Zoo Neural Network Classifier Test

The system is trained to infer an animal's type from a vector of mostly binary features. The training data is uci-zoo-data.xlsx. A symbolic representation of this data is zoo.json.txt.

Each time a test is run, five items are randomly withheld from the training set, and used to test the classifier. Incorrect predictions are marked in pink. Instances says how many examples of the correct type were available. The fewer the instances, the more likely there will errors and low confidence.