Recommender systems are seen as an effective tool to address information overload, but it is widely known that the presence of various biases makes direct training on large-scale observational data result in sub-optimal prediction performance. In contrast, unbiased ratings obtained from randomized controlled trials or A/B tests considered be golden standard, costly and small scale reality. To exploit both types data, recent works proposed use correct parameters propensity imputation models...

Data-driven recommender systems have demonstrated great success in various Web applications owing to the extraordinary ability of machine learning models recognize patterns (ie correlation) from users' behaviors. However, they still suffer several issues such as biases and unfairness due spurious correlations. Considering causal mechanism behind data can avoid influences In this light, embracing modeling is an exciting promising direction.

Recommender systems should answer the intervention question "if recommending an item to a user, what would feedback be", calling for estimating causal effect of recommendation on user feedback. Generally, this requires blocking confounders that simultaneously affect and To mitigate confounding bias, strategy is incorporating propensity into model learning. However, existing methods forgo possible unmeasured (e.g., financial status), which can result in biased propensities hurt performance....

Post-click conversion rate (CVR) prediction is an essential task for discovering user interests and increasing platform revenues in a range of industrial applications. One the most challenging problems this existence severe selection bias caused by inherent self-selection behavior users item process systems. Currently, doubly robust (DR) learning approaches achieve state-of-the-art performance debiasing CVR prediction. However, paper, theoretically analyzing bias, variance generalization...

Counterfactual inference aims to estimate the counterfactual outcome at individual level given knowledge of an observed treatment and factual outcome, with broad applications in fields such as epidemiology, econometrics, management science. Previous methods rely on a known structural causal model (SCM) or assume homogeneity exogenous variable strict monotonicity between variable. In this paper, we propose principled approach for identifying estimating outcome. We first introduce simple...

Effective personalized incentives can improve user experience and increase platform revenue, resulting in a win-win situation between users e-commerce companies. Previous studies have used uplift modeling methods to estimate the conditional average treatment effects of users' incentives, then placed by maximizing sum estimated under limited budget. However, some will always buy whether are given or not, they actively collect use if provided, named "Always Buyers". Identifying predicting...

Current guidelines for treatment decision making largely rely on data from randomized controlled trials (RCTs) studying average effects. They may be inadequate to make individualized decisions in real-world settings. Large-scale electronic health records (EHR) provide opportunities fulfill the goals of personalized medicine and learn rules (ITRs) depending patient-specific characteristics patient data. In this work, we tackle challenges with EHRs propose a machine learning approach based...

Recommender systems learn personalized user preferences from feedback like clicks.However, is usually biased towards partially observed interests, leaving many users' hidden interests unexplored.Existing approaches typically mitigate the bias, increase recommendation diversity, or use bandit algorithms to balance exploration-exploitation trade-offs.Nevertheless, they fail consider potential rewards of recommending different categories items and lack global scheduling allocating top-...

In evolutionary multi-objective optimization, it has been illuminated that guide search with neighboring solutions improve the quality of new trial and accelerate algorithms convergence by regularity property continuous optimization problems (MOPs). Very recently, clustering learning-based mating strategies have popular for establishing reproduction operators solutions. However, existing may be more reasonable full consideration utilization property. The current restrictions excessively...

Understanding the posttreatment prognosis of skin lesions in patients with psoriasis is essential for improving patients' treatment satisfaction.To model after 3 types therapy.This prospective cohort study included who visited a dermatologist and were enrolled platform Psoriasis Standardized Diagnosis Treatment Center China from August 2020 to December 2021.Biologic, traditional, systemic therapy psoriasis.Skin measured by Investigator's Global Assessment (IGA) scale subsumed into 4 stages...

There is growing interest in exploring causal effects target populations via data combination. However, most approaches are tailored to specific settings and lack comprehensive comparative analyses. In this article, we focus on a typical scenario involving source dataset dataset. We first design six under covariate shift conduct analysis by deriving the semiparametric efficiency bounds for ATE population. then extend new that incorporate both posterior drift. Our study uncovers key factors...

Modern learnable collaborative filtering recommendation models generate user and item representations by deep learning methods (e.g. graph neural networks) for modeling user-item interactions. However, most of them may still have unsatisfied performances due to two issues. Firstly, some assume that the users or items are fixed when interactions with different objects. a interests in items, an also attractions users. Thus should depend on their contexts extent. Secondly, existing learn...

Motivated by the testing of genetic pleiotropy, we discuss a general class hypothesis testing, exclusive test (EHT). A is an EHT if null can be divided into set sub‐hypotheses, and main difficulty for calculation p ‐value. To address this problem, propose weighted procedure develop two methods, one likelihood‐based other Bayesian information criterion (BIC)‐based, determining corresponding weights. Furthermore, show that BIC‐based method control asymptotic type I error. We conduct extensive...

Understanding how treatment effects vary on several key characteristics is critical in the practice of personalized medicine.To estimate these conditional average effects, non-parametric estimation often desirable, but few methods are available due to computational difficulty.Existing such as inverse probability weighting have limitations that hinder their use many practical settings where values propensity scores close 0 or 1.We propose score regression (PSR) allows a wide context.PSR...

In time to event data analysis, it is often of interest predict quantities such as <i>t</i>-year survival rate or the function over a continuum time. A commonly used approach relate covariates by semiparametric regression model and then use fitted for prediction, which usually results in direct estimation conditional hazard estimating equation. Its prediction accuracy, however, relies on correct specification covariate-survival association difficult practice, especially when patient...

In biomedical studies, estimating drug effects on chronic diseases requires a long follow-up period, which is difficult to meet in randomized clinical trials (RCTs).The use of short-term surrogate replace the long-term outcome for assessing effect relies stringent assumptions that empirical studies often fail satisfy.Motivated by kidney disease study, we investigate outcomes combining an RCT without observation and observational study observed but unmeasured confounding may exist.Under mean...