Multiclass Boosting with Adaptive Group-Based kNN and Its Application in Text Categorization

Algorithm 1

Pseudocode of DIWC Double iterative weighted cascading.

 Input: training set 𝑆={𝑠₁,𝑠₂,…,𝑠_𝑚} and weak

     classifier 𝐶={𝑐₁,𝑐₂,…,𝑐_𝑛}

 OutPut: strong classifier 𝐻

1   begin

2  test 𝑐₁ with 𝑠₁

3   for (𝑖=2; 𝑖≤𝑚; 𝑖++)

4    for (𝑗=2; 𝑗≤𝑛; 𝑗++)

5    test 𝑐_𝑗 with 𝑠, and 𝑒_𝑗2−1

6  calculate 𝑊¹_𝑗↲,

7  test 𝑐₁ with 𝑠_𝑚 and 𝑒_𝑚

8   for (𝑖=𝑚−1; 𝑖≥1; 𝑖−−)

9    for (𝑗=2; 𝑗≤𝑛; 𝑗++)

10    test 𝑐_𝑗 with 𝑠_𝑖−1 and 𝑒²_𝑖

11  calculate 𝑤²_𝑖

12  for (𝑗=1; 𝑗≤𝑚; 𝑗++)

13   if |𝑤¹_𝑖−𝑤²_𝑖|≤𝜀&&𝑤¹_𝑖,𝑤²_𝑖≥𝜎_ℎ↲

14    𝑤^′_𝑖=𝑤_MAX

15   else  if  |𝑤¹_𝑖−𝑤²_𝑖|≤𝜀&&𝑤¹_𝑖,𝑤²_𝑖≤𝜎₁

16    𝑤^′_𝑖=𝑤_MIN↲

17   else  if  𝜀≤|𝑤¹_𝑖−𝑤²_𝑖|≤𝛿

18    𝑊^′_𝑖=(𝑤¹_𝑖−𝑤²_𝑖)/2

19   else

20    int𝑡=0;𝑡++

21    while (𝑡≤1 )

22        goto  2↲

23    |𝑤¹_𝑖−𝑤²_𝑖|≤𝛿?𝑤^′_𝑖=(𝑤¹_𝑖−𝑤²_𝑖)/2∶𝑤^′_𝑖=𝑤_MIN

24   end