This module uses neural network model to classify non-numerical (categorial, or factor) response Z based on one or more numerical predictors X. The response has usually text values, like “PETER”, “JOHN”, etc. called “levels”. The levels have no order (one cannot say whether one level is greater than another). The factor must have at least two levels, for example: (A, B, C, D, E); (YES, NO); (green, blue, yellow); (“Elytrigia repens”, “Lolium perenne”, “Phleum pratense”), etc. If there are numbers in the factor columns, they are interpreted as text with no numerical value. The goal of the classification module is to find any possibly existing relationship between the predictors and the response. Examples of applications is industrial fault identification, diagnosis support from a blood analysis, and many applications in psychology, chemical and environmental research, biology and life sciences, assessing and predicting activity of drugs in pharmaceuticals, and so on. .

ANN Classification - Pdf manual
Neural network - Pdf manual

Protokol: Task name Data Independent variable Transformation type Dependent variable Prediction Layer / Neurons Sigmoid steepness Moment Training speed Terminate when error < Training data (%) Termination conditions Optimization report No of iterations Max training error Mean training error Max validation error Mean validation error Number of cases Number of weights Number of rows Number of levels Total sum of squares Residual sum of squares Explained sum of squares F-statistic F-crit P-value Classification probabilities Prediction, Data, Misclass Misclassification table Weights Relative influence Prediction

Grafy: Plot of predicted and actual levels of the factor Graphical representation of the network architecture Plot of the training (network optimization) process Relative influence Relative predictability

Plot of predicted and actual levels of the factor

Graphical representation of the network architecture

Plot of the training (network optimization) process

Relative influence

Relative predictability