Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

Cryptocurrency Price Analysis With Artificial Intelligence

Chittala Madhuri Venkata Phani Sree,
Sri.V.Bhaskara Murthy,
MCA Student, Associate Professor
Dept Of MCA
B.V.Raju College, Bhimavaram

ABSTRACT an algorithm to record a block of transactions to

public ledger named blockchain and the
Cryptocurrency is playing an increasingly cryptocurrency is created when the block is
important role in reshaping the financial system added to the blockchain. It allows people to store
due to its growing popular appeal and mechant and transfer through encryption protocol and
acceptance. While many people are making distributed network. Mining is a necessary and
investments in Cryptocurrency, the dynamical competitive component of the cryptocurrency
features, uncertainty, the predictability of system. The miner with more computational
Cryptocurrency are still mostly unknown, which power has a better chance of finding a new coin
dramatically risk the investments. It is a matter than that of less. Bitcoin is the first and one of
to try to understand the factors that infiuence the the leading digital currencies (its market
value formation. In this study, we use advanced capitalisation had more than $ 7 billion in 2014,
artificial intelligence frameworks of fully and then it increased significantly to $ 29 billion
connected Artificial Neural Network (ANN) and in 2017) which was first introduced by Satoshi
Long Short-Term Memory (LSTM) Recurrent Nakamoto in 2008. Among many features of
Neural Network to analyse the price dynamics of bitcoin, the most impressive one is
Bitcoin, Etherum, and Ripple. We find that decentralisation that it can remove the
ANN tends to rely more on long-term history involvement of traditional financial sectors and
while LSTM tends to rely more on short-term monetary authorities effectively due to its
dynamics, which indicate the efficiency of blockchain network features . In addition, the
LSTM to utilise useful information hidden in electronic payment system of Bitcoin is based on
historical memory is stronger than ANN. cryptographic proof rather than the trust between
However, given enough historical information each other as its transaction history cannot be
ANN can achieve a similar accuracy, compared changed unless redoing all proof of work of all
with LSTM. This study provides a unique blockchain, which play a critical role of being a
demonstration that Cryptocurrency market price trust intermediary and this can be widely used in
is predictable. However, the explanation of the reality such as recording charitable contribution
predictability could vary depending on the to avoid corruption. Moreover, bitcoin has
nature of the involved machine-learning model. introduced the controllable anonymity scheme,
and this enhances users’ safety and anonymity
I. INTRODUCTION
by using this technology, for instance, we can
Cryptocurrency is the peer-to-peer digital take advantage of this property of blockchain to
moneyory and payment system that exist online make identification cards, and it not only can
via a controlled algorithm. When a miner cracks protect our privacy but verify our identity.

www.jespublication.com Page 971

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

Nowadays, investing in cryptocurrencies, like success. In an Multi-Layer Perceptron (MLP)

Bitcoin, is one of the efficient ways of earning based prediction model was presented to
money. For example, the rate of Bitcoin forecast the next day price of bitcoin by using
significant rises in 2017, from a relatively low two sets of input: the first type of inputs: the
point 963 USD on January 1ST 2017, to its peak opening, minimum, maximum and closing price
19186 USD on December 17th 2017, and it and the second set of inputs: Moving Average of
closed with 9475 USD at the end of the year. both short (5,10,20 days) and long (100, 200
Consequently, the rate of return of bitcoin days) windows. During validation, their model
investment for 2017 was over 880%, which is an was proved to be accurate at the 95% level.
impressive and surprising scenery for most There has been many academic researches
investors. While an increasing number of people looking at exchang rate forecasting, for example,
are making investments in Cryptocurrency, the the monetary and portfolio balance models
majority of investors cannot get such profit for examined by Meese and Rogoff (1983, 1988) .
being inconsiderable to cryptocurrencies’ Significant efforts have been made to analyse
dynamics and the critical factors that influence and predict the trends of traditional financial
the trends of bitcoins. Therefore, raising markets especially the stock market however,
people’s awareness of vital factors can help us to
be wise investors. Although market prediction is predicting cryptocurrencies market prices is still
demanding for its complex nature the dynamics at an early stage. Compared to these stock price
are predictable and understandable to some prediction models, traditional time series
degree. For example, when there is a shortage of methods are not very useful as cryptocurrencies
the bitcoin, its price will be increased by their are not precisely the same with stocks but can be
sellers as investors who regard bitcoin as a deemed as a complementary good of existing
profitable investment opportunity will have a currency system with sharp fluctuations features.
strong desire to pay for bitcoin. Furthermore, the Therefore, it is urgently needed to understand
price of bitcoin may be easily influenced by the dynamics of cryptocurrencies better and
some influential external factors such as political establish a suitable predictive modelling
factors . Although existing efforts on framework. In this study, we hypothesise that
Cryptocurrency analysis and prediction is time series of cryptocurrencies exhibits a clear
limited, a few studies have been aiming to internal memory, which could be used to help
understand the Cryptocurrency time series and the memory-based time series model to works
build statistical models to reproduce and predict more appropriately if the length of internal
price dynamics. For example, Madan et al. memory could be quantified. We aim to use two
collected bitcoins price with the time interval of artificial intelligence modelling frameworks to
0.5, 1and 2 hours, and combined it with the understand and predict the most popular
blockchain network, the underlying technology cryptocurrencies price dynamics, including
of bitcoin. Their predictive model leveraging Bitcoin, Ethereum, and Ripple.
random forests and binomial logistic regression
II. LITERATURE SURVEY
classifiers , and the precision of the model is
around 55% in predicting bitcoin’s price. Shah
1) Using the Bitcoin Transaction Graph to
et al. used Bayesian regression and took
Predict the Price of Bitcoin AUTHORS:
advantages of high frequency (10-second) prices
Greaves, A., & Au, B.
data of Bitcoin to improve investment strategy
of bitcoin . Their models had also achieved great

www.jespublication.com Page 972

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

with different objectives. A pre-defined set of expenses or a maximum of the train-ing level
minimum qualification levels should be with a limitation of the budget. First, a
distributed between the crew members with description of the cosmonaut training process is
minimum training time differences, training given. Then four models are considered for the
expenses or a maximum of the train-ing level volume planning problem. The objective of the
with a limitation of the budget. First, a first model is to minimize the differences
description of the cosmonaut training process is between the total time of the preparation of all
given. Then four models are considered for the crew members, the objective of the second one
volume planning problem.The objective of the is to minimize the training expenses with a
first model is to minimize the differences limitation of the training level, and the objective
between the total time of the preparation of all of the third one is to maximize the training level
crew members, the objective of the second one with a limited budget. The fourth model
is to minimize the training expenses with a considers the problem as an 𝑛-partition problem.
limitation of the training level, and the objective Then two models are considered for the calendar
of the third one is to maximize the training level planning problem Abstract We consider the
with a limited budget. The fourth model problem of planning the ISS cosmonaut training
considers the problem as an 𝑛-partition problem. with different objectives. A pre-defined set of
Then two models are considered for the calendar minimum qualification levels should be
planning problem. We consider the problem of distributed between the crew members with
planning the ISS cosmonaut training with minimum training time differences, training
different objectives. A pre-defined set of expenses or a maximum of the train- ing level
minimum qualification levels should be with a limitation of the budget. First, a
distributed between the crew members with description of the cosmonaut training process is
minimum training time differences, training given. Then four models are considered for the
expenses or a maximum of the train-ing level volume planning problem. The objective of the
with a limitation of the budget. First, a first model is to minimize the differences
description of the cosmonaut training process is between the total time of the preparation of all
given. Then four models are considered for the crew members, the objective of
volume planning problem. The objective of the the second one is to minimize the training
first model is to minimize the differences expenses with a limitation of the training level,
between the total time of the preparation of all and the objective of the third one is to maximize
crew members, the objective of the second one the training level with a limited budget. The
is to minimize the training expenses with a fourth model considers the problem as an 𝑛-
limitation of the training level, and the objective partition problem. Then two models are
of the third one is to maximize the training level considered for the calendar planning problem
with a limited budget. The fourth model Bitcoin is the world’s leading cryptocurrency,
considers the problem as an 𝑛-partition problem. allowing users to make transactions securely and
Then two models are considered for the calendar anonymously over the Internet. In recent years,
planning problem Abstract The Bitcoin the ecosystem has gained the
We consider the problem of planning the ISS attention of consumers, businesses, investors and
cosmonaut training with different objectives. A speculators alike. While there has been
pre-defined set of minimum qualification levels significant research done to analyze the network
should be distributed between the crew members topology of the Bitcoin network, limited
with minimum training time differences, training research has been performed to analyze the

www.jespublication.com Page 973

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

network’s influence on overall Bitcoin price. In examining 66 of the most used such 'coins'. A
this paper, we investigate the predictive power regression model was estimated that points to
of blockchain network-based features on the three main drivers of cryptocurrency value: the
future price of Bitcoin. As a result of difficulty in 'mining 'for coins; the rate of unit
blockchain-networkbased feature engineering production; and the cryptographic algorithm
and machine learning optimization, we obtain employed. These amount to relative differences
up-down Bitcoin price movement classification in the cost of production of one coin over
accuracy of roughly 55%. another at the margin, holding all else equal.
Bitcoin-denominated relative prices were used,
We consider the problem of planning the ISS avoiding much of the price volatility associated
cosmonaut training with different objectives. A with the dollar exchange rate. The resulting
pre-defined set of minimum qualification levels regression model can be used to better
should be distributed between the crew members understand the drivers of relative value observed
with minimum training time differences, training in the emergent area of cryptocurrencies. Using
expenses or a maximum of the train- ing level the above analysis, a cost of production model is
with a limitation of the budget. First, a proposed for valuing bitcoin, where the primary
description of the cosmonaut training process is input is electricity. This theoretical model
given. Then four models are considered for the produces useful results for both an individual
volume planning problem.The objective of the producer, by setting breakeven points to start
first model is to minimize the differences and stop production, and for the bitcoin
between the total time of the preparation of all exchange rate on a macro level. Bitcoin
crew members, the objective of the second one production seems to resemble a competitive
is to minimize the training expenses with a commodity market; in theory miners will
limitation of the training level, and the objective produce until their marginal costs equal their
of the third one is to maximize thetraining level marginal product.
with a limited budget. The fourth model
considers the problem as an 𝑛-partition problem.
3. Economic prediction using neural
Then two models are considered for the calendar
networks: the case of IBM daily stock returns
planning problem. For the volume planning
problem, two algorithms are presented. The first
AUTHORS: H. White
one is a heuristic with a complexity of
(𝑛) operations. The second one consists of a
A report is presented of some results of an
heuristic and exact parts, and it is based on
ongoing project using neural-network modeling
the 𝑛-partition problem appro
and learning techniques to search for and decode
nonlinear regularities in asset price movements.
2) CRYPTOCURRENCY VALUE
The author focuses on the case of IBM common
FORMATION: AN EMPIRICAL
stock daily returns. Having to deal with the
ANALYSIS LEADING TO A COST OF
salient features of economic data highlights the
PRODUCTION MODEL FOR VALUING
role to be played by statistical inference and
BITCOIN
requires modifications to standard learning
AUTHORS: Hayes, A. S.
techniques which may prove useful in other
This paper aims to identify the likely source(s) contexts.
of value that cryptocurrencies exhibit in the
marketplace using cross sectional empirical data

www.jespublication.com Page 974

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

III. EXISTING SYSTEM of blockchain to make identification

cards, and it not only can protect our
Although existing efforts on Cryptocurrency privacy but verify our identity.
analysis and prediction is limited, a few studies  By using ANN and LSTM We can
have been aiming to understand the predict the future price of
Cryptocurrency time series and build statistical cryptocurrency
models to reproduce and predict price dynamics.  and time series also successful We can
Madan et al. collected bitcoins price with the find the 100% result.
time interval of 0.5, 1and 2 hours, and combined
it with the blockchain network, the underlying
technology of bitcoin. Their predictive model
leveraging random forests and binomial logistic V. SYSTEM ARCHITECTURE
regression classifiers and the precision of the
model is around 55% in predicting bitcoin’s
price. While an increasing number of people are
making investments in Cryptocurrency, the
majority of investors cannot get such profit for
being inconsiderable to cryptocurrencies’
dynamics and the critical factors that influence
the trends of bitcoins.

IV. PROPOSED SYSTEM

Among many features of bitcoin, the most

VI. IMPLEMENTATION
impressive one is decentralisation that it can
User
remove the involvement of traditional financial
Registration of User details next admin activated
sectors and monetary authorities effectively due
by the user. go to user page and login. User
to its block chain network features. In proposed
view some fields [start trading, bit bucket,
system we are using ANN algorithm and LSTM
prediction, logout] now open start treding page
easily predict the time series of crypto currency
contains sale available cryptocurrencies and user
prices. we use an ANN model to predict the
buy the currency.user views the transaction
price of Bitcoin one day into the future using
history details and user view the prediction for
five different lengths of memory. While LSTM
available datasets.
is intentionally designed to model the internal
Agent
memory flow and its impact on future
First registration of the agent page. next
prediction, therefore, both ANN and LSTM are
activated by the admin. Login agent page. agent
suitable for the crypto currencies price time
view some fields[buy, bitbucket, block bucket,
series prediction
prediction, logout]. here three types of digital
ADVANTAGES OF PROPOSED SYSTEM: currency’s are there [ bitcoins, ripple, Etherum]
agent can buy digital currency (any one or all
 The bitcoin has introduced the three).Agent buying the crypto currency. agent
controllable anonymity scheme, and this transaction history also available.
enhances users’ safety and anonymity Admin
by using this technology, for instance, The aim of admin is to approve the users and
we can take advantage of this property agents .admin contains some

www.jespublication.com Page 975

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

fields[user,agents,crypto,bitblock].admin view There are a large number of cryptocurrencies

user and agent registered details and in crypto sprang up after Bitcoin such as Ethereum and
field user update the currecy rate and view the Ripple. Because of the significant uncertainty in
recently crypto currency changes list.admin its prices, many people hold them as a means of
view the current transaction details. When a speculation. Therefore, it is critically important
miner cracks an algorithem to record a block of to understand the internal features and
transactions to public ledger named blockchain predictability of those cryptocurrencies. In this
and the cryptocurrency is created when the block study, we use two distinct artificial intelligence
is added to the blockchain. It allows people to frameworks, namely, fully-connected Artificial
store and transfer through encryption protocol Neural Network (ANN) and Long-Short-Term-
and distributed network. Mining is a necessary Memory (LSTM) to analyse and predict the
and competitive component of the price dynamics of Bitcoin, Etherum, and Ripple.
cryptocurrency system. The miner with more We showed that the ANN and LSTM models are
computational power has a better chance of comparable and both reasonably well enough in
finding a new coin than that of less . Bitcoin is price prediction, although the internal structures
the first and one of the leading digital currencies are different. Then we further analyse the
(its market capitalisation had more than $ 7 influence of historical memory on model
billion in 2014, and then it increased prediction. We find that ANN tends to rely more
significantly to $ 29 billion in 2017) which was on long-term history while LSTM tends to rely
first introduced by Satoshi Nakamoto in 2008. more on short-term dynamics, which indicate the
Among many features of bitcoin, the most efficiency of LSTM to utilise useful information
impressive one is decentralisation that it can hidden in historical memory is stronger than
remove the involvement of traditional financial ANN. However, given enough historical
sectors and monetary authorities effectively due information ANN can achieve a similar
to its blockchain network features. accuracy, compared with LSTM. This study
Artificial intelligence provides a unique demonstration that
The application of advanced digital, smart Cryptocurrency market price is predictable.
technologies, robotic systems, new materials and However, the explanation of the predictability
design techniques, creation of large data could vary depending on the nature of the
processing systems, computer-aided learning involved machine-learning model.
and artificial intelligence (AI) are relevant for
various branches of science and technology, REFERENCES
including manned space programs. Some [1] Greaves, A., & Au, B. (2015). Using the
technology concepts and pilot systems based on bitcoin transaction graph to predict the price of
the AI (3-D computer vision, automated systems bitcoin. No Data.
for planning and evaluating the activities of [2] Hayes, A. S. (2017). Cryptocurrency value
cosmonauts, inquiry and communications formation: An empirical study leading to a cost
system) were developed in the industry over of production model for valuing bitcoin.
several decades . Telematics and Informatics, 34(7), 1308-1321.

VII. CONCLUSION [3] Shah, D., & Zhang, K. (2014, September).

Bayesian regression and Bitcoin. In
Cryptocurrency, such as Bitcoin, has established Communication, Control, and Computing
itself as the leading role of decentralisation.

www.jespublication.com Page 976

Journal of Engineering Sciences Vol 13 Issue 07,2022, ISSN:0377-9254

(Allerton), 2014 52nd Annual Allerton International Conference on. IEEE, 1988, pp.
Conference on (pp. 409-414). IEEE. 451–458 .

[4] Indra N I, Yassin I M, Zabidi A, Rizman Z I. [12] Cheung, Y. W., Chinn, M. D., & Pascual,
Non-linear autoregressive with exogenous input A. G. (2005). Empirical exchange rate models of
(mrx) bitcoin price prediction model using so- the nineties: Are any fit to survive?. Journal of
optimized parameters and moving average international money and finance, 24(7), 1150-
technical indicators. J. Fundam. Appl. Sci., 1175.
2017, 9(3S), 791-808`

[5] Adebiyi AA, Ayo C K, Adebiyi MO, Otokiti

SO. Stock price prediction using a neural
network with hybridized market indicators.
Journal of Emerging Trends in Computing and
Information Sciences, 2012, 3(1):1-9

[6] Adebiyi AA, Ayo C K, Adebiyi MO, Otokiti

SO. Stock price prediction using a neural
network with hybridized market indicators.
Journal of Emerging Trends in Computing and
Information Sciences, 2012, 3(1):1-9

[7] Ariyo AA, Adewumi AO, Ayo CK. Stock

price prediction using the ARIMA model. In
UKSim-AMSS 16th IEEE International
Conference on Computer Modelling and
Simulation (UKSim), 2014, pp. 106-112

[8] Ron, D., & Shamir, A. (2013, April).

Quantitative analysis of the full bitcoin
transaction graph. In International Conference
on Financial Cryptography and Data Security
(pp. 6-24). Springer, Berlin, Heidelberg.

[9] H. White, “Economic prediction using neural

networks: The case of ibm daily stock returns,”
in Neural Networks, 1988., IEEE International
Conference on. IEEE, 1988, pp. 451–458

[10] Kaastra and M. Boyd, “Designing a neural

network for forecasting financial and economic
time series,” Neurocomputing, vol. 10, no. 3, pp.
215–236, 1996.

[11] H. White, “Economic prediction using

neural networks: The case of ibm daily stock
returns,” in Neural Networks, 1988., IEEE

www.jespublication.com Page 977