Dependency Update Strategies and Package Characteristics

This is the replication package for our paper on predicting dependency update strategies.

Here is a short description of what is contained in this package:

Pre-processing Code

The data_preparation.py file filters the initial libraries.io dataset to only include relevant columns for the npm packages. The feature_extraction.py file includes the extraction of model features. The feature_process.py file includes the majority of preprocessing scripts to derive new features, clean-up missing values and prepare the data for the models.

ML Models

The models.py file contains the scripts for training, validating and evaluating the random forest model and the two baselines (stratified random and SemVer only models) used for the study.

Datasets

The raw dataset can be downloaded from libraries.io. The Processed_Project_Features[SP51][RT].csv dataset is the result of all preprocessing steps and is used in the model feed function.

Visualization Scripts

The visualizer.py file includes the visualization scripts used for the paper.

Sampled Packages

The complete set of visualizations for the 160 sampled packages in RQ3.