AiFeatureEval/Src/Correlations.mqh

//+------------------------------------------------------------------+
//|                                                 Correlations.mqh |
//|                                             Copyright 2025, Leo. |
//|                     https://www.mql5.com/es/users/nique_372/news |
//+------------------------------------------------------------------+
#property copyright "Copyright 2025, Leo."
#property link      "https://www.mql5.com/es/users/nique_372/news"
#property strict

#ifndef AIFEATUREEVAL_CORRELATIONS_MQH
#define AIFEATUREEVAL_CORRELATIONS_MQH

//+------------------------------------------------------------------+
//| Include                                                          |
//+------------------------------------------------------------------+
#include <Math\\Stat\\Math.mqh>
#include <Graphics\Graphic.mqh>
#include <TSN\\MQLArticles\\Utils\\File.mqh>

//+------------------------------------------------------------------+
//|                Estrcturas \ Defines \ Enums                      |
//+------------------------------------------------------------------+
//---
struct pack(sizeof(double)) CorrData
 {
  double             origValue;
  double             absValue;
  int                index;

  CorrData()
    :                origValue(0.0), absValue(0.0), index(0) {}

  CorrData(const CorrData& other)
   {
    this = other;
   }
 };

//---
struct pack(sizeof(double)) CorrelationSort
 {
  double             value_correlation;
  int                index_col;
 };

//---
struct CorrelationAr
 {
  string             cols_name;
  double             values[];
 };

//---
#define CORR_SALIDA_ULTIMO_ELEMENTO -1

enum ENUM_CORRELATION_FUNCTION
 {
  F_CORRELATION_PEARSON = 0,
  F_CORRELATION_SPEARMAN,
  F_CORRELATION_KENDALL
 };

//+------------------------------------------------------------------+
//|                       Funciones                                  |
//+------------------------------------------------------------------+
typedef double (*funcion_correlacion)(const double &x[], const double &y[]);

//+------------------------------------------------------------------+
double CorrPearson(const double &x[], const double &y[])
 {
  if(VerifyCorrelation(x, y) == false)
    return 0;

  double r;
  MathCorrelationPearson(x, y, r);
  return r;
 }

//+------------------------------------------------------------------+
double CorrSpearman(const double &x[], const double &y[])
 {
  if(VerifyCorrelation(x, y) == false)
    return 0;

  double r;
  MathCorrelationSpearman(x, y, r);
  return r;
 }
//+------------------------------------------------------------------+
double CorrKendall(const double &x[], const double &y[])
 {
  if(VerifyCorrelation(x, y) == false)
    return 0;

  double r;
  MathCorrelationKendall(x, y, r);
  return r;
 }



//+------------------------------------------------------------------+
bool VerifyCorrelation(const double & x[], const double &y[])
 {
  if(x.Size() < 1)
   {
    FastLog(FUNCION_ACTUAL, ERROR_TEXT, "El tamaño del array x es invalido");
    return false;
   }

  if(y.Size() < 1)
   {
    FastLog(FUNCION_ACTUAL, ERROR_TEXT, "El tamaño del array y es invalido");
    return false;
   }

  if(x.Size() != y.Size())
   {
    FastLog(FUNCION_ACTUAL, ERROR_TEXT, StringFormat("El tamño del array x = %u y el tamaño del array y = %u, no coinciden", x.Size(), y.Size()));
    return false;
   }

  return true;
 }

//+------------------------------------------------------------------+
funcion_correlacion GetRelationFunction(ENUM_CORRELATION_FUNCTION mode)
 {
  const static funcion_correlacion  funciones[3] = {CorrPearson, CorrSpearman, CorrKendall};
  return funciones[int(mode)];
 }

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
class CCorrelations : public CLoggerBase
 {
private:
  //--- Target
  double             target_data[]; //data del target
  string             target_data_col_name;

  //--- Data
  double             arr_out_correlation[]; //correlacion de cada columna
  CorrelationAr      correlation_matrix[]; //matrix de correlacoin (para plot, aqui se guarda la data y el nombre de la col)
  CorrelationSort    correlations_sort[];

  //--- Correlacion
  funcion_correlacion func;


  //--- Plots
  CGraphic           graph;

public:
                     CCorrelations(void);
                    ~CCorrelations(void) { Clear(); }

  //--- Setters principales
  bool               Set(string file_name, bool common_flag, ushort csv_separator,  bool csv_tiene_cabezera, int salida_col_index = CORR_SALIDA_ULTIMO_ELEMENTO);
  bool               Set(matrix &mtx, vector &out);
  bool               Set(CorrelationAr &data[], double &outputs[]);

  //--- Setters extra
  void               CorrelationFuction(ENUM_CORRELATION_FUNCTION new_value) { this.func = GetRelationFunction(new_value); }

  //--- Funciones generales
  void               GetBestCorrelations(int &indexes_cols[], int num_bests);
  void               Run();

  //--- Limpieza
  void               Clear();

  //--- Plot
  void               TargetColName(string new_value) { this.target_data_col_name = new_value; }
  void               PlotAnalisis(string prefix, string file_folder, int width, int height, color clr_out, int &idxs_column[], bool scale_1_0, double min_val = 0.0, double max_val = 1.0);
 };

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
CCorrelations::CCorrelations(void)
  : func(NULL), target_data_col_name("Target")
 {
  Clear();
 }

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
void CCorrelations::GetBestCorrelations(int &indexes_cols[], int num_bests)
 {
  if(num_bests > (int)correlation_matrix.Size())
   {
    LogError(StringFormat("El numero de correlaciones esperadas = %d, es mayor a las disponibles = %u", num_bests, correlation_matrix.Size()), FUNCION_ACTUAL);
    return;
   }

  ArrayResize(indexes_cols, 0);
  int start = MathMax((ArraySize(correlations_sort) - num_bests), 0);

  for(int i = start; i < ArraySize(correlations_sort) ; i++)
   {
    int val = correlations_sort[i].index_col;
    Print("Correlation: ", correlations_sort[i].value_correlation);
    Print("Col name: ", this.correlation_matrix[val].cols_name);
    AddArrayNoVerification(indexes_cols, val, 0);
   }
 }


//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
void CCorrelations::PlotAnalisis(string prefix, string file_folder, int width, int height, color clr_out, int &idxs_column[], bool scale_1_0, double min_val = 0.0, double max_val = 1.0)
 {
  if(idxs_column.Size() < 1)
   {
    LogError(StringFormat("El tamaño de las columnas a plotear = %u, es invalido", idxs_column.Size()), FUNCION_ACTUAL);
    return;
   }

//---
  string file_name_corr = file_folder + prefix + "_graph_correlation.png";
  string name_grap_corr = prefix + "graph_corr";

//--- Seteamos la curva out
  graph.Create(0, name_grap_corr, 0, 0, 0, width, height);

  if(scale_1_0)
   {
    double temp_arr[];
    ArrayCopy(temp_arr, this.target_data);
    ScaleArrayCustom(temp_arr, ArraySize(temp_arr), min_val, max_val);
    graph.CurveAdd(temp_arr, clr_out, CURVE_POINTS_AND_LINES, target_data_col_name);
   }
  else
    graph.CurveAdd(this.target_data, clr_out, CURVE_POINTS_AND_LINES, target_data_col_name);

//--- Seteamos las curvas restantes
  for(int i = 0; i < ArraySize(idxs_column); i++)
   {
    int col = idxs_column[i];

    if(scale_1_0)
     {
      double temp_arr[];
      ArrayCopy(temp_arr, correlation_matrix[col].values);
      ScaleArrayCustom(temp_arr, ArraySize(temp_arr), min_val, max_val);
      graph.CurveAdd(temp_arr, CURVE_POINTS_AND_LINES, correlation_matrix[col].cols_name);
      continue;
     }

    graph.CurveAdd(correlation_matrix[col].values, CURVE_POINTS_AND_LINES, correlation_matrix[col].cols_name);
   }

//--- Seteamos los axis
  graph.XAxis().Name("Numero de datos");
  graph.XAxis().NameSize(12);
  graph.YAxis().Name("Value");
  graph.YAxis().NameSize(12);
  graph.YAxis().ValuesWidth(15);
  graph.CurvePlotAll();
  graph.Update();


  bool res = false;
  if(ChartSetInteger(0, CHART_SHOW, false))
    res = ChartScreenShot(0, file_name_corr, width, height);
  ChartSetInteger(0, CHART_SHOW, true);

  Sleep(2000);
  graph.Destroy();
 }

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
void CCorrelations::Run()
 {
  if(arr_out_correlation.Size() < 1)
   {
    LogError(StringFormat("El tamaño del array de correlaciones = %u, es invalido", arr_out_correlation.Size()), FUNCION_ACTUAL);
    return;
   }

//---
  int size = (int)this.arr_out_correlation.Size();

//---
  CorrData corrArray[];
  ArrayResize(corrArray, size);

  for(int i = 0; i < size; i++)
   {
    corrArray[i].absValue = fabs(arr_out_correlation[i]);
    corrArray[i].origValue = arr_out_correlation[i];
    corrArray[i].index = i;
   }

//---
  for(int i = 0; i < size - 1; i++)
   {
    for(int j = i + 1; j < size; j++)
     {
      if(corrArray[i].absValue > corrArray[j].absValue)
       {
        CorrData temp = corrArray[i];
        corrArray[i] = corrArray[j];
        corrArray[j] = temp;
       }
     }
   }

//---
  ArrayResize(this.correlations_sort, 0);
  for(int i = 0; i < size; i++)
   {
    CorrelationSort cs;
    cs.index_col         = corrArray[i].index;
    cs.value_correlation = corrArray[i].origValue;
    AddArrayNoVerification(correlations_sort, cs, 0);
   }
 }


//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
bool CCorrelations::Set(matrix &mtx, vector &out)
 {
  if(mtx.Rows() != out.Size())
   {
    LogError(StringFormat("Las filas de la matriz de data = %I64u, es diferente al tamaño del vector de target = %I64u", mtx.Rows(), out.Size()), FUNCION_ACTUAL);
    return false;
   }

//---
  ArrayResize(correlation_matrix, (int)out.Size());
  ArrayResize(arr_out_correlation, (int)out.Size());
  VectorToDoubleArray(out, this.target_data);

//---
  for(ulong i = 0; i < mtx.Cols(); i++)
   {
    double temp_arr[];
    VectorToDoubleArray(mtx.Col(i), temp_arr);
    arr_out_correlation[i]  = func(temp_arr, this.target_data);
    ArrayCopy(correlation_matrix[i].values, temp_arr);
    correlation_matrix[i].cols_name = IntegerToString(i);
   }
  return true;
 }


//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
bool CCorrelations::Set(CorrelationAr &data[], double &outputs[])
 {
//---
  ArrayResize(correlation_matrix, data.Size());
  ArrayResize(arr_out_correlation, data.Size());
  ArrayCopy(this.target_data, outputs);

//---
  for(ulong i = 0; i < data.Size(); i++)
   {
    arr_out_correlation[i] = func(data[i].values, this.target_data); //Obtenemos correlacion de la columna -> vector
    correlation_matrix[i] = data[i];
   }
  return true;
 }

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
bool CCorrelations::Set(string file_name, bool common_flag, ushort csv_separator, bool csv_tiene_cabezera, int salida_col_index = -1)
 {
  matrix mtx;
  string header;
  if(!CSVToMatrix(file_name, mtx, common_flag, csv_tiene_cabezera, csv_separator, header))
   {
    LogError(StringFormat("No se pudo cargar el archivo = %s", file_name), FUNCION_ACTUAL);
    return false;
   }

//--- Declaraciones iniciales
  ulong cols = mtx.Cols();
  ulong rows = mtx.Rows();

//---
  if(salida_col_index == CORR_SALIDA_ULTIMO_ELEMENTO)
    salida_col_index = (int)cols - 1;

//---
  vector output = mtx.Col(salida_col_index);
  VectorToDoubleArray(output, this.target_data);

//---
  ArrayResize(this.correlation_matrix, (int)cols - 1); //omitimos una columnas de salida
  ArrayResize(this.arr_out_correlation, (int)cols - 1);

//--- Cols Name
  if(csv_tiene_cabezera) //si tiene cabezera los nombres de las columnas que coincidadn con el nombre del csv
   {
    string res[];
    int s;
    if((s = StringSplit(header, csv_separator, res)) != cols)
     {
      LogError(StringFormat("El numero de columnas = %I64u es diferente al tamaño del heador = %d", cols, s), FUNCION_ACTUAL);
      return false;
     }

    ulong curr_idx = 0;
    for(ulong i = 0; i < cols - 1; i++)
     {
      if(i == salida_col_index)
       {
        this.target_data_col_name = res[i];
        continue;
       }

      correlation_matrix[curr_idx].cols_name = res[i];
      curr_idx++;
     }
   }
  else //si no solo asignamos el indice
   {
    ulong curr_idx = 0;
    for(ulong i = 0; i < cols; i++)
     {
      if(i == salida_col_index)
        continue;

      correlation_matrix[curr_idx].cols_name = IntegerToString(i);
      curr_idx++;
     }
   }

//--- Finalizamos obteniendo la correalcion de cada columna
  ulong curr_idx = 0;
  for(ulong i = 0; i < cols; i++)
   {
    if(i == salida_col_index)
      continue;



    //---
    double arr[];
    VectorToDoubleArray(mtx.Col(i), arr);
    ArrayCopy(correlation_matrix[curr_idx].values, arr);

    //---
    double correlacion = func(arr, this.target_data);
    this.arr_out_correlation[curr_idx] = correlacion;
    curr_idx++;
   }
  return true;
 }

//+------------------------------------------------------------------+
//|                                                                  |
//+------------------------------------------------------------------+
void CCorrelations::Clear(void)
 {
  ArrayFree(this.arr_out_correlation);
  ArrayFree(this.correlations_sort);
  ArrayFree(correlation_matrix);
  ArrayFree(target_data);
 }
//+------------------------------------------------------------------+
void ScaleArrayCustom(double &arr[], int size, double target_min = 0.0, double target_max = 1.0)
 {
  if(size <= 1)
    return;

//---
  double data_min = arr[0];
  double data_max = arr[0];

  for(int i = 1; i < size; i++)
   {
    if(arr[i] < data_min)
      data_min = arr[i];
    if(arr[i] > data_max)
      data_max = arr[i];
   }

//---
  double data_range = data_max - data_min;
  if(data_range == 0.0)
   {
    double mid_point = (target_min + target_max) / 2.0;
    for(int i = 0; i < size; i++)
      arr[i] = mid_point;

    return;
   }

//---
  double target_range = target_max - target_min;
  for(int i = 0; i < size; i++)
   {
    double normalized = (arr[i] - data_min) / data_range;
    arr[i] = target_min + (normalized * target_range);
   }
 }
//+------------------------------------------------------------------+
#endif // AIFEATUREEVAL_CORRELATIONS_MQH