🗳 ML_PackagingMonitor Project

📘 Overview

This project demonstrates a Voting Ensemble machine learning pipeline for predicting a binary target on an imbalanced dataset.
The dataset contains a mix of numeric and categorical features, requiring preprocessing, one-hot encoding, and scaling. The main goal is to predict disruption events (0/1) accurately.

🧠 Project Highlights

• Type: Classification (Binary)
• Language: Python 🐍
• Libraries: Pandas, NumPy, Scikit-learn, Imbalanced-learn, Matplotlib, Seaborn
• Dataset: disruption_dataset.csv — mix of numeric & categorical features

⚙️ Workflow

1. Data Cleaning & Preprocessing – handle missing values, encode categorical features.
2. EDA (Exploratory Data Analysis) – visualize distributions, density plots, boxplots, and heatmaps.
3. Train/Test Split – separate features (X) and target (y) for evaluation.
4. Spot-Check Models – baseline performance with LR, LDA, SVM, and KNN.
5. Voting Ensemble – combine LR, LDA, SVM, and Random Forest with hard/soft voting.
6. SMOTENC Oversampling – balance classes to improve minority class prediction.
7. Threshold Tuning – adjust decision threshold for better class recall.
8. Model Evaluation – Accuracy, Confusion Matrix, Classification Report.

🚧 Challenges

• Imbalanced dataset led to poor recall for minority class.
• Categorical + numeric mix required careful encoding and handling in SMOTENC.
• Hyperparameter tuning of ensemble weights needed to optimize predictions.

📊 Results

✅ Best Model: Voting Ensemble (LR + LDA + SVM + Random Forest)
🎯 CV Accuracy: ~73.8%
📉 Test Accuracy: ~72.9%
⚖️ Minor class prediction still low due to extreme imbalance; SMOTENC improved balance but careful thresholding is necessary.

💡 Skills Used

🐍 Python
🤖 Machine Learning
📊 Data Science
📈 Ensemble Learning & Model Optimization
📉 Handling Imbalanced Data

📁 Files Included

• packaging_disruption.ipynb → Main notebook with preprocessing, training, and evaluation
• packaging_center_records.csv → Dataset (numeric + categorical)

📬 Contact

📧 amirhossin6825@gmail.com
💬 Telegram: @AmirHossin6825