Streamlit ML Playground

Interactive machine learning app built with Streamlit and scikit-learn.
Train, compare, and visualize ML models directly in your browser — no backend server required.

🚀 Live Demo: ml-playground-visura.streamlit.app

---

💡 Overview

Training and comparing multiple machine learning models usually requires writing repetitive boilerplate code. The Streamlit ML Playground is a zero-code, interactive web application that allows users to instantly benchmark regression, classification, and clustering algorithms on any tabular dataset. It is designed for rapid prototyping, educational purposes, and quick data exploration.

---

🛠️ Tech Stack

• Frontend: Streamlit (Multi-page app architecture)
• Backend/ML: Python, Scikit-learn, Pandas, NumPy
• Visualization: Plotly / Matplotlib / Seaborn

Run Locally

pip install -r requirements.txt
streamlit run app.py

The app opens at http://localhost:8501

---

Pages

📁 Upload

• Upload your own CSV or load a built-in sample dataset
• Instant data preview — row count, column types, missing values
• Three sample datasets ready to use: House Prices, Loan Approval, Clustering

📉 Regression

• Trains 5 models in one click: Linear, Scaled Linear, Polynomial, Ridge, Lasso
• MSE comparison bar chart with best model highlighted
• Actual vs Predicted scatter plot
• Plain-English interpretation of scaling, polynomial, and regularization results

🎯 Classification

• Models: Logistic Regression, Decision Tree (tunable depth), Random Forest
• Metrics: Accuracy, Precision, Recall, F1 Score
• Confusion matrix heatmaps for each model
• Train vs Test accuracy comparison — overfitting detection built in

🔍 Unsupervised

• KMeans: Adjustable k, cluster distribution chart, 2D scatter by cluster
• PCA: Configurable components, explained variance chart, PC1 vs PC2 scatter
• All numeric features used automatically — no configuration needed

---

Sample Datasets

Dataset	Rows	Features	Best For
🏠 House Prices	50	6 numeric	Regression
💳 Loan Approval	50	6 numeric, binary target	Classification
🔵 Clustering	50	4 numeric	KMeans / PCA

---

Folder Structure

streamlit_app/
├── app.py                  # Home page
├── requirements.txt        # Dependencies
├── pages/
│   ├── 1_Upload.py         # Upload & preview
│   ├── 2_Regression.py     # Regression models
│   ├── 3_Classification.py # Classification models
│   └── 4_Unsupervised.py   # KMeans & PCA
└── utils/
    ├── api_client.py       # ML service layer (direct scikit-learn calls)
    └── sample_data.py      # Sample dataset loader

---

Screenshots

Home

Upload

Regression

Classification

Unsupervised

---

Troubleshooting

CSV upload failed — make sure the file is .csv format with numeric columns. Try a sample dataset first to confirm the app is working.

Classification target error — choose a column with a small number of unique values (binary or categorical). Continuous numeric columns won't work well as classification targets.

Clustering needs numeric data — if your dataset has only text columns, the unsupervised page will flag it.

---

👤 Connect with Me Built by Visura Rodrigo