🌟 Continuous Attribute Discretization Algorithm

Implementation of a top-down discretization algorithm with a greedy criterion, which maximizes the number of object pairs with different labels (classes) separated by the split.

---

🚀 How does it work?

The algorithm performs the following steps:

1. Data loading: Input CSV file.
2. Continuous attribute analysis:
   • Generate all possible split points (averages between unique values).
   • Calculate the number of object pairs from different classes that are separated for each split point.
   • Select the split point with the maximum "separation gain."
3. Recursion: Repeat the process in subintervals until the desired number of intervals (n_bins) is reached.
4. Save the result: Create a new CSV file with discretized values.

---

🧪 Example

Input CSV file (test_data_large.csv):

distance,label
3.75,near
9.51,far
...

Running the algorithm (for 3 intervals):

python main.py

Terminal output:

2025-05-03 15:52:44 - INFO - Selected split 3.95 with gain 12
2025-05-03 15:52:44 - INFO - Selected split 7.8 with gain 12
...

Output CSV file: diskretized_test_data.csv

---

🛠️ Project structure

📂 Project
├── main.py                   # Main file to run the algorithm
├── utils.py                  # Helper functions: logging, I/O, timing
├── test_data.csv             # Example small dataset
├── test_data_large.csv       # Generated larger dataset for testing
├── diskretized_test_data.csv # Algorithm output

---

⚙️ How to use?

1. Place your CSV file in the project folder.

2. Ensure the distance and label columns exist (or adjust them in main.py).

3. Run the algorithm:

   python main.py

4. You can change the number of intervals (n_bins) in the discretization_alg() function.

---

📄 License

Project created for educational purposes. You are free to modify and use it. 🌱