https://github.com/samuele-lolli/data-analytics-techniques

A practical approach to data analytics pipeline.
https://github.com/samuele-lolli/data-analytics-techniques

numpy pandas pytorch scikit-learn

Last synced: 2 months ago
JSON representation

A practical approach to data analytics pipeline.

• Host: GitHub
• URL: https://github.com/samuele-lolli/data-analytics-techniques
• Owner: samuele-lolli
• Created: 2025-01-10T03:30:46.000Z (over 1 year ago)
• Default Branch: master
• Last Pushed: 2025-01-10T19:55:08.000Z (over 1 year ago)
• Last Synced: 2025-03-15T14:12:36.708Z (over 1 year ago)
• Topics: numpy, pandas, pytorch, scikit-learn
• Language: Jupyter Notebook
• Homepage:
• Size: 10.5 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Data Analytics Techniques

This project uses a subset of the Million Song Database, which contains acoustic features and a `year` column. The objective of this task is to predict the year of release of a track based on its acoustic features.

The project covers the complete standard data analytics pipeline, including the following steps:

1. **Data Visualization**

2. **Preprocessing**

   - Handling outliers (Winsorization and other techniques)

   - Various types of scaling and normalization

   - Principal Component Analysis (PCA)

3. **Training Classical Machine Learning Models**

   - Random Forest

   - K-Nearest Neighbors (KNN)

   - Support Vector Machine (SVM)

   - Linear Regression

4. **Training a Feedforward Neural Network using PyTorch**

5. **Using PyTorch Tabular Models**

   - TabNet

   - TabTransformers

6. **Hyperparameter Tuning**

7. **Model Evaluation on Test Set**

For more information, you can view the following Jupyter Notebooks:

- `data-visualization.ipynb`

- `ml-sklearn.ipynb`

- `feedforward-pytorch.ipynb`

- `tabular-pytorch.ipynb`

This project demonstrates a comprehensive approach to data analytics, from preprocessing to model evaluation, using both classical machine learning models and more advanced neural network models.