Machine Learning. Stochastic Method of Solving the Problem of Classification with a supervised imbalanced training Set

In machine learning with a large class imbalance in a supervised training set, a small class can be poorly found. We use Anderson's discriminant class estimates of posterior class probabilities to solve the imbalance problem. We do not change the proportions of classes in the training set, but choose the cost of classification errors from the ranges calculated from the posterior probabilities at points in the training set of a small class. Thus, we solve the problem in the framework of the stochastic method.