1 #include "opencv2/ml/ml.hpp"
2 #include "opencv2/core/core_c.h"
3 #include "opencv2/core/utility.hpp"
10 "\nThis sample demonstrates how to use different decision trees and forests including boosting and random trees:\n"
14 "CvERTrees ertrees;\n"
15 "CvGBTrees gbtrees;\n"
16 "Call:\n\t./tree_engine [-r <response_column>] [-c] <csv filename>\n"
17 "where -r <response_column> specified the 0-based index of the response (0 by default)\n"
18 "-c specifies that the response is categorical (it's ordered by default) and\n"
19 "<csv filename> is the name of training data file in comma-separated value format\n\n");
23 static int count_classes(CvMLData& data)
25 cv::Mat r = cv::cvarrToMat(data.get_responses());
26 std::map<int, int> rmap;
27 int i, n = (int)r.total();
28 for( i = 0; i < n; i++ )
30 float val = r.at<float>(i);
31 int ival = cvRound(val);
36 return (int)rmap.size();
39 static void print_result(float train_err, float test_err, const CvMat* _var_imp)
41 printf( "train error %f\n", train_err );
42 printf( "test error %f\n\n", test_err );
46 cv::Mat var_imp = cv::cvarrToMat(_var_imp), sorted_idx;
47 cv::sortIdx(var_imp, sorted_idx, CV_SORT_EVERY_ROW + CV_SORT_DESCENDING);
49 printf( "variable importance:\n" );
50 int i, n = (int)var_imp.total();
51 int type = var_imp.type();
52 CV_Assert(type == CV_32F || type == CV_64F);
54 for( i = 0; i < n; i++)
56 int k = sorted_idx.at<int>(i);
57 printf( "%d\t%f\n", k, type == CV_32F ? var_imp.at<float>(k) : var_imp.at<double>(k));
63 int main(int argc, char** argv)
70 const char* filename = 0;
72 bool categorical_response = false;
74 for(int i = 1; i < argc; i++)
76 if(strcmp(argv[i], "-r") == 0)
77 sscanf(argv[++i], "%d", &response_idx);
78 else if(strcmp(argv[i], "-c") == 0)
79 categorical_response = true;
80 else if(argv[i][0] != '-' )
84 printf("Error. Invalid option %s\n", argv[i]);
90 printf("\nReading in %s...\n\n",filename);
100 CvTrainTestSplit spl( 0.5f );
102 if ( data.read_csv( filename ) == 0)
104 data.set_response_idx( response_idx );
105 if(categorical_response)
106 data.change_var_type( response_idx, CV_VAR_CATEGORICAL );
107 data.set_train_test_split( &spl );
109 printf("======DTREE=====\n");
110 dtree.train( &data, CvDTreeParams( 10, 2, 0, false, 16, 0, false, false, 0 ));
111 print_result( dtree.calc_error( &data, CV_TRAIN_ERROR), dtree.calc_error( &data, CV_TEST_ERROR ), dtree.get_var_importance() );
113 if( categorical_response && count_classes(data) == 2 )
115 printf("======BOOST=====\n");
116 boost.train( &data, CvBoostParams(CvBoost::DISCRETE, 100, 0.95, 2, false, 0));
117 print_result( boost.calc_error( &data, CV_TRAIN_ERROR ), boost.calc_error( &data, CV_TEST_ERROR ), 0 ); //doesn't compute importance
120 printf("======RTREES=====\n");
121 rtrees.train( &data, CvRTParams( 10, 2, 0, false, 16, 0, true, 0, 100, 0, CV_TERMCRIT_ITER ));
122 print_result( rtrees.calc_error( &data, CV_TRAIN_ERROR), rtrees.calc_error( &data, CV_TEST_ERROR ), rtrees.get_var_importance() );
124 printf("======ERTREES=====\n");
125 ertrees.train( &data, CvRTParams( 18, 2, 0, false, 16, 0, true, 0, 100, 0, CV_TERMCRIT_ITER ));
126 print_result( ertrees.calc_error( &data, CV_TRAIN_ERROR), ertrees.calc_error( &data, CV_TEST_ERROR ), ertrees.get_var_importance() );
128 printf("======GBTREES=====\n");
129 if (categorical_response)
130 gbtrees.train( &data, CvGBTreesParams(CvGBTrees::DEVIANCE_LOSS, 100, 0.1f, 0.8f, 5, false));
132 gbtrees.train( &data, CvGBTreesParams(CvGBTrees::SQUARED_LOSS, 100, 0.1f, 0.8f, 5, false));
133 print_result( gbtrees.calc_error( &data, CV_TRAIN_ERROR), gbtrees.calc_error( &data, CV_TEST_ERROR ), 0 ); //doesn't compute importance
136 printf("File can not be read");