//+------------------------------------------------------------------+
//|                                            Linear Regression.mq5 |
//|                                               Gamuchirai Ndawana |
//|                    https://www.mql5.com/en/users/gamuchiraindawa |
//+------------------------------------------------------------------+
#property copyright "Gamuchirai Ndawana"
#property link      "https://www.mql5.com/en/users/gamuchiraindawa"
#property version   "1.00"

//+------------------------------------------------------------------+
//| System constants                                                 |
//+------------------------------------------------------------------+
#define TOTAL_INPUTS 6

//+------------------------------------------------------------------+
//| System Inputs                                                    |
//+------------------------------------------------------------------+
int bars = 90;//Number of historical bars to fetch
int horizon = 1;//How far into the future should we forecast
int MA_PERIOD = 2; //Moving average period
ENUM_TIMEFRAMES TIME_FRAME = PERIOD_D1;//User Time Frame
ENUM_TIMEFRAMES RISK_TIME_FRAME = PERIOD_D1;
double sl_size = 2;

//+------------------------------------------------------------------+
//| Dependencies                                                     |
//+------------------------------------------------------------------+
#include <Trade\Trade.mqh>
#include <VolatilityDoctor\Time\Time.mqh>
#include <VolatilityDoctor\Trade\TradeInfo.mqh>

//+------------------------------------------------------------------+
//| Global Variables                                                 |
//+------------------------------------------------------------------+
int        ma_close_handler,ma_high_handler,ma_low_handler;
double     ma_close[],ma_high[],ma_low[];
Time       *Timer;
TradeInfo  *TradeInformation;
vector     bias,temp,temp_2,temp_3,temp_4,temp_5,Z1,Z2;
matrix     X,y,prediction,b;
int        time;
CTrade Trade;
int state;
int atr_handler;
double atr[];

//+------------------------------------------------------------------+
//| Expert initialization function                                   |
//+------------------------------------------------------------------+
int OnInit()
  {
//---
   Timer                   = new Time(Symbol(),TIME_FRAME);
   TradeInformation        = new TradeInfo(Symbol(),TIME_FRAME);
   ma_close_handler        = iMA(Symbol(),TIME_FRAME,MA_PERIOD,0,MODE_SMA,PRICE_CLOSE);
   ma_high_handler        = iMA(Symbol(),TIME_FRAME,MA_PERIOD,0,MODE_SMA,PRICE_HIGH);
   ma_low_handler        = iMA(Symbol(),TIME_FRAME,MA_PERIOD,0,MODE_SMA,PRICE_LOW);

   bias = vector::Ones(TOTAL_INPUTS);
   Z1 = vector::Ones(TOTAL_INPUTS);
   Z2 = vector::Ones(TOTAL_INPUTS);
   X = matrix::Ones(TOTAL_INPUTS,bars);
   y = matrix::Ones(1,bars);
   time = 0;
   state = 0;
   atr_handler = iATR(Symbol(),RISK_TIME_FRAME,14);
//---
   return(INIT_SUCCEEDED);
  }

//+------------------------------------------------------------------+
//| Expert deinitialization function                                 |
//+------------------------------------------------------------------+
void OnDeinit(const int reason)
  {
//---
   IndicatorRelease(atr_handler);
   IndicatorRelease(ma_close_handler);
  }

//+------------------------------------------------------------------+
//| Expert tick function                                             |
//+------------------------------------------------------------------+
void OnTick()
  {
//---
   if(Timer.NewCandle())
     {
      CopyBuffer(atr_handler,0,0,1,atr);
      CopyBuffer(ma_close_handler,0,0,1,ma_close);
      CopyBuffer(ma_low_handler,0,0,1,ma_low);
      CopyBuffer(ma_high_handler,0,0,1,ma_high);

      setup();

      double c =  iClose(Symbol(),TIME_FRAME,0);
      double buy_sl = (TradeInformation.GetBid() - (sl_size * atr[0]));
      double sell_sl = (TradeInformation.GetAsk() + (sl_size * atr[0]));


      Comment("Forecasted MA: ",prediction[0,0],"\nForecasted MA High: ",prediction[1,0],"\nForecasted MA Low: ",prediction[2,0],"\nForecasted Price: ",prediction[3,0]);

      if(PositionsTotal() == 0)
        {
         state = 0;
         if((prediction[0,0] > c) && (prediction[2,0] > buy_sl) && (prediction[3,0] > c) && (prediction[2,0] > ma_low[0]) && (prediction[1,0] > ma_high[0]))
           {
            Trade.Buy(TradeInformation.MinVolume(),Symbol(),TradeInformation.GetAsk(),buy_sl,0);
            state = 1;
           }
         if((prediction[0,0] < c) && (prediction[1,0] < sell_sl) && (prediction[3,0] < c) && (prediction[2,0] < ma_low[0]) && (prediction[1,0] < ma_high[0]))
           {
            Trade.Sell(TradeInformation.MinVolume(),Symbol(),TradeInformation.GetBid(),sell_sl,0);
            state = -1;
           }
        }

      if(PositionsTotal() > 0)
        {
         if(((state == -1) && (prediction[0,0] > c)) || ((state == 1)&&(prediction[0,0] < c)))
            Trade.PositionClose(Symbol());
         if(PositionSelect(Symbol()))
           {
            double current_sl = PositionGetDouble(POSITION_SL);

            if((state == 1) && ((ma_close[0] - (2 * atr[0]))>current_sl))
              {
               Trade.PositionModify(Symbol(),(ma_close[0] - (2 * atr[0])),0);
              }

            else
               if((state == -1) && ((ma_close[0] + (1 * atr[0]))<current_sl))
                 {
                  Trade.PositionModify(Symbol(),(ma_close[0] + (2 * atr[0])),0);
                 }
           }
        }
     }
  }

//+------------------------------------------------------------------+
//| Prepare the training data for our model                          |
//+------------------------------------------------------------------+
void prepare_data(void)
  {
//--- Reshape the matrix
   X = matrix::Ones(TOTAL_INPUTS,bars);

//--- Store the Z-scores
   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_OPEN,horizon,bars);
   Z1[0] = temp.Mean();
   Z2[0] = temp.Std();
   temp = ((temp - Z1[0]) / Z2[0]);
   X.Row(temp,1);

//--- Store the Z-scores
   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_HIGH,horizon,bars);
   Z1[1] = temp.Mean();
   Z2[1] = temp.Std();
   temp = ((temp - Z1[1]) / Z2[1]);
   X.Row(temp,2);

//--- Store the Z-scores
   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_LOW,horizon,bars);
   Z1[2] = temp.Mean();
   Z2[2] = temp.Std();
   temp = ((temp - Z1[2]) / Z2[2]);
   X.Row(temp,3);

//--- Store the Z-scores
   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_CLOSE,horizon,bars);
   Z1[3] = temp.Mean();
   Z2[3] = temp.Std();
   temp = ((temp - Z1[3]) / Z2[3]);
   X.Row(temp,4);

//--- Store the Z-scores
   temp.CopyIndicatorBuffer(ma_close_handler,0,horizon,bars);
   Z1[4] = temp.Mean();
   Z2[4] = temp.Std();
   temp = ((temp - Z1[4]) / Z2[4]);
   X.Row(temp,5);

//--- Labelling our targets
   temp.CopyIndicatorBuffer(ma_close_handler,0,0,bars);
   temp_2.CopyIndicatorBuffer(ma_high_handler,0,0,bars);
   temp_3.CopyIndicatorBuffer(ma_low_handler,0,0,bars);
   temp_4.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_CLOSE,0,bars);

   y.Reshape(4,bars);

   y.Row(temp,0);
   y.Row(temp_2,1);
   y.Row(temp_3,2);
   y.Row(temp_4,3);
  }

//+------------------------------------------------------------------+
//| Fit our model                                                    |
//+------------------------------------------------------------------+
void fit(void)
  {
//--- Fit the model
   matrix OB_U,OB_VT,OB_SIGMA;
   vector OB_S;

   PrintFormat("Computing Singular Value Decomposition of %s Data using OpenBLAS",Symbol());
   X.SingularValueDecompositionDC(SVDZ_S,OB_S,OB_U,OB_VT);
   OB_SIGMA.Diag(OB_S);
   b = y.MatMul(OB_VT.Transpose().MatMul(OB_SIGMA.Inv()).MatMul(OB_U.Transpose()));
   Print("OLS Solutions: ");
   Print(b);
  }

//+------------------------------------------------------------------+
//| Get a prediction from our multiple output model                  |
//+------------------------------------------------------------------+
void predict(void)
  {
//--- Prepare to get a prediction
//--- Reshape the data
   X = matrix::Ones(TOTAL_INPUTS,1);

//--- Get a prediction
   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_OPEN,0,1);
   temp = ((temp - Z1[0]) / Z2[0]);
   X.Row(temp,1);

   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_HIGH,0,1);
   temp = ((temp - Z1[1]) / Z2[1]);
   X.Row(temp,2);

   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_LOW,0,1);
   temp = ((temp - Z1[2]) / Z2[2]);
   X.Row(temp,3);

   temp.CopyRates(Symbol(),TIME_FRAME,COPY_RATES_CLOSE,0,1);
   temp = ((temp - Z1[3]) / Z2[3]);
   X.Row(temp,4);

   temp.CopyIndicatorBuffer(ma_close_handler,0,0,1);
   temp = ((temp - Z1[4]) / Z2[4]);
   X.Row(temp,5);

   Print("Prediction Inputs: ");
   Print(X);

//--- Get a prediction
   prediction.Reshape(1,2);
   prediction = b.MatMul(X);
   Print("Prediction");
   Print(prediction);
  }

//+------------------------------------------------------------------+
//| Obtain a prediction from our model                               |
//+------------------------------------------------------------------+
void setup(void)
  {
   prepare_data();
   fit();

   Print("Training Input Data: ");
   Print(X);

   Print("Training Target");
   Print(y);

   predict();
  }
//+------------------------------------------------------------------+

#undef TOTAL_INPUTS
//+------------------------------------------------------------------+