source: proiecte/Parallel-DT/R8/Src/st-thresh.c @ 22

/*************************************************************************/
/*                                                                       */
/*      Soften thresholds for continuous attributes                      */
/*      -------------------------------------------                      */
/*                                                                       */
/*************************************************************************/


#include "defns.i"
#include "types.i"
#include "extern.i"


Boolean *LHSErr,        /*  Does a misclassification occur with this value of an att  */
        *RHSErr;        /*  if the below or above threshold branches are taken  */

ItemNo  *ThreshErrs;    /*  ThreshErrs[i] is the no. of misclassifications if thresh is i  */

float   *CVals;         /*  All values of a continuous attribute  */


#define Below(v,t)      (v <= t + 1E-6)


/*************************************************************************/
/*                                                                       */
/*  Soften all thresholds for continuous attributes in tree T            */
/*                                                                       */
/*************************************************************************/


    SoftenThresh(T)
/*  ------------  */
    Tree T;
{
    CVals = (float *) calloc(MaxItem+1, sizeof(float));
    LHSErr = (Boolean *) calloc(MaxItem+1, sizeof(Boolean));
    RHSErr = (Boolean *) calloc(MaxItem+1, sizeof(Boolean));
    ThreshErrs = (ItemNo *) calloc(MaxItem+1, sizeof(ItemNo));

    InitialiseWeights();

    ScanTree(T, 0, MaxItem);

    cfree(ThreshErrs);
    cfree(RHSErr);
    cfree(LHSErr);
    cfree(CVals);
}



/*************************************************************************/
/*                                                                       */
/*  Calculate upper and lower bounds for each test on a continuous       */
/*  attribute in tree T, using data items from Fp to Lp                  */
/*                                                                       */
/*************************************************************************/


    ScanTree(T, Fp, Lp)
/*  --------  */
    Tree T;
    ItemNo Fp, Lp;
{
    short v;
    float Val, Se, Limit, Lower, Upper, GreatestValueBelow();
    ItemNo i, Kp, Ep, LastI, Errors, BaseErrors;
    ClassNo CaseClass, Class1, Class2, Category();
    Boolean LeftThresh=false;
    Description CaseDesc;
    Attribute Att;
    void Swap();

    /*  Stop when get to a leaf  */

    if ( ! T->NodeType ) return;

    /*  Group the unknowns together  */

    Kp = Group(0, Fp, Lp, T);

    /*  Soften a threshold for a continuous attribute  */

    Att = T->Tested;

    if ( T->NodeType == ThreshContin )
    {
        printf("\nTest %s <> %g\n", AttName[Att], T->Cut);

        Quicksort(Kp+1, Lp, Att, Swap);

        ForEach(i, Kp+1, Lp)
        {
            /*  See how this item would be classified if its
                value were on each side of the threshold  */

            CaseDesc = Item[i];
            CaseClass = Class(CaseDesc);
            Val = CVal(CaseDesc, Att);
                
            Class1 = Category(CaseDesc, T->Branch[1]);
            Class2 = Category(CaseDesc, T->Branch[2]);

            CVals[i] = Val;
            LHSErr[i] = (Class1 != CaseClass ? 1 : 0);
            RHSErr[i] = (Class2 != CaseClass ? 1 : 0);
        }

        /*  Set Errors to total errors if take above thresh branch,
            and BaseErrors to errors if threshold has original value  */

        Errors = BaseErrors = 0;
        ForEach(i, Kp+1, Lp)
        {
            Errors += RHSErr[i];

            if ( Below(CVals[i], T->Cut) )
            {
                BaseErrors += LHSErr[i];
            }
            else
            {
                BaseErrors += RHSErr[i];
            }
        }

        /*  Calculate standard deviation of the number of errors  */

        Se = sqrt( (BaseErrors+0.5) * (Lp-Kp-BaseErrors+0.5) / (Lp-Kp+1) );
        Limit = BaseErrors + Se;

        Verbosity(1)
        {
            printf("\t\t\tBase errors %d, items %d, se=%.1f\n",
                   BaseErrors, Lp-Kp, Se);
            printf("\n\tVal <=   Errors\t\t+Errors\n");
            printf("\t         %6d\n", Errors);
        }

        /*  Set ThreshErrs[i] to the no. of errors if the threshold were i  */

        ForEach(i, Kp+1, Lp)
        {
            ThreshErrs[i] = Errors = Errors + LHSErr[i] - RHSErr[i];

            if ( i == Lp || CVals[i] != CVals[i+1] )
            {
                Verbosity(1)
                    printf("\t%6g   %6d\t\t%7d\n",
                        CVals[i], Errors, Errors - BaseErrors);
            }
        }

        /*  Choose Lower and Upper so that if threshold were set to
            either, the number of items misclassified would be one
            standard deviation above BaseErrors  */

        LastI = Kp+1;
        Lower = Min(T->Cut, CVals[LastI]);
        Upper = Max(T->Cut, CVals[Lp]);
        while ( CVals[LastI+1] == CVals[LastI] ) LastI++;

        while ( LastI < Lp )
        {
            i = LastI + 1;
            while ( i < Lp && CVals[i+1] == CVals[i] ) i++;

            if ( ! LeftThresh &&
                 ThreshErrs[LastI] > Limit &&
                 ThreshErrs[i] <= Limit &&
                 Below(CVals[i], T->Cut) )
            {
                Lower = CVals[i] -
                        (CVals[i] - CVals[LastI]) * (Limit - ThreshErrs[i]) /
                        (ThreshErrs[LastI] - ThreshErrs[i]);
                LeftThresh = true;
            }
            else
            if ( ThreshErrs[LastI] <= Limit &&
                 ThreshErrs[i] > Limit &&
                 ! Below(CVals[i], T->Cut) )
            {
                Upper = CVals[LastI] +
                        (CVals[i] - CVals[LastI]) * (Limit - ThreshErrs[LastI]) /
                        (ThreshErrs[i] - ThreshErrs[LastI]);
                if ( Upper < T->Cut ) Upper = T->Cut;
            }

            LastI = i;
        }

        T->Lower = Lower;
        T->Upper = Upper;

        Verbosity(1) printf("\n");

        printf("\tLower = %g, Upper = %g\n", T->Lower, T->Upper);
    }

    /*  Recursively scan each branch  */

    ForEach(v, 1, T->Forks)
    {
        Ep = Group(v, Kp+1, Lp, T);

        if ( Kp < Ep )
        {
            ScanTree(T->Branch[v], Kp+1, Ep);
            Kp = Ep;
        }
    }
}