public void addParameter(string label, Color color, int H)
{
ParameterVisualizer newParameterVisualizer = new ParameterVisualizer(pictureBox, form1, label, color);
newParameterVisualizer.needToRefresh.Stop();
newParameterVisualizer.needToRefresh.Close();
newParameterVisualizer.needToRefresh.Enabled = false;
newParameterVisualizer.needToRefresh.Dispose();
newParameterVisualizer.H = H;
parameters.Add(newParameterVisualizer);
for (int i = 0; i < parameters.Count; i++)
{
parameters[i].yDownGap = 10;
parameters[i].yUpGap = 35;
if (i == 0)
{
parameters[i].Ymin = 0;
}
else
{
parameters[i].Ymin = parameters[i - 1].Ymin + parameters[i - 1].H;
}
if (i == parameters.Count - 1)
{
parameters[i].Ymax = H;
}
parameters[i].multy = true;
}
// parameters[parameters.Count - 1].multy = false;
}
public MyAlgorithmOptimization(MainForm form1, Algorithm algorithm)
{
this.algorithm = algorithm;
explored = new List <TargetFunctionPoint>();
this.form1 = form1;
variablesVisualizer = new MultiParameterVisualizer(form1.picBox, form1);
variablesVisualizer.addParameter("max Q", Color.Cyan, 400);
variablesVisualizer.addParameter("target_function", Color.LightCyan, 800);
//VVVVVVVVVVV
// 1)разобраться с косяками на графике всех индивидов
// 2)почему при нахождении центра получаются близкие, но разные числа
// 3)добиться одинакового числа точек в currentHypercube связано с (1)
// 4) прикрутить красивые графики везде, где уместно
// -параллелизм
//for (int i = 0; i < population_value; i++)
// variablesVisualizer.parameters[1].functions.Add(new Function(" [" + i.ToString() + "]", valueToColor(0, population_value, population_value - i - 1)));
report_file_name = form1.pathPrefix + "/MyAlgorithmOptimization/Optimization report " + DateTime.Now.ToString().Replace(':', '-') + ".txt";
P = algorithm.h.Clone();
string reportHeader = "";
for (int i = 0; i < P.nodes.Count; i++)
{
if (P.nodes[i].getAttributeValue("variable") == "numerical")
{
reportHeader += P.nodes[i].getAttributeValue("name") + ';';
variablesCount++;
}
}
variablesVisualizer.enableGrid = false;
int fucnCount = Convert.ToInt32(Math.Pow(3.0, Convert.ToDouble(variablesCount)));
for (int i = 0; i < fucnCount; i++)
{
variablesVisualizer.parameters[1].functions.Add(new Function(" [" + i.ToString() + "]", ParameterVisualizer.valueToColor(0, fucnCount, fucnCount - i - 1)));
}
File.WriteAllText(report_file_name, reportHeader + "Q;" + '\n');
initPool();
variablesVisualizer.refresh();
}
public void cyclicPrediction(bool runCyclePredScript)
{
if (runCyclePredScript)
{
List <Task> tasks = new List <Task>();
foreach (Algorithm algorithm in algorithms)
{
tasks.Add(new Task(() =>
{
mainForm.I.executePythonScript(mainForm.pathPrefix + @"\Экспертная система\Экспертная система\Алгоритмы прогнозирования\" + algorithm.name + @"\cyclic_prediction.py", "--json_file_path \"" + algorithm.getValueByName("json_file_path") + '\"');
}));
tasks[tasks.Count - 1].Start();
}
foreach (Task task in tasks)
{
task.Wait();
}
}
mainForm.vis.clear();
mainForm.vis.enableGrid = true;
mainForm.vis.addParameter("Ensemble", Color.Red, 900);
mainForm.vis.parameters[0].functions.Add(new Function("_real", Color.Red, 3));
mainForm.vis.parameters[0].functions.Add(new Function("Ensemble prediction", Color.Cyan, 4));
mainForm.vis.parameters[0].functions.Add(new Function("Графики должны совпадать", Color.DarkGray));
int predicted_column_index = Convert.ToInt16(algorithms[0].getValueByName("predicted_column_index"));
List <double[]> ensemble_predictions = new List <double[]>();
foreach (Algorithm algorithm in algorithms)
{
int steps_forward = int.Parse(algorithm.h.getValueByName("steps_forward"));
var predictionsCSV = File.ReadAllLines(algorithm.getValueByName("save_folder") + "\\cyclic_prediction.txt");
var features = predictionsCSV[1].Split(';');
var n = algorithm.getValueByName("model_name");
string alg_name = algorithm.getValueByName("model_name");
mainForm.vis.addParameter(alg_name, ParameterVisualizer.valueToColor(0, algorithms.Count, algorithms.Count - algorithms.IndexOf(algorithm) - 1), 300);
mainForm.vis.parameters[mainForm.vis.parameters.Count - 1].functions.Add(new Function(alg_name + "real", Color.Red, 1));
mainForm.vis.parameters[mainForm.vis.parameters.Count - 1].functions.Add(new Function(alg_name + " prediction", Color.Cyan, 1));
double[] algorithm_predictions = new double[predictionsCSV.Length - steps_forward - 1];
for (int i = 1; i < predictionsCSV.Length - steps_forward; i++)
{
double predictedValue;
double realValue;
predictedValue = Convert.ToDouble(predictionsCSV[i].Split(';')[features.Length - 2].Replace('.', ','));
realValue = Convert.ToDouble(predictionsCSV[i + steps_forward].Split(';')[predicted_column_index].Replace('.', ','));
algorithm_predictions[i - 1] = predictedValue;
mainForm.vis.addPoint(realValue, alg_name + "real");
mainForm.vis.addPoint(predictedValue, alg_name + " prediction");
if (algorithms.IndexOf(algorithm) == 0)
{
mainForm.vis.addPoint(realValue, "_real");
}
}
ensemble_predictions.Add(algorithm_predictions);
}
for (int j = 0; j < ensemble_predictions[0].Length; j++)
{
double sum = 0;
int count = 0;
for (int i = 0; i < ensemble_predictions.Count; i++)
{
if (ensemble_predictions[i].Length > j)
{
sum += ensemble_predictions[i][j];
count++;
}
}
mainForm.vis.addPoint(sum / count, "Ensemble prediction");
}
mainForm.vis.refresh();
}
public void stepByStepPrediction()
{
bool pred_chart_must_match_original_chart = false;
mainForm.vis.clear();
mainForm.vis.enableGrid = true;
mainForm.vis.addParameter("Ensemble", Color.Red, 900);
mainForm.vis.parameters[0].functions.Add(new Function("real", Color.Red, 3));
mainForm.vis.parameters[0].functions.Add(new Function("Ensemble prediction", Color.Cyan, 4));
if (pred_chart_must_match_original_chart)
{
mainForm.vis.parameters[0].functions.Add(new Function("Графики должны совпадать", Color.DarkGray));
}
mainForm.vis.parameters[0].showLastNValues = true;
mainForm.vis.parameters[0].window = 50;
int predicted_column_index = Convert.ToInt16(algorithms[0].getValueByName("predicted_column_index"));
List <double[]> ensemble_predictions = new List <double[]>();
foreach (Algorithm algorithm in algorithms)
{
int steps_forward = 0;
if (pred_chart_must_match_original_chart)
{
steps_forward = int.Parse(algorithm.h.getValueByName("steps_forward"));
}
string model_name = algorithm.getValueByName("model_name");
var predictionsCSV = File.ReadAllLines(algorithm.getValueByName("save_folder") + "\\predictions.txt");
predictionsCSV = Expert.skipEmptyLines(predictionsCSV);
var features = predictionsCSV[1].Split(';');
var n = model_name;
mainForm.vis.parameters[0].functions.Add(
new Function(model_name,
ParameterVisualizer.valueToColor(0, algorithms.Count, algorithms.Count - algorithms.IndexOf(algorithm) - 1))
);
double[] algorithm_predictions = new double[predictionsCSV.Length - 1];
for (int i = 1; i < predictionsCSV.Length - steps_forward; i++)
{
double predictedValue;
double realValue;
predictedValue = Convert.ToDouble(predictionsCSV[i].Split(';')[features.Length - 1].Replace('.', ','));
realValue = Convert.ToDouble(predictionsCSV[i + steps_forward].Split(';')[predicted_column_index].Replace('.', ','));
/* if (i == 1)
* {
* // добавление сдвига к графику прогноза, чтобы он совпадал с реальным значением
* for (int q = 0; q < steps_forward; q++)
* {
* mainForm.vis.addPoint(predictedValue, model_name);
* }
* }*/
algorithm_predictions[i - 1] = predictedValue;
if (algorithms.IndexOf(algorithm) == 0)
{
mainForm.vis.addPoint(realValue, "real");
}
mainForm.vis.addPoint(predictedValue, model_name);
}
ensemble_predictions.Add(algorithm_predictions);
}
/* for (int q = 0; q < int.Parse(algorithms[0].h.getValueByName("steps_forward")); q++)
* {////////////////////////////////////////////////////////////
* /////////////// и это лишнее
* // добавление сдвига к графику прогноза, чтобы он совпадал с реальным значением
* mainForm.vis.addPoint(0, "Ensemble prediction")
* ////////////////////////////////////////////////////////////////////////////////////////////////////////////
* }*/
for (int j = 0; j < ensemble_predictions[0].Length; j++)
{
double sum = 0;
int count = 0;
for (int i = 0; i < ensemble_predictions.Count; i++)
{
if (ensemble_predictions[i].Length > j)
{
sum += ensemble_predictions[i][j];
count++;
}
}
mainForm.vis.addPoint(sum / count, "Ensemble prediction");
}
mainForm.vis.refresh();
}