Attention enhanced long short-term memory network with multi-source heterogeneous information fusion: An application to BGI Genomics

Abstract

The recent availability of enormous amounts of both data and computing power has created new opportunities for predictive modeling. This paper compiles an analytical framework based on multiple sources of data including daily trading data, online news, derivative technical indicators, and time-frequency features decomposed from closing prices. We also provide a real-life demonstration of how to combine and capitalize on all available information to predict the stock price of BGI Genomics. Moreover, we apply a long short-term memory (LSTM) network equipped with an attention mechanism to identify long-term temporal dependencies and adaptively highlight key features. We further examine the learning capabilities of the network for specific tasks, including forecasting the next day’s price direction and closing price and developing trading strategies, comparing its statistical accuracy and trading performance with those of methods based on logistic regression, support vector machine,gradient boosting decision trees, and the original LSTM model. The experimental results for BGI Genomics demonstrate that the attention enhanced LSTM model remarkably improves prediction performance through multi-source heterogeneous information fusion, highlighting the significance of online news and time-frequency features, as well as exemplifying and validating our proposed framework.

A graph-based approach to multi-source heterogeneous information fusion in stock market

“…The performance of the proposed stock market data fusion analysis forecasting method is measured according to references [ 19 , 23 – 25 ] based on the accuracy (accuracy), F1 value and Mathews correlation coefficient (MCC). The accuracy, F1 value and Mathews correlation coefficient are shown in Eqs ( 11 )–( 13 ).…”

“…The news text information in traditional multisource data fusion studies comprises features extracted from only a single perspective [18,19]. Second, traditional multisource heterogeneous data fusion in the stock market is generally performed in a tensor or vector manner [22][23][24][25], and data redundancy is substantial given the current extremely large data volume and extremely fast information dissemination, which limits the efficiency of later data processing and analysis. When processing and analyzing multisource heterogeneous information, traditional econometric models cannot effectively capture and quantify the complex effects of nonlinear factors on asset volatility; furthermore, the black-box nature of machine learning algorithms has limitations in mining the implicit semantic information of complex financial data.…”

A scientific research topic trend prediction model based on multi‐LSTM and graph convolutional network

“…Zhang et al compile an analytical framework 33,34 based on multisource heterogeneous information fusion, which combines and capitalizes on all available information and applies an LSTM network equipped with attention mechanism to identify long‐term temporal dependencies and adaptively highlight key features. The effectiveness of the model has been verified on several prediction tasks, including forecasting the next day's price direction and closing price and developing trading strategies 35 . In this study, research topic trend prediction is essentially time series prediction problem.…”

“…The effectiveness of the model has been verified on several prediction tasks, including forecasting the next day's price direction and closing price and developing trading strategies. 35 In this study, research topic trend prediction is essentially time series prediction problem.…”

Progress and challenges for the application of machine learning for neglected tropical diseases

“…In Nguyen's study, they integrated more environmental variables irrespective of time-lagged (namely evaporation, humidity, rainfall, sunshine hours, and temperature) together with DF incidences for their dengue forecast model compared to Panja's and colleague's work. The authors proposed LSTM-ATT to be the best performing model when compared to CNN and regular LSTM because of having integrated an additional step of attention mechanism layer right after the LSTM network step(Zhang, Yang & Zhou 2021). enumerating the eggs-per-gram (EPG) metric.…”