Machine-Learned Fix Verification Procedures from Historical Defects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying and verifying corrections to defects in computing environments is challenging due to insufficient fix verification procedures, often resulting in additional support costs and customer downtime, as developers and fix verification agents may lack access to the same tools and knowledge.
Innovation Solution
A problem fix verification component generates recommended fix verification procedures based on problem analysis data and historical data using rules-based techniques and machine learning, ranking these procedures to optimize the verification process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If developers manually create fix verification procedures, then the procedures can be tailored to specific defects, but the process is time-consuming and increases support costs
Solution Approach 1:
The system performs preliminary action by automatically generating fix verification procedures before developers need to review them. The machine learning model creates initial verification procedures based on historical data and defect patterns, allowing developers to focus only on reviewing and refining the generated procedures rather than creating them from scratch.
Solution Approach 2:
The system copies successful verification procedures from historical defects and adapts them to current defects. By analyzing historical problem analysis data and using machine learning to identify patterns, the system replicates effective verification approaches from past defects, reducing the time needed to create new verification procedures while maintaining reliability.
2Ease of operation
If developers communicate back-and-forth with fix verification agents, then verification procedures can be refined, but downtime and computational resources increase
Solution Approach 1:
The system enables self-service by allowing the machine learning model to autonomously generate and rank verification procedures without requiring continuous human intervention. The model independently analyzes problem analysis data, identifies relevant historical defects, and creates verification procedures that can be directly implemented, reducing the need for back-and-forth communication and minimizing downtime.
Solution Approach 2:
The system incorporates feedback mechanisms where the machine learning model continuously learns from verified procedures and adjusts its generation process. By analyzing which generated procedures are accepted or modified, the model refines its algorithms to produce more accurate verification procedures in subsequent iterations, reducing the need for corrective communication cycles.
3Reliability
If comprehensive verification procedures are created, then defect corrections can be thoroughly verified, but computational resources and complexity increase
Solution Approach 1:
The system applies parameter changes by dynamically adjusting the depth and scope of verification procedures based on the specific defect characteristics. The machine learning model analyzes defect severity, type, and historical resolution patterns to determine the appropriate level of verification detail, avoiding unnecessary complexity for minor defects while ensuring thorough verification for critical issues.
Solution Approach 2:
The system implements local quality by tailoring verification procedures to the specific characteristics of each defect rather than applying a uniform comprehensive approach. The machine learning model identifies which verification steps are most relevant for each specific defect type and prioritizes those, creating focused verification procedures that maintain reliability without unnecessary complexity.
Data Source
AI summary
Methods and apparatus for generating problem fix verification procedures for varying corrections to defects are described. An example method includes obtaining problem analysis data associated with a defect of a first computing system. A first representational vector corresponding to the problem analysis data is generated. Second representational vectors corresponding to respective historical defects associated with one or more second computing systems are obtained. A set of historical problem fix verification procedures associated with the defect is determined, based on the first representational vector and at least one of the second representational vectors. A recommended set of problem fix verification procedures for verifying a correct to the defect of the first computing system is provided to a third computing system. The recommended set of problem fix verification procedures include the ranked set of historical problem fix verification procedures.