independence.cpp

Go to the documentation of this file.

#include "common.hpp"
#include "io/input/transaction_reader/brBufferedTransactionReader.hpp"
#include "io/input/transaction_reader/SortedTransactionReader.hpp"

#include "io/codec/decoder/df/CacheDFDecoder.hpp"
#include "io/output/StatOutput.hpp"

#include "util/StreamParser.hpp"
#include "util/FrequentFilter.cpp"


#include <vector>
#include <iostream>
#include <string>


std::string file_format;

void init()
{
   file_format = "File format:";
   file_format += "\n\nThe transactionfile is a plan text file. Each row ";
   file_format += "represents a transaction. \n";
   file_format += "A transaction is a set of items seperated by a nonnumeric ";
   file_format += "character.\nIt can be for example a white space, comma, ";
   file_format += "colon, etc.\n";
   file_format += "Items are nonnegative integers.\n";
}
void usage()
{
   std::cerr << "\nUsage: independence transactionfile min_supp ";
   std::cerr << "outcomefile [options]\n";
   std::cerr << "\n transactionfile  file, that contains the transactions of items";
   std::cerr << "\n outcomefile\t  file to write the outcome";
   std::cerr << "\n min_supp\t  absolute support threshold";

   std::cerr << file_format;
   std::cerr << "\n\nHave a succesful mining ;-)"<<std::endl<<std::endl;
}

int process_arguments( int argc, char *argv[],
                       counter_t& min_supp, bool &isrel, double &relminsupp )
{
   if ( argc < 4 )
   {
      usage();
      log_err(0,"There are 3 mandatory arguments.");
      return 2;
   }
   std::string mins=argv[2];
   if (mins[mins.size()-1]=='%') {
      mins.erase(mins.size()-1);
      isrel=true;
      relminsupp=atof(mins.c_str());
      relminsupp/=100;
      log_info(0,"Using relative minimum support of %lg",relminsupp);
      return 0;
   }
   isrel=false;

   int min_supp_i;
   try
   {
      convert(argv[2], min_supp_i);
      if ( min_supp_i <= 0  )
      {
         log_err(0,"%s cannot be converted to a positive integer.",argv[2]);
         return 3;
      }
   }
   catch(BadConversion e)
   {
      log_err(0,"min_supp conversion problem.");
      return 3;
   }
   min_supp = static_cast<counter_t>(min_supp_i);
   log_info(0,"min_supp is set to %d", min_supp);
   return 0;
}
void findFrequentSets(const std::vector<double>& frequencies,
                      const double min_freq, const counter_t nr_of_transactions,
                      double prefixfrequency,
                      size_t index, StatOutput< CacheDFDecoder< >, true >& df_decoder )
{
   double new_prefixfrequency;
   while(index < frequencies.size())
   {
      new_prefixfrequency = prefixfrequency * frequencies[index];
      if( new_prefixfrequency >= min_freq)
      {
         df_decoder.pushItemWithWrite(
            index, static_cast<counter_t>( new_prefixfrequency * nr_of_transactions));
         findFrequentSets( frequencies, min_freq, nr_of_transactions,
                           new_prefixfrequency, ++index, df_decoder );
         df_decoder.popItem();
      }
      else
         break;
   }
}
int main( int argc, char *argv[] )
{
   init();
   counter_t min_supp;
   bool relative;
   double relminsupp;
      
   {
      int return_val = 
         process_arguments( argc, argv, min_supp, relative, relminsupp );
      if(return_val)
         return return_val;
   }

   char* input_file = argv[1];
   char* output_file = argv[3];

   try
   {
      // We assume that the transactions does not contain duplicates!!!
      typedef brBufferedTransactionReader< > T_R;
      // Otherwise uncmment this:
      // typedef SortedTransactionReader<brBufferedTransactionReader< >, true> T_R;

      T_R::params_t par_i;
      par_i.file_name = input_file;
      par_i.mode=FileReprBase::READ;
      par_i.file_buffer_size = 16 * 1024;
      T_R tr_reader(&par_i);
      std::vector< std::pair<counter_t, item_t> > freq_items_with_counters;
      counter_t nr_of_transactions;
      // The first step of each algorithms is determining the frequent items.
      FrequentFilter<T_R>
         fr_filter(tr_reader);
      log_status(0,"Finding frequent items.");
      fr_filter.findFrequentItems( freq_items_with_counters,   
                                   nr_of_transactions, min_supp);

      sort( freq_items_with_counters.begin(),
            freq_items_with_counters.end() );
      reverse( freq_items_with_counters.begin(),
               freq_items_with_counters.end() );
      std::vector<item_t> code_inverse;
      code_inverse.reserve(freq_items_with_counters.size());
      std::vector<double> frequencies;
      frequencies.reserve(freq_items_with_counters.size());
      std::vector< std::pair<counter_t, item_t> >::size_type index;
      for( index = 0; index < freq_items_with_counters.size(); ++index )
      {
         code_inverse.push_back( freq_items_with_counters[index].second );
         frequencies.push_back( 
            static_cast<double>(freq_items_with_counters[index].first)/ nr_of_transactions);
      }
      log_status(0,"Doing decoder.");
      typedef StatOutput< CacheDFDecoder< >, true >DF_D;

      DF_D::params_t par_d;
      par_d.file_name = output_file;
      par_d.mode=FileReprBase::WRITE;
      par_d.numfreq = freq_items_with_counters.size(); // If StatOutput is used!!!
      DF_D df_decoder(&par_d);
      df_decoder.setCodeInverse(code_inverse);

      if( min_supp<= nr_of_transactions)
      {
         df_decoder.write(nr_of_transactions);
         findFrequentSets( frequencies,
                           static_cast<double>(min_supp)/nr_of_transactions,
                           nr_of_transactions, 1.0, 0, df_decoder );
      }
   }
   catch (std::ios_base::failure e)
   {
      log_err(0,"Exiting the program due to IO exception");
      return 1;
   }
   return 0;
}

---