[Verse 1]
Starting with a dataset messy and raw
Split the chaos using entropy's law
Information gain reveals the perfect cut
Which attribute makes the cleanest strut
ID3 algorithm hunting for the best
Calculate the gain, forget about the rest
Categorical features only in this game
Numeric values need a different frame
[Chorus]
Entropy down, information up
Split until you fill that learning cup
Gain ratio when the bias gets loud
C4.5 handles what ID3 won't allow
Branch by branch we carve the knowledge tree
Every leaf holds pure certainty
[Verse 2]
C4.5 evolved beyond the basic scheme
Continuous attributes living the dream
Binary splits on numeric thresholds tight
Pruning overgrown branches left and right
Gain ratio fixes the bias that we hate
When attributes have values by the crate
Missing data handled with a weighted trick
Probability splits keep the logic slick
[Chorus]
Entropy down, information up
Split until you fill that learning cup
Gain ratio when the bias gets loud
C4.5 handles what ID3 won't allow
Branch by branch we carve the knowledge tree
Every leaf holds pure certainty
[Bridge]
Stop condition triggers when the gain runs dry
Minimum samples or the purity's high
Recursive splitting like a fractal bloom
Each node spawning children in the room
Pessimistic pruning cuts the deadweight back
Keep the model sharp and on the track
[Verse 3]
Gini index alternative to entropy's rule
Different metrics but the same basic tool
Majority class labels every final node
Confidence measures how the model flowed
Cross-validation tests the tree's true might
Accuracy metrics prove the splits were right
[Outro]
From root to leaves the wisdom crystallized
Dataset transformed and classified
ID3 foundation, C4.5 refined
Decision trees with algorithmic mind