public static void testSVMPredictOnlyPackage()
{
string data_folder = @"D:\My_Dropbox\UProjects\OriBrainLearner\TestData\";
string data_file = data_folder + "svmguide1.t";
string range_file = data_folder + "svmguide1.range";
string svm_file = data_folder + "svmguide1.scale.model";
GuiPreferences.Instance.setLog("THIS PACKAGE ONLY PREDICTS!");
//Read in SVM model and feature range data
LibSVM svm = new LibSVM(); svm.LoadModel(svm_file);
double target_min, target_max; double[] features_min, features_max;
LibSVM.ReadRange(range_file, out target_min, out target_max, out features_min, out features_max);
//Read in Ground Truth data
List <int> labels_g = new List <int>();
List <LibSVMNode[]> samples = new List <LibSVMNode[]>();
using (StreamReader sr = new StreamReader(data_file))
{
string line; LibSVMNode[] sample; double label;
while ((line = sr.ReadLine()) != null)
{
LibSVM.ToLibSVMFormat(line, out sample, out label);
samples.Add(sample);
labels_g.Add((int)label);
}
}
//Convert to Array and List format
List <double[]> samples_array = new List <double[]>();
List <List <double> > samples_list = new List <List <double> >();
for (int i = 0; i < samples.Count; i++)
{
LibSVMNode[] sample = samples[i];
List <double> sample_list = new List <double>();
for (int k = 0, l = 0; k < sample.Length; k++)
{
int index = sample[k].index; if (index == -1)
{
break;
}
double value = sample[k].value;
for (int m = l; m < index - 1; m++)
{
sample_list.Add(double.NaN);
}
sample_list.Add(value); l = index;
}
samples_list.Add(sample_list);
samples_array.Add(sample_list.ToArray());
}
int count; double[] probs, dec_values;
//Scale the original data
LibSVMNode[][] samples_scaled = new LibSVMNode[samples.Count][];
GuiPreferences.Instance.setLog("Test scale data from SVMNode[] to SVMNode[]");
for (int i = 0; i < samples.Count; i++)
{
samples_scaled[i] = LibSVM.ScaleData(samples[i], target_min, target_max, features_min, features_max);
}
GuiPreferences.Instance.setLog("Test Predict functions. ");
count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test scale data from double[] to SVMNode[]");
for (int i = 0; i < samples_array.Count; i++)
{
samples_scaled[i] = LibSVM.ScaleData(samples_array[i], target_min, target_max, features_min, features_max);
}
GuiPreferences.Instance.setLog("Test Predict functions. ");
count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test scale data from list<double> to SVMNode[]");
for (int i = 0; i < samples_list.Count; i++)
{
samples_scaled[i] = LibSVM.ScaleData(samples_list[i], target_min, target_max, features_min, features_max);
}
GuiPreferences.Instance.setLog("Test Predict functions. ");
count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test scale data from double[][] to SVMNode[][]");
samples_scaled = LibSVM.ScaleData(samples_array.ToArray(), target_min, target_max, features_min, features_max);
GuiPreferences.Instance.setLog("Test Predict functions. "); count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test scale data from List<List<double>> to SVMNode[][]");
samples_scaled = LibSVM.ScaleData(samples_list, target_min, target_max, features_min, features_max);
GuiPreferences.Instance.setLog("Test Predict functions. "); count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
//Convert scaled data to double[] and list<double> format
List <double[]> samples_scaled_array = new List <double[]>();
List <List <double> > samples_scaled_list = new List <List <double> >();
for (int i = 0; i < samples_scaled.Length; i++)
{
LibSVMNode[] sample = samples_scaled[i];
List <double> sample_list = new List <double>();
for (int k = 0, l = 0; k < sample.Length; k++)
{
int index = sample[k].index; if (index == -1)
{
break;
}
double value = sample[k].value;
for (int m = l; m < index - 1; m++)
{
sample_list.Add(double.NaN);
}
sample_list.Add(value); l = index;
}
samples_scaled_list.Add(sample_list);
samples_scaled_array.Add(sample_list.ToArray());
}
GuiPreferences.Instance.setLog("Test Predict(SVMNode[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.Predict(samples_scaled[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test Predict(List<double>) functions. "); count = 0;
for (int i = 0; i < samples_scaled_list.Count; i++)
{
double label = svm.Predict(samples_scaled_list[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test Predict(double[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled_array.Count; i++)
{
double label = svm.Predict(samples_scaled_array[i]);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictProb(SVMNode[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.PredictProb(samples_scaled[i], out probs);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictProb(List<double>) functions. "); count = 0;
for (int i = 0; i < samples_scaled_list.Count; i++)
{
double label = svm.PredictProb(samples_scaled_list[i], out probs);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictProb(double[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled_array.Count; i++)
{
double label = svm.PredictProb(samples_scaled_array[i], out probs);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictValues(SVMNode[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled.GetLength(0); i++)
{
double label = svm.PredictValues(samples_scaled[i], out dec_values);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictValues(List<double>) functions. "); count = 0;
for (int i = 0; i < samples_scaled_list.Count; i++)
{
double label = svm.PredictValues(samples_scaled_list[i], out dec_values);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
GuiPreferences.Instance.setLog("Test PredictValues(double[]) functions. "); count = 0;
for (int i = 0; i < samples_scaled_array.Count; i++)
{
double label = svm.PredictValues(samples_scaled_array[i], out dec_values);
if ((int)label == labels_g[i])
{
count++;
}
}
GuiPreferences.Instance.setLog("Accuracy = " + ((double)count / labels_g.Count).ToString());
}
