Kalman filter (generalized)

Author: © mladen, 2019
0 Views
0 Downloads
0 Favorites
Kalman filter (generalized)
ÿþ//------------------------------------------------------------------

#property copyright "© mladen, 2019"

#property link      "mladenfx@gmail.com"

//------------------------------------------------------------------

#property indicator_chart_window

#property indicator_buffers 4

#property indicator_plots   1

#property indicator_label1  "Kalman filter"

#property indicator_type1   DRAW_COLOR_LINE

#property indicator_color1  clrDarkGray,clrMediumSeaGreen,clrOrangeRed

#property indicator_width1  2



//

//--- input parameters

//



input double             inpPeriod = 1;           // Period/smoothing ratio

input ENUM_APPLIED_PRICE inpPrice  = PRICE_CLOSE; // Price



//

//--- indicator buffers

//



double val[],valc[];



//------------------------------------------------------------------

// Custom indicator initialization function

//------------------------------------------------------------------ 

//

//

//



int OnInit()

{

   //

   //--- indicator buffers mapping

   //

      SetIndexBuffer(0,val,INDICATOR_DATA);

      SetIndexBuffer(1,valc,INDICATOR_COLOR_INDEX);

         iKalmanFilter.init(inpPeriod);

   //      

   //--- indicator short name assignment

   //

   IndicatorSetString(INDICATOR_SHORTNAME,"Kalman filter ("+(string)inpPeriod+")");

   return (INIT_SUCCEEDED);

}

void OnDeinit(const int reason) { }



//------------------------------------------------------------------

// Custom indicator iteration function

//------------------------------------------------------------------

//

//---

//



int  OnCalculate( const int        rates_total,

                  const int        prev_calculated,

                  const int        begin,

                  const double&    price[])

{

   int i= prev_calculated-1; if (i<0) i=0; for (; i<rates_total && !_StopFlag; i++)

   {

      double _velocity;

      val[i]  = iKalmanFilter.calculate((price[i]==EMPTY_VALUE ? 0 : price[i]),_velocity,i,rates_total);

      valc[i] = (_velocity>0) ?  1 :(_velocity<0) ? 2 : 0;

   }

   return(i);

}



//------------------------------------------------------------------

// Custom functions

//------------------------------------------------------------------

//

//---

//



class CKalmanFilter

{

   private :

      double m_period;

      double m_coeff;

      struct sKalmanFilter

      {

         double filter;

         double velocity;

      };

      sKalmanFilter m_array[];

      int           m_arraySize;

      

   public :      

      CKalmanFilter() : m_arraySize(-1) {};

     ~CKalmanFilter()                   {};

     

      //

      //

      //

      

      void init (double period)

      {

         m_coeff  = (period>0 ? period : 1.0)/100.0;

         m_period = MathSqrt(m_coeff);

      }

      double calculate(double value,double& velocity, int i, int bars)

      {

         if (m_arraySize<bars) m_arraySize = ArrayResize(m_array,bars+500);

         

         if (i>0)

         {

            double _distance = value-m_array[i-1].filter;

            double _error    = m_array[i-1].filter+_distance*m_period;

                               m_array[i].velocity = m_array[i-1].velocity+_distance*m_coeff;

                               m_array[i].filter   = _error+m_array[i].velocity;

         }

         else { m_array[i].filter = value; m_array[i].velocity = 0; }

         velocity = m_array[i].velocity; 

             return(m_array[i].filter);

      }

};

CKalmanFilter iKalmanFilter;

//------------------------------------------------------------------

Comments