The Google slide summary for our project was created in Canva here.
Target audience for our project is the Codeup Data Science cohort and instructors.
- google slides that:
- summarize results from exploration and modeling
- visuals that help describe our findings
- a github repository containing:
- final notebook containing steps, analysis and findings
- a README.md that contains description and steps to reproduce
- README.md should also contain hyperlink to google slides
- a 5-6 minute presentation of findings
- What are the top 5 programming languages when searching for 'Game' repos in github?
- What are the most common words for some of these languages?
- How do the most common words for a language compare to the highest TF-IDF scores for that language?
- What are common bigrams for some of these languages? Are they significant to the kind of programming language?
- Acquire Web scrapping of Github search results. The first 500 repos with the keyword of 'games' on the search date of July 22, 2022 were used.
- Clean and prepare data for the explore phase.
- Create wrangle.py to store functions I created to automate the cleaning and preparation process.
- Separate train, validate, test subsets and scaled data.
- Explore the data through visualization using natural language processing methods.
- Clearly define our hypotheses and questions.
- Document findings and takeaways.
- Perform modeling:
- Identify model evaluation criteria (what is our baseline?)
- Create at least three different classification models.
- Evaluate models on appropriate data subsets.
- Create Final Report notebook with a curtailed version of the above steps.
- Create and review README. Ensure it contains:
- Data dictionary
- Project summary and goals
- Initial Hypothesis
- Executive Summary
Through classification models, we were able to beat our baseline of 25.38% accuracy with a K-Nearest Neighbor model that had 38% accuracy. This is a small win, but I do believe there is room for improvement to have an even more accurate model. Project Goal: The goal of this project was to use natural language processing and classification models to identify terms for predicting a readme's primary language on Github. Key Findings:
Recommendations & Next Steps:
- Coding language name and libraries associated with that language seem to be an indicator of language. This is most apparent using the TF-IDF score.
- Without a deeper understanding of common terms for each language, it is more difficult for us to identify.
- KNN performed best on our validate subset with an accuracy of 47% and a Precision of 75%. However, these scores dropped notably on our test subset with an accuracy of 38% and precision of 43%. The KNN accuracy outperformed baseline by 12.65%.
- With more time, we would further tune our models. We would like to try additional random forest and decision tree classifiers with a greater depth given the number of features.
- We would also like to perform this modeling with a larger dataset. Some of the less common languages appeared less than 20 times, making it a relatively small amount of data to train the model on. This also likely adversely affected the performance of our KNN model for the smaller datasets (validate and test).
| Target | Definition |
|---|---|
| Language | The primary coding language identified in the GitHub repository |
| Feature | Definition |
|---|---|
| content | Content's of the repository's readme.md (String) |
| repo | partial repo URL |
You can reproduce this project with the following steps:
- Read this README
- Clone the repository. Alternatively, you can download the .py, .json and Final_Report from the main folder.
- Run the Final_Report notebook or explore the other notebooks for greater insight into the project.
If you want to recreate this with your own list of repos:
- Clone the repository
- Remove the repos.csv and data2.json files.
- Read through the acquire.py functions and update the search URL with any keyword changes or increase the number of results.
- run "python acquire.py" from your terminal to generate the csv and json files.
- Step through the Final Report and other notebooks. You may want to modify which languages are viewed.