private static void write_about_Estimates(XmlWriter writer, c_Fuzzy_System Classifier)
{
writer.WriteStartElement("Estimates");
if (Classifier.Test_Samples_set != null)
{
writer.WriteAttributeString("Count", XmlConvert.ToString(2));
writer.WriteStartElement("Estimate");
writer.WriteAttributeString("Table", Classifier.Learn_Samples_set.File_Name.Remove(Classifier.Learn_Samples_set.File_Name.Length - 4, 4));
writer.WriteAttributeString("Type", "PrecisionPercent");
writer.WriteAttributeString("Value", XmlConvert.ToString(Classifier.Classify_Learn_Samples()));
writer.WriteEndElement();
writer.WriteStartElement("Estimate");
writer.WriteAttributeString("Table", Classifier.Test_Samples_set.File_Name.Remove(Classifier.Learn_Samples_set.File_Name.Length - 4, 4));
writer.WriteAttributeString("Type", "PrecisionPercent");
writer.WriteAttributeString("Value", XmlConvert.ToString(Classifier.Classify_Test_Samples()));
writer.WriteEndElement();
}
else
{
writer.WriteAttributeString("Count", XmlConvert.ToString(1));
writer.WriteStartElement("Estimate");
writer.WriteAttributeString("Table", Classifier.Learn_Samples_set.File_Name.Remove(Classifier.Learn_Samples_set.File_Name.Length - 4, 4));
writer.WriteAttributeString("Type", "PrecisionPercent");
writer.WriteAttributeString("Value", XmlConvert.ToString(Classifier.Classify_Learn_Samples()));
writer.WriteEndElement();
}
writer.WriteEndElement();
}
public override Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System TuneUpFuzzySystem(Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System Classifier, Abstract_learn_algorithm_conf conf)
{
int count_iteration = ((Term_Config_Random_Search_conf)conf).Количество_итераций;
int count_c_Rules = ((Term_Config_Random_Search_conf)conf).Количество_генерируемых_баз_правил_за_итерацию;
c_Fuzzy_System result = Classifier;
for (int i = 0; i < count_iteration; i++)
{
int temp_prev_count_c_Rule = result.Rulles_Database_Set.Count;
double temp_best_result = result.Classify_Learn_Samples();
int temp_best_index = 0;
for (int j = 0; j < count_c_Rules; j++)
{
Knowlege_base_CRules temp_c_Rule = new Knowlege_base_CRules(result.Rulles_Database_Set[0]);
result.Rulles_Database_Set.Add(temp_c_Rule);
int temp_index = result.Rulles_Database_Set.Count - 1;
for (int k = 0; k < result.Rulles_Database_Set[temp_index].Terms_Set.Count; k++)
{
result.Rulles_Database_Set[temp_index].Terms_Set[k] =
randomize_term(result.Rulles_Database_Set[temp_index].Terms_Set[k]);
}
bool success = true;
double current_score = 0;
try
{
current_score = result.Classify_Learn_Samples(temp_index);
}
catch (Exception)
{
success = false;
}
if (success && (current_score >= temp_best_result))
{
temp_best_result = current_score;
temp_best_index = temp_index;
}
}
result.Rulles_Database_Set[0] = result.Rulles_Database_Set[temp_best_index];
result.Rulles_Database_Set.RemoveRange(temp_prev_count_c_Rule, result.Rulles_Database_Set.Count - temp_prev_count_c_Rule);
}
return(result);
}
public override Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System TuneUpWeigth(Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System Classifier, Abstract_weigth_config_conf conf)
{
int count_iteration = ((Weigth_Config_Random_Search_conf)conf).Количество_итераций;
int count_generation_by_iteration =
((Weigth_Config_Random_Search_conf)conf).Количество_генерируемых_векторов_веса_за_итерацию;
c_Fuzzy_System result = Classifier;
for (int i = 0; i < count_iteration; i++)
{
double [][] weigth = new double[count_generation_by_iteration + 1][];
weigth[0] = Classifier.Rulles_Database_Set[0].Weigth;
double best_result = result.Classify_Learn_Samples();
int best_index = 0;
for (int j = 1; j < count_generation_by_iteration + 1; j++)
{
weigth[j] = new double[weigth[0].Count()];
for (int k = 0; k < weigth[0].Count(); k++)
{
weigth[j][k] = rand.NextDouble();
}
result.Rulles_Database_Set[0].Weigth = weigth[j];
double current_result = result.Classify_Learn_Samples();
if (current_result > best_result)
{
best_result = current_result;
best_index = j;
}
result.Rulles_Database_Set[0].Weigth = weigth[best_index];
}
}
return(result);
}
public override Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System TuneUpFuzzySystem(Fuzzy_system.Class_Pittsburgh.c_Fuzzy_System Classifier, Abstract_learn_algorithm_conf conf)
{
count_iteration = ((Term_Config_PSO_Search_conf)conf).Количество_итераций;
c1 = ((Term_Config_PSO_Search_conf)conf).Коэффициент_c1;
c2 = ((Term_Config_PSO_Search_conf)conf).Коэффициент_c2;
w = 1;
count_particle = ((Term_Config_PSO_Search_conf)conf).Особей_в_популяции;
c_Fuzzy_System result = Classifier;
Knowlege_base_CRules[] X = new Knowlege_base_CRules[count_particle];
Knowlege_base_CRules[] V = new Knowlege_base_CRules[count_particle];
Knowlege_base_CRules[] Pi = new Knowlege_base_CRules[count_particle];
Knowlege_base_CRules Pg = new Knowlege_base_CRules();
double[] Errors = new double[count_particle];
double[] OldErrors = new double[count_particle];
double minError = 0;
Random rnd = new Random();
for (int i = 0; i < count_particle; i++)
{
Knowlege_base_CRules temp_c_Rule = new Knowlege_base_CRules(result.Rulles_Database_Set[0]);
X[i] = temp_c_Rule;
Errors[i] = result.Classify_Learn_Samples(0);
OldErrors[i] = Errors[i];
Pi[i] = new Knowlege_base_CRules(X[i]);
V[i] = new Knowlege_base_CRules(X[i]);
//
for (int j = 0; j < V[i].Terms_Set.Count; j++)
{
for (int k = 0; k < c_Fuzzy_System.Count_Params_For_Term(V[i].Terms_Set[j].Term_Func_Type); k++)
{
if (i == 0)
{
V[i].Terms_Set[j].Parametrs[k] = 0;
}
else
{
V[i].Terms_Set[j].Parametrs[k] = rnd.NextDouble() - 0.5;
}
}
double[] bf = new double[V[i].Weigth.Length];
for (int k = 0; k < V[i].Weigth.Length; k++)
{
if (i == 0)
{
bf[k] = 1;
}
else
{
//System.Windows.Forms.MessageBox.Show(rnd.NextDouble().ToString());
bf[k] = rnd.NextDouble() / 200;
}
}
V[i].Weigth = bf;
}
}
Pg = new Knowlege_base_CRules(result.Rulles_Database_Set[0]);
minError = Errors[0];
for (int i = 0; i < count_iteration; i++)
{
for (int j = 0; j < count_particle; j++)
{
w = 1 / (1 + Math.Exp(-(Errors[j] - OldErrors[j]) / 0.01));
for (int k = 0; k < X[j].Terms_Set.Count; k++)
{
for (int q = 0; q < c_Fuzzy_System.Count_Params_For_Term(X[j].Terms_Set[k].Term_Func_Type); q++)
{
double bp = Pi[j].Terms_Set[k].Parametrs[q];
V[j].Terms_Set[k].Parametrs[q] = V[j].Terms_Set[k].Parametrs[q] * w + c1 * rnd.NextDouble() * (bp - X[j].Terms_Set[k].Parametrs[q]) +
c2 * rnd.NextDouble() * (Pg.Terms_Set[k].Parametrs[q] - X[j].Terms_Set[k].Parametrs[q]);
X[j].Terms_Set[k].Parametrs[q] += V[j].Terms_Set[k].Parametrs[q];
}
}
double[] bf = new double[V[j].Weigth.Length];
double[] bfw = new double[V[j].Weigth.Length];
for (int k = 0; k < V[j].Weigth.Length; k++)
{
bfw[k] = V[j].Weigth[k] * w + c1 * rnd.NextDouble() * (Pi[j].Weigth[k] - X[j].Weigth[k]) +
c2 * rnd.NextDouble() * (Pg.Weigth[k] - X[j].Weigth[k]);
double sw = X[j].Weigth[k] + bfw[k];
if (sw > 0 && sw <= 2)
{
bf[k] = sw;
}
else
{
bf[k] = X[j].Weigth[k];
bfw[k] = V[j].Weigth[k];
}
}
X[j].Weigth = bf;
V[j].Weigth = bfw;
double newError = 0;
result.Rulles_Database_Set.Add(X[j]);
int temp_index = result.Rulles_Database_Set.Count - 1;
bool success = true;
try
{
newError = result.Classify_Learn_Samples(temp_index);
}
catch (Exception)
{
success = false;
}
result.Rulles_Database_Set.RemoveAt(temp_index);
if (success && (newError > Errors[j]))
{
OldErrors[j] = Errors[j];
Errors[j] = newError;
Pi[j] = new Knowlege_base_CRules(X[j]);
}
if (minError < newError)
{
minError = newError;
Pg = new Knowlege_base_CRules(X[j]);
}
}
}
result.Rulles_Database_Set[0] = Pg;
return(result);
}
