logfit.hpp

orange源码 数据挖掘技术

/*
    This file is part of Orange.

    Orange is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    (at your option) any later version.

    Orange is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with Orange; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

    Authors: Martin Mozina, Janez Demsar, Blaz Zupan, 1996--2002
    Contact: [e-mail]
*/

#include "root.hpp"
#include "orvector.hpp"
#include "examplegen.hpp"

#ifndef __LOGFIT_HPP
#define __LOGFIT_HPP

// WRAPPERS
WRAPPER(LogRegFitter)

// abstract class for LR fitters. 
// New fitters should be derived from this one
class ORANGE_API TLogRegFitter : public TOrange {
public:
	__REGISTER_ABSTRACT_CLASS

  // Don't change the order (<= Divergence means that model is fitted, > means error)
	enum {OK, Infinity, Divergence, Constant, Singularity};

	// main function call, fits LR, returns coefficients and their 
	// corres. standard errors
	virtual PAttributedFloatList operator()(PExampleGenerator, const int &, PAttributedFloatList &, float &, int &, PVariable &)=0;

	// transforms orange PExampleGenerator attributes in a classic C double 2D array
	// returns number of examples and number of attributes as well
	virtual double** generateDoubleXMatrix(PExampleGenerator, long &nexamples, long &nattr);

	// transforms orange PExampleGenerator class in a classic C double array 
	virtual double* generateDoubleYVector(PExampleGenerator, const int &);
	virtual double* generateDoubleTrialsVector(PExampleGenerator, const int &);
};


// output values computed in logistic fitter
class ORANGE_API LRInfo {
public:
	LRInfo();	
	~LRInfo();

   int nn, k;
   double chisq;			// chi-squared
   double devnce;			// deviance
   int    ndf;				// degrees of freedom
   double *beta;			// fitted beta coefficients
   double *se_beta;			// beta std.devs
   double *fit;				// fitted probabilities for groups
   double **cov_beta;		// approx covariance matrix
   double *stdres;   		// residuals
   int    *dependent;		// dependent/redundant variables
   int	  error;
};

// input values for logistic fitter
class ORANGE_API LRInput {
public:
	LRInput();
	~LRInput();

	long nn;
	long k;
	double **data;    //nn*k
	double *success;     //nn
	double *trials;
};


// Logistic regression fitter via minimization of log-likelihood
// orange integration of Aleks Jakulin version of LR
// based on Alan Miller's(1992) F90 logistic regression code
class ORANGE_API TLogRegFitter_Cholesky : public TLogRegFitter {
public:
	__REGISTER_CLASS

/*	int maxit; //maximum no. iterations
	double offset; //offset on the logit scale
	double tol; //  tolerance for matrix singularity
	double eps; //difference in `-2  log' likelihood for declaring convergence.
	double penalty; //penalty (scalar), substract from ML beta'譸enalty譩eta. Set if 
				    //model doesnt converge */

	// constructor
	TLogRegFitter_Cholesky();
	TLogRegFitter_Cholesky(bool showErrors);

	// Public main function, use it for fitting LR
	virtual PAttributedFloatList operator()(PExampleGenerator, const int &, PAttributedFloatList &, float &, int &, PVariable &);


private:
	static const char *errors[];
};

#endif