BMLP Final Project - Clustering & Classification

End-to-end machine learning notebooks for unsupervised clustering and supervised classification. The project includes data loading, preprocessing, PCA + KMeans clustering, Decision Tree and Random Forest classifiers, evaluation metrics/plots, hyperparameter tuning, and persisted artifacts.

Features

• Data loading from CSV datasets (data_clustering.csv, data_clustering_inverse.csv).
• Preprocessing: LabelEncoder for categorical features and StandardScaler for numeric scaling.
• Clustering: PCA dimensionality reduction + KMeans; includes silhouette score and cluster analysis.
• Classification: DecisionTreeClassifier and RandomForestClassifier with train/test split.
• Evaluation: Accuracy, Precision, Recall, F1-score, and confusion matrix visualizations (seaborn/matplotlib).
• Hyperparameter tuning: RandomizedSearchCV for Random Forest; best estimator saved.
• Artifacts (saved via joblib using .h5 filenames):
  • decision_tree_model.h5
  • explore_random_forest_classification.h5
  • tuning_classification.h5
  • model_clustering.h5
  • PCA_model_clustering.h5

Tech Stack

• Python 3 (tested with 3.9–3.11)
• Jupyter Notebook / JupyterLab
• NumPy, pandas
• scikit-learn
• seaborn, matplotlib
• joblib

Installation

1. Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

2. Install dependencies:

pip install --upgrade pip
pip install jupyter numpy pandas scikit-learn seaborn matplotlib

How to Run

Option A - VS Code

• Open this folder in VS Code.
• Open [Klasifikasi]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb or [Clustering]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb.
• Select a Python 3 kernel, then “Run All”.

Option B - Jupyter (CLI)

jupyter lab
# or
jupyter notebook

Then open:

• Classification: [Klasifikasi]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb
• Clustering: [Clustering]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb

Notes for running:

• Keep the working directory at the repository root so relative data paths resolve.
• The notebooks save models automatically to the repo root (files listed above).

Project Structure (reference)

.
├─ [Klasifikasi]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb
├─ [Clustering]_Submission_Akhir_BMLP_Nelson_Ahli.ipynb
├─ data_clustering.csv
├─ data_clustering_inverse.csv
├─ decision_tree_model.h5
├─ explore_random_forest_classification.h5
├─ tuning_classification.h5
├─ model_clustering.h5
├─ PCA_model_clustering.h5
└─ README.md

Notes

• The model artifacts use the .h5 extension for consistency with prior naming conventions, but are actually saved using joblib.dump(...) (not HDF5 format). This is a legacy naming choice from the notebooks.
• Exact metrics and plots are produced inside the notebooks.