[PYTHON] Bitcoin Price Forecast on TensorFlow (LSTM)

What is Bitcoin?

Bitcoin is a type of virtual currency. There are various definitions of virtual currency, but according to the Japanese Bitcoin exchange "bitFlyer" Below.

It is the currency in the online game that is easy to imagine as a virtual currency. If you pay yen or dollars to get the currency used in the game, you can buy items at in-game stores where the currency used in the game is "usable". A virtual currency that can only be used in-game. Source * 1: What is Bitcoin?

Bitcoin has the highest unit price and large transaction volume among virtual currencies, so investors are competing to invest.

What is TensorFlow (LSTM)?

Bitcoin's price forecast is a draft of the following article. For more information on TensorFlow and LSTM, please refer to the following articles.

Source * 2: [Stock Price Forecast with TensorFlow (LSTM) ~ Stock Forecast Part 1 ~](http://qiita.com/tsunaki/items/a5f3f975a31dc45fc9c9#%E3%81%AF%E3%81%98%E3% 82% 81% E3% 81% AB)

According to the article of the source * 2, the neural network can calculate the information of the next step in a chain by referring to the information of the N step before. Using this principle, you can easily predict the data of the next step of time series data.

About the data to be handled

When I searched for an HP that has provided time series data of Bitcoin prices over the past few months, I arrived at the following blog.

You can also get all the data owned by Bitcoincharts at once. If you access the URL below, the transaction history of various exchanges and currencies is listed, and you can easily obtain the data by selecting the data, saving it, and decompressing it locally. 　　　　　　　　　　　 ●Bitcoincharts 　　　　 Source * 3: Get and visualize the past transaction history of bitcoin with Bitcoin arts

result

Let's feed the Bitcoin price data to the code of the source * 2 quickly. 'predict' is the predicted price and'actual' is the actual price.

I feel like I can predict it.