Modern e-commerce recommendation systems aim to improve customer experience by ranking content on search results page (SRP). However, displaying content is not always beneficial for customers across all contexts; even top-ranked content can be irrelevant, misleading, or redundant in certain scenarios. In this work, we propose a robust content suppression mechanism to selectively suppress content when necessary. Our approach leverages causal effect learning to measure the incremental value of showing versus suppressing content. Prior work uses Conditional Average Treatment Effect (CATE) to estimate treatment effect without considering the inherent uncertainty in uplift predictions, resulting in over-suppression and degraded customer experience. Our work introduces a novel Conformal Confidence (CoCo) suppressor that explicitly accounts for prediction uncertainty in uplift-based modeling. We first evaluate it on synthetic datasets with controlled noise, bias, and contexts. Results demonstrate superior performance compared to baseline approaches. Subsequently, our online traffic tests show statistically significant improvements in revenue and profit compared to existing methods.

E-commerce Search Results Pages (SRPs) are evolving from linear lists to complex, non-linear layouts, rendering traditional position-biased ranking models insufficient. Moreover, existing optimization frameworks typically maximize short-term signals (e.g., clicks, same-day revenue) because long-term satisfaction metrics (e.g., expected two-week revenue) involve delayed feedback and challenging long-horizon credit attribution. To bridge these gaps, we propose a novel Whole-Page Experience Optimization Framework. Unlike traditional list-wise rankers, our approach explicitly models the interplay between item relevance, 2D positional layout, and visual elements. We use a causal framework to develop metrics for measuring long-term user satisfaction based on quasi-experimental data. We validate our approach through industry-scale A/B testing, where the model demonstrated a 1.86% improvement in brand relevance (our primary customer experience metric) while simultaneously achieving a statistically significant revenue uplift of +0.05%.

Shared autonomous electric mobility has attracted significant interest in recent years due to its potential to save energy consumption, enhance mobility accessibility, reduce air pollution, mitigate traffic congestion, etc. Although providing convenient, low-cost, and environmentally-friendly mobility, there are still some roadblocks to achieve efficient shared autonomous electric mobility, e.g., how to enable the accessibility of shared autonomous electric vehicles in time. To overcome these roadblocks, in this article, we designSafari, an efficientSharedAutonomous electric vehicleFleet mAnagement system with jointRepositioning and chargIng based on dynamic deadlines to improve both user experience and operating profits. OurSafariconsiders not only the highly dynamic user demand forvehicle repositioning(i.e., where to relocate) but also many practical factors like the time-varying charging pricing forcharging scheduling(i.e., where to charge). To perform the two tasks efficiently, inSafari, we design a dynamic deadline-based deep reinforcement learning algorithm, which generates dynamic deadlines via usage prediction combined with an error compensation mechanism to adaptively learn the optimal decisions for satisfying highly dynamic and unbalanced user demand in real time. More importantly, we implement and evaluate theSafarisystem with 10-month real-world shared electric vehicle data, and the extensive experimental results show that ourSafariachieves 100% of accessibility and effectively reduces 26.2% of charging costs and reduces 31.8% of vehicle movements for energy saving with a small runtime overhead at the same time. Furthermore, the results also showSafarihas a great potential to achieve efficient and accessible shared autonomous electric mobility during its long-term expansion and evolution process.

The modern e-commerce web pages have brought better customer experience and more profitable services by whole page optimization at different granularity, e.g., page layout optimization, item ranking optimization, etc. Generating the proper page layout per customer’s request is one of the vital tasks during the web page rendering process, which can directly impact customers’ shopping experience and their decision-making. In this paper, we formulate the request-rendering interactions as a Markov decision process (MDP) and solve it by deep reinforcement learning (RL). Specifically, we present the design and implementation of applying offline Deep Q-Learning (DQN) to the contextual page layout optimization problem. Through the offline evaluation method, we demonstrate the effectiveness of the proposed framework, i.e., the RL agent has the potential to perform better than the baseline ranker by learning from the offline data set, e.g., the RL agent can improve the average cumulative rewards up to 36.69% comparing to the baseline ranker.

Electric carsharing, i.e., electric vehicle sharing, as an emerging mobility-on-demand service, has been proliferating worldwide recently. Though providing convenient, low-cost, and environmentally-friendly mobility, there are also some potential roadblocks in electric carsharing services due to existing inefficient fleet management strategies, which relocate the vehicles using predefined periodic schedules without self-adapting to the highly dynamic user demand, and many practical factors like time-variant charging pricing also have not been fully considered. To remedy these problems, in this paper, we design Record, an effective fleet management system with joint Repositioning and Charging for electric carsharing based on dynamic deadlines to improve its operating profits and also satisfy users' real-time pickup and return demand. Record considers not only the highly dynamic user demand for vehicle repositioning (i.e., where to relocate) but also the time-varying charging pricing for charging scheduling (i.e., where to charge). To perform the two tasks efficiently, in Record, we design a dynamic deadline-based distributed deep reinforcement learning algorithm, which generates dynamic deadlines via usage prediction combined with an error compensation mechanism to adaptively search and learn the optimal locations for satisfying highly dynamic and unbalanced user demand in real time. We implement and evaluate the Record system with 10-month real-world electric carsharing data, and the extensive experimental results show that our Record effectively reduces 25.8% of charging costs and reduces 30.2% of vehicle movements by workers, and it also satisfies user demand and achieves a small runtime overhead at the same time.

Urban traffic sensing has been investigated extensively by different real-time sensing approaches due to important applications such as navigation and emergency services. Basically, the existing traffic sensing approaches can be classified into two categories by sensing natures, i.e., explicit and implicit sensing. In this article, we design a measurement framework called EXIMIUS for a large-scale data-driven study to investigate the strengths and weaknesses of two sensing approaches by using two particular systems for traffic sensing as concrete examples. In our investigation, we utilize TB-level data from two systems: (i) GPS data from five thousand vehicles, (ii) signaling data from three million cellphone users, from the Chinese city Hefei. Our study adopts a widely used concept called crowdedness level to rigorously explore the impacts of contexts on traffic conditions including population density, region functions, road categories, rush hours, holidays, weather, and so on, based on various context data. We quantify the strengths and weaknesses of these two sensing approaches in different scenarios and then we explore the possibility of unifying two sensing approaches for better performance by using a truth discovery-based data fusion scheme. Our results provide a few valuable insights for urban sensing based on explicit and implicit data from transportation and telecommunication domains.

Recent research explores incorporating knowledge graphs (KG) into e-commerce recommender systems, not only to achieve better recommendation performance, but more importantly to generate explanations of why particular decisions are made. This can be achieved by explicit KG reasoning, where a model starts from a user node, sequentially determines the next step, and walks towards an item node of potential interest to the user. However, this is challenging due to the huge search space, unknown destination, and sparse signals over the KG, so informative and effective guidance is needed to achieve a satisfactory recommendation quality. To this end, we propose a CoArse-to-FinE neural symbolic reasoning approach (CAFE). It first generates user profiles as coarse sketches of user behaviors, which subsequently guide a path-finding process to derive reasoning paths for recommendations as fine-grained predictions. User profiles can capture prominent user behaviors from the history, and provide valuable signals about which kinds of path patterns are more likely to lead to potential items of interest for the user. To better exploit the user profiles, an improved path-finding algorithm called Profile-guided Path Reasoning (PPR) is also developed, which leverages an inventory of neural symbolic reasoning modules to effectively and efficiently find a batch of paths over a large-scale KG. We extensively experiment on four real-world benchmarks and observe substantial gains in the recommendation performance compared with state-of-the-art methods.

Mobile edge networks are pervasive now due to the ubiquitous 4G networks and coming 5G networks, broad edge computing applications are enabled in the meantime, such as mobile bus WiFi. In this paper, we focus on the caching problem in the mobile edge networks and use bus WiFi as an example to further investigate. Mobile bus WiFi is a newly emerged service in modern cities, which provides convenience for citizens and gains certain benefits for operators via commercial advertisements and other services. While the vital challenge for the bus WiFi industry is the high cost of cellular traffic considering the massive number of users and longtime running hours of the bus system. To tackle this, we investigate the caching problem in a nation-scale bus WiFi network deployed in 22 cities of China with 34,377 WiFi devices. We then delve a fundamental question, i.e., how can historical visiting records help us predict future visiting events and further save the cellular traffic by caching. In detail, we propose a Deep Neural Networks (DNN) based method by considering bus WiFi users' historical visits to cached contents to save cellular traffic data for WiFi providers. We implement our method via the city-scale bus WiFi data and compare with a series of state-of-the-art models, the results show that our method achieves the best performance.

Urban traffic sensing has been investigated extensively by different real-time sensing approaches due to important applications such as navigation and emergency services. Basically, the existing traffic sensing approaches can be classified into two categories, i.e., explicit and implicit sensing. In this paper, we design a measurement framework called EXIMIUS for a large-scale data-driven study to investigate the strengths and weaknesses of these two sensing approaches by using two particular systems for traffic sensing as concrete examples, i.e., a vehicular system as a crowdsourcing-based explicit sensing and a cellular system as an infrastructure-based implicit sensing. In our investigation, we utilize TB-level data from two systems: (i) vehicle GPS data from 3 thousand private cars and 2 thousand commercial vehicles, (ii) cellular signaling data from 3 million cellphone users, from the Chinese city Hefei. Our study adopts a widely-used concept called crowdedness level to rigorously explore the impacts of various spatiotemporal contexts on real-time traffic conditions including population density, region functions, road categories, rush hours, etc. based on a wide range of context data. We quantify the strengths and weaknesses of these two sensing approaches in different scenarios then we explore the possibility of unifying these two sensing approaches for better performance. Our results provide a few valuable insights for urban sensing based on explicit and implicit data from transportation and telecommunication domains.