The project is a content-based filtering approach for suggesting tasks and projects to GitHub newcomers.
- Update cosine similarity 6.1 and 8.5 to calculate recommendation scores using known commit authors to validate
- Commits used for validation seems off, TODO: recalculate
- Add 6.1. Cosine Similarity (including VSM) to recommend issues for users.
- Simplified the code.
- For 1.1., now we also collected "referenced commits" with issues.
- Update K-Means, Decision Tree and Random Forest.
- Add TF-IDF analysis for issues.
| Data | Description |
|---|---|
all_issues_REPO-NAME.json |
list of repository issues (also includs it's pr, commits) that are bug fixes and/or "easy pick" |
users_REPO-NAME.json |
list of users from issues that are bug fixes and/or "easy pick" (THIS FILE MAY BE TOO HUGE TO OPEN) |
users_REPO-NAME_filtered.json |
filtered user json file. include different ages with same user. |
data_users_REPO-NAME_ready_to_analysis.csv |
csv format file of users_REPO-NAME_filtered.json |
data_users_cluster_with_results.csv |
K-Means result table |
issue_text_REPO-NAME.json |
textual content of each issue |
- Simplified the code.
- For 1.1., now only concern users who submit pr and commits whth related the issues with "Easy Pick" label
- For 1.1., now collect all issue data and its related pr and commits data in order to save needed time when future usage.
- For 1.2., now collect users' whole data in order to save needed time when future usage.
- For 1.3., simplified the user data extraction process.
- For 2.1. and 2.1.2., modified the code to fit latest version of data files.
- For 1.1., modify the process logic in order to reduce the time needed.
- For 2.1., add column "newcomer" in order to verify the newcomer.
- For 3.1., modify KMeans in order to get more accuracy clustering result.
- For 3.2., add "Silhouette Analysis" to determine the number of clusters.
- For 4.1., move Decision Tree to this file.
- For 4.2., move Random Forest to this file.
Next:
- Finalising how many cluster we need to use.
- Starting issue classify.
- Param/Model outpts from Nov 4 RESULTS.md
- Rewrite data extraction and user extration in order to get more data and increase predict precision.
- Added "User Classification" file to predict newcomer.
- Saved "User Decision Tree Model" and "User Random Forest Model" files for future usage.
- Symfony data set, MSR 14 https://github.com/symfony/symfony
- Rewrite data extraction (in order to get more data)
Next:
- Getting user data and train user model to determine what charateristics that newcomers should have.
No license. All rights reserved by Jonathan Lam (@jonlamca) and Lance Wang (@ycpss91303).