A Financial Deep Learning Framework

Abstract

Prediction of stock price movement is regarded as a challenging task of financial time series prediction. Due to the complexity and massive financial market data, the research of deep learning approaches for predicting the future price is very difficult. This study attempted to develop a novel framework, named 13f-LSTM, where the AutoRegressive Integrated Moving Average (ARIMA), for the first time, as one of the technical features, Fourier transforms for trend analysis and Long-Short Term Memory (LSTM), including its variants, to forecast the future’s closing prices. Thirteen historical and technical features of stock were selected as inputs of the proposed 13f-LSTM model. Three typical stock market indices in the real world and their corresponding closing prices in 30 trading days are chosen to examine the performance and predictive accuracy of it. The experimental results show that the 13f-LSTM model outperforms other proposed models in both profitability performance and predictive accuracy.

Research on Intelligent Computing Methods for Multistep Process Capability

“…Reference [7] proposed the first end-to-end CCI learning method based on convolutional neural networks, forming a string representation of chemical structures. Reference [8] proposes the use of Fourier transform for trend analysis and long short-term memory neural network prediction of financial time series, and selects three typical stock market indices in the real world and their closing prices within 30 trading days to test their performance and prediction accuracy. Reference [9] used feature frames of different sizes to input into the LSTM network for efficient transportation mode detection, achieving a classification accuracy of up to 98%.…”

Deep Learning for Financial Time Series Prediction: A State-of-the-Art Review of Standalone and Hybrid Models

“…These techniques can be categorized based on the linearity of the model (linear vs. non-linear) or the prediction type (regression vs. classification). Linear models include statistical analysis methods and traditional machine learning ones, such as linear regression (LR), autoregression (AR) and its variants like autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA) [1], and autoregressive conditional heteroscedasticity (ARCH), and support vector machines (SVM) [2], etc. Non-linear models include decision trees (DT), k-nearest neighbors (KNN) [3], random forests (RF) [4], support vector regression (SVR) [5], artificial neural networks (ANN) [6], and variants of ANN like multilayer perceptron (MLP), convolutional neural network (CNN) [7], and long shortterm memory (LSTM) [8].…”