MPST: Movie Plot Synopses with Tags

This project was developed to solve MPST: Movie Plot Synopses with Tags multi-label text classification problem. It's a complex dataset which deal with imbalanced data and 71 differents tags to predict!

Some technnologies of Machine Learning and Data Science used:

Dealing with imbalanced data

• Tune class weights
• Up-sampling with sklearn

Models

• Logistic Regression
• XGBoost Classifier

API

• FastAPI

Running

To clone and run this application:

# Clone this repository
$ git clone https://github.com/BruBel/MPST-Movie-Plot-Synopses-with-Tags

# Go into the repository
$ cd MPST-Movie-Plot-Synopses-with-Tags

# Install dependencies
$ pip install -r requirements.txt

# Go into API repository
$ cd code/api

# Run api
$ uvicorn main:app --host 0.0.0.0 --port 5058

Now there are two options:

• Go to http://localhost:5058/docs#/default/predict_category_predict_post
• Click in "Try it out" and fill in the JSON. Execute.

OR

• In terminal:

curl -X POST "http://localhost:5058/predict" -H "accept: application/json" -H "Content-Type: application/json" -d "{"synopsis":"SYNOPSISEHERE"}"