基于大数据技术的电子商务个性化信息自动推荐算法应用

当前个性化推荐算法应用因未构建增量更新推荐模型,对电子商务平台的大数据进行处理,导致推荐结果不能及时更新,严重影响用户体验.因此,提出基于大数据技术的电子商务个性化信息自动推荐算法.首先对平台中用户与项目的相异度进行计算,利用计算后的相异度矩阵构建数据增量更新模型;然后,将该模型作为个性化推荐模型,引入IU-UserCF推荐算法,提高预测准确性;最后以电商物流为应用对象,将所提方法引入应用进行探索.实验对比结果表明,所提方法在电子商务应用中有着更加出色的推荐效果.     

基于用户聚类和移动上下文的矩阵分解推荐算法研究

随着移动互联网技术的快速发展,越来越多的用户通过移动设备获取移动信息和服务,导致信息过载问题日益凸出。针对目前上下文感知推荐算法中存在的数据稀疏性差、上下文信息融入不够、用户相似性度量被忽略等问题,提出一种基于用户聚类和移动上下文的矩阵分解推荐算法。该算法通过利用k-means对用户聚类找到偏好相似的用户簇,求出每簇中并对 用户所处上下文之间的相似度并对其进行排序,由此找出与目标用户偏好和上下文均相似的用户集合,借助该集合改进传统矩阵分解模型损失函数,并以此为基准进行评分预测和推荐。仿真实验结果表明,所提算法可有效提高预测评分的准确度。     

融合朋友关系和标签信息的张量分解推荐算法

针对大众标注网站项目推荐系统中存在数据矩阵稀疏性影响推荐效果的问题,考虑矩阵奇异值分解(SVD)能有效地平滑数据矩阵中的数据,以及朋友圈能够反映出一个人的兴趣爱好,提出了一种融合朋友关系和标签信息的张量分解推荐算法。首先,利用高阶奇异值分解(HOSVD)方法对用户-项目-标签三元组信息进行潜在语义分析和多路降维,分析用户、项目、标签三者间关系;然后,再结合用户朋友关系、朋友间相似度,修正张量分解结果,建立三阶张量模型,从而实现推荐。该模型方法在两个真实数据集上进行了实验,结果表明,所提算法与高阶奇异值分解的方法比较,在推荐的召回率和精确度指标上分别提高了2.5%和4%,因此,所提算法进一步验证了结合朋友关系能够提高推荐的准确率,并扩展了张量分解模型,实现用户个性化推荐。     

基于User Profile的微博用户推荐

传统微博用户推荐算法采用的数据来源单一,模型简单,导致推荐准确率不高。针对这一问题,本文提出一种基于标签的User Profile推荐算法,根据微博数据的特点,深入研究“微博文本”、“标签”、“社交关系”和“用户自身基本信息”等因素对微博个性化推荐的影响,通过训练LDA主题模型和SVM分类器将它们转换为标签,并赋予权重来描述用户兴趣,进行用户推荐以提高推荐准确性。实验结果表明,与传统VSM模型方法相比,该算法进行用户推荐效果更佳。     

基于多主题追踪的网络新闻推荐

针对网络新闻推荐系统推荐准确率偏低的问题,提出一种基于多主题追踪的网络新闻推荐算法。基于多主题追踪的推荐算法采用多个用户模型表示用户对不同主题的兴趣,并动态更新用户模型以动态反映用户的兴趣变化。实现了网络新闻推荐系统的核心推荐算法,并在标准路透社新闻数据集(RCV1)上验证了算法的有效性,有效提升了新闻推荐的准确率。     

基于局部模型加权融合的Top-N电影推荐算法

为了解决传统推荐算法使用单一模型无法准确捕获用户偏好的问题, 将稀疏线性模型作为基本推荐模型,提出了基于用户聚类的局部模型加权融合算法来实现电影的Top-N个性化推荐。同时,为了实现用户聚类,文中利用LDA主题模型和电影的文本内容信息,提出了语义层次用户特征向量的计算方法,并基于此来实现用户聚类。在豆瓣网电影数据集上的实验验证结果表明,所提局部加权融合推荐算法提升了原始基模型的推荐效果,同时又优于一些传统的经典推荐算法,从而证明了该推荐算法的有效性。     

基于弹幕情感分析和主题模型的视频推荐算法

针对互联网上大量自制视频缺少用户评分、推荐准确率不高的问题,提出一种融合弹幕情感分析和主题模型的视频推荐算法（VRDSA）。首先,对视频的弹幕评论进行情感分析,得到视频的情感向量,之后基于情感向量计算视频之间的情感相似度;同时,基于视频的标签建立主题模型来得到视频标签的主题分布,并使用主题分布计算视频之间的主题相似度;接着,对视频的情感相似度和主题相似度进行融合得到视频间的综合相似度;然后,结合视频间的综合相似度和用户的历史记录得到用户对视频的偏好度;同时通过视频的点赞量、弹幕量、收藏数等用户互动指标对视频的大众认可度进行量化,并结合用户历史记录计算出视频的综合认可度;最后,基于用户对视频的偏好度和视频的综合认可度预测用户对视频的认可度,并生成个性化推荐列表来完成视频的推荐。实验结果表明,与融合协同过滤和主题模型的弹幕视频推荐算法（DRCFT）以及嵌入LDA主题模型的协同过滤算法（ULR-itemCF）相比,所提算法推荐的准确率平均提高了17.1%,召回率平均提高了22.9%,F值平均提高了22.2%。所提算法对弹幕进行情感分析,并融合主题模型,以此来完成对视频的推荐,并且充分挖掘了弹幕数据的情感性,使得推荐结果更加准确。     

融合知识图谱与注意力机制的推荐算法

为了解决信息过载问题,提出了一种融合知识图谱与注意力机制的推荐模型.在该模型中,将知识图谱作为辅助信息进行嵌入,可以缓解传统推荐算法数据稀疏和冷启动问题,并且给推荐结果带来可解释性.为了提升推荐准确率以及捕捉用户兴趣的动态变化,再结合深度学习中的神经网络以及注意力机制生成用户自适应表示,加上动态因子来更好地捕捉用户动态...     

基于好友关系和标签的混合协同过滤算法

针对传统推荐算法存在数据稀疏影响推荐效果的问题,考虑到社交网络中的链路预测能够综合考虑用户节点之间的拓扑结构,以及好友关系能反映用户的兴趣爱好,提出了一种融合好友关系和标签信息的推荐算法。首先,借助网络资源分配算法对社交网络的结构信息进行特征提取;然后,利用TF-IDF构建合理的社会化标签模型;最后,利用线性模型融合两方面的信息,从而实现推荐。在Last.fm和Delicious数据集上的实验表明,与传统算法相比,所提算法在推荐的召回率和准确率指标上有显著提高。     

多特征融合的兴趣点推荐算法

基于位置社交网络的兴趣点推荐越来越受到工业界和学术界的关注。由于用户签到数据集的稀疏性以及签到地理位置的聚集性，使得目前的推荐算法效率普遍不高，特别是当用户外出到新的地点时，推荐效果更是急剧下降。因此本文提出了一种基于用户-区域-内容主题的多特征联合推荐算法（UCRTM），以隐主题模型为基础，在统一的框架下利用隐含因子关联性融合了用户的偏好、兴趣点的内容以及兴趣点所属地理区域主题等信息来进行推荐，使得用户无论身处何地，都能获得理想的推荐服务。本文在两种真实的数据集上进行了实验，结果表明该方法不仅能够克服数据的稀疏性以及弱语义性等问题，而且与其他方法相比具有更高的推荐准确率。     

Multi-objective mobile app recommendation: A system-level collaboration approach

The tremendous increase of mobile apps has given rise to the significant challenge of app discovery. To alleviate such a challenge, recommender systems are employed. However, the development of recommender systems for mobile apps is at a slow pace. One main reason is that a general framework for efficient development is still missing. Meanwhile, most existing systems mainly focus on single objective recommendations, which only reflect monotonous app needs of users. For such reasons, we initially present a general framework for developing mobile app recommender systems, which leverages the multi-objective approach and the system-level collaboration strategy. Our framework thus can satisfy ranges of app needs of users by integrating the strengths of various recommender systems. To implement the framework, we originally introduce the method of swarm intelligence to the recommendation of mobile apps. To be detailed, we firstly present a new set based optimization problem which is originated from the collaborative app recommendation. We then propose a novel set based Particle Swarm Optimization (PSO) algorithm, namely, the Cylinder Filling Set based PSO, to address such a problem. Furthermore, we implement the algorithm based on three popular mobile app recommender systems and conduct evaluations. Results verify that our framework and algorithm are with promising performance from both the effectiveness and efficiency.     

Personalized app recommendation based on app permissions

With the development of science and technology, the popularity of smart phones has made exponential growth in mobile phone application market. How to help users to select applications they prefer has become a hot topic in recommendation algorithm. As traditional recommendation algorithms are based on popularity and download, they inadvertently fail to recommend the desirable applications. At the same time, many users tend to pay more attention to permissions of those applications, because of some privacy and security reasons. There are few recommendation algorithms which take account of apps’ permissions, functionalities and users’ interests altogether. Some of them only consider permissions while neglecting the users’ interests, others just perform linear combination of apps’ permissions, functionalities and users’ interests to implement top-N recommendation. In this paper, we devise a recommendation method based on both permissions and functionalities. After demonstrating the correlation of apps’ permissions and users’ interests, we design an app risk score calculating method ARSM based on app-permission bipartite graph model. Furthermore, we propose a novel matrix factorization algorithm MFPF based on users’ interests, apps’ permissions and functionalities to handle personalized app recommendation. We compare our work with some of the state-of-the-art recommendation algorithms, and the results indicate that our work can improve the recommendation accuracy remarkably.     

Leveraging app usage contexts for app recommendation: a neural approach

The large volume and variety of apps pose a great challenge for people to choose appropriate apps. As a consequence, app recommendation is becoming increasingly important. Recently, app usage data which record the sequence of apps being used by a user have become increasingly available. Such data record the usage context of each instance of app use, i.e., the app instances being used together with this app (within a short time window). Our empirical data analysis shows that a user has a pattern of app usage contexts. More importantly, the similarity in the two users’ preferences over mobile apps is correlated with the similarity in their app usage context patterns. Inspired by these important observations, this paper tries to leverage the predictive power of app usage context patterns for effective app recommendation. To this end, we propose a novel neural approach which learns the embeddings of both users and apps and then predicts a user’s preference for a given app. Our neural network structure models both a user’s preference over apps and the user’s app usage context pattern in a unified way. To address the issue of unbalanced training data, we introduce several sampling methods to sample user-app interactions and app usage contexts effectively. We conduct extensive experiments using a large real app usage data. Comparative results demonstrate that our approach achieves higher precision and recall, compared with the state-of-the-art recommendation methods.

     

嵌入LDA主题模型的协同过滤推荐算法

协同过滤推荐算法由于其推荐的准确性和高效性已经成为推荐领域最流行的推荐算法之一。该算法通过分析用户的历史评分记录来构建用户兴趣模型,进而为用户产生一组推荐。然而,推荐系统中用户的评分记录是极为有限的,导致传统协同过滤算法面临严重的数据稀疏性问题。针对此问题,提出了一种改进的嵌入LDA主题模型的协同过滤推荐算法(ULR-CF算法)。该算法利用LDA主题建模方法在用户项目标签集上挖掘潜在的主题信息,进而结合文档-主题概率分布矩阵和评分矩阵来共同度量用户和项目相似度。实验结果表明,提出的ULR-CF算法可以有效缓解数据稀疏性问题,并能显著提高推荐系统的准确性。     

Diversified recommendation method combining topic model and random walk

As one of the most widely used algorithms in recommendation field, collaborative filtering (CF) predicts the unknown rating of items based on similar neighbors. Although many CF-based recommendation methods have been proposed, there still be room for improvement. Firstly, the data sparsity problem still remains a big challenge for CF algorithms to find similar neighbors. Secondly, there are many redundant similar items in the recommendation list generated by traditional CF algorithms, which cannot meet the user wide interest. Therefore, we propose a diversified recommendation method combining topic model and random walk. A weighted random walk model is presented to find all direct and indirect similar neighbors on the sparse data, improving the accuracy of rating prediction. By taking both users’ behavior data and items’ lags into account, we give a diversity measurement method based on the topic distribution of items discovered by Linked-LDA model. Furthermore, a diversified ranking algorithm is developed to balance the accuracy and diversity of recommendation results. We compare our method with six other recommendation methods on a real-world dataset. Experimental results show that our method outperforms the other methods and achieves the best personalized recommendation effect.     

Recommendation algorithm of the app store by using semantic relations between apps

In this paper, we propose a personalized recommendation system for mobile application software (app) to mobile user using semantic relations of apps consumed by users. To do that, we define semantic relations between apps consumed by a specific member and his/her social members using Ontology. Based on the relations, we identify the most similar social members from the reasoning process. The reasoning is explored from measuring the common attributes between apps consumed by the target member and his/her social members. The more attributes shared by them, the more similar is their preference for consuming apps. We also develop a prototype of our system using OWL (Ontology Web Language) by defining ontology-based semantic relations among 50 mobile apps. Using the prototype, we showed the feasibility of our algorithm that our recommendation algorithm can be practical in the real field and useful to analyze the preference of mobile user.     

Forensic taxonomy of android productivity apps

Android productivity apps have provided the facility of having a constantly accessible and productive workforce to the information and work capabilities needed by the users. With hundreds of productivity apps available in the Android app market, it is necessary to develop a taxonomy for the forensic investigators and the end users to allow them to know what personal data remnants are available from the productivity apps. In this paper, 30 popular Android productivity apps were examined. A logical extraction of the Android phone was collected by using a well-known mobile forensic tool- XRY to extract various information of forensic interest such as user email ID and list of tasks. Based on the findings, a two-dimensional taxonomy of the forensic artefacts of the productivity apps is proposed with the app categories in one dimension and the classes of artefacts in the other dimension. The artefacts identified in the study of the apps are summarised using the taxonomy. In addition, a comparison with the existing forensic taxonomies of different categories of Android apps is provided to facilitate timely collection and analysis of evidentiary materials from mobile devices.     

移动互联网音视频类协议识别技术研究与实现

音视频类型的应用程序是应用市场中下载的热点,针对传统互联网的协议识别技术已经相对成熟,但对于移动网络中的音视频应用的识别研究还刚开始受到关注,通过对此类应用识别,运营商可以收集用户在线观看音视频等行为习惯,进而为提供用户差异化服务,也可以服务于对应用的安全审计。文章主要研究了移动互联网音视频类协议识别技术,从对应用协议数据的分析中获取应用特征值,利用特征值实现协议识别;通过开发识别程序和大量实验,实现了对移动互联网音视频类协议的自动识别,并进一步提高识别的准确率及效率。     

基于LDA主题模型的移动应用相似度构建方法

随着移动互联网的快速发展,如何从大量的移动应用中抽取有效的描述信息继而为移动用户提供有效准确的推荐策略变得尤为迫切。目前,移动应用市场对应用的推荐策略相对传统,大多是根据应用的单一属性进行推荐,如下载量、应用名称、应用分类等。针对推荐粒度过粗和推荐不准确的问题,提出了一种基于潜在狄利克雷分布（LDA）主题模型的移动应用相似度构建方法。该方法从应用的标签入手,构造应用的主题模型分布矩阵,利用该主题分布矩阵构建移动应用的相似度矩阵,同时提出了将移动应用相似度矩阵转化为可行的存储结构的方法。实验结果表明该方法是有效的,相比现有的360应用市场推荐的应用其相似度提升130%。该方法解决了移动应用推荐过程中推荐粒度过粗的问题,可使推荐结果更加准确。     

数据驱动的移动应用用户接受度建模与预测

应用市场（app market）已经成为互联网环境下软件应用开发和交付的一种主流模式.相对于传统模式,应用市场模式下,软件的交付周期更短,用户的反馈更快,最终用户和开发者之间的联系更加紧密和直接.为应对激烈的竞争和动态演变的用户需求,移动应用开发者必须以快速迭代的方式不断更新应用,修复错误缺陷,完善应用质量,提升用户体验.因此,如何正确和综合理解用户对软件的接受程度（简称用户接受度）,是应用市场模式下软件开发需考量的重要因素.近年来兴起的软件解析学（software analytics）关注大数据分析技术在软件行业中的具体应用,对软件生命周期中大规模、多种类的相关数据进行挖掘和分析,被认为是帮助开发者提取有效信息、作出正确决策的有效途径.从软件解析学的角度,首先论证了为移动应用构建综合的用户接受度指标模型的必要性和可行性,并从用户评价数据、操作数据、交互行为数据这3个维度给出基本的用户接受度指标.在此基础上,使用大规模真实数据集,在目标用户群体预测、用户规模预测和更新效果预测等典型的用户接受度指标预测问题中,结合具体指标,提取移动应用生命周期不同阶段的重要特征,以协同过滤、回归融合、概率模型等方法验证用户接受度的可预测性,并讨论了预测结果与特征在移动应用开发过程中可能提供的指导.