Incremental Optimization Mechanism for Constructing a Decision Tree in Data Stream Mining

Table 2

The comparison between VFDT and OVFDT.

Approach	VFDT	OVFDT
Testing	Sort new data by current HT; Update the sufficient statistics; FTL of MC, NB classifiers; Assign a predicted class by FTL	Sort new data by current HT; Update the sufficient statistics; FTL of MC, NB, and WNB classifiers; Assign a predicted class by FTL; Update incremental; Sequential  error
Training	Check node splitting by HB; Check node splitting by fixed 𝜏; Tree model HT update	Check node splitting by HB; Check node splitting by  adaptive 𝜏; Check node splitting by  statistical error; Tree model HT update

∗FTL is functional tree leaf; MC is Majority Class, NB is Naïve Bayes, and WNB is Weighted Naïve Bayes; HT is the decision tree using a Hoeffding bound.