rmontanana/Platform
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add train classification report

Browse Source
This commit is contained in:
Ricardo Montañana Gómez
2024-05-14 11:45:54 +02:00
parent 99c9c6731f
commit 5c190d7c66
6 changed files with 85 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
src/main/Experiment.cpp
View File

@@ -102,6 +102,7 @@ namespace platform {
auto edges = torch::zeros({ nResults }, torch::kFloat64);
auto num_states = torch::zeros({ nResults }, torch::kFloat64);
json confusion_matrices = json::array();
json confusion_matrices_train = json::array();
std::vector<std::string> notes;
Timer train_timer, test_timer;
int item = 0;
@@ -150,8 +151,12 @@ namespace platform {
train_time[item] = train_timer.getDuration();
double accuracy_train_value = 0.0;
// Score train
if (!no_train_score)
accuracy_train_value = clf->score(X_train, y_train);
if (!no_train_score) {
auto y_predict = clf->predict(X_train);
Scores scores(y_train, y_predict, states[className].size(), labels);
accuracy_train_value = scores.accuracy();
confusion_matrices_train.push_back(scores.get_confusion_matrix_json(true));
}
// Test model
if (!quiet)
showProgress(nfold + 1, getColor(clf->getStatus()), "c");
@@ -183,6 +188,8 @@ namespace platform {
partial_result.setNodes(torch::mean(nodes).item<double>()).setLeaves(torch::mean(edges).item<double>()).setDepth(torch::mean(num_states).item<double>());
partial_result.setDataset(fileName).setNotes(notes);
partial_result.setConfusionMatrices(confusion_matrices);
if (!no_train_score)
partial_result.setConfusionMatricesTrain(confusion_matrices_train);
addResult(partial_result);
}
}

1
src/main/PartialResult.h
View File

@@ -28,6 +28,7 @@ namespace platform {
return *this;
}
PartialResult& setConfusionMatrices(const json& confusion_matrices) { data["confusion_matrices"] = confusion_matrices; return *this; }
PartialResult& setConfusionMatricesTrain(const json& confusion_matrices) { data["confusion_matrices_train"] = confusion_matrices; return *this; }
PartialResult& setHyperparameters(const json& hyperparameters) { data["hyperparameters"] = hyperparameters; return *this; }
PartialResult& setSamples(int samples) { data["samples"] = samples; return *this; }
PartialResult& setFeatures(int features) { data["features"] = features; return *this; }

34
src/main/Scores.cpp
View File

@@ -126,24 +126,28 @@ namespace platform {
oss << std::setw(dlen) << std::setprecision(ndec) << std::fixed << recall << " ";
}
oss << std::setw(dlen) << std::setprecision(ndec) << std::fixed << f1_score << " "
<< std::setw(dlen) << std::right << support << std::endl;
<< std::setw(dlen) << std::right << support;
return oss.str();
}
std::string Scores::classification_report(std::string color)
std::vector<std::string> Scores::classification_report(std::string color, std::string title)
{
std::stringstream oss;
std::vector<std::string> report;
for (int i = 0; i < num_classes; i++) {
label_len = std::max(label_len, (int)labels[i].size());
}
oss << Colors::GREEN() << "Classification Report" << std::endl;
oss << "=====================" << std::endl << color;
oss << std::string(label_len, ' ') << " precision recall f1-score support" << std::endl;
oss << std::string(label_len, ' ') << " ========= ========= ========= =========" << std::endl;
report.push_back("Classification Report using " + title + " dataset");
report.push_back("=========================================");
oss << std::string(label_len, ' ') << " precision recall f1-score support";
report.push_back(oss.str()); oss.str("");
oss << std::string(label_len, ' ') << " ========= ========= ========= =========";
report.push_back(oss.str()); oss.str("");
for (int i = 0; i < num_classes; i++) {
oss << classification_report_line(labels[i], precision(i), recall(i), f1_score(i), confusion_matrix[i].sum().item<int>());
report.push_back(classification_report_line(labels[i], precision(i), recall(i), f1_score(i), confusion_matrix[i].sum().item<int>()));
}
oss << std::endl;
report.push_back(" ");
oss << classification_report_line("accuracy", 0, 0, accuracy(), total);
report.push_back(oss.str()); oss.str("");
float precision_avg = 0;
float recall_avg = 0;
float precision_wavg = 0;
@@ -159,10 +163,11 @@ namespace platform {
recall_wavg /= total;
precision_avg /= num_classes;
recall_avg /= num_classes;
oss << classification_report_line("macro avg", precision_avg, recall_avg, f1_macro(), total);
oss << classification_report_line("weighted avg", precision_wavg, recall_wavg, f1_weighted(), total);
oss << std::endl << Colors::GREEN() << "Confusion Matrix" << std::endl;
oss << "================" << std::endl << color;
report.push_back(classification_report_line("macro avg", precision_avg, recall_avg, f1_macro(), total));
report.push_back(classification_report_line("weighted avg", precision_wavg, recall_wavg, f1_weighted(), total));
report.push_back("");
report.push_back("Confusion Matrix");
report.push_back("================");
auto number = total > 1000 ? 4 : 3;
for (int i = 0; i < num_classes; i++) {
oss << std::right << std::setw(label_len) << labels[i] << " ";
@@ -171,10 +176,9 @@ namespace platform {
oss << std::setw(number) << confusion_matrix[i][j].item<int>() << " ";
if (i == j) oss << color;
}
oss << std::endl;
report.push_back(oss.str()); oss.str("");
}
oss << Colors::RESET();
return oss.str();
return report;
}
json Scores::get_confusion_matrix_json(bool labels_as_keys)
{

4
src/main/Scores.h
View File

@@ -17,7 +17,7 @@ namespace platform {
float precision(int num_class);
float recall(int num_class);
torch::Tensor get_confusion_matrix() { return confusion_matrix; }
std::string classification_report(std::string color = "");
std::vector<std::string> classification_report(std::string color = "", std::string title = "");
json get_confusion_matrix_json(bool labels_as_keys = false);
void aggregate(const Scores& a);
private:
@@ -30,7 +30,7 @@ namespace platform {
int total;
std::vector<std::string> labels;
torch::Tensor confusion_matrix; // Rows ar actual, columns are predicted
int label_len = 12;
int label_len = 16;
int dlen = 9;
int ndec = 7;
};