251

Bitcoin Price Forecasting

Attributes count and Attribute’s list:

Eight numerical attributes are selected, as follows:

• Timestamp.

• Open.

• High.

• Low.

• Close.

• Volume_(BTC).

• Volume_(Currency).

• Weighted price.

Class count and class list: In this experiment, continuous numerical values are pre­

dicted. The Instance count: A total of 3,161,057 instances approximately selected

for training and testing. The different steps involved in this experimentation are

data preprocessing, important feature selection and loading training, and testing data

loaded with neural network and backpropagation network is applied to minimize the

error rate. The results are discussed in the next section.

15.5  RESULTS AND DISCUSSION

The results obtained using two models, RNN with ARIMA and RNN with LSTM,

are depicted as a graph in Figure 15.1 and Figure 15.2 at different points of time

(a particular date is considered).

FIGURE 15.1  Bitcoin price prediction on 24 August 2017 – actual price, RNN with ARIMA

and RNN with LSTM.