Hazardous Content Classifier

This project implements a classifier to predict whether a content is hazardous or not. It uses a machine learning model trained on a dataset of text with binary labels (0 for non-hazardous content and 1 for hazardous content). The model combines TF-IDF vectorization with an SVM classifier to make predictions.

Features

• Automatic Training: The system supports automatic model training using a new dataset, combining it with the original dataset to improve accuracy.
• Retraining: The model can be retrained with new data to adapt to new trends or changes in data.
• Content Prediction: The API allows you to submit content for predictions on whether it is hazardous or not.
• Prediction Cache: It uses a caching system to store previous predictions and accelerate the classification process for similar content.
• Model Statistics Report: Generates model statistics reports such as accuracy, recall, and F1-Score to evaluate model performance.
• Automatic Cleanup: After training is complete, the system automatically deletes temporary data files to free up space.

Project Structure

• classifier.py: Main file containing the Flask API for handling training and classification requests.
• datasets/: Directory containing datasets for training and storing results.
• modelsft/: Directory storing trained models.
• cache.pkl: Cache file storing previous predictions to improve performance.

How to Use

1. Training:
   • Send a POST request to /train with a CSV file containing the new dataset.
   • The system will combine the dataset with the original one and retrain the model.
2. Prediction:
   • Send a POST request to /mlclassifier with a list of contents to predict.
   • You will receive a JSON response with the contents and their corresponding predictions.
3. Model Verification:
   • Send a GET request to /getModelName to obtain information about the most recent trained model, (Version and model name).

Requirements

• Python 3.11
• Flask
• Pandas
• scikit-learn
• colorama (For information messages on terminal)
• tqdm
• joblib
• pickle
• torch (optional, if using PyTorch)

Installation

Install dependencies using pip:

pip install -r requirements.txt

Running the App

Run the Flask application:

python classifier.py

The application will run on http://127.0.0.1:5000/.

Notes

• Ensure you have a requirements.txt file that contains the necessary dependencies.
• Keep datasets updated to improve model performance.

Authors

• @AlejandroChitivaC