neurojit 1.0.2

A Python package for calculating the commit understandability features of Java projects.

neurojit

neurojit is a Python package designed to enhance Just-In-Time (JIT) defect prediction by leveraging insights from both neurophysiological and empirical developer reactions to code. Modern developers make changes based on their understanding of the existing code context, and the difficulty in comprehending these changes can increase the likelihood of human error, potentially introducing defects. NeuroJIT captures the understandability of each commit through features that correlate with developers’ cognitive and empirical responses to different code segments. These features have been shown to improve JIT defect prediction models by identifying defect-inducing commits more effectively. For more information, see our research paper "NeuroJIT: Improving Just-In-Time Defect Prediction Using Neurophysiological and Empirical Perceptions of Modern Developers."

Installation

You can install the NeuroJIT package from PyPI using pip:

$ pip install neurojit

To install with additional dependencies for replicating the research results:

$ pip install neurojit[replication]

For more information about the replication, see our NeuroJIT replication package

Example Usage

1. Filtering Commits and Saving Method Changes

   from neurojit.commit import MethodChangesCommit, Mining, Method
   
   mining = Mining()
   target_commit = mining.only_method_changes(repo="activemq", commit_hash="8f40a7")
   if target_commit is not None:
       mining.save(target_commit)
   

2. Calculating Commit Understandability Features

   Compute commit understandability features from the saved MethodChangesCommit instances.

   from neurojit.cuf.metrics import CommitUnderstandabilityFeatures
   
   cuf_calculator = CommitUnderstandabilityFeatures(target_commit)
   features = ["HV","DD", "MDNL", "NB", "EC", "NOP", "NOGV", "NOMT", "II", "TE", "DD_HV"]
   for feature in features:
       value = getattr(cuf_calculator, feature)
       print(f"{feature}: {value}")
   

3. Splitting the Dataset Chronologically with KFoldDateSplit

   Use KFoldDateSplit to split the dataset into training and testing sets, considering chronological order, verification latency, and concept drifts.

   from neurojit.tools.data_utils import KFoldDateSplit
   
   data = pd.read_csv("...your jit-sdp dataset...")
   data["date"] = pd.to_datetime(data["date"])
   data = data.set_index(["date"])
   
   splitter = KFoldDateSplit(
       data, k=20, start_gap=3, end_gap=3, is_mid_gap=True, sliding_months=1
   )
   
   for i, (train, test) in enumerate(splitter.split()):
       X_train, y_train = train[features], train["buggy"]
       X_test, y_test = test[features], test["buggy"]
   

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.