Random Forest Based Coarse Locating and KPCA Feature Extraction for Indoor Positioning System

Algorithm 1

Initialization: setting the number of class ; number of RPs ; number of features (i.e., number of APs);

number of features at a node of decision tree, where .

FOR Each decision tree

Selecting a subset (with replacement) of radio map dataset

randomly with known label of class (i.e., to randomly select RPs with its class labels, where ).

The rest part of radio map is reserved to test the error rate.

  FOR each node of the tree

  Selecting features randomly to make the criterion at the node

  Calculating the best split accordingly

  END FOR

END FOR

END