Algorithm for the Formation of a Training Set for a Neural Network on the Recognition of Chart Patterns

Abstract

© 2018 IEEE. The article gives an example of an algorithm for forming a training set for a neural network, predicting time series in the stock market through the recognition of one of the classic reversal chart pattern. The forecast is formed by an ensemble of neural networks. Each of the networks of the ensemble is trained on its own training set, created on the basis of the influence of a certain factor. At the same time, the entire total sample was formed according to the analytical description of the chart pattern used by traders.