An approach to identifying minimal and equivalent mutants from relevant mutant features and properties

Mutation Testing

This project presents a tool that apply an approach that aims to reduce the cost of Mutation Testing.

Our work intends to collect data from mutants and use Machine Learning Algorithms to classify a new mutant (without classification) as minimal, or not, as equivalent, or not.

Step by Step

• Execute Proteum to generate and execute mutants and generate informations about them (status, program graph node, operator, offset, etc)
• Gather generated informations in CSV Files
• Import CSV files into ML algorithms
• Preprocess data before classify the mutants
• Train ML algorithms
• Evaluate ML algorithms

How to run

python3 experiment.py executionMode - executionMode = 1 Run Proteum and Analyze Data 2 Just run Proteum 3 Just analyze

python3 ML/ML_Mutants.py {parameters} --column {targetColumn} --classifier {classifier} - The parameters for execution are: --all - Execute all programs, all target columns and all classifiers. --allPbp - Execute all target columns, all classifiers and all programs, but with one execution for each one. --column | The targetColumn to be classified. Could be 'MINIMAL' or 'EQUIVALENT' --classifier | The classifier used to classify. Could be 'KNN' for K Nearest Neighbors, 'DT' for Decision Tree, 'RF' for Random Forest, 'SVM' for Support Vector Machine, 'LDA' for Linear Discriminant Analysis, 'LR' for Logistic Regression and 'GNB' for Gaussian Naive Bayes --program | The specified program to classify the target column --pbp | Execute program by program --best | Indicating that will be execute the classifiers with the best parameters

• Possible executions --all --all --best --allPbp --allPbp --best --column {column} --classifier {classifier} --column {column} --classifier {classifier} --best --column {column} --classifier {classifier} --program {programName} --column {column} --classifier {classifier} --program {programName} --best --column {column} --classifier {classifier} --pbp --column {column} --classifier {classifier} --pbp --best

• For each ML algorithm and classification (minimal or equivalent) will be created a result CSV File

TU notes:

1. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier DT
2. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier KNN
3. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier RF
4. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier SVM
5. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier LDA
6. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier LR
7. python3 ML/ML_Mutants.py --column EQUIVALENT --classifier GNB