Non-parametric probabilistic neural network: statistics summary of 100 repetitions

SamplesC1C2C3C4C5C6
Training sample
    Mean correct classification98.9192.4587.3080.9574.4259.50
    Standard deviation3.005.896.453.639.7311.67
    Coeff variation (%)3.046.377.394.4813.0819.61
    Minimum84.6173.3372.4170.6947.6237.04
    Maximum100.00100.00100.0086.2195.0077.78
    Range15.3826.6727.5815.5247.3840.74
    Conf interv upper99.7694.1289.1381.9877.1962.82
    Conf interv low98.06190.7885.4679.9271.6656.18
Validation sample
    Mean correct classification89.2089.0375.2186.6079.4158.67
    Standard deviation6.416.426.416.673.4910.47
    Coeff variation (%)7.198.3012.997.704.3917.80
    Minimum72.7373.2040.0669.0072.2238.46
    Maximum100.00100.0094.1296.7086.7978.57
    Range27.2726.8054.0427.7014.5740.11
    Conf interv upper91.0391.5275.5188.5080.4161.65
    Conf interv low87.3889.0267.1584.7078.4359.60

Stability estimates and confidence intervals of the classification function were tested by 100 bootstrap repetitions with random selection of the replaced samples. The average correct classification was higher, and the function stable for C1, C2, C3, and C4, while C6 had large standard deviations and coefficients of variation during both training and validation. C: cluster