Add Thesis

The Influence of Bitcoin on Ethereum Price Predictions

Written by A. Caldegren

Paper category

Bachelor Thesis

Subject

Computer Science

Year

2018

Abstract

Thesis Cryptocurrencies: Cryptocurrencies is a new technology that has exploded in popularity and economic value in recent years [1]. The technology is based on cryptography and aims to achieve a decentralized currency. Except that cash is not decentralized [2], cryptocurrencies can be compared with digital versions of cash. By eliminating all middlemen, compared with transferring funds through the use of conventional financial institutions, it is possible to use international transactions with little or no fees [6]. In addition, by eliminating middlemen, cryptocurrency effectively eliminates the complete control of any institution, government, group, or individual. This makes most cryptocurrencies completely decentralized. Cryptocurrency does not rely on external participants to track every transaction, but on the so-called blockchain, also known as a distributed ledger [1, 2, 7]. Blockchain is one of the pillars of cryptocurrency and is distributed to any user who owns cryptocurrency software. Verification and authentication of transactions is completely left to the network users to perform. When a new transaction is made between two users, the staff on the network will verify and authenticate it, and insert it into the blockchain for other people on the network to use. See [8]. Since everyone knows every transaction, it becomes impossible to trick the system into giving someone an unearned coin, or to use the same coin for two transactions at the same time. Although transactions usually charge a small fee in the form of cryptocurrency, this fee is used to reward the staff who made the identity verification process possible. When a new cryptocurrency is released to the public, control of the cryptocurrency is released by the creator, and the network is run by the users of the cryptocurrency. In addition, cryptocurrency attempts to achieve anonymity to the greatest extent [9]. No transaction should be traced back to anyone. This is achieved by having an anonymous wallet that can be created by anyone and has nothing to do with anyone. These cryptocurrency wallets belong to people who know the password and have the hardware to store the wallet. Most cryptocurrencies are open source, anyone can understand how it works and why [10]. In addition, it allows anyone to create their own cryptocurrency. This has led to the creation of hundreds of cryptocurrencies, of which the largest and most influential is Bitcoin, which is worth US$144 billion as of the time of writing [11]. Every non-bitcoin cryptocurrency is classified as an alternative currency, also known as an altcoin [10]. There are two ways to obtain cryptocurrency: mining and trading. Supervised learning is the process of making neural networks learn by providing examples. These examples are marked in such a way that the correct answers have been provided with a set of inputs. Although supervised learning can be used for many different algorithms, this section will focus on how to apply it to neural networks. By training the labeled data set, the neural network can adjust its future prediction results to better reflect reality. If the neural network does not correctly guess the output of the input set, the training algorithm will look at the degree of difference between the guessed answer and the correct answer. The algorithm then adjusts the prediction function to provide better results in the future. The goal of this process is to achieve as many correct guesses as possible. After the training phase is completed, the data without a given answer will pass through the network and make predictions for the user. The data set used in this article does contain input and correct output. Most of these data should be used for training, while the other part is used to test the accuracy of the network. Instead of using the results of running a neural network on test data for training, the predictions are compared with labeled results from the data set to provide the accuracy of the network. However, when the data marked with the correct answer is incorrect, there are two main alternative methods available: unsupervised learning and reinforcement learning. In unsupervised learning, neural networks, and possibly creators, do not know what the correct output is from the beginning. Therefore, there is no way to evaluate the accuracy of the network. In reinforcement learning, as the name implies, behavior is reinforced by providing the network with rewards for guesses that lead to good results and penalizing the network for guesses that lead to bad results. However, since the data used in this paper has the correct data for training output, these methods will not be studied further. 2.3. Artificial Neural Network Artificial Neural Network, also known as ANN, is a type of algorithm brain that tries to replicate human functions [17]. By using neurons to create models, usually in multiple layers with input layers, one or more hidden and output layers, and synapses (also called connections), artificial neural networks can simulate a small part of the brain. Figure 1 shows a single neuron with synaptic connections. If the network has multiple hidden layers, it is classified as a deep neural network. Each layer in the neural network has any number of neurons. Each neuron is connected to each neuron at the next level through a weighted synapse. The weight of these synapses will affect the final result. Read Less