Automated Source Code Alarm Classification via AST Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual classification of alarm types in static analyzers for source code errors is resource-intensive and prone to false alarms, as static analyzers may incorrectly determine errors, leading to unnecessary resource allocation for classification processes.
Innovation Solution
An automated method for classifying error detection alarms using an alarm type classifying apparatus that converts source code into an abstract syntax tree, removes unnecessary sub-trees, generates feature vectors based on preset patterns, and clusters alarms using the K-means algorithm to identify and categorize error detection alarms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual classification of alarm types is performed by developers, then alarm type classification can be done with simple tools, but resource consumption increases and time is wasted
Solution Approach 1:
The system performs automatic alarm type classification without requiring developer intervention. The alarm type classifying apparatus autonomously receives alarm information, converts it to AST, extracts features, and classifies alarms using clustering algorithms, making the system self-sufficient in the classification task.
Solution Approach 2:
The manual mechanical classification process by developers is replaced with an automated computational system. The apparatus uses computer-based technologies including AST conversion, feature extraction, and clustering algorithms to substitute human manual classification work.
2Reliability
If static analyzers perform comprehensive error detection, then detection coverage is improved, but false alarms increase leading to resource waste
Solution Approach 1:
The system provides feedback by classifying alarms into types including false alarms, and enabling additional analysis for specific alarm types. This feedback mechanism helps identify and respond to false alarms, allowing the static analyzer to improve its detection accuracy over time.
Solution Approach 2:
The system changes the parameter of alarm analysis by introducing alarm type classification. By categorizing alarms into different types based on extracted features and clustering, the system transforms raw alarm data into structured information that can be selectively analyzed, reducing the impact of false alarms.
3Measurement precision
If alarm type classification is performed for all alarms, then analysis accuracy is improved, but resource consumption increases
Solution Approach 1:
Instead of performing comprehensive additional analysis on all alarms, the system applies partial action by classifying alarms into types and enabling additional analysis only for specific alarm types that benefit from it. This selective approach maintains necessary analysis accuracy while conserving resources.
Solution Approach 2:
The system applies different levels of analysis to different alarm types based on their characteristics. By extracting specific features and applying clustering algorithms locally to relevant alarm categories, the system optimizes resource allocation by focusing computational effort where it is most needed rather than uniformly processing all alarms.
Data Source
AI summary
The present invention relates to a method for classifying alarm types in detecting source code errors, a computer program therefor, and a recording medium thereof. The method for classifying alarm types in detecting source code errors includes: receiving input of alarm path information about an occurring error detection alarm and source code information that is an object associated with the occurring alarm, the alarm path information being information about an execution path related to the error detection; converting the source code into an abstract syntax tree (AST); removing, from the AST, an unnecessary sub-tree that is not related to the error detection alarm; obtaining a feature vector of the AST having the unnecessary sub-tree removed therefrom based on a preset feature pattern set; and classifying, by types, the error detection alarm associated with the feature vector by clustering the obtained feature vector using a preset method.