Comparison of classification algorithms C4.5 and C5.0

This work compares features of tree decision algorithms C4.5 and C5.0, which are the most effective data mining classification tool. We considered two software tools: analytics platform Deductor and system See5. Three data sets were tested to improve comparative analysis accuracy. First is conventional Fisher’s iris data set, second contains information about US Congress deputy votes (distribution Deductor), and third includes information about applicants of the one of Russian Federation universities. According to test results, C5.0 builds more compact decision trees, but its operation speed is almost the same to C4.5 under reducing of classification model validity. However, we do not preclude that these results can be explained by using of See5 system demo version that provides only files processing with no more 400 entries.