Bigdata team project 'Drug Use prediction'

Team project on the subject “Architecture and Technologies of Big Data Systems” at the National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”.

The goal of the project is to build a prediction model that will determine whether a person uses drugs (such as cocaine, crack, or marijuana)

The dataset used for training was the National Survey of Drug Use and Health (2015-2019)

Models used:

1. LGBMRegressor: Uses gradient boosting based on LightGBM
2. XGBRegressor: Uses gradient boosting based on XGBoost
3. Ridge Regressor: Linear regression with regularization using Ridge
4. Gradient Boosting Regressor: Uses gradient boosting to improve model accuracy
5. Random Forest Regressor: Uses an ensemble of decision trees for prediction
6. StackingCVRegressor: A model that combines the predictions of several underlying models using cross-validation

Repository contents

• data/: The folder with the preprocessed datasets
• doc/: Іnformation about the dataset is taken from SAMHSA site
• demo/: Images with the results of model testing
• models/: Saved models and pipeline
• notebooks/: A Jupyter notebook for data visualization, models training, and analysis of results
• requirements.txt: List of required Python packages for installation

Results

Model	Accuracy	Precision	Recall	F1-Score
LightGBM	0.912226	0.992937	0.830966	0.904760
XGBoost	0.909194	0.999785	0.819188	0.900521
Ridge Regression	0.909039	0.999677	0.818968	0.900344
Gradient Boosting	0.912580	0.992683	0.831892	0.905203
Random Forest	0.908441	1.000000	0.817512	0.899595
Stacking	0.913221	0.989709	0.835730	0.906225

Installation and use

Clone the repository:

git clone https://github.com/TokenRR/Bigdata_university_course.git

cd Bigdata_university_course

Install the required packages:

pip install -r requirements.txt

---

You can use the notebooks from the notebooks/ folder to research and analyze the results.
If you would like to contribute to this project, please create a pull request or open a new issue