Table 3 |
||||||
|
Accuracy of final classifiers for 'simultaneous' events in the cdc15 dataset |
||||||
|
Gene name |
cdc15, 20 genes |
cdc15, cyclins |
||||
|
|
||||||
|
10-CV |
cdc28 |
α |
10-CV |
cdc28 |
α |
|
|
CLN1 |
82.8% |
76.5% |
94.4% |
91.1% |
76.5% |
94.4% |
|
CLN2 |
69.9% |
88.2% |
77.8% |
73.5% |
88.2% |
77.8% |
|
CLB1 |
95.8% |
88.2% |
88.9% |
95.3% |
88.2% |
88.9% |
|
CLB2 |
95.8% |
88.2% |
77.8% |
95.0% |
88.2% |
77.8% |
|
CLB5 |
76.0% |
94.1% |
83.3% |
76.0% |
94.1% |
83.3% |
|
CLB6 |
83.7% |
88.2% |
88.9% |
84.4% |
88.2% |
88.9% |
|
SWI5 |
73.7% |
88.2% |
83.3% |
|||
|
|
||||||
|
Estimates are shown for C4.5 by Quinlan with wrappers by Kohavi on the cdc15 dataset with continuous features discretized by the Fayyad and Irani method. 10-CV, 10-fold cross-validation; cdc28 and α, accuracy estimations where cdc8 and alpha-factor datasets were used as test sets. See Materials and methods for the algorithm description. |
||||||
|
Soinov et al. Genome Biology 2003 4:R6 doi:10.1186/gb-2003-4-1-r6 |
||||||
