AI/ML & Recommendation Systems
Revolutionizing Shared-Account Recommendations with Dynamic Latent User Inference
Shared-account usage is common on streaming and e-commerce platforms, where multiple users share one account. Existing shared-account sequential recommendation (SSR) methods often assume a fixed number of latent users per account, limiting their ability to adapt to diverse sharing patterns and reducing recommendation accuracy. Recent latent reasoning technique applied in sequential recommendation (SR) generate intermediate embeddings from the user embedding (e.g, last item embedding) to uncover users' potential interests, which inspires us to treat the problem of inferring the number of latent users as generating a series of intermediate embeddings, shifting from inferring preferences behind user to inferring the users behind account. However, the last item cannot be directly used for reasoning in SSR, as it can only represent the behavior of the most recent latent user, rather than the collective behavior of the entire account. To address this, we propose DisenReason, a two-stage reasoning method tailored to SSR. DisenReason combines behavior disentanglement stage from frequency-domain perspective to create a collective and unified account behavior representation, which serves as a pivot for latent user reasoning stage to infer the number of users behind the account. Experiments on four benchmark datasets show that DisenReason consistently outperforms all state-of-the-art baselines across four benchmark datasets, achieving relative improvements of up to 12.56% in MRR@5 and 6.06% in Recall@20.
Executive Impact & Key Findings
DisenReason introduces a novel two-stage reasoning framework that significantly advances Shared-Account Sequential Recommendation (SSR) by adaptively inferring the dynamic number of latent users behind a shared account. By moving beyond the static assumption of a fixed user count and leveraging frequency-domain behavior disentanglement and progressive residual reasoning, DisenReason achieves superior predictive accuracy. Its ability to construct a unified account-level representation from mixed interaction sequences, then iteratively uncover individual user preferences, leads to more personalized and effective recommendations in complex multi-user environments. This framework addresses a critical gap in existing SR and SSR models, which often struggle with the heterogeneity of shared accounts and the limitations of last-item-based reasoning.
0%
MRR@5 Improvement
0%
Recall@20 Improvement
0
Benchmark Datasets Tested
0
Novel Reasoning Framework
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Existing Shared-Account Sequential Recommendation (SSR) methods often assume a fixed number of latent users per account, which is misaligned with real usage patterns where user counts vary dynamically. Conventional latent reasoning paradigms rely on the last item as the reasoning pivot, capturing only the most recent user's intent, not the collective behavior of the entire account. This limits adaptability and recommendation accuracy in multi-user scenarios.
DisenReason employs a novel two-stage reasoning framework. Stage one, 'Behavior Disentanglement for Pivot,' uses Fast Fourier Transform (FFT) to transform the mixed interaction sequence into the frequency domain, disentangling distinct behavioral patterns. These patterns are adaptively fused using a Mixture-of-Experts scheme to create a unified, collective account behavior representation (the 'pivot'). Stage two, 'Progressive Residual Reasoning for Latent User,' leverages this pivot to sequentially uncover latent users. At each step, an intermediate embedding represents an identified user, which is then removed from the pivot using a residual technique to prevent redundant reasoning. This iterative process continues until consecutive inferred users become semantically similar, adaptively determining the total number of latent users.
Experiments on four benchmark datasets (HvideoE, HvideoV, HamazonM, HamazonB) consistently show DisenReason outperforms all state-of-the-art baselines. It achieves relative improvements of up to 12.56% in MRR@5 and 6.06% in Recall@20. The model demonstrates superior stability and robustness across varying sequence lengths and training data proportions, and effectively infers the dynamic number of latent users per account, adapting to diverse sharing patterns.
Theoretically, DisenReason advances SSR by introducing dynamic latent user inference and a novel two-stage reasoning framework. Practically, it offers a scalable, robust, and user-centric recommendation service for real-world shared-account scenarios like streaming media and e-commerce. Its adaptive nature addresses the variability in user counts, leading to more personalized recommendations without explicit user identification or manual configuration, thus enabling practitioners to build more effective multi-user platforms.
12.56%
Relative MRR@5 Improvement in Shared-Account Recommendation
DisenReason achieves a significant boost in recommendation accuracy, outperforming existing baselines by up to 12.56% in MRR@5. This highlights its effectiveness in predicting relevant items within complex shared-account environments.
Enterprise Process Flow
Behavior Disentanglement for Pivot
→
Frequency-Domain Transformation (FFT)
→
Adaptive Fusion of Behavioral Patterns
→
Progressive Residual Reasoning for Latent User
→
Iterative User Uncovering & Residual Subtraction
→
Adaptive Termination
DisenReason's innovative two-stage reasoning framework ensures comprehensive account representation and dynamic user inference, addressing the core challenges of shared-account sequential recommendation.
| Component | Contribution to Performance |
|---|---|
| LightGCN |
|
| Behavior Disentanglement |
|
| Adaptive Fusion |
|
| Residual Operation |
|
An ablation study confirms the critical role of each component in DisenReason. Behavior disentanglement, especially, is vital for constructing a robust account-level pivot, driving the most significant performance gains.
Adaptive Latent User Inference
Our in-depth analysis of accounts from HV-E and HV-V datasets (e.g., IDs 10293, 5673 vs. 6304, 2643) visually demonstrates DisenReason's ability to adaptively infer the dynamic number of latent users. Unlike fixed-assumption models, DisenReason successfully distinguishes varying complexities in shared-account scenarios, from two siblings to several household members, based on distinct interaction patterns. This adaptability is crucial for real-world shared accounts where user counts are not predefined, validating the model's robustness and practical utility.
DisenReason's ability to dynamically infer the correct number of latent users is a game-changer for real-world shared-account scenarios, moving beyond rigid assumptions to truly reflect user heterogeneity.
Calculate Your Potential ROI
Estimate the significant efficiency gains and cost savings your enterprise could achieve by implementing DisenReason's advanced AI recommendation systems.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your Enterprise AI Implementation Roadmap
A typical phased approach to integrate DisenReason's capabilities into your existing infrastructure.
Phase 1: Discovery & Strategy
Comprehensive assessment of current recommendation systems, data infrastructure, and business objectives. Develop a tailored AI strategy and roadmap aligned with your enterprise goals.
Phase 2: Data Integration & Model Training
Secure and efficient integration of user behavior data. Train DisenReason models, focusing on behavior disentanglement and latent user reasoning to optimize for your specific shared-account scenarios.
Phase 3: System Deployment & Optimization
Seamless deployment of the DisenReason framework into your production environment. Continuous monitoring, A/B testing, and iterative optimization to maximize recommendation accuracy and user satisfaction.
Phase 4: Performance Monitoring & Scaling
Establish robust monitoring and reporting dashboards. Scale the solution across diverse platforms and user segments, ensuring sustained performance and adaptive inference of latent users as your business evolves.
Ready to Transform Your Recommendations?
DisenReason offers a unique advantage in handling complex shared-account behaviors. Connect with our AI specialists to explore how this innovative framework can drive more personalized and accurate recommendations for your enterprise.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat