Fuzzy Crow Search Algorithm-Based Deep LSTM for Bitcoin Prediction

Fuzzy Crow Search Algorithm-Based Deep LSTM for Bitcoin Prediction

Chandrasekar Ravi
Copyright: © 2020 |Pages: 19
DOI: 10.4018/IJDST.2020100104
(Individual Articles)
No Current Special Offers


Prediction of stock market trends is considered as an important task and is of great attention as predicting stock prices successfully may lead to attractive profits by making proper decisions. Stock market prediction is a major challenge owing to non-stationary, blaring, and chaotic data and thus, the prediction becomes challenging among the investors to invest the money for making profits. Initially, the blockchain network is fed to the blockchain network bridge from which the bitcoin data is acquired that is followed with the bitcoin prediction. Bitcoin prediction is performed using the proposed FuzzyCSA-based Deep Long short-term memory (LSTM). At first, the flow strength indicators are extracted based on Double exponential moving average (DEMA), Rate of Change (ROCR), Average True Range (ATR), Simple Moving Average (SMA), and Moving Average Convergence Divergence (MACD) from the blockchain data. Based on the extracted features, the prediction is done using FuzzyCSA-based Deep LSTM, which is the combination of FuzzyCSA with Deep LSTM. Then, the CSA is modified using the fuzzy operator for determining the optimal weights in Deep LSTM. The experimentation of the proposed method is performed from the openly available dataset. The analysis of the method in terms of Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) reveals that the proposed FuzzyCSA-based Deep LSTM acquired a minimal MAE of 0.4811, and the minimal RMSE of 0.3905, respectively.
Article Preview

1. Introduction

The advancements in stock price prediction have gained significant importance among expert analysts and investors. The stock market is defined as the collection of markets and interactions in which the trading and issuing of commodities, stocks, derivatives, and other sorts of securities takes place. The complexities of the stock prices adapt certain factors that involve quarterly earnings’ reports, market news, and varying changing behaviours (Nakamoto, 2008). Several researches have been devised for introducing a predictive model effectively to support traders to make wise investment decisions (Næs & Skjeltorp, 2006). The accurate prediction of stock trends is interesting and a complex task in the changing industrial world. Several aspects, which affect the behaviour of stock trends, are non-economic and economic factors and are taken into consideration. Thus, stock market prediction is considered as a major challenge for increasing the production (Amjad & Shah, 2017). Market direction is predicted by various methods, like technical analysis, statistical analysis, and fundamental analysis. Several investors are presently seeing cryptocurrencies as a promising asset class. Bitcoin is a cryptocurrency that would be the disrupting technology into traditional financial systems because the blockchain techniques ignore the involvement of banking sector and monetary authorities from financial transactions. Bitcoin plays a very important role in commerce and recognition of digital cryptocurrencies, because of its limited supply; cost of transaction is limited, ability to easily transfer value across state borders, and capability to perform as a store of value against unpredictable fiat currencies (Campbell & Thompson, 2007).

Bitcoin is located from peer-to-peer network system through World Wide Web. When new transactions are introduced in the system, the verification is performed using network nodes and recording is done in a public distributed ledger known as blockchain (Indera et al., 2017). Blockchain is utilized for creating unchangeable, permanent, and transparent record of exchange and processing without relying on a central authority (Guo et al., 2018). The fundamental characteristics of blockchain technology are decentralized system. The blockchain mechanism has the massive defect from a distributed consensus in which the verification is done for each online transaction without compromising the privacy of the digital assets and parties involved in any time (Crosby et al., 2015). Blockchain technology employs broadcast mechanism to perform distributed consensus as message transmission. Here, the broadcast approach introduces unnecessary data and messages to the network system, which affects the high message overhead (Poonpakdee et al., 2018).

Various models applied for predicting the Bitcoin are managed using the time series models that involve Auto-Regressive Conditional Heteroscedastic (ARCH) model, Generalized Auto-Regressive Moving Average (GARCH), and Auto-Regressive Moving Average (ARMA) (Katsiampa, 2017; Kumar, 2006). In (Ciaian et al., 2016), bitcoin price prediction is done based on linear algorithm that is categorized into several features, such as investor attraction features, features of market competitive force, and global economic features. Here, the first two features have a great impact on Bitcoin prices. In (McNally, 2016), prediction of Bitcoin is done using some machine learning technologies, like recurrent neural network, long, and short-term memory. After that, Autoregressive Integrated Moving Average (ARIMA) is introduced for lower prediction accuracy. Concept in (Madan et al., 2015) conducts a comparative discussion on the performance of Bitcoin price predictive based on random forest, binomial logistic regression, and support vector machine. Relatively some studies focussed on forecasting and time-series modeling of Bitcoin price prediction. Some work utilizes the optimization algorithms (Remmiya & Abisha, 2018; Dennis & Muthukrishnan, 2014; Binu, 2015) and machine learning methods (Darekar & Dhande; Menaga & Revathi; Krishnamoorthy et al., 2016) for bitcoin price prediction. Bayesian neural network (BNN) method is employed for forecasting as well as modeling the Bitcoin prices by considering its capacity of dealing with a huge number of relevant features. BNN is also utilized for avoiding the overfitting issue based on regularization function (Dennis & Muthukrishnan, 2014; Zheng, 2018).

Complete Article List

Search this Journal:
Volume 14: 2 Issues (2023)
Volume 13: 8 Issues (2022)
Volume 12: 4 Issues (2021)
Volume 11: 4 Issues (2020)
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing