I have developed a program that finds the probabilities of around 500 classes based on some training data that involves a few thousand features.

It works by training about 500 logistic regression models that take these few thousand features and find the probability of a single class. Each model finds the probability for a different class, so I am able to find to find the probability for each of the classes.

Since these are all different models, I have been able to find the accuracy for each model and have the mean accuracy by averaging all of these values.

My problem is that right now I have these 500 or so data points and the average and I don't really know how to represent them graphically. I can't really plot them with a line graph since there isn't much relation between the classes and the ROC curves don't work since this isn't a binary classification.

Does anyone have any suggestions on ways I can graph this data? Thank you!

CodePudding user response：

Are you saying that you have 500 classes which each have been assigned 500 probabilities? Does the mean accuracy refer to the mean accuracy for a particular class ( i.e. there are 500 different mean accuracies ) or is the mean accuracy refer to the accuracy of the probability assigned to each class?

Suggestion: Assign an arbitrary unique index of 1 to 500 to each class. Plot the class index against the class probability, with a vertical error bar