>
Algorithm Comparison
Algorithm Comparison
Question
OneR
NaiveBayes
ID3
J48
Prism
Linear Regression
IB-K
Can
Predict Categories
Yes
Yes
Yes
Yes
Yes
No, modification possible
00perform regression for each class, predict highest
Yes
Can
make numeric predictions
No
No
No
No
No, no chance
Yes
Yes
Can
use nominal attributes in training
Yes
yes
Yes
Yes
Yes
No. See Logistic for handling
nominals
Yes
Can
use numeric attributes in training
Discretizes them using own
algorithm
Yes, using probability distribution
function
No
Yes
No, but modification
possible
Yes
yes
Can
handle missing values
Yes
Yes
No
Yes
No
Yes
Yes
Resistant
to 00oise00/b>
no
no
no
yes
No, somewhat better if
reduce accuracy threshold below 100%
yes
Yes, with k > 1
download Algorithm Comparison
