Personalized Task-Time Estimates Through Quantile Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing software applications fail to provide personalized estimates of time needed to complete a task, such as filling out electronic documents, as they do not account for user-specific preferences for aggressive or conservative estimates, leading to inefficient resource utilization and inadequate interventions.
Innovation Solution
A machine learning-based approach that utilizes multiple time-to-complete models trained at different quantile levels, a recommender model selects an appropriate model based on user features, and adjusts its recommendations using feedback data to provide personalized estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single average time estimate is provided to all users, then the estimation process is simple and fast, but the estimate does not account for individual user preferences leading to inaccurate recommendations
Solution Approach 1:
The patent segments the user base into different groups based on their preferences for time estimates (aggressive vs. conservative). Multiple TTC models are trained at different quantile levels to serve different user segments. The machine learning model selects the appropriate segment and corresponding model for each user based on their characteristics, thereby providing accurate personalized estimates without requiring a completely complex system redesign.
Solution Approach 2:
The patent changes the parameter of quantile level in the TTC models to generate different types of time estimates (aggressive, average, conservative). By training multiple models at different quantile levels and selecting the appropriate one based on user preferences, the system achieves high measurement precision while managing complexity through parameter-based differentiation rather than fundamentally different model architectures.
2Measurement precision
If multiple TTC models are trained and selected based on user preferences, then personalized accurate estimates are provided, but processor resources are increased
Solution Approach 1:
The patent performs preliminary action by pre-training multiple TTC models at different quantile levels before deployment. The machine learning model is also pre-trained to predict user preferences and select appropriate models. During actual operation, the system only needs to execute the selection process and run the selected model, rather than training multiple models on-the-fly, thereby reducing real-time processor resource consumption while maintaining personalization accuracy.
Solution Approach 2:
The patent creates multiple copies of TTC models trained at different quantile levels (aggressive, average, conservative). Instead of using one complex model that tries to handle all cases, the system uses simpler model copies that are optimized for specific preference types. The machine learning selector routes users to the appropriate model copy, reducing the computational burden on any single model and optimizing overall processor resource usage.
3Productivity
If interventions are provided based on average estimates, then resource allocation is simplified, but unnecessary or inadequate interventions occur
Solution Approach 1:
The patent implements feedback mechanisms where user responses to time estimates and interventions are collected and used to refine the machine learning model's ability to predict user preferences. This feedback loop allows the system to learn from actual user behavior and improve the accuracy of personalized estimates over time, thereby reducing unnecessary interventions and improving intervention efficiency without losing user preference information.
Data Source
AI summary
A method for automatically recommending personalized estimates of amounts of time needed to complete a task include includes providing a plurality time-to-complete (TTC) models trained at different quantile levels to a machine learning model configured to select one TTC model of the plurality of TTC models as a selected TTC model to generate an estimated amount of time needed for a current user to complete the task. The method includes obtaining the estimated amount of time from the selected TTC model. The method includes obtaining feedback data regarding the estimated amount of time obtained from the selected TTC model. The method includes training the machine learning model based, at least in part, on the feedback data.