Predicting User Visits via Time-Sensitive Intervention Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current models for predicting user engagement on social networks, such as click-through rate models, are not time-sensitive and fail to determine the effectiveness of interventions, leading to challenges in deciding when and how to send notifications to maximize user visits and engagement.
Innovation Solution
The development of a pVisit model using an accelerated failure time model constrained by a Weibull distribution to predict the probability of user visits, considering historical session counts, notification counts, app installation, and last visit dates, to determine the optimal timing and type of interventions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional click-through rate models are used to predict user engagement, then the prediction process is simple, but the models are not time-sensitive and fail to determine the effectiveness of interventions
Solution Approach 1:
The patent transforms the prediction model from a static click-through rate model to a dynamic time-sensitive model by introducing time parameters and using accelerated failure time models with Weibull distribution. This allows the model to capture temporal patterns in user behavior and intervention effectiveness while maintaining mathematical tractability through parametric assumptions.
2Productivity
If interventions are sent frequently to maximize user visits, then user engagement increases, but user annoyance and churn risk increase
Solution Approach 1:
The patent applies preliminary action by predicting the probability of user visit before sending interventions. The system uses the accelerated failure time model to estimate when users are likely to visit naturally, and only sends interventions when the predicted probability is low, thereby preventing unnecessary notifications that would annoy users while still maximizing engagement when interventions are sent.
Solution Approach 2:
The patent implements feedback by continuously updating the prediction model with actual user visit data and intervention responses. The model learns from past user behavior patterns and intervention effectiveness, adjusting future intervention timing and targeting to balance engagement maximization with user annoyance minimization based on observed user responses.
3Measurement precision
If detailed user behavior data is collected to improve prediction accuracy, then the precision of visit probability prediction increases, but data privacy concerns and processing complexity increase
Solution Approach 1:
The patent transforms complex user behavior data into standardized parametric forms suitable for accelerated failure time modeling. By assuming Weibull distribution for visit times and using parametric regression models, the system can process detailed user behavior data through a unified mathematical framework that balances prediction precision with computational efficiency and data management simplicity.
Data Source
AI summary
A method can include determining a first probability that a first member of members of a website will visit the website within a specified time window if the first member is provided an intervention at a specified time, determining a second probability that the first member will visit the website within the specified time window without being provided the intervention, determining a difference between the first and second probability, and in response to determining the difference is greater than a first specified threshold, providing the intervention at the specified time.