Issue Analysis

The Issue Analysis module provides two main functionalities:

• Issue report classification: A binary and a multiclass classifier are integrated. The binary classifier can distinguish bugs and non-bugs. The multiclass classifier supports the types bug, improvement and feature request.
• Bug report quality estimation: For a given bug report, a score can be calculated based on its content.

Integration

This project can be added as Maven dependency. A snapshot repository is provided. In your project's pom add:

<repositories>
    <repository>
        <id>issue-analysis-mvn-repo</id>
        <url>https://raw.github.com/curtys/issue-analysis/mvn-repo/</url>
        <snapshots>
            <enabled>true</enabled>
            <updatePolicy>always</updatePolicy>
        </snapshots>
    </repository>
</repositories>

<dependencies>
   <dependency>
       <groupId>ch.unibe.scg.curtys</groupId>
       <artifactId>issue-analysis</artifactId>
       <version>1.0-SNAPSHOT</version>
   </dependency>
</dependencies>

Source and Javadocs can also be downloaded from the Maven repsitory.

Usage

Two tools are provided to analyse issues:

• Issue type classifier
• Quality estimator

Classifier

There are two issue classifiers: A binary classifier, which predicts if the issue in question is a bug or not, and a multi-class classifier with finer type distinction. Use the binary classifier if you just want to know whether or not the issue in question represents a bug.

// instantiate a binary classifier
Classifier binaryClassifier = new BinaryClassifier();

// or a multiclass classifier
Classifier multiclassClassifier = new MulticlassClassifier();

Prediction prediction = binaryClassifier.query(issue);

// get the label with the highest probability
String type = prediction.getBestClassLabel();

// get the probability of a given label
float probability = prediction.probabilityOf(label);

// get all class labels
Set<String> labelSet = prediction.labels();

// get a map with all labels and their probabilities
Map<String, Float> propabilitiesMap = prediction.getProbabilities();

Quality Estimator

Use the quality estimation to calculate a score for a bug report:

QualityEstimator estimator = new QualityEstimator();

// calculate the score
int score = estimator.score(issue);

// get a map with the features as keys and a value 
// of either 1 (if feature is present) or 0 (otherwise)
Map<String, Integer> featureMap = 
   	estimator.activationFeatures(vector);

License

MIT