Little Bits of Artificial Intelligence

Step 1: Prepare our Data

• Organize our data as a table: features → target value
• Features can be numeric (e.g., area) or categorical (e.g., location)
• Target must be a continuous numeric value (e.g., price, rent)

Step 2: Choose a Splitting Criterion

• For every feature:
  • Try possible split points (e.g., area > 1200)
  • Divide the data into left/right groups
• Calculate the total squared error after the split:
  

  Error=∑(actual−predicted)2

Step 3: Find the Best Split

• Choose the feature and value that minimizes total error
• This becomes a decision node

Step 4: Recursively Split the Data

• Apply the same logic to each child node (left/right)
• Keep splitting until:
  • Maximum depth reached, or
  • Number of samples is small, or
  • Target values are similar (node is “pure”)

Step 5: Assign Prediction at Leaf Nodes

• For each leaf (end of a branch):
  • Set the prediction as the mean of target values in that group

Step 6: Make Predictions

For any new input:
• Start at the root
• Traverse the tree based on feature comparisons
• Reach a leaf, return the prediction stored there

Optional Enhancements

• Add pruning to prevent overfitting
• Use categorical feature encoding
• Combine with Random Forest for ensemble learning