Deep state space model for predicting cryptocurrency price

作     者：Sharma, Shalini Majumdar, Angshul 

作者机构：Indraprastha Inst Informat Technol Delhi India 

出 版 物：《INFORMATION SCIENCES》 (信息科学)

年 卷 期：2022年第618卷

页      面：417-433页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Infosys Center for Artificial Intelligence at Indraprastha Institute of Information Technology-Delhi  (IIIT Delhi) 

主　　题：Time series analysis Deep state space models Deep matrix factorization Kalman filtering Bayesian smoothing EM algorithm Cryptocurrency forecasting Dynamic recurrent network 

摘      要：There are two fundamental contributions of this work. On the application side, one of the most challenging problems is tackled, predicting day-ahead crypto-currency prices. On the theoretical front, a new dynamical modeling approach is proposed. The proposed approach keeps the probabilistic formulation of the State-Space Model that yields point estimates along with the uncertainty about the estimate and the function approximation ability of the deep neural network. We call the proposed approach the deep state-space model. The experiments are carried out on established cryptocurrencies (obtained from Yahoo Finance). The goal of the work has been to predict the price for the next day. Benchmarking has been done with both state-of-the-art and classical dynamical modeling techniques. Results show that the proposed approach yields the best overall results in terms of accuracy.(c) 2022 Elsevier Inc. All rights reserved.