Analyze and reconstruct order book to develop a model to predict cryptocurrency assets price by Mikhail Lyashenko

Chapter 1

Introduction
In this chapter, we go through motivation behind the thesis, introduce the reader to the
overall theme, state the problem and define the goals.
1.1 Motivation
Motivation behind this thesis is to inspire anyone, who is reading this thesis, to make
researches in the world of cryptocurrency as a whole and continue the work of predicting
the price of cryptocurrencies. This is an ambitious project, considering its relative novelty.
It may not have given the results, that were hoped for, but it surely gave a contribution
for future work.
1.2 Introduction to the theme and problem
Statement
As the world evolves, also does its financial side. In the year 2009, the first decentralized
was introduced by pseudonymous developer Satoshi Nakamoto. It was Bitcoin, which now
has an estimated market capitalization of 747 billion dollars
. At first decentralization,
that was brought by , was what interested people about it. Removing enormous control
of banks and huge organization. It is the technology, that also has a great security, which
increases, as more people get involved. Then in the year 2011, a first altcoin was born,
called Namecoin. Compared to Bitcoin, many altcoins brought a completely different
spectrum of functionality, from faster transactions to technology platforms intended to
help manage assets and data easily and inexpensively. One of the first altcoins and now
the second biggest in terms of market capitalization is Ethereum
, which is most popular

because of its smart contracts functionality. One of the phenomenons in financial terms are
memecoins, which usually has some humorous background and does not have any utility.
Dogecoin is one of them, which takes the 10th place in terms of market capitalization
. The
question stands: is it possible to predict prices of , as a fairly new thing? Bitcoin, Ethereum
and Dogecoin, these three completely different are going to a part of this research. Machine
learning is used for making predictions, in particular recurrent neural networks, such as
LSTM and GRU. The research is going to be made solely on the market. Order flow is
going to be the key thing in this thesis: decision stands between limit order books and
executed transaction, what works best for the prediction?
1.3 Goals of the Bachelor Thesis
1. Collect data and analyze it.
2. Build LSTM and GRU to predict future volumes.
3. Build deep neural network to predict price based on volumes.
4. Combine step 2. and 3. in a pursue to predict price based on predicted volumes.
1.4 Structure of the Bachelor Thesis
The thesis is organized into chapters as follows:
1. Introduction: Gives a gentle introduction to the subject and set out goals of the
thesis.
2. Background and State-of-the-Art: Introduces the necessary theoretical background
and surveys the current state-of-the-art.
3. Overview of Our Approach: States approach, used in the thesis.
4. Main Results: Describes results main results, considering volume prediction and
results, considering future price predictions.
5. Conclusions: Summarizes results received, designates contribution of the thesis and
proposes approaches for future research.

Analyze and reconstruct order book to develop a model to predict cryptocurrency assets price by Mikhail Lyashenko