Sampling Object Determination for Mobile App Performance Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The evaluation of mobile application performance is challenged by fluctuations in user data due to different update orders and speeds across various operating systems and device models, leading to inaccurate statistical representations of new version performance.
Innovation Solution
A sampling object determination method that identifies representative users based on a sampling proportion and total user count, determining sampling objects from high to low version orders, ensuring accurate reflection of performance changes without being overwhelmed by historical version data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data from all versions is collected for evaluation, then comprehensive performance analysis is achieved, but statistical fluctuations increase due to mixed version data
Solution Approach 1:
The patent segments the user base by version order, dividing users into distinct groups (first version order, second version order, etc.) based on their application update sequences. This segmentation isolates performance data from different version adoption patterns, preventing statistical contamination between groups while maintaining comprehensive evaluation coverage across all versions.
Solution Approach 2:
The patent extracts and isolates performance data from specific version order groups, separating the data from users who updated in the first version order from those who updated in subsequent orders. This extraction allows independent analysis of each group's performance characteristics without the confounding effects of mixed version data.
2Measurement precision
If sampling is performed on latest version only, then new version performance is highlighted, but insufficient data volume leads to high statistical variance
Solution Approach 1:
The patent performs preliminary classification of users by version order before sampling. By pre-identifying users in the first version order group (who updated to the latest version first), the system ensures that sampling is targeted at the most relevant user群体 while maintaining sufficient data volume through the stratification process.
Solution Approach 2:
The patent samples from multiple version order groups (first, second, and subsequent orders) rather than limiting to only the latest version users. This partial expansion to multiple groups provides excessive data volume that can be analyzed to confirm consistent performance patterns across different adoption sequences.
3Adaptability or versatility
If evaluation includes all operating systems and device models, then comprehensive coverage is achieved, but update order differences cause data inconsistency
Solution Approach 1:
The patent applies local quality by treating different operating systems and device models with differentiated evaluation approaches. Specifically, it recognizes that Android and iOS systems have different update propagation patterns, and analyzes version order groups separately for each system type, allowing customized analysis strategies for different platform characteristics.
Solution Approach 2:
The patent changes the evaluation parameter from raw performance metrics to version order-based performance patterns. By transforming the data into version order groups (first order, second order, etc.), the system standardizes comparison across different operating systems and device models, making update timing differences irrelevant to the core performance evaluation.
Data Source
AI summary
Provided are a sampling object determination method and apparatus, an electronic device, and a computer-readable storage medium. The sampling object determination method includes: determining a sampling number through a sampling proportion and a total number of users of an application; determining at least one version in order of versions of the application from high to low, where a sum of a number of users of the at least one version is greater than or equal to the sampling number; and determining the sampling number of users in the users of the at least one version as sampling objects.