ML Regression Model for ALM Work Item Time Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating the aggregate time to complete work items in Application Lifecycle Management (ALM) systems are either manual and inefficient or rely on averaging, which do not accurately predict completion times for uncompleted work items.
Innovation Solution
A machine learning-based regression model is trained on completed work items to predict the completion times of uncompleted work items, using feature extraction and pre-processing to generate feature vectors that are then used to estimate aggregate completion times, incorporating uncertainty and complexity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual estimation methods are used to estimate completion times for work items, then the process is simple to implement, but the accuracy and reliability of the estimates are poor
Solution Approach 1:
The patent replaces manual mechanical estimation processes with an automated machine learning system. The regression model automatically analyzes historical work item data, extracts features, and generates completion time predictions without human intervention, thereby improving accuracy while managing complexity through automation.
Solution Approach 2:
The patent introduces feature vectors as an intermediary between raw work item data and completion time predictions. These feature vectors capture essential characteristics of work items (complexity, priority, type) and serve as inputs to the regression model, enabling accurate predictions while maintaining system manageability.
2Measurement precision
If averaging methods are used to estimate completion times, then the calculation is simple, but the predictions do not accurately reflect individual work item characteristics
Solution Approach 1:
The patent applies local quality by analyzing individual work item characteristics through feature extraction rather than applying a uniform average to all items. Each work item receives a customized prediction based on its specific features (complexity, priority, type), improving accuracy while the automated regression model maintains efficiency.
Solution Approach 2:
The patent transforms the estimation approach by changing from a single average parameter to multiple feature parameters. The regression model uses various work item attributes (complexity score, priority level, work item type) as input parameters to generate predictions, capturing individual variations while maintaining computational efficiency through the trained model.
Data Source
AI summary
A technique includes extracting data from a historical data store representing completed work items and associated features of the work items. The work items are associated with a lifecycle stage of an application. The technique includes training a regression model to estimate a time for completing a given work item based at least in part on the features.