Multi-scenario recommender systems (MSRSs) have been increasingly used in real-world industrial platforms for their excellent advantages mitigating data sparsity and reducing maintenance costs. However, conventional MSRSs usually use all relevant features indiscriminately ignore that different kinds of varying importance under scenarios, which may cause confusion performance degradation. In addition, existing feature selection methods deep lack the exploration scenario relations. this paper,...

Click-through prediction (CTR) models transform features into latent vectors and enumerate possible feature interactions to improve performance based on the input set. Therefore, when selecting an optimal set, we should consider influence of both their interaction. However, most previous works focus either field selection or only select interaction fixed set produce The former restricts search space field, which is too coarse determine subtle features. They also do not filter useless...

Self-supervised learning (SSL) has recently achieved great success in mining the user-item interactions for collaborative filtering. As a major paradigm, contrastive (CL) based SSL helps address data sparsity Web platforms by contrasting embeddings between raw and augmented data. However, existing CL-based methods mostly focus on batch-wise way, failing to exploit potential regularity feature dimension. This leads redundant solutions during representation of users items. In this work, we...

Click-through rate (CTR) prediction model usually consists of three components: embedding table, feature interaction layer, and classifier. Learning table plays a fundamental role in CTR from the view performance memory usage. The is two-dimensional tensor, with its axes indicating number values dimension, respectively. To learn an efficient effective recent works either assign various dimensions for fields reduce embeddings respectively or mask parameters. However, all these existing cannot...

With the rapid development of mobile app ecosystem, apps have grown greatly popular. The explosive growth makes it difficult for users to find that meet their interests. Therefore, is necessary recommend user with a personalized set apps. However, one challenges data sparsity, as users’ historical behavior are usually insufficient. In fact, user’s behaviors from different domains in store regarding same relevant. we can alleviate sparsity using complementary information correlated domains....

As user behaviors become complicated on business platforms, online recommendations focus more how to touch the core conversions, which are highly related interests of platforms. These conversions usually continuous targets, such as watch time, revenue, and so on, whose predictions can be enhanced by previous discrete conversion actions. Therefore, multi-task learning (MTL) adopted paradigm learn these hybrid targets. However, existing works mainly emphasize investigating sequential...

Introduction The factors that significantly and negatively impact carbon dioxide (CO 2 ) emissions coastal water quality (CWQ) must be continuously monitored thoroughly evaluated. Among these, tourism (TR) volume stands out as one of the primary contributors to such effects. In contrast, green fiscal policy (GFP) fintech (FT) can considered proactive modern efforts contributing improvement these environmental indicators. Exploring whether impacts exhibit uniformity across quantiles will...

Click-through rate prediction is one of the core tasks in commercial recommender systems. It aims to predict prob-ability a user clicking particular item given and features. As feature interactions bring non-linearity, they are widely adopted improve performance CTR models. Therefore, effectively modelling has attracted much attention both research industry field. The current approaches can generally be categorized into three classes: (i) naïve methods, which do not model only use original...

Fraud behavior poses a severe threat to e-commerce platforms and anti-fraud systems have become indispensable infrastructure of these platforms. Recently, there been large number fraud detection models proposed monitor online purchasing transactions extract hidden patterns. Thanks models, we observed significant reduction committed frauds in the last several years. However, an increasing malicious sellers on platforms, according our recent statistics, who purposely circumvent by transferring...

Embedding tables are usually huge in click-through rate (CTR) prediction models. To train and deploy the CTR models efficiently economically, it is necessary to compress their embedding tables. this end, we formulate a novel quantization training paradigm embeddings from stage, termed low-precision (LPT). Also, provide theoretical analysis on its convergence. The results show that stochastic weight has faster convergence smaller error than deterministic LPT. Further, reduce accuracy...

In order to enhance market share and competitiveness, large banks are increasingly focusing on promoting marketing strategies. However, the traditional bank strategy often leads homogenization of customer demand, making it challenging distinguish among various products. To address this issue, paper presents a demand learning model based financial datasets optimizes distribution big data channels through induction rectify imbalance in transaction data. By comparing prediction models random...

Uplift modeling has been widely employed in online marketing by predicting the response difference between treatment and control groups, so as to identify sensitive individuals toward interventions like coupons or discounts. Compared with traditional conversion uplift modeling,revenue exhibits higher potential due its direct connection corporate income. However, previous works can hardly handle continuous long-tail distribution revenue modeling. Moreover, they have neglected optimize ranking...

Learning effective embedding has been proved to be useful in many real-world problems, such as recommender systems, search ranking and online advertisement. However, one of the challenges is data sparsity learning large-scale item embedding, users' historical behavior are usually lacking or insufficient an individual domain. In fact, user's behaviors from different domains regarding same items relevant. Therefore, we can learn complete user alleviate using complementary information...

Tabular data is one of the most common storage formats behind many real-world web applications such as retail, banking, and e-commerce. The success these largely depends on ability employed machine learning model to accurately distinguish influential features from all predetermined in tabular data. Intuitively, practical business scenarios, different instances should correspond sets features, set same instance may vary scenarios. However, existing methods focus global feature selection...

User Interface (UI) testing has become a common practice for quality assurance of industrial mobile applications (in short as apps). While many automated tools have been developed, they often do not satisfy two major requirements that make tool desirable in settings: high applicability across platforms (e.g., Android, iOS, AliOS, and Harmony OS) capability to handle apps with non-standard UI elements (whose internal structures cannot be acquired using platform APIs). Toward addressing these...

Embedding techniques have become essential components of large databases in the deep learning era. By encoding discrete entities, such as words, items, or graph nodes, into continuous vector spaces, embeddings facilitate more efficient storage, retrieval, and processing databases. Especially domain recommender systems, millions categorical features are encoded unique embedding vectors, which facilitates modeling similarities interactions among features. However, numerous vectors can result...

We present our work helping to adapt mobile apps be friendlier for elderly users. design actionable guidelines based on empirical investigations, shaping future practices of making a large number popular easier

As a key component in online marketing, uplift modeling aims to accurately capture the degree which different treatments motivate users, such as coupons or discounts, also known estimation of individual treatment effect (ITE). In an actual business scenario, options for may be numerous and complex, there correlations between treatments. addition, each marketing instance have rich user contextual features. However, existing methods still fall short both fully exploiting information mining...

Click-through Rate (CTR) prediction is essential for commercial recommender systems. Recently, to improve the accuracy, plenty of deep learning-based CTR models have been proposed, which are sensitive hyperparameters and difficult optimize well. General hyperparameter optimization methods fix these across entire model training repeat them multiple times. This trial-and-error process not only leads suboptimal performance but also requires non-trivial computation efforts. In this paper, we...