model-training

!

This repository trains a Restaurant Review Sentiment Analysis model using the Naive Bayes classifier.

Project Setup

1. Clone the Repository

git clone https://github.com/remla2025-team10/model-training.git
cd model-training

2. Set Up Virtual Environment

For Unix/macOS:

python3 -m venv venv
source venv/bin/activate

For Windows:

python -m venv venv
venv\Scripts\activate

3. Install Dependencies

Make sure you have pip up to date:

python -m pip install --upgrade pip

Then install the project dependencies:

pip install -r requirements.txt

4. Run the training script

With default parameters:

python -m restaurant_model_training.modeling.train

With custom parameters:

python -m restaurant_model_training.modeling.train \
    --data_p <path/to/raw_data.tsv> \
    --processed_p <path/to/processed_data.csv> \
    --bow_p <path/to/bow_model.pkl> \
    --model_p <path/to/classifier> \
    --bow_max_features 1000 \
    --test_size 0.15 \
    --random_state 10

5. (Optionally) Try out the prediction

With default parameters:

python -m restaurant_model_training.modeling.predict

With custom parameters

python -m restaurant_model_training.modeling.predict \
    --bow_p <path/to/bow_model.pkl> \
    --model_p <path/to/classifier> \

6. (Optionally) test the model

Install local package (restaurant_model_training)

pip install -e .

Run the tests manually

pytest tests/ -v --cov=. --cov-report=term-missing

7. Deactivate the Virtual Environment (When Done)

deactivate

Remote Model Repository

This repository uses Data Version Control (DVC) to manage machine learning datasets and model artifacts. We've configured a Google Drive remote storage to facilitate collaboration without duplicating large data files in Git.

Dependencies

Make sure you have DVC and dvc-gdrive installed. If not, you can run pip install -r requirements.txt to install all dependencies, including DVC and dvc-gdrive.

Remote Storage Configuration

We use Google Drive as our DVC remote storage. The configuration can be set with a script (Please reach out to us to get the bash file with credentials), and it will been set up in the .dvc/config file. You can check the current remote configuration with:

dvc remote list

Woring with DVC

After making changes to tracked data files, reproduce the pipeline:

dvc repro

Push your data changes to the remote:

dvc push

Pull the latest changes from the remote:

dvc pull

Structure

The structure follows the established Cookiecutter template for data science projects. Some of the structure still contains empty files/folders, as they were created according to the template and may be used in the future.

The directories you should pay attention to are the following:

• data/: The folder containing all the data files
  • raw/: Original raw data dumps
  • processed/: The processed data directly used by the model
• models/: Containes the models which have already been trained
• restaurant_model_training/: The main package (module) of this project
  • config.py: Contains the configurations such as default values and paths
  • dataset.py: Logic for loading and preprocessing the data
  • features.py: Logic for creating BOW features
  • modeling/: Module containing logic for model training (train.py) and predicting (predict.py)
• tests/: The test files for the model
  • test_data_features.py: Tests for data and features
  • test_infrastructure.py: Tests for infrastructure
  • test_model_development.py: Tests for model training, evaluation, robustness
  • test_monitoring.py: Tests for model monitoring
• requirements.txt: The project dependencies
• setup.py: The package setup file

Notes

• The code is tested on Python 3.12.

TEST3