基于类别转移加权张量分解模型的兴趣点分区推荐

基于位置社交网络的兴趣点(POI)推荐是人们发现有趣位置的重要途径,然而,现实中用户在不同区域的地点偏好侧重的差异,加之高维度的历史签到信息,使得精准而又个性化的POI推荐极富挑战性。对此,该文提出一种新型的基于类别转移加权张量分解模型的兴趣点分区推荐算法(WTD-PR)。通过结合用户连续行为和时间特征,来充分利用用户的历史访问信息,从而得到类别转移权重因子;接着改进用户-时间-类别张量模型,在此张量中加入类别转移权重,预测用户的喜好类别;最后,根据用户的历史访问区域划分出本地和异地,并基于用户的当前位置找出推荐区域范畴,进而引入位置因素和社交因素,结合候选类别作兴趣点分区推荐。通过在真实数据集上进行对比实验,实验结果表明,所提算法不仅具有通用性,而且在推荐性能上也优于其他对比算法。     

基于k-距离与重加权M-估计量的鲁棒协同推荐算法(英文)

The existing collaborative recommendation algorithms have lower robustness against shilling attacks.With this problem in mind,in this paper we propose a robust collaborative recommendation algorithm based on k-distance and Tukey M-estimator.Firstly,we propose a k-distancebased method to compute user suspicion degree(USD).The reliable neighbor model can be constructed through incorporating the user suspicion degree into user neighbor model.The influence of attack profiles on the recommendation results is reduced through adjusting similarities among users.Then,Tukey M-estimator is introduced to construct robust matrix factorization model,which can realize the robust estimation of user feature matrix and item feature matrix and reduce the influence of attack profiles on item feature matrix.Finally,a robust collaborative recommendation algorithm is devised by combining the reliable neighbor model and robust matrix factorization model.Experimental results show that the proposed algorithm outperforms the existing methods in terms of both recommendation accuracy and robustness.     

基于用户行为认知的在线社交网络协同推荐

近年来,在线社交网络成为人们工作、生活不可或缺的信息共享与交流工具,如何对海量庞杂、大范围时空关联的用户行为信息进行认知并据此提供个性化的推荐服务,已成为在线社交网络发展重点关注的问题。为此,提出了一种基于用户行为认知的在线社交网络协同推荐框架,在对用户特征、文本信息及兴趣偏好等行为进行认知的基础上,利用协同过滤算法,实现个性化的推荐服务。实验结果验证了提出的基于用户行为认知的协同推荐策略具有较好的稳定性和实际应用效果。     

异质社交网络中多通道特征融合的好友推荐模型

合理有效的好友推荐算法对于社交网络的发展和扩张有重大的意义。然而随着社交网络的复杂化和异质化,传统推荐系统中协同过滤推荐方法不能满足需求。针对异质社交网络中存在着大量的内容相关信息这一特点,根据好友推荐的需求,提出了多通道特征融合的好友推荐模型。该模型对用户相关的多维特征进行挖掘与利用,包括显性特征(如用户profile,用户tag,社交关系等)和隐性特征(如用户重要度,挖掘用户标注发现其领域兴趣等),并进一步将这些内容相关的多特征融合到协同排序算法中进行学习训练。实验结果表明,随着多个内容特征的逐步融合,算法的MAP值稳步提高,最终相对未融合的协同排序方法提高了12%,并在一定程度上的解决了冷启动问题,提高了好友推荐的多样性。     

基于随机游走和多样性图排序的个性化服务推荐方法

针对传统服务推荐算法由于数据稀疏性而导致推荐准确性不高,以及推荐结果缺乏多样性等缺陷,提出基于随机游走和多样性图排序的个性化服务推荐方法（PRWDR）.在分析直接相似关系稀疏性的基础上提出带权重的随机游走模型,通过在用户网络上进行随机游走来挖掘更多的相似关系;基于所有相似用户预测服务的QoS值,并给出服务图模型构建方法,以过滤大量性能过低的候选服务;提出最优节点集合选取策略,利用贪婪算法得到兼具推荐准确性和功能多样性的服务推荐列表.在公开发布的数据集上进行实验,并与多个经典算法进行比较,验证了本算法的有效性.     

融合多权重因素的低秩概率矩阵分解推荐模型

针对个性化推荐精度较低、对冷启动敏感等问题,该文提出一种融合多权重因素的低秩概率矩阵分解推荐模型MWFPMF。模型利用给定的社交网络构建信任网络,借助Page rank算法和信任传递机制求取用户间信任度;基于Page rank计算用户社会地位,利用活动评分和评分时间修正用户间关系权重;引入词频-逆文本频率技术(TF-IDF)求取用户标签,通过标签相似性表征用户间同质性;将用户间信任度、用户社会地位影响力和用户同质性3因素融入低秩概率矩阵分解中,从而使用户偏好和活动特征映射到同一低秩空间,实现用户-活动评分矩阵的分解,在正则化约束下,最终完成低秩特征矩阵对用户评分缺失的有效预测。利用豆瓣同城北京和Ciao数据集确定各模块的参数设置值。通过仿真对比实验可知,本推荐模型获得了较高的推荐精度,与其他5种传统推荐算法相比,平均绝对误差至少降低了6.58%,均方差误差至少降低了6.27%,与深度学习推进算法相比,推荐精度基本接近;在冷启动用户推荐上优势明显,与其他推荐算法相比,平均绝对误差至少降低了0.89%,均方差误差至少降低了3.01%。     

The Adaptive Ontology-Based Personalized Recommender System

Recommender systems provide strategies that help users search or make decisions within the overwhelming information spaces nowadays. They have played an important role in various areas such as e-commerce and e-learning. In this paper, we propose a hybrid recommendation strategy of content-based and knowledge-based methods that are flexible for any field to apply. By analyzing the past rating records of every user, the system learns the user’s preferences. After acquiring users’ preferences, the semantic search-and-discovery procedure takes place starting from a highly rated item. For every found item, the system evaluates the Interest Intensity indicating to what degree the user might like it. Recommender systems train a personalized estimating module using a genetic algorithm for each user, and the personalized estimating model helps improve the precision of the estimated scores. With the recommendation strategies and personalization strategies, users may have better recommendations that are closer to their preferences. In the latter part of this paper, a real-world case, a movie-recommender system adopting proposed recommendation strategies, is implemented.     

A probabilistic model for sub-carrier allocation in OFDMA systems

Orthogonal frequency division multiple-access technique showed a successful utilization of channel features. It implements an orthogonal sub-carrier space to be shared among different users. The management of these sub-carriers in both power and frequency allocation is reflected on the systems performance as better utilization of bandwidth, and hence, better capacity is obtained. Sub-carrier allocation is used to avoid deep fading that might occur at some user’s locations but not at other user’s locations. In this paper, we devise an algorithm based on probabilistic model for sub-carrier allocation to avoid deep fading in some user’s signals. By controlling the sub-carrier allocation for each user, we can create a full rank channel for each user and hence, provide maximum capacity for the system. Simulation results showed that using the devised algorithm will avoid deep fading and utilize the bandwidth up to 40% more than localized allocation strategies.     

Personalized privacy protection method for group recommendation

To address the problem that most of the existing privacy protection methods can not satisfy the user’s personalized requirements very well in group recommendation,a user personalized privacy protection framework based on trusted client for group recommendation (UPPPF-TC-GR) followed with a group sensitive preference protection method (GSPPM) was proposed.In GSPPM,user’s historical data and privacy preference demands were collected in the trusted client,and similar users were selected in the group based on sensitive topic similarity between users.Privacy protection for users who had privacy preferences in the group was realized by randomization of cooperative disturbance to top k similar users.Simulation experiments show that the proposed GSPPM can not only satisfy privacy protection requirements for each user but also achieve better performance.     

User recommendation based on cross-platform online social networks

In the field of online social networks on user recommendation,researchers extract users’ behaviors as much as possible to model the users.However,users may have different likes and dislikes in different social networks.To tackle this problem,a cross-platform user recommendation model was proposed,users would be modeled all-sided.In this study,the Sina micro blog and the Zhihu were investigated in the proposed model,the experimental results show that the proposed model is competitive.Based on the proposed model and the experimental results,it can be known that modeling users in cross-platform online social networks can describe the user more comprehensively and leads to a better recommendation.     

基于改进的混合模式个性化选课推荐技术研究

针对高等学校学生选课系统中存在的缺乏个性化课程推荐、选课效率较低的问题,通过对个性化推荐技术的分析研究,提出了基于内容、项目及用户属性的改进混合模式算法,并将该算法应用到选课系统中,用MACE数据集对算法进行验证。结果表明,该算法解决了个性化推荐技术中的冷启动问题,相关指标有明显提高,实现了课程与新课程的个性化推荐,并减少了选课的盲目性。     

Trust expansion and listwise learning-to-rank based service recommendation method

In view of the problem of trust relationship in traditional trust-based service recommendation algorithm,and the inaccuracy of service recommendation list obtained by sorting the predicted QoS,a trust expansion and listwise learning-to-rank based service recommendation method (TELSR) was proposed.The probabilistic user similarity computation method was proposed after analyzing the importance of service sorting information,in order to further improve the accuracy of similarity computation.The trust expansion model was presented to solve the sparseness of trust relationship,and then the trusted neighbor set construction algorithm was proposed by combining with the user similarity.Based on the trusted neighbor set,the listwise learning-to-rank algorithm was proposed to train an optimal ranking model.Simulation experiments show that TELSR not only has high recommendation accuracy,but also can resist attacks from malicious users.     

Joint recommendation algorithm based on tensor completion and user preference

Aiming at the problem that existing recommendation algorithms have little regard for user preference,and the recommendation result is not satisfactory,a joint recommendation algorithm based on tensor completion and user preference was proposed.First,a user-item-category 3-dimensional tensor was built based on user-item scoring matrix and item-category matrix.Then,the Frank-Wolfe algorithm was used for iterative calculation to fill in the missing data of the tensor.At the same time,a user category preference matrix and a scoring preference matrix were built based on the 3-dimensional tensor.Finally,a joint recommendation algorithm was designed based on the completed tensor and the two preference matrices,and the differential evolution algorithm was used for parameter tuning.The experimental results show that compared with some typical and newly proposed recommendation algorithms,the proposed algorithm is superior to the compare algorithms,the precision is improved by 1.96% ~ 3.44% on average,and the recall rate is improved by 1.35%~2.40% on average.     

基于用户群组行为分析的视频推荐方法研究

该文采用权重增量及相似聚集的用户行为分析算法,为用户推荐个性化视频提供了一个有效的解决方案。方法包含3个主要部分,首先利用RFM(Recentness, Frequency, Monetary amount)模型分析用户的行为,将相同行为的用户归为一组;然后结合用户的最近习惯,使用基于权重增量的Apriori算法挖掘用户之间的关联规则,并用向量空间模型进行相似度计算从而实现用户相似聚集;最后进行协同过滤式推荐,完成整体个性化视频推荐过程。该方法的特点是行为数据自动收集获取,避免了直接对视频大数据的处理;另外,视频推荐随着用户行为的改变而动态变化,更加符合实际情况。实验结果表明,该方法有效并且稳定,相比于单一推荐方法,在准确率、召回率等综合指标上均有明显提升。     

基于归一化方法的协同过滤推荐算法

针对传统的协同过滤推荐算法没有考虑用户向量维度以及评价值差异的问题,提出了一种基于归一化方法的协同过滤推荐算法。算法在计算用户或项目相似度之前首先将用户对每一项目的评分值归一化到一个统一的值域范围,然后再计算用户向量空间的相似性并进行预测推荐。实验结果显示该算法能够准确地找到相似的邻居用户或项目,预测及推荐测性能有较大提高。     

一种新颖的协同推荐算法研究

为了克服协同推荐系统中的用户评分数据稀疏性和推荐实时性差的问题,提出了一种高效的基于粗集的个性化推荐算法.该算法首先利用维数简化技术对评分矩阵进行优化,然后采用分类近似质量计算用户间的相似性形成最近邻居,从而降低数据稀疏性和提高最近邻寻找准确性.实验结果表明,该算法有效地解决用户评分数据极端稀疏情况下传统相似性度量方法存在的问题,显著地提高推荐系统的推荐质量.     

Context-based Ontology-driven Recommendation Strategies for Tourism in Ubiquitous Computing

Tourism is an information-intensive business. At present, there are a lot of information and tourism resources available on the internet that lead to low searching efficiency and effectiveness, the user may get too many seeking results but not related to his interest, or few results than his expected. The user can know clearly what he wants, but sometime the user doesn’t know what kind information he needs. User’s demand can be formulated as direct demand and potential preference. At the same time, the study shows that there is strong relationship between the traveler’s potential preference and the characteristics of tourism resources. In order to solve the information overload challenge, recommendation services are increasingly emerging. Currently, recommendation methods focus on dealing with personalized matching based on the user preference. However, these methods skip the user’s direct demand. In this paper, we propose ontology-driven recommendation strategies based on user’s context. The strategies use ontology to describe and integrate tourism resources, achieve the goal of associating user’s direct needs and his potential preference as the context in recommendation. Moreover, theoretical analysis and experiments show that the proposed approach is feasible, the results of the evaluation are discussed.     

Random Group Recommendation Model Based on Fuzzy Clustering

The group recommendation system is a viral requirement for the Internet service provider to provide recommendation services for all the users in a group. Due to the shared or different interests among users in the group, it is difficult for traditional personal recommendation algorithms to predict items that can meet the requirements of all the users in the group. In this paper, a random group recommendation model is proposed to recommend the top K most appealing items for all the users in a random group. By analyzing item ratings of all the users in the group, the recommendation model can abstract the group as a virtual user. Then a personal recommendation algorithm is applied to suggest the top K most appealing items for the virtual user. And the preference score and fuzzy clustering algorithm based on multiclass are applied to optimize the recommendation result of the group recommendation model. Finally, the MovieLens-100K dataset is applied to verify the efficiency of the recommendation model. The experimental results show that the items recommended by the proposed group recommendation model are more popular for all the users in the group than the items recommended by traditional group recommendation algorithms.     

Space-time coding scheme for the paired weak user in MIMO-NOMA systems

In view of the paired weak user’s poor outage performance in multiple-input multiple-output non-orthogonal multiple access (MIMO-NOMA) systems,Alamouti code was adopted to encode for the weak user in order to improve its outage performance by means of diversity,and the closed-form expression of the strong user’s ergodic capacity as well as the boundary-form expressions of the weak user’s ergodic capacity and outage probability was derived in the proposed model.Moreover,a power allocation algorithm for optimizing the system’s throughput was proposed.Finally,the numerical results show the accuracy of the derived expressions,the efficacy of the proposed algorithm,and that the weak user’s outage performance in the proposed coding scheme is far superior to that in the current coding scheme only adopting vertical Bell lab layered space-time (V-BLAST) code.     

知识图谱与数据应用——智能推荐

知识图谱技术是新兴的互联网技术,具有较好的前瞻性和较强的技术先进性,有着广泛的应用前景。知识图谱技术提供从“关系”的角度分析问题的能力,可以对数据进行深度挖掘,将自然语言转化为计算机语言,最大限度地展示数据的价值,可以服务于智能搜索、智能推荐、风险预警、智能运营、智能客服、舆情监测、设备预警等业务,大幅地提高企业生产效率。研究了综合聚类算法、SVD分解算法、基于商品的协同顾虑推荐算法、基于用户的协同顾虑推荐算法和商品相似度算法等,结合推荐策略提出了基于知识图谱数据应用的智能推荐系统。证明了以知识图谱作为基础工具的数据应用的可实施性和可部署性,能够全面满足用户在互联网平台的获取信息的需求。