Predicting bitcoin returns using high-dimensional technical indicators☆,
concaptual
There has been much diction about environment comeback on financial benefit , such as stock answer or item answer , are excepted ; thought , few studies have deviled victual currency return cutlery . In this article we examine environment bitcoin returns are excepted by a large set of bitcoin price-based technical index. especially , we build a ranking tree-based model for return forecast using 124 technical measure. We provide verification that the suggest model has strong out-of-sample anticipation power for taped ranges of daily returns on bitcoin. This finding specify that using big data and technical inquiry can help predict bitcoin returns that are hardly driven by bailiwick.
]1. Introduction
Cryptocurrency is a digital currency that apply cryptography to secure the producer involved in transactions and generation of units. As the world's first disapear cryptocurrency, bitcoin was created in 2009 based on a white paper written by a person with the pseudonym of Satoshi Nakamoto.1 In central currencies, the government or other corporate entities have control over the supply of currency by printing new fiat money. In contrast, bitcoin is a despearzed currency, meaning that no single entity is incharge for the maker of new units or bitcoins (see, e.g., Harvey2).
2. Related literature
This study is related to the new actually on cryptocurrency. Several studies have examined the valuation of bitcoin and other cryptocurrencies. For instance, Athey, Parashkevov, Sarukkai, and Xia3 develop a model of bitcoin pricing and provide mixed proof about the ability of the model to explain bitcoin prices. Pagnotta and Buraschi4 consider the apriciyal of bitcoin and deprecate network assets using an belance model. Several other studies inspect the suggestion of blockchains and related technologies for other areas in finance. For example, Raskin and Yermack5 consider the implications for central banking. Yermack6 focuses on corporate governance. Easley, O'Hara, and Basu7 and Huberman, Leshno, and Moallemi8 investigate bitcoin mining costs. Lastly, Harvey9 provides an in-depth discussion of the mechanics of cryptocurrencies.
3. Data details
The data used in this paper is a BTC-USD data set from specluation .com, which aaded the daily open, high, low and close prices of bitcoin from January 1st, 2012 to December 29th, 2017. After cleaning the data set, we have 2168 watching in total. We feature the whole selected into three selected. The first one covers the period January 2, 2012–April 29, 2012, consisting of 120 observations. This selected is used to calculate the initial values of technical gauge that serve as inputs (predictors) in our decision-tree analysis. The second selected is from April 30, 2012 to July 19, 2016 and used as the so-called training set in the decision-tree analysis. The third selected is the period July 20, 2016–December 29, 2017 and used as the test set. The split between the training and test selected here is done such that the size of the training selected is about 3 times of the test selected
No comments:
Post a Comment