"Feed the Machine - Language Analysis" (Project 3)

Jenni Davis, David Jimenez, Elizabeth Conway, Austin Olea, Susan Farago, & Catherine Poirier (June 2021)

For more information and detailed results, please refer to the Project Final Report!

Problem to Solve

Can we use machine learning to accurately and effectively evaluate a trainer’s responses to essay-style questions in order to predict a trainer’s training and facilitation skills?

Results / Findings

Based on the dataset we used, we were unable to reliably predict a trainer’s training and facilitation skills based on evaluating the trainer’s response to essay-style questions. The data was run against three different machine learning models: Scikit-Learn, Linear Regression, and Random Forest Regressor. Next steps: Reevaluate questions in order to gain better responses. Rerun models and analyze results. It's not the model, it's the data.

Database Information

• File #1

• Extracted from SalesForce & cleaned utilizing Tableau.

• Essay responses from 2,388 trainers to twenty open-ending questions from June 2020 - May 2021, resulting in 42,998 rows of data.

• File #2

• Extracted from SalesForce & cleaned utilizing Tableau.

• Student scoring on the respective trainer’s training and facilitation skills.

• 912 trainers taught courses of those 486 received scoring.

• Final cleaned data included 312 trainers and 6,240 records.

Tools Used

• Tableau
• Python Pandas
• Python NLTK
• Vader Sentiment Analysis
• Matplotlib
• Machine Learning Models:
• Scikit-learn
• Naive Bayes Classifier
• GaussianNB
• Linear Regression
• Random Forest Regressor
• HTML / CSS / Bootstrap
• GitHub Pages