A user can be represented as what he/she does along the history. common way to deal with modeling problem is manually extract all kinds of aggregated features over heterogeneous behaviors, which may fail fully represent data itself due limited human instinct. Recent works usually use RNN-based methods give an overall embedding a behavior sequence, then could exploited by downstream applications. However, this only preserve very information, or memories person. When application requires...

A user can be represented as what he/she does along the history. common way to deal with modeling problem is manually extract all kinds of aggregated features over heterogeneous behaviors, which may fail fully represent data itself due limited human instinct. Recent works usually use RNN-based methods give an overall embedding a behavior sequence, then could exploited by downstream applications. However, this only preserve very information, or memories person. When application requires...

Data-driven recommender systems that can help to predict users' preferences are deployed in many real online service platforms. Several studies show they vulnerable data poisoning attacks, and attackers have the ability mislead system perform as their desires. Considering realistic scenario, where is usually a black-box for complex algorithms may be them, how learn effective attack strategies on such still an under-explored problem. In this paper, we propose adaptive framework, PoisonRec,...

Product bundling, offering a combination of items to customers, is one the marketing strategies commonly used in online e-commerce and offline retailers. A high-quality bundle generalizes frequent interest, diversity across bundles boosts user-experience eventually increases transaction volume. In this paper, we formalize personalized list recommendation as structured prediction problem propose generation network (BGN), which decomposes into quality/diversity parts by determinantal point...

The aim of this study was to assess the ability acute physiology and chronic health evaluation II (APACHE II) score, poisoning severity score (PSS) as well sequential organ failure assessment (SOFA) combining with lactate (Lac) predict mortality in Emergency Department (ED) patients who were poisoned organophosphate. A retrospective review 59 stands-compliant carried out. Receiver operating characteristic (ROC) curves constructed based on APACHE PSS, SOFA or without Lac, respectively, areas...

This paper concentrates on group anomalies in general large-scale networks. Existing algorithms mainly focus homogeneous or bipartite networks, and thus are difficult to apply heterogeneous networks directly. Moreover, these follow the non-overlapping hypothesis of groups implicitly, which is improper many scenarios. For example, fraud users Alibaba E-commerce platform may join more than one organization at same time. In this paper, we introduce a novel algorithm called <italic...

Fraud sellers in e-commerce usually promote their products via fake transactions. Such behaviors damage the reputation of platform and jeopardize business environment platform. The search engine existing platforms mainly focuses on generating transactions by matching users' queries sellers' products. most common method to defense fraud is set up a blacklist based detection manual investigation, then punish those list, which inefficient can only cover small fraction potential sellers. In this...

SimRank is an important measure of vertex-pair similarity according to the structure graphs. Although progress has been achieved, existing methods still face challenges handle large Besides huge index construction and maintenance cost, may require considerable search space time overheads in online query. In this paper, we design a Monte Carlo based method, UniWalk, enable fast top-k computation over undirected UniWalk directly locates similar vertices for any single source vertex u via R...

Text retrieval has been widely-used in many online applications to help users find relevant information from a text collection. In this paper, we study new attack scenario against evaluate its robustness adversarial attacks under the black-box setting, which attackers want their own texts always get high relevance scores with different users' input queries and thus be retrieved frequently can receive large amounts of impressions for profits. Considering that most current methods only simply...

Impression regulation plays an important role in various online ranking systems, e.g. , e-commerce systems always need to achieve local commercial demands on some pre-labeled target items like fresh item cultivation and fraudulent counteracting while maximizing its global revenue. However, impression may cause “butterfly effects” the scale, e-commerce, price preference fluctuation initial conditions (overpriced or underpriced items) create a significantly different outcome, thus affecting...

Existing recommender systems usually generate personalized recommendation lists based on the estimation of preference scores over user-item pairs, while ignoring impacts entire display list that plays a central part in decision making process user. This leaves us an opportunity to better results by considering all offered choices. However, such extension cannot be handled efficiently traditional top-k methods, due dependency issue which means complete items is needed before we can precisely...