248

Blockchain Technology

The price of bitcoin has been influenced by many factors. Bitcoin price and inves­

tors attention are very closely related, it has been influenced by information-driven

and noise components (Ibikunle et al., 2020). Another interesting fact about bitcoin

has been identified: The close relationship between bitcoin users, bitcoin miners and

investors, and bitcoin price. These factors strongly influence the bitcoin price (Chen

et al., 2020b). In most cases, bitcoin is considered a speculative asset or investment

because of its rapid growth and sudden fluctuations. A quantitative framework has

been developed to predict bitcoin price, bitcoin price behaviour, identification of

strongly influential parameters, the sentiment of bitcoin users, investor interest, and

attention towards investment and selling. This model is a continuous-time model,

and it did not consider bitcoin as currency but rather, as a speculative asset because

of rapid changes and fluctuations in price (Cretarola et al., 2020).

The speculative nature of bitcoin has created many prediction models using differ­

ent kinds of techniques to predict the price of bitcoin. Statistical techniques (Khedr

et al., 2021), machine learning techniques (Jalali and Heidari, 2020), recurrent unit

approach (Dutta et al., 2020), grey system theory, ensembles of neural networks

(Sin & Wang, 2017), deep learning techniques (Ji et  al., 2019), Auto-Regressive

Integrated Moving Average (ARIMA) (Poongodi et  al., 2020) and other predic­

tion techniques have been applied to predict bitcoin price. This research proposes

a hybrid model in which recurrent loops in the activation function of the recurrent

neural network (RNN) model are combined with ARIMA.

In this chapter, Section 15.2 describes a literature review on stochastic models and

RNNs, Section 15.3 describes the research problem, Section 15.4 describes experi­

mentation and data sets used for bitcoin price prediction, Section 15.5 describes the

results with implications and Section 15.6 concludes the chapter.

15.2  LITERATURE REVIEW

The price of bitcoin changes every day, and the change can be considered as a daily

price change as well as a high-frequency price change. The bitcoin price change is

dominated by many factors such as property and network, trading and market, atten­

tion and gold spot price. These factors are considered high-dimensional features.

Also, the basic bitcoin trading feature exchange of cryptocurrency is considered to

predict the bitcoin price for a short interval of time (Chan et al., 2020c).

The price of bitcoin is affected by many factors (Chan et al., 2020c), and it is

measured using two phases. In the first phase, the different trends are associated

with bitcoin, and in the second phase, the available information about bitcoin is

considered to measure the price using machine learning techniques (Velankar et al.,

2018). Similarly, bitcoin price is predicted using a decision tree in which parameters

such as bitcoin’s open price, high price, low price and closing price are considered

(Rathan et al., 2019). The bitcoin price is also compared with time series analysis and

machine learning techniques (Felizardo et al., 2019).

Also, deep learning techniques such as long short-term memory (LSTM) and

gated recurrent unit (GRU) are applied to predict the bitcoin price. These tech­

niques were found to be very good at predicting the price of bitcoin very accurately