Developing Fine-Grained Actigraphies for Rheumatoid Arthritis Patients from a Single Accelerometer Using Machine Learning

. 2017 Sep 14;17(9):2113. doi: 10.3390/s17092113

Algorithm 1 Dichotomy mapped forest—Training

Input: X,Y
Output: trained model, map &

S e l_{n o d e}^{f e a t}

1:
for Each node in logic three according to the dichotomy in Figure 3 do
2:
Split X into $X_{A_{n o d e}}$ and $X_{B_{n o d e}}$ acording to the node label grouping in each node in Figure 3
3:
[ $X_{A_{n o d e}}^{s e l}$ , $X_{B_{n o d e}}^{s e l}$ , $S e l_{n o d e}^{f e a t}$ ] ← call SFFS (input=[ $X_{A_{n o d e}}$ , $X_{B_{n o d e}}$ ], labels = [ $A_{n o d e}$ , $B_{n o d e}$ ])
4:
if mapping == DL then
5:
[ $m a p (n o d e)$ , $m a p p e d_{A}$ , $m a p p e d_{B}$ ] ← call DAUT (input = [ $X_{A_{n o d e}}^{s e l}$ , $X_{B_{n o d e}}^{s e l}$ ], labels = [ $A_{n o d e}$ , $B_{n o d e}$ ])
6:
else if mapping == Metric then
7:
[ $m a p (n o d e)$ , $m a p p e d_{A}$ , $m a p p e d_{B}$ ] ← call MCM (input = [ $X_{A_{n o d e}}^{s e l}$ , $X_{B_{n o d e}}^{s e l}$ ], labels = [ $A_{n o d e}$ , $B_{n o d e}$ ])
8:
end if
9:
$m o d e l (n o d e)$ ← call train RF (input = [ $m a p p e d_{A}$ , $m a p p e d_{B}$ ], labels = [ $A_{n o d e}$ , $B_{n o d e}$ ])
10:
end for