The aim of this project was to build a multi-output regression pipeline to predict the coordinates (x, y) indicating where a particle of interest passed, based on data collected from signals in an RSD sensor (Resistive Silicon Detector).

A variety of preprocessing techniques, dimensionality reduction methods, and ML models were tested. A Random Forest Regressor achieved the best score, outperforming the competition baseline.

Best result: Euclidean Distance = 5.417 (baseline: 6.629)

Particles_RSD_Sensors_RegressionTask/
├── data/
│   ├── raw/                  ← development.csv, evaluation.csv (not included)
|
├── models/                   ← saved trained model (.pkl, generated on run)
├── notebooks/                ← EDA, experiments, model comparison
├── outputs/                  ← predictions CSV and hexbin plot (generated on run)
├── reports/                  ← full project report (PDF)
├── src/
│   ├── __init__.py
│   ├── data_loader.py        ← loading and cleaning
│   ├── preprocessor.py       ← feature selection and scaling
│   ├── trainer.py            ← model training and evaluation
│   └── predictor.py          ← inference and output generation
├── main.py                   ← full pipeline entry point
└── requirements.txt

Note: data/raw/, models/, and outputs/ are not included in the repository due to file size from the actual training. Add your own development.csv and evaluation.csv to data/raw/ before running (find here: https://drive.google.com/drive/folders/1O8rqMLWaUhkS8Cn6uicfMAGA1DME1JPH?usp=sharing). Models and outputs are populated automatically by main.py.

Base experiments, EDA, and model comparisons are documented in notebooks/Report_ParticlesinRSDSensors.ipynb. This includes all tested configurations: linear models (Linear Regression, Lasso, Ridge), PCA experiments, Decision Tree, and Random Forest variants.

The final best model pipeline is implemented as a clean, reproducible Python package in src/ and runs end-to-end via main.py.

# 1. Clone the repo
git clone https://github.com/siinvictus/RegressionTask_ParticlesinRSDSensors.git
cd Particles_RSD_Sensors_RegressionTask

# 2. Add your data
# Place development.csv and evaluation.csv in data/raw/

# 3. Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate

# 4. Install dependencies
pip install -r requirements.txt

# 5. Run the pipeline
python main.py

Output files will be saved to outputs/. Trained model will be saved to models/.

Python, Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn

Silva Bashllari & Baharak Qaderi — Politecnico di Torino

Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)