In this paper, we introduce a new setting for graph embedding, which considers embedding communities instead of individual nodes. We find that community is not only useful community-level applications such as visualization but also provide an exciting opportunity to improve detection and node classification. Specifically, consider the interaction between closed loop, through embedding. On one hand, can since detected are used fit other be optimize by introducing community-aware high-order...

As one of the most representative social media platforms, Twitter provides various real-life information on events in real time. Despite that event detection has been actively studied, tweet images, which appear around 36 percent total tweets, have not well utilized for this research problem. Most existing methods tend to represent an image as a bag-of-visual-words and then process these visual words same way textual words. This may fully exploit properties images. State-of-the-art features...

Most existing community-related studies focus on detection, which aim to find the community membership for each user from friendship links. However, alone, without a complete profile of what is and how it interacts with other communities, has limited applications. This motivates us consider systematically profiling communities thereby developing useful community-level In this paper, we first time formalize concept profiling. With rich information network, such as published content diffusion...

Tweet streams provide a variety of real-life and real-time information on social events that dynamically change over time. Although event detection has been actively studied, how to efficiently monitor evolving from continuous tweet remains open challenging. One common approach for text is use single-pass incremental clustering. However, this does not track the evolution events, nor it address issue efficient monitoring in presence large number events. In paper, we capture dynamics using...

Tweet streams provide a variety of real-time information on dynamic social events. Although event detection has been actively studied, most the existing approaches do not address issue efficient monitoring in presence large number events detected from continuous tweet streams. In this paper, we capture dynamics using four operations: creation, absorption, split and merge.We also propose novel indexing structure, named Multi-layer Inverted List (MIL), for acceleration large-scale search...

Various detection methods have been proposed for defense against group shilling attacks in recommender systems; however, these cannot effectively detect attack groups generated based on adversarial (e.g., GOAT) or mixed groups. In this study, we propose a two-stage method, called KC-GCN, which is <math xmlns="http://www.w3.org/1998/Math/MathML" id="M1"> <mi>k</mi> </math> -cliques and graph convolutional networks. First, construct user relationship graph, generate suspicious candidate...

Abstract At present, most of the personalized sequential recommendations utilize users’ implicit positive feedback (such as clicks) to predict user behavior, ignoring impact negative and explicit on accuracy recommendation results prediction. In this paper, we propose a robust sequence model based multi behavior denoising trusted neighbors, which utilizes multiple data for feature considers nearest neighbor information improve performance. Firstly, by learning representations interactions...

To solve the subjective and uncertainly about trust in e-commerce system, a new model based on multidimensional cloud is proposed, which basis of past research model. History evaluation data regard as quantitative universe discourse, entities' property weighted backward generation algorithm, so history behavior an entity can be reflected very well through three numerical characteristics The results experiments show that proposed paper reflect seller's more accurately provide favorable...

Directly publishing the original data of social networks may compromise personal privacy because relationship contain sensitive information about users. To protect relationships against inference attacks and achieve trade-off between utility, we propose a protection algorithm that combines friendship links central nodes (PPCN) in dynamic network. In preparation work, design two indices for user influence based on characteristics can identify (Definition 1) Operating effectively improve...

To protect recommender systems against shilling attacks, a variety of detection methods have been proposed over the past decade. However, these focus mainly on individual features and rarely consider lockstep behaviours among attack users, which suffer from low precision in detecting group attacks. In this work, we propose three-stage method based strong members behaviour for First, construct weighted user relationship graph by combining direct indirect collusive degrees between users....

With the rapid growth in popularity of social websites, event detection has become one hottest research topics. However, continuously monitoring events not been well studied. In this demo, we present a novel system called EventEye to effectively monitor evolving and visualize their paths, which are discovered from tweet streams. particular, four operations defined for our proposed stream clustering algorithm capture evolutions over time multi-layer indexing structure is designed support...

Community analysis is an important task in graph mining. Most of the existing community studies are detection, which aim to find membership for each user based on friendship links. However, alone, without a complete profile what and how it interacts with other communities, has limited applications. This motivates us consider systematically profiling communities thereby developing useful community-level In this paper, we introduce novel concept profiling, upon build SocialLens system1 enable...

In the open distributed environment, knowledge of belief is absent for entity with which we will contact, so recommendation trust very important system. this paper a model proposed based on encouragement and punishment. model, value any consists trading reputation value, factor affected by multi parameters can adjust dynamically recommend result. When recommender supports real information then increase otherwise decrease value.

Sequential recommendation can make predictions by fitting users’ changing interests based on the continuous historical behavior sequences. Currently, many existing sequential methods put more emphasis upon recent preference (i.e., short-term interests), but simplify or even ignore influence of long-term interests, resulting in important interest features users not being effectively mined. Moreover, real intentions may be fully captured only focusing their histories, because are diverse and...

Research on social networks is at its peak in the current era of big data, especially field computer research. Link prediction has attracted an increasing number researchers. However, most studies have focused visible relationships between users, ignoring existence invisible relationships. The same as relationships, are also indispensable part networks, and they can uncover more potential users. To better understand relationship, definition, types, characteristics relationship been...

Friend recall is an important way to improve Daily Active Users (DAU) in online games. The problem generate a proper lost friend ranking list essentially. Traditional methods focus on rules like intimacy or training classifier for predicting players' return probability, but ignore feature information of (active) players and historical events. In this work, we treat as link prediction explore several which can use features both active players, well Furthermore, propose novel Edge Transformer...

Most existing community-related studies focus on detection, which aim to find the community membership for each user from friendship links. However, alone, without a complete profile of what is and how it interacts with other communities, has limited applications. This motivates us consider systematically profiling communities thereby developing useful community-level In this paper, we first time formalize concept profiling. With rich information network, such as published content diffusion...

The researches of social networks are in full swing under the era big data, especially field computer research, and link prediction has attracted an increasing number researchers. However, most current studies had focused on visible relationship between users, ignoring existence invisible relationships. same as relationships, relationships also indispensable part networks, can discover more potential users. In order to do a better research relationship, definition, types characteristics have...