DTREG
  		• DTREG is the ideal tool for modeling business and medical data with categorical variables such as sex, race and marital status.

  		• Decision trees present a clear, logical model that can be understood easily by people who are not mathematically inclined.

  		• If you have a need for linear or nonlinear regression analysis, check out the NLREG program.

  		• You also should check out the News Rover program that automatically scans Usenet newsgroups, downloads messages of interest to you, decodes binary file attachments, reconstructs files split across multiple messages, and eliminates spam and duplicate files. News Rover also has a built-in MP3 music search engine and can quickly locate music files on any Usenet newsgroup.

    DTREG Benchmarks

    The following table shows the results for various types of predictive models applied to a large number of benchmarks. Note that different types of models work best for different types of data.

    The following predictive models were tested using DTREG:

    • Single tree - Classical single decision trees.
    • TreeBoost - DTREG implementation of Jerome Friedman's Stochastic Gradient Boosting ("MART").
    • Decision Tree Forests - DTREG implementation of Leo Breiman's "Random Forest"(tm) algorithm.
    • SVM - Support Vector Machine.
    • LDA - Linear Discriminant Analysis.
    • Logistic Regression - Regression adapted for binary classifications.
    • PNN - Probabilistic Neural Network.
    • CCNN - Cascade Correlation Neural Network.
    The numbers in the table are the percent misclassification (error) for each method on each benchmark. The misclassification percentages were computed using 10-fold cross-validation for all methods except for Decision Tree Forests which used Out-Of-Bag (OOB) validation and Probabilistic Neural Networks which used Leave-One-Out (LOO) validation.

    Logistic regression could not be performed for benchmark problems that have more than two target categories.

    Many of the benchmark data sets came from the UCI Machine Learning Repository.

    The best result for each benchmark is shown in bold face type. The worst result is in italics.

    Percent Misclassification Using Validation Data
    Benchmark Single tree TreeBoost Tree forest SVM LDA Logistic reg. PNN CCNN
    Abalone 46 45 45 45 45   44 44
    Ad 3 3 3 3 3   2 3
    AdultCensus 20 14 15 20 16   13 15
    Anneal steel 11 5 23 11 18   6 9
    Argentina 27 14 24 32 24 28 16 20
    Astroparticle 4 2 3 3 10 5 3 3
    Audiology 61 19 36 15 25   9 20
    AustralianCrabSex 14 8 7 3 4 4 4 2
    AustralianCredit 15 13 13 24 15 15 12 14
    Balance 22 15 18 0 14   10 4
    Bands 29 19 20 25 26   7 24
    Bioinformatics 38 18 21 16 17 21 16 18
    Bisbey 19 10 12 18 18   14 16
    Bridges 30 24 36 57 33   25 24
    Car 4 1 4 0 10   13 2
    ChurchParticipation 63 56 58 55 49   47 49
    ClevelandHeart14 58 46 44 48 40   39 41
    ColonTumor 19 8 16 20 23   10 24
    Contraception 44 44 44 46 49   45 45
    CreditApplication 14 9 13 14 14   12 14
    DeathPenalty 31 25 28 31 25 26 7 24
    Dermatology 4 3 3 0 3   0 4
    DNA 7 4 30 4 6   4 5
    Ecoli 22 14 14 18 13   12 13
    ElectroCardiogram 26 23 24 25 26 27 22 23
    EvansCounty 12 11 11 25 10 10 10 10
    Federalist 5 5 5 4 15 12 0 6
    Flags 45 32 46 36 36   25 44
    Fraud 26 29 31 45 29 29 14 33
    GermanCredit 33 24 25 23 24 24 20 24
    GfaNormaux 6 5 6 4 8 8 2 6
    Glass 30 24 22 34 36   12 34
    GymTutor 7 4 10 4 11   2 4
    Haberman 30 33 39 33 25 25 24 25
    Hayes-Roth 15 20 33 23 33   16 26
    Heart13 46 21 16 16 21 25 14 13
    Hepatitis 21 15 21 21 15 20 6 15
    HorseColic 20 17 18 16 18   7 21
    HOSLEM 50 33 33 34 31 33 25 31
    HouseVotes 5 4 3 3 4 3 2 3
    InsuranceFraud 27 18 31 29 28   5 19
    Ionosphere 9 7 6 5 14 13 8 10
    Iris 5 3 3 3 2   3 4
    Labor-neg 12 10 5 7 10 10 0 12
    Lenses 25 17 42 12 12   4 21
    Letter-recognition 14 4 0 2 30   2 20
    LibSvmVehicle 25 16 16 20 18   17 17
    LiverDisorder 32 29 26 29 30 34 30 29
    LowBwt 36 34 34 35 31 36 30 29
    LungCancer 50 50 87 47 50   6 56
    Lymphography 22 14 22 17 21   5 16
    Marketing 53 48 51 50 47   45 44
    Microchip 33 35 38 38 34 34 34 34
    Mushrooms 0 0 0 0 0   0 0
    Musk 24 10 10 5 18 18 4 13
    NLS 42 33 31 31 30   29 29
    Nursery 2 0 45 0 47   2 0
    NursingHome 20 7 6 16 6 16 4 5
    OilSpill 12 7 18 10 3   3 3
    Optdigits 10 2 2 1 4   1 3
    Pageblocks 8 5 2 8 5   3 3
    PenDigits 4 2 1 0 11   1 2
    P.I.-Diabetes 25 24 26 24 23 26 16 22
    PostOperative 58 42 43 51 38   29 37
    PrimaryTumor 74 59 68 60 55   55 52
    Reuters 11 5 4 3 23   11 7
    RingNorm 13 2 4 1 23 24 49 7
    SalesPlan 65 59 61 56 61   63 59
    Satellite 15 8 8 8 16   8 12
    SatImage 13 8 8 7 16   10 12
    Segment 5 2 2 0 8   3 5
    Shuttle 1 1 1 0 6   0 0
    Smoking 65 49 41 66 32   31 32
    Sonar 24 13 13 13 24 26 1 26
    SpamBase 7 6 5 6 10 7 9 7
    Spectf 27 17 20 21 41 39 6 22
    Splice DNA 5 4 35 3 5   3 5
    Spiral 47 42 47 8 51 51 24 57
    SvmTumor 46 25 24 24 33   60 33
    Tae 51 47 61 51 50   48 54
    Thyroid (ANN) 2 2 3 3 6   2 1
    Tic-tac-toe 6 1 1 0 2 2 2 2
    Tin 49 29 28 34 26 26 24 26
    Titanic 21 24 22 21 22 22 21 21
    TorchClassif 16 9 7 9 27 27 3 13
    Twonorm 15 2 3 2 2 2 2 2
    UTI 69 78 85 71 62   64 64
    Vehicle 28 24 24 14 22   25 22
    Vibration 60 60 60 61 51   48 49
    Vowel 17 7 3 2 28   0 11
    Waveform 22 15 15 13 14   15 13
    wbdc 7 5 6 4 7 7 6 7
    Wine 8 3 2 1 2   0 2
    Zoo 12 7 7 4 9   1 4
    Average error 24.80 18.61 22.19 19.79 22.21 20.42 15.38 19.26
    Median error 21.00 14.00 18.00 16.00 21.00 23.00 10.00 15.50
    Num. times best 4 14 8 25 5 2 53 16
    Num. times worst 42 2 17 14 24 8 3 9


    Return to DTREG home page

    Download demonstration copy of DTREG.

    Download manual for DTREG.

    Download manual for DTREG COM Library.

    Purchase DTREG.