Neural network training using flower pollination algorithm for bitcoin price prediction

Abstract

One of the important uses of blockchain technology is cryptocurrencies. Interest in cryptocurrencies as a financial investment tool has increased recently. With the increasing interest, many cryptocurrencies have taken place in the market. The most popular among cryptocurrencies is bitcoin. Predictions and analyzes about the future are important in order to profit in bitcoin and other cryptocurrencies. For this, different traditional and artificial intelligence-based methods are used in the literature. One of the artificial intelligence techniques used is artificial neural networks (ANNs). One of the important processes of the ANN is the training process. In order to obtain effective results with the ANN, an effective training algorithm is needed. The flower pollination algorithm (FPA), which models the pollination process in nature, is one of the popular optimization algorithms. It has been used in the solution of many problems and has been accepted in the literature. In this study, an approach based on FPA and ANN is proposed for the prediction of bitcoin price. The weights of the ANN are determined using the FPA. With the proposed approach, time series analysis is performed using historical bitcoin prices. Daily bitcoin data between 1 April 2022 and 30 June 2022 is utilized. The applications are realized on different network structures for effective bitcoin price prediction. The results show that the proposed approach based on FPA and ANN is effective for the prediction of bitcoin price.