Predicting Bitcoin Market Trend with Deep Learning Models
2
조회
0
다운로드
9
인용
0
참고문헌
키워드
초록
As bitcoin attracts public attention as a new method of investment, the bitcoin market becomes a field of research that should be technically analyzed. Since bitcoin market is very uncertain, time series associated with bitcoin prices are complex, nonstationary and chaotic. In this paper, we attempt to predict bitcoin market trend using deep neural network (DNN) and four types of recurrent neural networks (RNNs). Our data set consists of fourteen input variables related to the bitcoin prices recorded daily from September 1, 2013 to August 31, 2017. Thirteen input variables are related to bitcoin prices, which are date, market price, high price, low price and closing price, and eight technical indicators derived from bitcoin prices. The other is keyword feature value obtained through text mining for social network services (SNS) data. Empirical study shows that DNN outperforms four RNNs in terms of specificity, precision and accuracy. Bidirectional RNN outperforms the other four deep learning models in terms of sensitivity. As a whole, DNN works quite well compared with RNNs.
Abstract
As bitcoin attracts public attention as a new method of investment, the bitcoin market becomes a field of research that should be technically analyzed. Since bitcoin market is very uncertain, time series associated with bitcoin prices are complex, nonstationary and chaotic. In this paper, we attempt to predict bitcoin market trend using deep neural network (DNN) and four types of recurrent neural networks (RNNs). Our data set consists of fourteen input variables related to the bitcoin prices recorded daily from September 1, 2013 to August 31, 2017. Thirteen input variables are related to bitcoin prices, which are date, market price, high price, low price and closing price, and eight technical indicators derived from bitcoin prices. The other is keyword feature value obtained through text mining for social network services (SNS) data. Empirical study shows that DNN outperforms four RNNs in terms of specificity, precision and accuracy. Bidirectional RNN outperforms the other four deep learning models in terms of sensitivity. As a whole, DNN works quite well compared with RNNs.
참고문헌 (10)
- 강수지, "Deep Recommender System Using Purchase History Data", Quantitative Bio-Science, 36(1), pp.45~50 (2017), 자연과학연구소[DOI]
- Kingma DP, "Adam: A method for stochastic optimization", International Conference on Learning Representations (2015)
- Glorot X, "Understanding the difficulty of training deep feedforward neural networks", International Conference on Artificial Intelligence and Statistics, pp.249~256 (2010)
- Hinton GE, "A fast learning algorithm for deep belief nets", Neural Comput, 18, pp.1527~1554 (2006)[DOI]
- Elman J, "Finding structure in time", Cognitive Science, 14, pp.179~211 (1990)[DOI]
- Hochreiter S, "Long short-term memory", Neural Comput, 9, pp.1735~1780 (1997)[DOI]
- Cho K, "Learning phrase representations using RNN encoder-decoder for statistical machine translation", Conference on Empirical Methods in Natural Language Processing, pp.1724~1734 (2014)
- Schuster M, "Bidirectional recurrent neural networks", IEEE Transactions on Signal Processing, 45, pp.2673~2681 (1997)[DOI]
- Murphy J, "Technical analysis of the financial markets" (1999)
- Kim YB, "Predicting fluctuations in cryptocurrency transactions based on user comments and replies", PLoS ONE, 11, pp.e0161197~ (2016)[DOI]