Chase Thompson

We scraped data from GitHub repository README files using GitHub's API.

Goal

The primary goal for this project was to build an NLP model that can predict the primary language of REPO using the text in the README file.

As a secondary goal, we decided to pull repos from a specific sector to see if an industry we were interested in is utilizing a programming language we're familiar with.

Deliverables

A well-documented jupyter notebook that contains your… Show more

We scraped data from GitHub repository README files using GitHub's API.

Goal

The primary goal for this project was to build an NLP model that can predict the primary language of REPO using the text in the README file.

As a secondary goal, we decided to pull repos from a specific sector to see if an industry we were interested in is utilizing a programming language we're familiar with.

Deliverables

A well-documented jupyter notebook that contains your analysis.

https://github.com/github-nlp-project/nlp-project/blob/master/github_nlp_notebook.ipynb

A slide deck with high-level findings

https://www.canva.com/design/DAD-LM8dJPk/6-UD1UppB0-u5LeA9fjB2g/view?utm_content=DAD-LM8dJPk&utm_campaign=designshare&utm_medium=link&utm_source=publishsharelink

Technical Skills used

Python
Jupyter Notebook
VS Code
Various Data Science libraries (Pandas, Numpy, Matplotlib, Seaborn, Sklearn, etc.)
BeautifulSoup
nltk
Stats (Hypothesis testing, correlation tests)
Classification Models (Decision Tree, Random Forest, KNN)

SUMMARY

Github READMEs can be good predictors of the programming languages of the repos. With NLP models language is important, so it isn't a surprise that the longer and more in-depth a README is, the better we can get at predicting the primary language.

Whenever possible, words that are unique to a particular programming language, including the name of that language, make for key identifiers of a language.

As with most data work, NLP modeling is no different, imbalances in your data affect everything, most importantly model performance. To improve our models, we'd want to ensure a better language distribution throughout the corpus, and gathering more observations generally speaking will also possitively impact model performace.

Want to use our code?

You can use our python files to recreate the work done in our notebook. You'll also need to get create an env.py file that includes your GitHub username and GitHub token. Show less

The Notebook:

You can also view the notebook as it's own webpage here: https://chasethompson.github.io/#

Background:

We've used Linear Regression models to start to tackle the home value estimate. We will now work on improving Zillow's log error.

The audience for this project is a data science team, and we want to demo the discoveries we made and work we've have done related to uncovering what the drivers of the error in the Zestimate are.

Goals

We will… Show more

The Notebook:

You can also view the notebook as it's own webpage here: https://chasethompson.github.io/#

Background:

We've used Linear Regression models to start to tackle the home value estimate. We will now work on improving Zillow's log error.

The audience for this project is a data science team, and we want to demo the discoveries we made and work we've have done related to uncovering what the drivers of the error in the Zestimate are.

Goals

We will be using the ML Clustering Algorithm KMeans to find clusters within the data to improve our estimate of the log error. Once those clusters have been identified, we will use Regression models to test the efficacy of the clustering work.

Deliverables

A report (in the form of a detailed, high-level notebook)

The notebook should cover all the steps from acquisition through modeling.

A detailed README explaining the project and steps to reproduce the results.

Technical Skills used

Python
SQL
Sequel Pro
Jupyter Notebook
VS Code
Various data science libraries (Pandas, Numpy, Matplotlib, Seaborn, Sklearn, etc.)
Stats (Hypothesis testing, correlation tests)
Clustering Model (KMeans)
Regression Models (Linear Regression, Decision Tree Regressor, Random Forest Regressor)

Executive Summary

The hypothesis and conclusion are unclear. Our derived variables proved useful, but not significantly.

Our main drivers appeared to hover around the overarching geological data and clustering using the selected features associated with those data points.

The linear regression model performed quite poorly. However, the decision tree and random forest regressors did slightly better than the baseline.

We observed some statistical difference between log error with regards to these features:

Longitude/Latitude
Lot size
Square footage
Age of the home
It appears either more time is necessary to evaluate the different clustering opportunities within the data. Or that, perhaps, clustering is not the best approach for this data. Show less

Here's the link to the deliverable report:

https://docs.google.com/presentation/d/1nObaqS4VnmiCM28RdxW_XqJfoU9f-sAFba4UjRf3A00/edit?usp=sharing

Background:

Our team lead needs us to take a look at some of our recent customer data. We've been tasked with identifying areas that represent high customer churn.

Goals

Identify as many different customer subgroups that have a propensity to churn more than others. Our target audience is our team lead, however, she… Show more

Here's the link to the deliverable report:

https://docs.google.com/presentation/d/1nObaqS4VnmiCM28RdxW_XqJfoU9f-sAFba4UjRf3A00/edit?usp=sharing

Background:

Our team lead needs us to take a look at some of our recent customer data. We've been tasked with identifying areas that represent high customer churn.

Goals

Identify as many different customer subgroups that have a propensity to churn more than others. Our target audience is our team lead, however, she will be presenting these findings to the Senior Leadership Team. We will need to keep this final audience in mind with regards to report readability, etc. We will need to communicate in a more concise and clear manner.

Deliverables

A report in the form of a slide deck

A GitHub repository containing our jupyter notebook that walks through the pipeline

She will also need a CSV with the customer_id, probability of churn, and the prediction of churn

The development team will need a .py file that will take in a new dataset, and perform all the transformations necessary to run the model.

Technical Skills used

Python
SQL
Various data science libraries (Pandas, Numpy, Matplotlib, Sklearn, etc.)
Stats (Hypothesis testing, correlation tests)
Classification Models (Logistic Regression, Decision Tree, Random Forest, KNN)
PowerPoint

Executive Summary

The model we selected to move forward with is a Classification ML Algorithm, Logistic Regression.

The model is able to correctly predict 80% of customers who actually churn.

Customers most likely to churn:

Month to month contract
Senior Citizens
Utilize Premium offerings
The initial Prediction CSV is in this repo, the code outlined in the notebook will produce the CSV as well.

Moving forward we should focus our CRM and marketing efforts on transitioning customers away from month-to-month contracts, strengthen our relationships with our Senior citizen customers, and further utilize our premium services to retain customers. Show less

Here's the link to the deliverable report requested from the team. It contains the slides, but not the verbal portion of the presentation.

https://drive.google.com/file/d/1eMrsrx6Vs1rHDuicfFpt9uNJNoz8g3LU/view?usp=sharing
Background:

We want to be able to predict the values of single-unit properties that the tax district assesses using the property data from those whose last transaction was during the "hot months" (in terms of real estate demand) of May and June in 2017. We… Show more

Here's the link to the deliverable report requested from the team. It contains the slides, but not the verbal portion of the presentation.

https://drive.google.com/file/d/1eMrsrx6Vs1rHDuicfFpt9uNJNoz8g3LU/view?usp=sharing
Background:

We want to be able to predict the values of single-unit properties that the tax district assesses using the property data from those whose last transaction was during the "hot months" (in terms of real estate demand) of May and June in 2017. We also need some additional information outside of the model.
Zach lost the email that told us where these properties were located. Ugh, Zach :-/. Because property taxes are assessed at the county level, we would like to know what states and counties these are located in.

Goals

Your customer is the Zillow data science team. State your goals as if you were delivering this to Zillow. You will be communicating in a more concise way, and very clearly, the goals as you understand them and as you have taken and acted upon through your research.

Deliverables

What should the Zillow team expect to receive from you?

A report (in the form of a presentation, both verbal and through slides)
A GitHub repository containing your jupyter notebook that walks through the pipeline along with the .py files necessary to reproduce your model. The data dictionary can be found in this file.

Technical Skills used

Python
SQL
Various data science libraries (Pandas, Numpy, Matplotlib, etc)
Stats (Hypothesis testing, correlation tests)
Linear Regression (Single Linear Regression, Multiple Linear Regression)
PowerPoint
Photoshop

Executive Summary

- Everyone already knows, but Location! Location! Location! is so important in the final product for any real estate application
- Location, square footage and neighborhood (aort of location, but more specific ) are major drivers of home value
- Other classification methods should help improve the accuracy of models. Show less

We’re here to show you that you’re never alone in the note business. Join us every Wednesday to find out what we’re up to now.