wolffd@0: function [score,outputs] = evaluate(CPD, fam, data, ns, cnodes)
wolffd@0: % Evaluate evaluate the performance of the classification/regression tree on given complete data
wolffd@0: % score = evaluate(CPD, fam, data, ns, cnodes)
wolffd@0: %
wolffd@0: % fam(i) is the node id of the i-th node in the family of nodes, self node is the last one
wolffd@0: % data(i,m) is the value of node i in case m (can be cell array).
wolffd@0: % ns(i) is the node size for the i-th node in the whold bnet
wolffd@0: % cnodes(i) is the node id for the i-th continuous node in the whole bnet
wolffd@0: %  
wolffd@0: % Output
wolffd@0: % score is the classification accuracy (for classification) 
wolffd@0: %          or mean square deviation (for regression)
wolffd@0: %            here for every case we use the mean value at the tree leaf node as its predicted value
wolffd@0: % outputs(i) is the predicted output value for case i
wolffd@0: %
wolffd@0: % Author: yimin.zhang@intel.com
wolffd@0: % Last updated: Jan. 19, 2002
wolffd@0: 
wolffd@0: 
wolffd@0: if iscell(data)
wolffd@0:   local_data = cell2num(data(fam,:));
wolffd@0: else
wolffd@0:   local_data = data(fam, :);
wolffd@0: end
wolffd@0: 
wolffd@0: %get local node sizes and node types
wolffd@0: node_sizes = ns(fam);
wolffd@0: node_types = zeros(1,size(ns,2)); %all nodes are disrete
wolffd@0: node_types(cnodes)=1;
wolffd@0: node_types=node_types(fam);
wolffd@0: 
wolffd@0: fam_size=size(fam,2);
wolffd@0: output_type = node_types(fam_size);
wolffd@0: 
wolffd@0: num_cases=size(local_data,2);
wolffd@0: total_error=0;
wolffd@0: 
wolffd@0: outputs=zeros(1,num_cases);
wolffd@0: for i=1:num_cases
wolffd@0:   %class one case using the tree
wolffd@0:   cur_node=CPD.tree.root;  % at the root node of the tree
wolffd@0:   while (1)
wolffd@0:     if (CPD.tree.nodes(cur_node).is_leaf==1)
wolffd@0:       if (output_type==0) %output is discrete
wolffd@0:         %use the class with max probability as the output  
wolffd@0:         [maxvalue,class_id]=max(CPD.tree.nodes(cur_node).probs);
wolffd@0:         outputs(i)=class_id;
wolffd@0:         if (class_id~=local_data(fam_size,i))
wolffd@0:           total_error=total_error+1;
wolffd@0:         end
wolffd@0:       else   %output is continuous
wolffd@0:         %use the mean as the value
wolffd@0:         outputs(i)=CPD.tree.nodes(cur_node).mean;
wolffd@0:         cur_deviation = CPD.tree.nodes(cur_node).mean-local_data(fam_size,i);
wolffd@0:         total_error=total_error+cur_deviation*cur_deviation;
wolffd@0:       end
wolffd@0:       break;
wolffd@0:     end
wolffd@0:     cur_attr = CPD.tree.nodes(cur_node).split_id; 
wolffd@0:     attr_val = local_data(cur_attr,i);
wolffd@0:     if (node_types(cur_attr)==0)  %discrete attribute
wolffd@0:         % goto the attr_val -th child
wolffd@0:         cur_node = CPD.tree.nodes(cur_node).children(attr_val);
wolffd@0:     else
wolffd@0:         if (attr_val <= CPD.tree.nodes(cur_node).split_threshhold)
wolffd@0:           cur_node = CPD.tree.nodes(cur_node).children(1);
wolffd@0:         else
wolffd@0:           cur_node = CPD.tree.nodes(cur_node).children(2);  
wolffd@0:         end
wolffd@0:     end
wolffd@0:     if (cur_node > CPD.tree.num_node)
wolffd@0:       fprintf('Fatal error: Tree structure corrupted.\n');
wolffd@0:       return;
wolffd@0:     end
wolffd@0:   end
wolffd@0:   %update the classification error number
wolffd@0: end
wolffd@0: if (output_type==0)
wolffd@0:   score=1-total_error/num_cases;
wolffd@0: else
wolffd@0:   score=total_error/num_cases;
wolffd@0: end