Investment Model Based on LSTM Network Forecasting and Portfolio Investment

Abstract

In recent years, Bitcoin has attracted more and more investors' attention. Different from traditional financial assets, the price of Bitcoin is extremely volatile. Therefore, investors need to choose an appropriate asset portfolio to hedge the risk of bitcoin with traditional financial assets. Long Short Term Memory (LSTM) networks are the most advanced sequential learning methods in deep learning of time series prediction, which can accurately predict the prices of financial assets and thus provide traders with the most profitable asset portfolio. In this paper, LSTM-CNN asset price prediction models are developed with data of gold and bitcoin in 2016-2021 as research samples to obtain return maximizing trading strategy. The contribution of this study is to apply deep learning to financial asset price prediction to improve the prediction accuracy and use it as a basis for asset allocation.