End-to-End Student Dropouts Prediction

Features

Demographics

• Age
• Gender
• Ethnicity

Academic

• Full time status
• First generation
• Number of credits is attemtping
• Number of courses retaken
• New student or not
• Number of developmental math courses
• Number of developmental english courses
• Course modularity
• Past dropout rate
• Dropout history for the last 3 semesters
• High school GPA if a student is new to college
• Last term cumulative GPA
• Last term GPA
• Average precentage of absences
• TODO: data from Canvas

Data

Get Raw Data

Simply run the following command:

mlflow run ./src/get_data/

The data will be saved in data/raw/ folder.

Preprocess Data

To get the preprocessed data for week 10, simply run the following command:

mlflow run ./src/preprocess/ -P week_number=10

The preprocessed data named cleaned-10.csv will be saved in data/processed/ folder.

EDA

To get an overview of what the data look like.

Run the following command

mlflow run ./src/eda/

Model

• Decision Tree
• Gradient Boosting Decision Tree

Run the following command to train the XGBoost model and make predictions.

mlflow run ./src/model/