Model Training & Testing

How to Train an AI Model

Training is the process of teaching the model to recognize patterns in data.

Steps:

1. Collect Data

• Example: House prices dataset with features (size, rooms, location) and target (price).

2. Preprocess Data

• Handle missing values, scale numbers, encode categories.

3. Split Data

• Usually 80% training, 20% testing.

4. Choose an Algorithm/Model

• Regression (predict numbers),
• Classification (predict categories),
• Neural networks (deep learning tasks).

5. Fit the Model (Training)

• The model learns patterns from the training dataset.
• Example: A decision tree learns rules like “if size > 2000 sqft → price higher”.

How to Test an AI Model

Testing means evaluating the trained model on data it has never seen before.

Steps:

1. Use Test Data (20%)

• Provide unseen examples to the model.
• Example: Give house features → model predicts price.

2. Compare Predictions with Actual Values

• If actual price = 300,000 and model predicted 290,000, it’s close.
• Collect errors for all test samples.

3. Evaluate Using Metrics

• Accuracy (for classification).
• Precision/Recall/F1-score (for imbalanced classes).
• Mean Squared Error (MSE) or R² Score (for regression).

4. Training vs. Testing (Key Difference)

• Training = Model learns from known data.
• Testing = Model performance is measured on new, unseen data.

✅ Summary:

• Collect & preprocess data.
• Split into training & testing sets.
• Train (fit) the model on training data.
• Test it with unseen data.
• Evaluate with metrics.