disctable.h

vc的源代码程序的基于最小特征向量的元素的集合方法的程序源代码

#ifndef DISCTABLE_H
#define DISCTABLE_H

#include "dataTab.h"
#include <time.h>
#include <sys/resource.h>

class DiscreteTable : public SupervisedDataTable {

public:
	Description **DiscTable;
	int *DiscTableSize;
	static const int LARGE_NUM = 99999999;
	int *nValuesBefore;
	static const int NO_BEST_CUT = 2;

	Description MaxClass;

	DiscreteTable(Source &src) : SupervisedDataTable(src)
	{
		DiscTable = new Description *[Col()-1];
		DiscTableSize = new int[Col()-1];
		nValuesBefore = new int [Col()-1];
		MaxClass = 1+getMaxClass();
		// this is to have the actual maximal class considered
	}
	
	// check whether the class labels, between [l,r]
	// inclusively, are the same or not

	int isSameClass(int l, int r);

	Description **getDiscTablePtr() { return DiscTable; }
	int *getDiscTableSizePtr() { return DiscTableSize; }

	// get the first boundary in [left, right] interval based
	// on Elomaa & Rousu's definition (1996)
	// the boundary is the first after the "left"
	int getBoundary(int left, int right, int c);

	// returns an array, which is already allocated, consists
	// of the boundary index
	int buildBoundaryTable(int *disc, int c);

	// call this function to discretize the training data
	// table
	void discretizeTable();
	void discretizeToIntTable();
	
	// user define discretization process
	// but he needs to call "getCutsTable()" and he needs
	// to assign the cut points to the "DiscTable" and record
	// down the number of cut points in each column within the
	// "getCutsTable()".

	void discretizeToIntTableUDF();
	void getCutsTable() {}

	// return the numerically maximal class label 
	Description getMaxClass();

	// output the discrete cut table
	void printDiscreteCutTable(char *name = NULL);

	// output the discrete data table
	void printDiscreteDataTable(char *name = NULL);

	// discretize the specific column to real number format
	void discretizeOnCol(int c);

	// discretize to int number format
	void discretizeToIntOnCol(int c);

	// discretize to real number format for future visualization
	void discretizeTable(SupervisedDataTable &tt);

	// discretize to int number format
	void discretizeToIntTable(SupervisedDataTable &tt);

	void printAllAttributeInfo();
	int  getNumOfAttributeValues(int c);

	void getAttrInfoBeforeDisc();

	// this function will be implemented independently for
	// each algorithm
	virtual void getCutsOnCol(int c) = 0;
	void cleanCntV(int *cntV);
	void getClassCntV(int *cntV, int left, int right);
	int countNumOfClass(int *cntV);
	int countNumOfClass(int *lCntV, int *rCntV);
	
	// it will return the next boundary where the value is
	// still the same as the value at row: begin 
	int getSameValueBoundary(int begin, int c);

	// Input:- 
	//   begin: is the row where to start to check
	//   c    : is the column
	//   next : is the row where it has the same value as the
	//          row: "begin"
	//
	// Return value:
	//   on success, it returns the 1, ie. the value at
	//   value(begin, c) always has pure classes. And it also
	//   returns through parameter "next", whose value is the
	//   same as the row "begin".
	int isValueHasPureClass(int begin, int c, int &next);
};

#endif