The project's data can be found here: https://www.kaggle.com/c/asap-aes. The original project was hosted 8 years ago. In this project, I try to beat the highest score which at the time was ~81%. Machine learning has come a long way from what it was 8 years ago. As such, beating the original score does not mean much, but it is still a great project to practice machine learning skills. There are many people who could achieve more than 95% accuracy for this project using neuron network. However, I will not be using neuron network as it is harder to explain how the model produces the result to stakeholders. Instead I will use feature engineering and K-Nearest Neighbors.

• Folder Data: This contains the processed data that I use.
• Using TFIDF - SVD to predict.ipynb: This is the Jupyter Notebook for the project.

Additional resources can be found by following the links.

• TF-IDF: Term frequency-inverse document frequency
• SVD: Singular Value Decomposion
• K Nearest Neighbors

This model has an accuracy of 87%, when calculated with the cohen cappa quadratic.