---

CART Classifier with Streamlit Integration

This project implements a CART (Classification and Regression Tree) Classifier in Python, designed to handle classification tasks. It provides a robust framework for training, testing, and evaluating decision trees using gini or entropy criteria. A Streamlit web interface is integrated to provide a user-friendly platform for interacting with the classifier.

---

Features

Core Functionality

• Custom CART Implementation:
  • Supports both Gini Index and Entropy as splitting criteria.
  • Handles dataset partitioning, threshold calculation, and weighted impurity computations.
• Visualization:
  • Displays the confusion matrix and other metrics after evaluation.
• Interactive Tree Traversal:
  • Allows users to visualize the decision tree traversal logic.

Streamlit Integration

• User Interface:
  • Upload datasets via an intuitive Streamlit web app.
  • Configure parameters such as max_depth, min_samples, and criterion through sliders and dropdowns.
• Real-Time Evaluation:
  • Displays metrics like Precision, Recall, and F1-Score instantly after running the classifier.
  • Visualizes the decision tree structure and confusion matrix.

---

Installation

Prerequisites

• Python 3.8 or above

Steps

1. Clone the repository:

   git clone https://github.com/yourusername/cart-classifier.git
   cd cart-classifier

2. Install required libraries:

   The key libraries include:

   • Streamlit
   • numpy
   • pandas
   • scikit-learn
   • matplotlib

---

Usage

Running the CART Classifier

1. Modify the script CART_Classifier.py to load and train your dataset.
2. Train and evaluate the classifier by calling the methods:

   from CART_Classifier import CartClassifier
   from DataHandler import DataHandler
   
   data = DataHandler("path_to_dataset.csv", label_column="target_column")
   classifier = CartClassifier(max_depth=5, min_samples=2, criterion="gini")
   classifier.fit(data)
   precision, recall, f1 = classifier.evaluate(data, "conf_matrix.png")
   print(f"Precision: {precision}, Recall: {recall}, F1-Score: {f1}")

Running the Streamlit App

1. Start the Streamlit server:

   streamlit run app/main.py

2. Open your browser and navigate to http://localhost:8501.
3. Use the app to:
   • Upload a dataset.
   • Configure training parameters.
   • Train, evaluate, and visualize the decision tree.

---

Output

The Output generated is a code and an image representing the decision tree and the confusion matrix of the model

Example Code

 import sys
 
 def predict(x):
 
     if x[5] <= 9.795053004:
         if x[1] <= 9.0:
             if x[3] <= 21.0:
                 if x[0] <= 65.0:
                     return 0
                 else:
                     return 0
             else:
                 if x[0] <= 48.0:
                     return 1
                 else:
                     return 0
         else:
             if x[3] <= 13.0:
                 return 0
             else:
                 if x[3] <= 24.0:
                     return 1
                 else:
                     return 0
     else:
         if x[1] <= 7.0:
             if x[4] <= 18.0:
                 if x[4] <= 2.0:
                     return 1
                 else:
                     return 0
             else:
                 if x[0] <= 68.0:
                     return 1
                 else:
                     return 0
         else:
             if x[1] <= 11.0:
                 if x[4] <= 11.0:
                     return 1
                 else:
                     return 1
             else:
                 if x[3] <= 53.0:
                     return 1
                 else:
                     return 0
                 
 
 x = eval(sys.argv[1])
 result = predict(x)
 print(result)

---

Key Features in the Streamlit App

• Dataset Upload: Allows users to upload a CSV dataset.
• Model Configuration:
  • Adjust max_depth, min_samples, and criterion.
• Results Visualization:
  • View metrics (Precision, Recall, F1-Score).
  • Display and save the confusion matrix.
  • Visualize the decision tree traversal logic.

---

Example Workflow

1. Upload Dataset: Select a dataset via the Streamlit app.
2. Configure Parameters: Choose training parameters (e.g., max_depth = 5).
3. Train the Model: Run the CART classifier.
4. View Results: Check metrics, visualize the tree, and download outputs.

---

Future Enhancements

• Tree Visualization: Implement graphical tree visualization using libraries like Graphviz.
• Multi-Classifier Support: Add support for other decision tree algorithms.
• Model Export: Enable saving and loading of trained models.

Contact

For queries or contributions, reach out at aswajith707@gmail.com.

---