Pre-Reservation Access Prediction Using Segmented Sub-Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting pre-reservation access size for game launches are either inaccurate or complex, with methodologies using average conversion rates lacking precision and model-based approaches being difficult to train and requiring extensive time, especially when dealing with games dissimilar to previously launched titles.
Innovation Solution
A method and device that categorize pre-reservation information into sub-information groups and generate individual sub-models using various variable combinations to determine pre-reservation access probability, allowing for efficient prediction by selecting the appropriate sub-model for collected information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a personalized prediction method using cumulative personal information is used, then prediction accuracy is improved, but model complexity and training time increase significantly
Solution Approach 1:
The patent segments pre-reservation users into distinct groups based on their reservation behavior patterns (e.g., early reservers, late reservers, frequent reservers). By dividing the user base into segments and creating specific prediction models for each segment rather than using a single complex model for all users, the system achieves high prediction accuracy while reducing overall model complexity. Each segment's model can be trained independently with targeted features relevant to that specific user group.
2Measurement precision
If a complex prediction model is trained to achieve high accuracy, then prediction precision is improved, but training time and computational resources increase
Solution Approach 1:
By segmenting users into behavior-based groups, the patent reduces the training time required for each individual model compared to training one large complex model on all user data. Each segmented model can be trained more quickly with focused features relevant to that specific user segment, while the collective accuracy across all segments matches or exceeds what a single comprehensive model would achieve.
3Productivity
If average conversion rate method is used, then calculation speed is improved, but prediction accuracy decreases
Solution Approach 1:
Instead of using a single average conversion rate for all users, the patent segments users into behavior-based groups and calculates conversion rates specific to each segment. This allows the system to maintain fast calculations (similar to the average rate method) while significantly improving accuracy by applying the appropriate segment-specific conversion rate to each user group rather than using a generic average.
Solution Approach 2:
The patent applies different conversion rate characteristics to different user segments based on their specific behaviors and patterns. Rather than using a uniform average conversion rate across all users, each segment receives a locally optimized conversion rate that reflects their specific reservation tendencies, thereby improving overall prediction accuracy while maintaining computational efficiency.
Data Source
AI summary
Disclosed is a method for predicting a pre-reservation access size performed by a computing device. The method may include: obtaining a plurality of sub-information groups from first pre-reservation information for at least one first game based on a plurality of pre-reservation variable combinations; generating a prediction model group unit including a plurality of sub-prediction models based on the plurality of sub-information groups, wherein each of the plurality of the sub-prediction models is related to at least one of the plurality of sub-information groups; and determining a pre-reservation access probability of a second game by using at least one sub-prediction model corresponding to second pre-reservation information for the second game in the prediction model group unit.