is a new evolution of e-commerce that combines the commercial and social activities by deploying social technologies into e-commerce sites. Social commerce reintroduces the social aspect of shopping to e-commerce, increasing the degree of social presences in online environment. Drawing upon the social presence theory, this study theorizes the nature of social aspect in online SC marketplace by proposing a set of three social presence variables. These variables are then hypothesized to have positive impacts on trusting beliefs which in turn result in online purchase behaviors. The research model is examined via data collected from a typical e-commerce site in China. Our findings suggest that social presence factors grounded in social technologies contribute significantly to the building of the trustworthy online exchanging relationships. In doing so, this paper confirms the positive role of social aspect in shaping online purchase behaviors, providing a theoretical evidence for the fusion of social and commercial activities. Finally, this paper introduces a new perspective of e-commerce and calls more attention to this new phenomenon of social commerce.     

The influence of information cascades on online purchase behaviors of search and experience products

Online users usually observe or refer to others’ behaviors and discount their own information when purchasing products online. This research employed a fixed-effect regression model to elucidate how information cascades could influence online purchase behaviors and how they moderated the influence of online word-of-mouth and product prices. To uncover the underlying mechanisms behind informational cascades, we compare search products and experience products. In particular, we utilize publicly available data from a B2C e-commerce site in China, i.e., Tmall.com. Our results indicate that online users’ choice of products was heavily driven by changes in product rankings after controlling for cumulative sales, online user reviews and product price, as predicted by informational cascades theory. Due to the information cascades effect, review volume had no impact on online users’ choice of products with high rankings, whereas it did exert a significant positive impact on consumer purchase decisions of products with low rankings. User rating had no impact on online users’ purchase decisions. Product price had a significant and negative impact for products with high rankings, but had a significant and positive influence on users’ choice for products with low rankings. Moreover, information cascades were more prominent for experience goods than for search goods.