Optimized OCPC strategy improves Taobao's ad traffic allocation efficiency.
problem Inefficient bid matching between advertisers and traffic quality.
method Proposed OCPC strategy to automatically adjust bids for finer matching.
result Substantially better results compared to fixed bid methods in production tests.
TPG-DNN predicts user intent using GRU loss and multi-task learning.
problem Improving user experience and shopping efficiency in e-commerce.
method Adaptive GRU loss function with multi-task learning.
result TPG-DNN outperforms existing CTR models on Taobao datasets.
Deep User Perception Network learns universal user representations from multiple e-commerce tasks.
problem Lack of shared user information across diverse e-commerce tasks.
method Model user behavior sequences using LSTM and attention mechanism, sharing user representations across multiple tasks.
result Our approach consistently achieves better results in personalization across multiple e-commerce tasks.
Time-based model improves recommendation performance.
problem Efficiently handling sequences of varying lengths in recommendation systems.
method Embedding layer and TSL attention-like mechanism with inner products.
result Model outperforms complex models on Taobao User Behavior dataset.
ComiRec framework predicts user interests for personalized recommendations.
problem Predicting user interests from sequential behavior data.
method ComiRec framework captures multiple user interests and balances recommendation accuracy and diversity.
result ComiRec achieves significant improvements over state-of-the-art models in sequential recommendation.
Adaptive spectral RL method enhances RL performance and interpretability.
problem Balancing interpretability and performance in reinforcement learning.
method Spectral based linear RL approach with adaptive regularization.
result Near-optimal bounds for parameter estimation and generalization error.
Paper tackles cold-start problems in online recommendation with few-shot learning and meta learning.
problem Cold-start problems in practical recommendations with limited interaction data.
method Combines scenario-specific learning with sequential meta-learning to create an integrated end-to-end framework.
result Significant gains over state-of-the-arts for cold-start problems in online recommendation.
DBMTL optimizes multiple e-commerce targets using Bayesian networks.
problem Multi-target optimization in e-commerce platforms.
method Bayesian modeling of target events in a Bayesian network, with hidden layers parameterized by neural networks.
result Significant improvement in recommendation performance over other methods.
DSPN predicts advertiser satisfaction and intent for e-commerce platforms.
problem Understanding advertiser intent and satisfaction for e-commerce platforms.
method Two-stage Deep Satisfaction Prediction Network (DSPN) that models intent and satisfaction.
result DSPN outperforms state-of-the-art baselines and predicts advertiser satisfaction accurately.
Paper tackles exposure bias in recommender systems using contrastive learning.
problem Exposure bias in large-scale recommender systems.
method Contrastive learning to reduce exposure bias via inverse propensity weighting.
result Contrastive learning effectively reduces exposure bias in recommender systems.
Adaptive display exposure improves e-commerce platform revenue.
problem Optimize ad display for better revenue and user experience.
method Constrained Markov Decision Process (psCMDP) and two-level reinforcement learning.
result Our approach increases platform revenue under constraints.
ESMM models CVR over entire space, overcoming sample selection bias and data sparsity.
problem Sample selection bias and data sparsity in CVR modeling.
method Entire Space Multi-task Model (ESMM) using sequential pattern of user actions.
result ESMM significantly outperforms competitive methods on Taobao dataset.
Optimizes auction mechanisms in e-commerce search ads to balance revenue and user experience.
problem Optimizing auction mechanisms in e-commerce search ads while maintaining quality users and ROI.
method Developed a practical convex optimization formulation and auction simulation system to estimate business indicators.
result Proper entropy regularization can maximize revenue while constraining other business indicators.
Secure submodel learning protects privacy in federated learning.
problem Efficiency and privacy in federated learning for resource-constrained clients.
method Designing a secure federated submodel learning scheme with randomized response, secure aggregation, and Bloom filter.
result Demonstrated the feasibility and scalability of the scheme with practical evaluations.
DIEN predicts CTR by evolving user interests from behavior data.
problem Capturing dynamic user interests for accurate CTR prediction.
method DIEN captures temporal and evolving user interests using interest extractor and evolving layers with attention mechanisms.
result DIEN significantly outperforms state-of-the-art solutions in CTR prediction.
Designs efficient factorial experiments for product design under budget constraints.
problem Designing effective experiments for product design with limited traffic and overlapping experiments.
method Two-stage design: first stage samples and infers performance, second stage selects a final policy.
result The method outperforms one-shot tensor completion and unstructured best-arm benchmarks.
FIVES generates high-order interactive features efficiently and effectively.
problem Automating the generation of high-order interactive features in tabular data.
method Formulates interactive feature generation as edge search on a feature graph, using a GNN and adjacency tensor.
result FIVES outperforms state-of-the-art methods in various datasets and real-world applications.
Proposes a tree-based method to efficiently predict user interests in large recommender systems.
problem Efficiently predicting user-item preferences in large recommender systems with high calculation costs.
method Predicts user interests from coarse to fine using a tree structure, which can incorporate deep neural networks.
result Significantly outperforms traditional methods in both training and prediction.
This paper optimizes search experiences in two-sided marketplaces by balancing multiple conflicting metrics.
problem Balancing conflicting business metrics in two-sided marketplaces like eBay and Etsy.
method Joint optimization of market-level metrics using Evolutionary Strategies.
result Demonstrated effectiveness of the proposed method on Etsy data.
RNE tackles scalable recommendation for billion-scale scenarios.
problem Designing a scalable recommendation system for diverse and dynamic user interests.
method RNE uses a diversity- and dynamics-aware neighbor sampling method for scalable network embedding.
result RNE achieves high-quality and diverse results on a billion-scale user-item graph.