Waste Cost Prediction Model 🏭💰

📊 Project Overview

This project implements a machine learning model to predict waste costs in manufacturing processes. By analyzing various production factors, the model aims to help industries optimize their processes and reduce waste-related expenses.

🚀 Features

• Data preprocessing and cleaning
• Feature engineering and selection
• Implementation of stacking regressor model
• Model evaluation and cross-validation
• Visualization of results
• Configurable parameters
• Comprehensive error handling and logging

🛠️ Installation

git clone https://github.com/dimplefrancis/waste-cost-prediction.git cd waste-cost-prediction pip install -r requirements.txt

🏃‍♂️ Usage

python main.py

📁 Project Structure

waste-cost-prediction/ │ ├── data/ │ ├── data_batch.csv │ └── data_fail.csv │ ├── data_preparation.py ├── feature_engineering.py ├── model.py ├── evaluation.py ├── utils.py ├── main.py └── README.md

📊 Results

The model's performance is evaluated using various metrics, including R-squared score, Mean Squared Error (MSE), and cross-validation scores. Visualizations of actual vs. predicted values are generated for both transformed and original scales.

Model Performance

• Cross-validation scores:

  • Fold 1: 0.95945306
  • Fold 2: 0.95269916
  • Fold 3: 0.92957072
  • Fold 4: 0.94759019
  • Fold 5: 0.94246247

• Mean Cross-validation R2 Score: 0.9463551220934384

• Standard Deviation of Cross-validation Scores: 0.010101455359369406

Model Performance (Original Scale)

• Mean Squared Error (MSE): 8162.65227882406
• Root Mean Squared Error (RMSE): 90.34739774240356
• Mean Absolute Error (MAE): 36.0736490097619
• R-squared (R2): 0.9676199484432447

Visualization of Results

The plot shows a strong correlation between actual and predicted values, indicating good model performance.

LIME Explanation of the Prediction

The LIME explanation provides insights into feature importance for individual predictions:

• Top positive contributors: Waste in ML, % Waste loss, Theoretical Yield Yield_Efficiency
• Top negative contributors: Theoretical Yield, G.R.Qty

This explanation helps understand which features are most influential in the model's predictions for specific instances.

Interpretation

1. High R2 Score: The model explains approximately 96.76% of the variance in the target variable, indicating excellent predictive power.
2. Consistent Performance: Cross-validation scores show consistent performance across different subsets of the data, with a high mean R2 of 0.9464.
3. Accurate Predictions: Low RMSE and MAE values suggest the model's predictions are close to actual values on average.
4. Feature Importance: The LIME explanation reveals that waste-related features and theoretical yield are crucial in predicting total waste cost.
5. Outlier Handling: The "After Outlier Removal" note in the plot title suggests that data preprocessing has improved model performance.

Implications

• The model demonstrates strong potential for accurate waste cost predictions in manufacturing processes.
• It can be a valuable tool for identifying factors that contribute most to waste costs, enabling targeted process improvements.
• The consistent performance across cross-validation folds indicates good generalization to unseen data.

🔧 Configuration You can adjust various parameters of the model and data processing in the config.py file, including:

Data file paths Missing value handling strategies Feature engineering parameters Model hyperparameters Logging settings

🐛 Error Handling The project implements comprehensive error handling and logging. Errors are categorized into DataError and ModelError, and all functions are decorated with an error_handler that logs any exceptions.

📝 License This project is MIT licensed.