Evaluation of Vulnerability to Bullying ML Model

A random forest classifier model that attempts to evaluate vulnerability to bullying based on the data from a school-based survey conducted in Argentina in 2018.

This model is created solely for demonstration purposes and does not claim to provide any insights on the issue.

The repository contains both notebooks and a simple front-end/back-end for demonstration.

• To run locally see instructions below.

• To view the final notebook, which takes input and outputs predictions using saved models, open input_predict_notebook.ipynb from the ML folder with Jupyter Notebook or an IDE that supports a Jupyter Python kernel.

No additional configuration should be required when using a conda environment and a conda Python kernel.

All files pertaining to the initial data set, data cleaning, and model creation can be found in the /ML folder of the repository:

• ML/data folder: initial data set, data sets created during data cleaning, input template;
• ML/data_cleaning folder: notebooks with the data cleaning process (divided into two steps);
• ML/models: notebooks with the model creation process (four models for different target variables);
• ML/models/saved_models: corresponding saved models for further usage in the back-end and final notebook;

Preview

• When running the project locally, you can view the home and about pages for information about the project:

• To test the model, switch to the evaluate tab, fill in the form, and click 'evaluate':

• After processing the inputs from the form using the Python script (model), a popup with the results will open:

• When viewing the project in input_predict_notebook.ipynb, restart the kernel and run all cells. The function in the last cell will prompt for input and output the results:

Run front-end and back-end locally

At this time, front-end CSS is not adaptive and is designed for screens with a width of 1024px or more.

Note: Since some dependencies used, such as scikit-learn, may not be the latest version, it is recommended to set up a virtual environment before installing dependencies from step two below.

To set up a virtual environment, open the terminal/bash, go to the root directory of the cloned repository, and follow these steps:

• mac: $ python3 -m venv venv / windows: $ python -m venv venv
• mac: $ source venv/bin/activate / windows: $ source venv/Scripts/activate
• upgrade pip by $ python -m pip install --upgrade pip

1. Clone the repository.
2. Install dependencies:
   • while in the root directory, run $ pip install -r requirements.txt for python script (/backend/predict.py) dependencies
   • switch to /backend directory and run $ npm i for back-end dependencies
   • switch to /frontend directory and run $ npm i to install front-end dependencies
3. While in /frontend directory, launch the client by $ npm run start
   The application will run on http://localhost:3000/ (should open automatically)
4. In another terminal window, while in /backend directory, launch the local server via the terminal by $ npm run start
   The server will run locally at http://localhost:3001/

Dataset

The School-based survey in Argentina can be found on Kaggle.