(WIP) Pattern Matching Algorithm using Stock Tick Data: Uses FooPlot for Rendering Graphs

a_stock_pattern_matching.cpp

#include <iostream>
#include <vector>
#include <fstream>
#include <string>
#include <cmath>

#include "functional.h"
#include "graph.h"

// DATA PARSING
#define PATTERN_DISTANCE 30
typedef std::vector< long long > DateVector;
typedef std::vector< long double > DataVector;
typedef std::vector< DataVector > PatternVector;

struct DataSet {
  DataSet( long long approx_size = 0 ) {
    date_times.reserve( approx_size );
    bid_prices.reserve( approx_size );
    asking_prices.reserve( approx_size );
  }
  
  DateVector date_times;
  DataVector bid_prices;
  DataVector asking_prices;
  
  DataVector average_prices() {
    
    // Cached?
    if( !m_average_prices.empty() ) return m_average_prices;
    
    // Make
    m_average_prices.reserve( bid_prices.size() );
    
    for( int i = 0; i < bid_prices.size(); ++i ) {
      auto average = ( bid_prices[i] + asking_prices[i] ) / 2;
      m_average_prices.push_back( average );
    }
    
    return m_average_prices;
  }
  
  private:
    DataVector m_average_prices;
};

DataSet parse_file( std::string path, long long approximate_size = 0 ) {
  DataSet data_set( approximate_size );
  
  std::ifstream file( path );
  
  std::string buffer;
  while( std::getline(file, buffer) ) {
    
    // First 14 is for date
    auto date = buffer.substr( 0, 14 );
    
    // Reset the buffer
    buffer = buffer.substr( 15 );
    size_t index = buffer.find(',');
    
    // Get data points
    auto data_point_1 = buffer.substr( 0, index );
    auto data_point_2 = buffer.substr( index + 1 );
    
    // Save the data points
    data_set.date_times.push_back( std::stoll( date ) );
    data_set.bid_prices.push_back( std::stod( data_point_1 ) );
    data_set.asking_prices.push_back( std::stod( data_point_2 ) );
  }
  
  return data_set;
}

// ALGORITHM

long double percent_change( long double start, long double current ) {
  // Refactor me!!
  try {
    long double result = ( ( current - start ) / ( std::fabs(start) ) ) * 100;
    
    if( result == 0 ) return 0.0000001;
    else return result;
    
  } catch(...) {
    return 0.0000001;
  }
}

PatternVector pattern_storage( DataVector average_line ) {

  PatternVector all_patterns;
  
  long long y = PATTERN_DISTANCE;
  long long x = average_line.size() - (2 * PATTERN_DISTANCE);
  
  while( y < x ) {
    DataVector pattern;
    
    // Save All Patterns
    for( int i = ( PATTERN_DISTANCE - 1 ); i >= 0; --i ) {
      auto index = y - PATTERN_DISTANCE;
      auto point = percent_change( average_line[ index ] , average_line[y - i] );
      pattern.push_back( point );
    }
    
    all_patterns.push_back( pattern );
    
    // Outcome Range and Current Point
    DataVector outcome_range;
    long long end_range = y + PATTERN_DISTANCE;
    long long i = ( y + 20 );
    
    for( ; i < end_range; ++i) {
      outcome_range.push_back( average_line[i] );
    }
    
    long double current_point = average_line[y];
    
    // Average Outcome
    auto fn_outcome_range = Functional< DataVector >( outcome_range );
    auto average_outcome = fn_outcome_range.average();
    
    // Future Outcome
    auto future_outcome = percent_change( current_point , average_outcome );

    // FINISH ME!
    
    ++y; // Inc
  }
  
  return all_patterns;
}

DataVector current_pattern( DataVector average_line ) {
  DataVector pattern;
  pattern.reserve( PATTERN_DISTANCE );
  
  for( int i = (PATTERN_DISTANCE - 1); i >= 0; --i ) {
    
    auto result = percent_change( average_line[ PATTERN_DISTANCE ], average_line[i] );
    pattern.push_back( result );
  }
  
  return pattern;
}

void pattern_recognition( PatternVector pattern_vec, DataVector current_pattern ) {
  
  for( auto& pattern : pattern_vec ) {
  
    long double similarity_total = 0;
  
    for( int i = 0; i < PATTERN_DISTANCE; ++i ) {
      long double _change = percent_change( pattern[i], current_pattern[i] );
      similarity_total += ( 100 - std::fabs( _change ) );
    }
    
    long double similarity_average = similarity_total / PATTERN_DISTANCE;
    
    if( similarity_average > 60 ) {
      std::cout << "Similarity: " << similarity_average << "%" << std::endl;
      
      Graph< DataVector > graph;
      
      graph.add_line( pattern, "FF0000" );
      graph.add_line( current_pattern, "001DAD" );
      
      std::cout << graph.to_fooplot() << std::endl;
    }
    
  }
  
}


// TESTER

int main(int argc, const char * argv[]) {
  
  std::cout << "Start" << std::endl;
  auto data_set = parse_file( "/tmp/GBPUSD/GBPUSD1d.txt", 1000 );
  std::cout << "Done" << std::endl;
  
  // Just do it (TM)
  auto average_line = data_set.average_prices();
  
  auto patterns = pattern_storage( average_line );
  auto current = current_pattern( average_line );
  pattern_recognition( patterns, current );
  
  // insert code here...
  std::cout << "Hello, World!\n";
  return 0;
}

b_functional.hpp

#include <functional>
#include <vector>

template< typename T >
  class Functional {
    public:
    
      typedef typename T::value_type V;
    
      Functional( T collection ): m_collection(collection) {}
    
      V sum() {
        V total = 0;
        for( auto val : m_collection ) {
          total += val;
        }
        return total;
      }
    
      V average() {
        return sum() / m_collection.size();
      }
    
    private:
      T m_collection;
  };

c_graph.hpp

#include <vector>
#include <sstream>

#include "base64.h"

template< class T >
  class Graph {
    public:
      Graph() {}
    
      void add_line( T data_points, std::string color = "000000" ) {
        std::ostringstream outstream;
        
        outstream << "{\"type\":3,\"eq\":[";
        for(int i = 0; i < data_points.size(); ++i) {
          // Json comma
          if( i > 0 ) {
            outstream << ",";
          }
          
          // Point
          outstream << "[\"" << (i+1) << "\",\"" << data_points[i] << "\"]";
        }
        outstream << "]";
        outstream << ",\"color\":\"#" << color << "\"}";
        
        m_chart_lines.push_back( outstream.str() );
      }
    
      std::string to_json() {
        std::ostringstream outstream;
        
        outstream << "[";
        for( auto line : m_chart_lines ) {
          outstream << line << ",";
        }
        outstream << "{\"type\":1000,\"window\":[\"2.427692307692307\",\"31.427692307692304\",\"-0.01092341430854167\",\"0.008419398805292345\"]}]";
        
        return outstream.str();
      }
    
      std::string to_fooplot() {
        std::string json = to_json();
        
        auto b64_input_json = reinterpret_cast<const unsigned char*>( json.c_str() );
        
        auto b64_encoded_json = base64_encode( b64_input_json , (unsigned int) json.length() );
      
        std::ostringstream outstream;
        outstream << "http://fooplot.com/#";
        outstream << b64_encoded_json;
        
        return outstream.str();
      }
    
    private:
      std::vector< std::string > m_chart_lines;
  };

d_base64.hpp

// Source: https://github.com/ReneNyffenegger/development_misc/tree/master/base64
// Only change:

    while((i++ < 3))
      ret += '=';
      
// Becomes

    while((i++ < 3))
      ret += '-';
      
// To match the fooplot b64 encoder

e_output.md

Similarity: 60.2744%

http://fooplot.com/#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

Data available here: http://sentdex.com/GBPUSD.zip