Glowny
about
Valid XHTML 1.1 Valid CSS!

RSS

On the previous page random forest...

Overview:

The aim of this test was to present a set of databases that were used to perform further tests. It contains also comparison between test set errors using various tree growth amgorithms.On each data set 25% instances was set as testing set. On remaining data forest algorithms were run three times, growing and combining 25 trees.

NOMINAL

  • CAR DATABASE

    Train:
    Number of attributes: 6 + 1 class attribute
    Number of instances: 1296
    Number of classes: 4
    • unacc: 918 instances
    • acc: 282 instances
    • vgood: 48 instances
    • good: 48 instances

    Test:
    Number of attributes: 6 + 1 class attribute
    Number of instances: 432
    Number of classes: 4
    • unacc: 292 instances
    • acc: 102 instances
    • vgood: 17 instances
    • good: 21 instances

    7.71605%- random forest
    26.6204%- random tree
    24.537%- det. forest
  • NURSARY DATABASE

    Train:
    Number of attributes: 8 + 1 class attribute
    Number of instances: 9720
    Number of classes: 5
    • priority: 3136 instances
    • not_recom: 3240 instances
    • recommend: 1 instances
    • very_recom: 228 instances
    • spec_prior: 3115 instances

    Test:
    Number of attributes: 8 + 1 class attribute
    Number of instances: 3239
    Number of classes: 5
    • spec_prior: 928 instances
    • priority: 1130 instances
    • very_recom: 100 instances
    • not_recom: 1080 instances
    • recommend: 1 instances

    10.9396%- random forest
    22.9906%- random tree
    14.0475%- det. forest
  • AUDIOLOGY.STANDARDIZED DATABASE

    Train:
    Number of attributes: 70 + 1 class attribute
    Number of instances: 200
    Number of classes: 24
    • cochlear_unknown: 48 instances
    • mixed_cochlear_age_fixation: 1 instances
    • mixed_cochlear_age_otitis_media: 4 instances
    • cochlear_age: 46 instances
    • normal_ear: 20 instances
    • cochlear_poss_noise: 16 instances
    • cochlear_age_and_noise: 18 instances
    • acoustic_neuroma: 1 instances
    • mixed_cochlear_unk_ser_om: 3 instances
    • conductive_discontinuity: 2 instances
    • retrocochlear_unknown: 2 instances
    • conductive_fixation: 6 instances
    • bells_palsy: 1 instances
    • cochlear_noise_and_heredity: 2 instances
    • mixed_cochlear_unk_fixation: 5 instancess
    • mixed_poss_noise_om: 2 instances
    • otitis_media: 4 instances
    • possible_menieres: 8 instances
    • possible_brainstem_disorder: 4 instances
    • cochlear_age_plus_poss_menieres: 1 instances
    • mixed_cochlear_age_s_om: 2 instances
    • mixed_cochlear_unk_discontinuity: 2 instances
    • mixed_poss_central_om: 1 instances
    • poss_central: 1 instances

    Test:
    Number of attributes: 70 + 1 class attribute
    Number of instances: 26
    Number of classes: 6
    • cochlear_age: 11 instances
    • cochlear_age_and_noise: 4 instances
    • cochlear_poss_noise: 4 instances
    • mixed_cochlear_unk_fixation: 4 instances
    • normal_ear: 2 instances
    • mixed_cochlear_age_fixation: 1 instances

    28.2051%- random forest
    48.7179%- random tree
    57.6923%- det. forest
  • BREAST CANCER DATABASE (Missing attributes)

    Train:
    Number of attributes: 9 + 1 class attribute
    Number of instances: 214
    Number of classes: 2
    • no-recurrence-events: 153 instances
    • recurrence-events: 61 instances

    Test:
    Number of attributes: 9 + 1 class attribute
    Number of instances: 71
    Number of classes: 2
    • no-recurrence-events: 48 instances
    • recurrence-events: 23 instances

    16.8582%- random forest
    45.5399%- random tree
    19.5077%- det. forest

CONTINUOUS

  • BALANCE-SCALE DATABASE

    Train:
    Number of attributes: 4 + 1 class attribute
    Number of instances: 469
    Number of classes: 3
    • B: 36 instances
    • R: 212 instances
    • L: 221 instances

    Test:
    Number of attributes: 4 + 1 class attribute
    Number of instances: 156
    Number of classes: 3
    • L: 67 instances
    • R: 76 instances
    • B: 13 instances

    18.1624%- random forest
    30.5556%- random tree
    24.359%- det. forest
    16.0256%- forest RC
  • IONOSPHERE DATABASE

    Train:
    Number of attributes: 34 + 1 class attribute
    Number of instances: 264
    Number of classes: 2
    • g: 174 instances
    • b: 90 instances

    Test:
    Number of attributes: 34 + 1 class attribute
    Number of instances: 87
    Number of classes: 2
    • g: 51 instances
    • b: 36 instances

    16,8582%- random forest
    24.9042%- random tree
    18,0077%- det. forest
    16,8582%- forest RC


On the next page extended tests...

Overview | Nominal | Continuous

Back

Visit my Guestbook, read my CV.
Disclaimer.

This site is a member of WebRing. To browse visit here.
Powered by elektroda.pl.
Forum elektronika, schematy, porady, naprawa