Multi-scale convolutional neural network-based foreign exchange transaction prediction model

Abstract

The present invention relates to a multi-scale convolutional neural network-based foreign exchange transaction prediction model. The implementation process of the model includes the following steps of: first step, data processing: foreign exchange transaction real-time price data are converted into price curve images; second step, convolutional neural network system establishment; and third step, parallel feature learning implementation: price features are included into a local context graph so as to be adopted as the input of two shared hidden layers which are completely connected with each other, and parallel feature learning is performed so as to generate the output of foreign exchange price change prediction. The convolutional neural network system establishment step includes the following steps that: preprocessing is carried out, namely, price curve graphs at each time period are photographed, and price curve graph images are converted into grayscale graphs, and preprocessing features are obtained based on embedded features; and a sliding window and a correction linear unit are adopted to process a convolutional neural network so as to extract local contexts, local context sets are processed through a kernel, and the processed local context sets are connected, so that the local context graph can be generated. With the system adopted, loss can be stopped quickly, optimal operation can be achieved, the problem of over-fitting can be solved, and the limitations and mistakes of manual operation can be avoided.

Description

technical field [0001] The invention relates to the financial field, in particular to a foreign exchange transaction prediction model based on a multi-scale convolutional neural network. Background technique [0002] Convolutional Neural Networks (CNNs) is a feed-forward neural network whose artificial neurons can respond to surrounding units within a part of the coverage area, and have excellent performance for large-scale image processing. It includes alternating convolutional layers and pooling layers. [0003] Convolutional neural networks are mainly used to recognize two-dimensional graphics that are invariant to displacement, scaling, and other forms of distortion. Since the feature detection layer of the convolutional neural network learns through the training data, when using the convolutional neural network, it avoids explicit feature extraction and learns from the training data implicitly; The weights of the neurons on the network are the same, so the network can...

AI Technical Summary

Problems solved by technology

However, due to the differences in age, experience, experience and prediction methods of experts, their operational suggestions often have certain limitations.

In addition, experts usually predict the foreign exchange trend within half an hour to one or two days. For the impact of some emergencies, the response is often not timely enough, which may easily lead to operational errors and huge losses.

Images