Follow these steps to complete the Customer Churn Prediction simulation using Decision Tree and Random Forest:

Step 1: Dataset Exploration

• Review the Feature Dictionary to understand the attributes (Tenure, Contract, Monthly Charges) used to predict customer behavior.
• Explore the Raw Dataset table to identify patterns in the target variable (Churn).
• Click "Next Step: Preprocessing" at the bottom of the navigation panel to continue.

Step 2: Data Preprocessing

• Click "Run Full Preprocessing" on the left panel to begin the automated data transformation pipeline.
• The simulation will sequentially execute the following steps:
  • Handle Missing Values: Removes or imputes missing data points to ensure dataset integrity.
  • Label Encoding: Converts categorical variables (like "Contract") into numerical representations.
  • Feature Standardization: Scales numerical features to a uniform range for optimal model performance.
• Observe the Data Transformation Pipeline animation to visualize these changes in real-time.
• Click "Next Step: Data Splitting" once the process is 100% complete.

Step 3: Dataset Splitting

• Use the Split Ratio Slider to divide your data into Training and Testing sets.
• A common ratio is 80:20 (80% for training the model and 20% for evaluating it).
• Observe the Live Calculation results to see how the records are distributed between the two sets.
• Click "Next Step: DT Parameters" to begin model configuration.

Step 4: Decision Tree Configuration

• Configure the Maximum Depth and other Splitting Criteria like Gini or Entropy.
• Adjust the Tree Depth and Pruning controls to manually tune model complexity and reduce overfitting risks.
• Observe the Tree Structure Preview to see how your depth selection affects the potential branching of the tree.
• Click "Next Step: DT Training" to proceed to the training phase.

Step 5: Decision Tree Training

• Click "Start Training" to execute the recursive partitioning algorithm on your training data.
• Visualize the resulting Decision Tree Structure and see how the model makes specific decisions at each node.
• Review the Feature Importance scores to identify which factors (like "Contract" or "Tenure") impact churn the most.
• Click "Next Step: RF Parameters" to compare this with ensemble methods.

Step 6: Random Forest Configuration

• Set the Number of Trees (n_estimators) and the Maximum Depth per individual tree.
• Select a Feature Strategy (like SQRT) to ensure feature diversity and improve the stability of the ensemble.
• Observe the Random Forest Structure diagram to understand how bootstrap samples are assigned to different trees.
• Click "Next Step: RF Training" to begin the training process.

Step 7: Random Forest Training

• Click "Start Training" to begin the ensemble voting process for the forest.
• Observe the Tree Voting Simulation as multiple trees independently predict outcomes for a specific test customer.
• Check the Prediction Confidence and the probability breakdown between "Stay" and "Churn" classes.
• Click "Next Step: Final Evaluation" to view the summary results.

Step 8: Model Comparison & Evaluation

• Review the Accuracy Comparison and analyze the Confusion Matrix Legend (TP, TN, FP, FN) to identify the top performer.
• Compare the Performance Metrics (Precision and Recall) to evaluate the prediction quality of both models.
• Click "Finished" to complete the laboratory session and review the final conclusion.