🔹Exploratory Data Analysis (EDA) using Pandas, Matplotlib, and Seaborn
🔹Feature Engineering and Data Preprocessing including one-hot encoding and dataset transformation
🔹Predictive Modeling with Scikit-learn
🔹Classification Algorithms
- Logistic Regression
- Random Forest Classifier
- Gradient Boosting Classifier
🔹Model Evaluation Techniques
- Accuracy Score
- Classification Report
- Confusion Matrix
- ROC Curve Analysis
🔹Feature Importance Analysis to identify the key factors driving supply chain disruptions
Supply chains depend on coordination between suppliers, manufacturing, transportation, and inventory systems. Delays or inefficiencies in any stage can disrupt operations, increase logistics costs, and affect customer satisfaction.
Organizations often detect these disruptions only after they occur, even though operational data frequently contains early warning signals.
This project builds a machine learning model to predict supply chain disruption risk using operational logistics data. The analysis uses variables such as lead times, supplier performance, defect rates, production volumes, transportation modes, and inventory levels to identify patterns linked to delays.
Through data exploration, visualization, and predictive modeling, the project addresses two key questions:
- Which operational factors drive supply chain disruptions?
- Can historical supply chain data predict disruption risk before delays occur?
Using Python and scikit-learn, the model identifies high-risk scenarios and highlights the variables that most strongly influence disruption risk, providing insights that support more proactive supply chain management.
The dataset contains operational supply chain metrics such as:
- Product type
- SKU
- Price
- Availability
- Number of products sold
- Revenue generated
- Stock levels
- Lead times
- Order quantities
- Shipping times
- Shipping carriers
- Supplier name
- Location
- Production volumes
- Manufacturing lead time
- Manufacturing costs
- Inspection results
- Defect rates
- Transportation modes
- Routes
- Shipping costs
These variables describe supplier performance, logistics operations, production efficiency, and demand patterns.
The project follows a structured machine learning workflow.
🔹 Exploratory Data Analysis
Explored the dataset using summary statistics, distribution plots, and scatterplots to understand relationships between logistics variables such as lead time and shipping costs. Supplier-level comparisons and correlation analysis helped identify patterns in operational performance.
🔹 Disruption Label Creation
The dataset did not contain a disruption label, so a binary target variable Disruption_Risk was created. Records with lead times above the dataset average were labeled as disruption events.
🔹 Data Preparation
Identifier fields such as SKU were removed. Categorical variables (supplier, transport mode, location, etc.) were converted into numerical features using one-hot encoding to make them usable for machine learning models.
🔹 Model Development
The dataset was split into training and testing sets (80/20). Multiple classification models were trained, including Random Forest, Logistic Regression, and Gradient Boosting, to predict disruption risk.
🔹 Model Evaluation
Model performance was evaluated using accuracy, classification reports, confusion matrices, and ROC curves. This allowed comparison of model performance and detection of misclassification patterns.
🔹 Model Interpretation
Feature importance analysis from the Random Forest model identified the variables that contribute most strongly to disruption risk, highlighting operational factors such as lead time, defect rates, and manufacturing costs.
🔹 Lead Time Drives Disruption Risk
Lead time emerged as the strongest indicator of disruption. Shipments with longer lead times were consistently classified as high-risk, confirming that delays in supplier fulfillment directly affect supply chain stability.
🔹 Supplier Performance Varies
Supplier-level analysis showed noticeable differences in shipping costs and operational efficiency. Some suppliers consistently exhibited higher disruption risk, indicating that supplier selection plays a significant role in logistics reliability.
🔹 Quality Issues Affect Reliability
Higher defect rates showed a relationship with disruption risk. Quality problems can introduce delays in production, inspection, and shipment, increasing the likelihood of operational disruptions.
🔹 Production and Manufacturing Costs Matter
Manufacturing cost variables contributed to disruption predictions, suggesting that inefficient or expensive production processes may correlate with slower fulfillment cycles.
🔹 Inventory Availability Impacts Stability
Lower product availability and stock levels appeared in higher-risk disruption cases, highlighting the importance of inventory management in preventing supply chain delays.
Multiple classification models were trained and compared, including Random Forest, Logistic Regression, and Gradient Boosting.
The Random Forest model produced the strongest results on the dataset and was used for feature importance analysis.
Model evaluation included:
- Accuracy score
- Classification report
- Confusion matrix
- ROC curve analysis
These metrics provided both performance measurement and insight into prediction errors.
Supply chain disruptions rarely occur without warning. Operational metrics such as lead time, supplier performance, production costs, and defect rates reveal clear signals that can help anticipate delays.
This project demonstrates how machine learning can transform operational supply chain data into predictive insight. By identifying the factors most associated with disruptions, organizations can detect risks earlier and make more informed logistics and supplier decisions.
Predictive models like this allow supply chain teams to shift from reactive problem-solving to proactive risk management.