/Movie-Genere-CI-CD-Pipeline-Prediction

Primary LanguageJupyter Notebook

Movie Genre Classification With CI/CD Implementation in Heroku

This repository contains a movie genre classification project that predicts the genres of movies based on their overviews. It includes a trained model, a Flask web application, and instructions for local setup and usage.

Prerequisites

Python 3.8 or higher Docker (optional)

Table of Contents

  1. Introduction
  2. Setup
    • Local Setup
    • Docker Setup
  3. Usage
    • Web Application
    • API Endpoint
  4. Model Training
  5. Model Evaluation
  6. CI/CD
  7. License

Introduction

The movie genre classification project uses natural language processing techniques to predict the genres of movies based on their overviews. It employs a machine learning model trained on a labeled dataset of movie overviews and corresponding genres.

Setup

To use the movie genre classification project, you can follow the instructions below to set it up in your local environment or as a Docker container.

Local Setup

  1. Clone the repository to your local machine:

    git clone https://github.com/nursnaaz/Movie-Genere-CI-CD-Pipeline-Prediction.git

  2. Navigate to the project directory:

    cd Movie-Genere-CI-CD-Pipeline-Prediction

  3. Install the required dependencies:

    pip install -r requirements.txt

  4. Run the Flask web application:

    python app.py

Docker Setup

  1. Install Docker on your machine. Refer to the official Docker documentation for instructions specific to your operating system.

  2. Clone the repository to your local machine:

    git clone https://github.com/nursnaaz/Movie-Genere-CI-CD-Pipeline-Prediction.git

  3. Navigate to the project directory:

    cd Movie-Genere-CI-CD-Pipeline-Prediction

  4. Build the Docker image:

    docker build -t movie-genre-classification .

  5. Run the Docker container:

    docker run -p 5555:5555 movie-genre-classification

Usage

Once you have set up the movie genre classification project, you can use it in two ways: through the web application or via the API endpoint.

Web Application

Screenshot 2023-05-15 at 7 15 18 PM

Screenshot 2023-05-15 at 7 15 05 PM Enter a movie overview in the provided input field.

Click the "Predict" button to get the predicted genres for the movie.

Local API Endpoint

You can also make predictions using the API endpoint.

Endpoint: http://localhost:5555/predict_api Method: POST Request Payload: Parameter name: overview Parameter value: A vengeful New York transit cop decides to steal a trainload of subway fares; his foster brother, a fellow cop, tries to protect him.

Heroku Endpoint

Endpoint: https://movie-genere.herokuapp.com/predict_api Method: POST Request Payload: Parameter name: overview Parameter value: A vengeful New York transit cop decides to steal a trainload of subway fares; his foster brother, a fellow cop, tries to protect him.

Example using cURL:

curl -d "overview=A vengeful New York transit cop decides to steal a trainload of subway fares; his foster brother, a fellow cop, tries to protect him." -X POST https://movie-genere.herokuapp.com/predict_api

Screenshot 2023-05-15 at 7 00 38 PM

curl -d "overview=A vengeful New York transit cop decides to steal a trainload of subway fares; his foster brother, a fellow cop, tries to protect him." -X POST http:///localhost:5555/predict_api

Screenshot 2023-05-15 at 7 01 47 PM

Model Training

The model for movie genre classification was trained using the provided code. Here's a summary of the training process:

  • The dataset used for training is the movies_metadata.csv file, which can be downloaded from Kaggle (34.45MB file within the 239MB zip file).
  • The data was preprocessed and cleaned by extracting the 'overview' and 'genres' columns from the dataset.
  • A TextPreprocessor class was implemented to clean the text data by removing special characters, converting to lowercase, and removing stopwords.
  • The dataset was split into training and testing sets using a 80% ratio.
  • The target labels were binarized using MultiLabelBinarizer.
  • The movie genre classification model was built using a pipeline that includes text preprocessing, TF-IDF vectorization, and a logistic regression classifier.
  • The model was trained using the training dataset.
  • The trained model pipeline and the MultiLabelBinarizer were saved as model_pipeline.pkl and mlb.pkl, respectively.

Model Evaluation

After training the model, it was evaluated using the testing dataset. Here are the evaluation results:

  • Accuracy: 0.16206766206766207
  • F1-score: 0.5600685836094441 (micro-average F1-score for multi-label classification)

These metrics provide an indication of the model's performance in predicting the genres of movies based on their overviews.

CI/CD

This repository implements CI/CD (Continuous Integration/Continuous Deployment) using GitHub Actions and Heroku. The CI/CD pipeline ensures that whenever a code commit is made, the code is automatically built, tested, and deployed to Heroku.

The workflow in the .github/workflows/main.yml file defines the CI/CD pipeline. It includes steps for installing dependencies, running tests, and deploying the application to Heroku.

To set up CI/CD for your own repository, you can follow these steps:

  • Create a Heroku account (if you don't have one already) and create a new app.

  • Set up the Heroku CLI on your local machine and log in to your Heroku account.

  • Add the necessary Heroku environment variables in your GitHub repository's secrets. These variables may include the Heroku API key, Heroku app name, etc.

  • Push the code to the GitHub repository, and the CI/CD pipeline will automatically trigger. The workflow will build, test, and deploy the code to your Heroku app.

Screenshot 2023-05-15 at 7 20 04 PM

Future Developement

  • Improving Model Performance: Experiment with different machine learning algorithms and techniques to improve the model's accuracy and F1-score. This could include exploring ensemble methods, deep learning models (e.g., LSTM or Transformer), or advanced feature engineering techniques.

  • Handling Imbalanced Data: Address the issue of imbalanced data in the movie genres dataset by applying techniques such as oversampling, undersampling, or utilizing class weights during model training. This can help improve the model's performance for genres with limited representation.

  • Enhancing Text Preprocessing: Refine the text preprocessing steps to improve the quality of the input data. This could involve experimenting with different techniques for removing noise, handling misspelled words, or incorporating part-of-speech tagging to capture more meaningful features.

  • Deployment Scalability: Optimize the deployment process to handle larger datasets and increased user traffic. This can involve exploring options like deploying the application on cloud platforms such as AWS, Google Cloud, or Azure, or implementing containerization with technologies like Kubernetes for easier scalability and resource management.

  • Multi-Language Support: Extend the model to support multiple languages for genre classification, allowing for a broader range of movies from different regions and languages to be categorized accurately.

  • User Interface Enhancements: Improve the user interface of the web application to provide a more intuitive and visually appealing experience. This can include adding search functionalities, filtering options, or displaying additional movie details along with the predicted genres.

License

The movie genre classification project is released under the MIT License. You are free to use, modify, and distribute the code for personal and commercial purposes.