Algorithm 2. Construction of Random Forest

Input: Originally collected CSI data packet$\mathit{D}$, each data packet contains$\mathit{j}$data for$\mathit{N}$ test cases

Input: the size of the forest: s

Output: random forest: F

1. generate training dataset $D_{train}$ by Wavelet transform

2. for $i$ = 1 to s do

3. Generate new training dataset $D_{train}^i(i=1,2,\cdots s)$ by bootstrap aggregating

4. set $m={\log }_{2}M$

5. randomly select m attributes from $D_{train}^i$

6. use $D_{train}^i$ train the $i_{th}$ $DT$ based Algorithm1

7. end

8. combine the s Decision Trees on the basic thought of voting method.

9. return $F$