Decision Trees from large Databases: SLIQ
Decision Trees from large Databases: SLIQ
Decision Trees from large Databases: SLIQ
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Compare: Information Gain vs. Gini Index<br />
• Which split is better?<br />
[29 +, 35 -]<br />
yes<br />
x 1<br />
x 2<br />
no<br />
• H L, y = − 29<br />
64 log 2 29<br />
64 + 35<br />
64 log 2 35<br />
64<br />
= 0.99<br />
[29 +, 35 -]<br />
[21+, 5 -] [8+, 30 -] [18+, 33 -] [11+, 2 -]<br />
• IG L, x 1 = 0.99 − 26<br />
64 H L x 1 =yes, y + 38<br />
64 H L x 1 =no, y ≈ 0.26<br />
• IG L, x 2 = 0.99 − 51<br />
64 H L x 2 =yes, y + 13<br />
64 H L x 2 =no, y ≈ 0.11<br />
1 − 8 38<br />
• Gini x1 L = 26<br />
64 Gini L x 1 =yes + 38<br />
64 Gini L x 1 =no ≈ 0.32<br />
• Gini x2 L = 51<br />
64 Gini L x 2 =yes + 13<br />
64 Gini L x 2 =no ≈ 0.42<br />
43<br />
yes<br />
2<br />
+<br />
30<br />
38<br />
2<br />
no<br />
≈ 0.33<br />
Sawade/Landwehr/Prasse/Scheffer, Maschinelles Lernen