Deep Video Prediction for Nonlinear Market Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time series forecasting methods, such as ARIMA and VAR, struggle to capture nonlinear patterns in market data, leading to insufficient forecasting accuracy.
Innovation Solution
Convert time-series data into images using two-dimensional visualization techniques and apply a convolutional neural network (CNN) and Long Short-Term Memory (LSTM) algorithm with stochastic latent residual video prediction (SRVP) to generate future market data predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional statistical methods like ARIMA are used for time series forecasting, then the methods are simple and widely adopted, but they cannot capture nonlinear patterns in market data, leading to insufficient forecasting accuracy
Solution Approach 1:
The patent transforms univariate time series data into bivariate sequences by creating adjacent time frame pairs. Each input frame contains historical data points and each output frame contains future data points to be predicted. This dimensional transformation enables the use of CNNs originally designed for image processing, allowing the model to capture spatial patterns in temporal data and nonlinear relationships that conventional ARIMA methods miss.
2Loss of information
If visualizations are used to represent time series data, then spatial structural information is provided which helps human understanding, but the data must be converted from numerical format to image format
Solution Approach 1:
The patent introduces an intermediary visualization layer that converts numerical time series data into image representations. Each time frame is visualized as an image where data points are positioned according to their temporal sequence and valued by their magnitude. This intermediary visual representation serves as a bridge between raw numerical data and the CNN model, enabling the network to perceive spatial patterns and relationships that are not apparent in tabular form.
Data Source
AI summary
A method for forecasting a change in a market is provided. The method includes: using historical market data to generate a plurality of first images that correspond to a predetermined time sequence; generating, based on the plurality of first images, second images that correspond to a future time point with respect to the predetermined time sequence; and determining a prediction of future market data based on the second images. The generation of the second images and the prediction of the future market data are implemented by applying a convolutional neural network (CNN) algorithm that implements a stochastic latent residual video prediction (SRVP) technique with respect to a group of financial assets.