A Dynamic Ensemble Framework for Mining Textual Streams with Class Imbalance

Algorithm 3

Adaptive selection method.

Output: The optimal CF

Input: 𝐷_𝑡: The data chunk at each time stamp

     𝑀: The number of CTs in CF

 (1) At the 𝑡-th time stamp:

 (2) Obtain current original clustering forest 𝑓^𝐸 relying on Algorithm 1;

 (3) For𝑓_𝑖∈𝑓^𝐸 do:

 (4) Estimate the accuracy of 𝑓_𝑖 using 𝐷_𝑡;

 (5) Compute the accuracy weight 𝜔_𝑖 of 𝑓_𝑖 with (1);

 (6) If 𝜔_𝑖>𝜔_𝜃:

 (7) REMOVE(𝑓_𝑖);

   UPDATE(𝑓^𝐸);

 (8) Endif