Content uploaded by Cristian Mihail Păuna
Author content
All content in this area was uploaded by Cristian Mihail Păuna on Jul 25, 2018
Content may be subject to copyright.
Content uploaded by Cristian Mihail Păuna
Author content
All content in this area was uploaded by Cristian Mihail Păuna on Jul 21, 2018
Content may be subject to copyright.
22 Automated Trading Software - Design and Integration in Business Intelligence Systems
Automated Trading Software - Design and Integration in Business
Intelligence Systems
Cristian PĂUNA
Economic Informatics Doctoral School - Academy of Economic Studies
cristian.pauna@ie.ase.ro
After the introduction of the electronic execution systems in all main stock exchanges in the
world, the role of the automated trading software in the business intelligence systems of any
financial or investment company became significant. Designing of reliable trading software to
build and send automated orders based on quantitative mathematical models applied in the
historical and real-time price data is a challenge for nowadays. Algorithmic trading and
high-frequency trading engines become today a relevant part of any trading system and their
specific characteristics related with the fast execution trading process and capital
management involves specific measures to be used. Smart integration of the trading software
in the business intelligence systems is also a sensitive theme for any financial and investment
activity, a plenty of functional, control and execution issues being subjects of researches for
future improvements. This paper wants to gather together more particular aspects on this
subject, based on the experience of last years, opening the way for future topics.
Keywords: automated trading software (ATS), business intelligence systems (BIS), business
process management (BPM), algorithmic trading (AT), high-frequency trading (HFT).
Introduction
Trading shares in an organized stock
market is already a stand-alone activity
for hundreds of years. Since the transfer of
an equity from a seller to a buyer for the
money is a sustained process, it underwent
a whole series of significant
transformations.
After some organized markets appeared
between the 15th and 16th century in
Belgium and Netherlands, "the United East
India Company was chartered in 1602" in
Amsterdam and "a new bourse was begun
in 1608" [1]. It was founded the first
company with shares that can be bought
and sold by the shareholders and this new
trading shares activity does not affect the
main activity of the company. "Altogether,
1,143 investors subscribed to the initial
capital of the of the Company's Amsterdam
chamber" [2]. "The Dutch pioneering
several financial innovations that helped
lay the foundations of modern financial
system" [3]. It was only a step to the
continuous trading of shares in the
incipient stock market. A new era has
begun, more stock exchanges was
developed worldwide.
It is well known the image of a trading
floor in a stock exchange, where the
communications where made in open
outcry language until recently. Even "until
2009 trades on the floor of the New York
Stock Exchange always involved a face-to-
face interaction" [4]. In this system all bids
and offers were made out of the open
market, giving all participants the chance
to compete with the best price. All
estimations were based on the last closed
price, on the whole price history using the
latest fundamental news, but no estimation
was possible to be made based on the real-
time price movement as we are used today.
In the last decades, introducing the
electronic systems in the stock exchange
activity was a significant step. Electronic
trading (ET) is the method that use
information technology to bring together
buyers and sellers in a virtual market place
using an electronic trading platform and a
network that links all participants. The
huge difference is now that the real-time
price is available, which makes possible a
more accurate prediction. The second
important difference is that all orders are
executed electronically, in a very fast
1
Database Systems Journal, vol. IX/2018 23
execution process, fact that affect the
volatility and the speed of the price
movement.
ET was born practically in 1971 when
National Association of Securities Dealers
Automated Quotations (NASDAQ) was
founded. "NASDAQ was the world's first
electronic stock market" [5]. After that, in
1987 Globex trading system was founded
by CME Group, "conceived in 1987 and
launched fully in 1992" [5]. It allowed
access for ET of treasuries bonds,
commodities and foreign exchange. A little
bit later, the rival ET system CBOT
(Chicago Board of Trade) was
implemented, "an electronic trading
platform that allowed for trading to take
place alongside that took place in the
CBOT pits" [5].
The fast IT development was a key factor
in the ET domain. In the years of 2010 and
2011 massive investments were made in
technology in all main stock exchanges in
the world. This process determined the
majority or the classical trading floors to
be changed in ET systems, the classical
brokers were removed from the trading
chain and all trading orders were made and
processed electronically. The impact was
tremendous in reducing the costs of
transactions, gather the liquidity and
increasing the transparency of the price
movement.
In this new context, the activity of any
financial or investment company is driven
by a BIS which is developed and optimized
individually and related with the objectives
of each institution. ET allowed a BPM to
be automatized. In this process, using an
ATS has become today a necessity for any
trading and investment systems, to permit a
quick adaptability to the market conditions
that are changing practically every second.
Using technology to analyze real-time
price has allowed the development of a
new type of trading based on mathematical
and statistical algorithms: algorithmic
trading (AT). Computer software generate
automatically buy and sell orders without
any human intervention to trade individual
financial instruments. High-frequency
trading (HFT) term "has gained some
significant attention due to the flash crash
in the U.S. on May 6, 2010" [6]. It is a
specialized form of AT characterized by
high turnover and especially by high order-
to-trade involved ratios, generating a huge
number of trades. We will see below the
place occupied by AT and HTF through
the ATS in the BIS of a financial or
investment company.
2 Business Intelligence Systems using
Automated Trading Software
It is well known that BIS comprises the
strategies and informational technologies
for data analysis. Once applied IT in the
trading activity, the BIS that includes ET
turned into a special architecture, as a
result of the main role played by the ATS.
The real-time data analysis for prediction
and risk management in the ET systems
place ATS to be the main engine in the BIS
of a financial or investment company.
Because the profit is obtained directly from
the actions of the ATS, the real-time data
mining processes in ET have become the
priority research subjects in the latest
years. The low-latency data flux together
with the fast speed data mining processes
allows a very fast build for the buy and sell
orders. ATS is linked directly with the
liquidity and capital infrastructure to obtain
the profit.
Even the real-time data are continuously
stored in the data warehouse of the BIS
together with the historical data, the real-
time flux is used in parallel by the ATS to
ensure the rapidity of the trading decisions
and actions. After the data mining process
and order execution, all resulting data are
reversed in the data flow to be stored in the
data warehouse. After that the BIS is
aligned again with the usual configuration,
all BIS results being obtained with the data
already stored. However some reports are
automatically generated by the ATS.
The particularity of the BIS with ATS is
the special management of the real-time
data flux to ensure a low-latency
24 Automated Trading Software - Design and Integration in Business Intelligence Systems
functionality. The real-time data flux
includes the price time series provided by
data sources linked directly with the stock
exchanges, real-time fundamental data and
news provided by specialized external
sources and the real-time capital and
liquidity data provided by external sources
linked directly with the bank and
brokerage system.
The key of success using BIS with ATS in
a financial or investment company, is how
the real-time data flux is organized and
used, how the real-time data mining
process is made together with the risk and
capital management and how fast can be
sent and executed the trading orders in the
brokerage system. In Figure 1. is presented
the data floe of BIS with ET:
Fig. 1. Data flow of the Business Intelligence System using Automated Trading Software
It is well known today that "researchers
believe the Cloud is a big part of the future
of business intelligence" [7]. In the new
time of the analytic data management of
BIS, to solve the most problems regarding
the complexity, costs and inflexibility, the
cloud computing is a solution for majority
systems. It is found that due to the
specificity of information managed in BIS
with ATS, there is a significant latency in
the process of adoption of the cloud
technologies in the companies using ET.
There is no documented argumentation for
this assertion except the personal
experience in this field in the last ten years.
Due to the specificity of the processed
Database Systems Journal, vol. IX/2018 25
data, due to the fact that the data mining
procedures and the obtained results are
strictly confidential and because of the
speed and low-latency requested in the BIS
with ATS, the cloud technology adoption
is systematic delayed in many BIS in the
financial and investment field. The
classical "in house secured" BIS solution is
still preferred.
2.1 Algorithmic trading
Algorithmic trading (AT) term refers to
"any form of trading using sophisticated
algorithms (programmed systems) to
automate all or some part of the trade" [8].
Triggering from the notion of algorithm, as
to be a finite set of precise instructions
performed in a prescribed sequence to
achieve a goal, especially a mathematical
rule or procedure used to compute a
desired result, the algorithmic trading
activity ca be defined as to be the process
of using computers programmed to follow
a defined set of instructions for placing
trades in order to generate profit at a speed
and frequency that is impossible for a
human trader. The main characteristics of
the AT is the complexity and large
computing volume ensuring in the same
time a very high speed for the process of
data mining, price prediction and building
and sending the trading orders, which is
also the intended purpose. Considering the
different aspects of the processes involved
in algorithmic trading, we can have some
several categories of systems in this field:
- depending on the level of automation:
semiautomated and automated trading
systems;
- depending on the algorithm type used:
statistical, neural or arbitrage trading
systems;
- depending on the frequency of orders:
long term, medium term and short term
systems.
The long term trading systems are usual
named as investments systems, being
characterized by a small number of trades
made to be profitable on long term (several
weeks, even months). The short term
trading systems are usual named and
assimilated as to be high-frequency trading
systems (HFT). These systems implied a
very high number of orders, rapid order
execution and cancelation, very short
holding periods (seconds, minutes, hours),
low-latency data and speed data-mining
required and a specific focus on a high
liquid equities [6].
2.2 Data mining in financial trading
In AT data is processed to find a profitable
solution to buy or to sell an equity in a
specified moment of time. The basic
principle is very simple. Computers do
everything using well-established
algorithms. However, developing a stable
and profitable system is a very laborious
job, paying a very particular attention to
the used algorithms.
Since statistics became widespread in the
context of computer emergence, we can
see several types of analytic tools:
"descriptive analytics focus on reports of
what has happened", "predictive analytics
extend statistical and/or artificial
intelligence to provide forecasting
capability", "diagnostic analytics can apply
analysis to sensor input to direct control
systems automatically" and "prescriptive
analytics applies quantitative models to
optimize systems, or at least to identify
improved systems" [9]. Considering the
specifics of each methodology, "data
mining includes descriptive and predictive
modelling", "operations research includes
all four" categories [9].
In financial trading we use descriptive
modelling to analyse the past evolution of
the price movement of an equity. This
process is sometimes called "historical
data-mining" (HDM) and the results help
to understand the general trend and
movement circumstances of a price. The
HDM process use all available historical
data in the warehouse (Figure 1.) and it is a
significant part of the data management
procedures in any trading system.
Predictive analysis which is involving
different forecasting models in practically
26 Automated Trading Software - Design and Integration in Business Intelligence Systems
present in majority fields of the human
activity. In financial trading the predictive
analytics is practically the main engine of
each trading system. Using the real-time
price data and the historical data, the
predictive models try to determine the
probability for a price increasing or
decreasing. Because this process is using
the real-time data flux, sometimes is called
as "real-time data mining" (RTDM). The
process is using the real time data without
any pre-cleaning or validation filter and
must provide a fast and reliable response to
the low-latency data management module
of the trading system (Figure 1.). In this
process the algorithms used are very
important but we know also that the "speed
is nature of the scripting or programming
language used" [10]. This means a very
significant role in the process of design and
development of a trading system the
optimization of all used resources.
In the RTDM are included different types
of mathematical algorithms for prediction.
Some of them are using typical statistical
functions together with different
mathematical functions and price
transformations. Classical algorithms
based on moving averages or regressions
are well known. However for a better
efficiency, very complex algorithms are
designed in this section and this is the key
of success in any trading or investment
company. Large amount of resources and
effort are spent here, a better algorithm
being what any investor want.
Other algorithms in RTDM are based on
neural networks methods. On this chapter
the algorithms try to use a continuous
optimization to find the best way to the
optimal solution. Usual this type of data-
mining algorithms are combined with the
statistical functions.
The third category in RTDM is the
arbitrage type. These methods are based
only on the real-time data price and are
used in all trading procedures regarding
equities with non-centralized price. Today
is well known the huge interest in the
arbitrage algorithms adapted for all
cryptocurrencies, big difference between
the sell and the buy price of a
cryptocurrency in two different exchanges
being a huge opportunity for profit.
Important investment are made latterly in
this field to develop algorithms and to reap
all these opportunities.
3. Automated trading software
An ATS is a software which is receiving
the real-time and historical price data of an
equity, generates the signals for buying and
selling of the equity based on well-
determined algorithms, sets the volume of
trading based on the capital liquidity and
the capital a defined risk level, builds the
trading orders and send them to the
brokerage account without any human
intervention.
The purposed objective of the ATS is to
generate profit into the investor's account.
Once established the algorithms and the
functional parameters, the ATS will run
continuous and automatically, being the
main part of any automated trading system.
The design of an ATS must take care about
some specific requirements: a). low-
latency data processing; b). fast trading
orders processing; c). profitable trading
strategies with measurable capital exposure
level; d). automated capital risk
management.
The low-latency data processing and fast
data orders processing are required to
ensure that the trading orders arrive in the
brokerage account system before the price
is changed. This is a defining requirement
for ATS because an order with an obsolete
price is ignored and will not be executed.
Consequently the processing speed is the
first design direction for an ATS.
The second design direction is the
measurable risk involved by the ATS. Not
all trading strategies are suitable, some of
the strategies can not measure priori the
risk involved. In an ATS can be included
any profitable strategy with a measurable
capital exposure level. With this parameter
the capital risk management will be
possible to be automatically made, in this
Database Systems Journal, vol. IX/2018 27
way the trading strategy will be integrated
with the risk management procedures.
Receiving the real-time and historical data,
ATS applies different data-mining methods
to extract information and to calculate
probabilities for an increasing or
decreasing of the price based on different
trading strategies. Depending on the time
frame used for each strategy, there are
several categories between high-frequency
trading and investment strategies: very-
short term, short term, medium term and
long term strategies.
Once a buy or a sell opportunity was
found, a trading signal is generated. The
trading signal is a record that includes the
action (buy/sell), the price and the code of
the equity. This signal can not be yet
executed because the volume of the
transaction is still missing.
In the capital and risk management
module, using the information about the
available capital liquidity (C), based on the
capital exposure level for each (i=1, 2, ...
n) used trading strategy (ξi) we can
calculate the volume of the current
transaction (Vi). The total capital exposure
level (R) of the ATS is the maximum limit
of the exposure level:
Vi = C * ξi with R ≥ ξ1 + ξ2 + ... + ξn (1)
In this way the total capital exposure of
any ATS can be controlled to the maximal
value (R). With an automated stop loss at
that maximal value, the risk can be limited
and controlled.
4. Conclusions
In the context of ET, the ATS occupies a
central place in the BIS of any financial or
investment company. ATS has become
today a necessity in this environment, to
permit a quick adaptability to the market
conditions that are changing very fast. The
computational and data processing speed is
the first design requirements of any ATS.
Based on more trading strategies, applied
on very-short, short, medium or long term,
the ATS require a real-time price and
capital liquidity data flux together with the
historical price data. The data mining
process, where the trading strategies are
included, is organized in two individual
modules: the real-time data mining module
and the historical data mining module,
depending on what data type are mined.
The results are processed by a low-latency
data management module which
incorporates the trading signals. Based on
the information provided by the capital and
risk management module, the trading
signals are transformed into trading orders
and are sent into the brokerage account
system. The speed for building and sending
orders to the capital account is very
important because the trading orders must
arrive in the brokerage system before the
price is changing, otherwise the obsolete
price orders will be ignored and not
executed.
The second design direction for an ATS is
the measurable risk involved by the trading
system. It is required by the automated
capital and risk management process. This
requirement is obtained by incorporating
into the ATS only trading strategies with
measurable capital exposure level.
Knowing the capital exposure involved by
each trading strategy, the total risk
exposure involved by the ATS can be in
real-time calculated, and therefore can be
limited as needed. Together with the real-
time capital liquidity informations it is
possible to build an automated capital
management process that will manage in
real time the capital and the risk involved.
Because the profit in any financial or
investment company is made directly by
the actions of the ATS, the design and
integration in BIS has gained a special
attention. Thinking that we have only
about ten-twelve years of ATS in the
context of a history of hundreds of years of
financial trading in the stock exchanges,
we can say that we are still on the
beginning in this domain, large research
and improvements are still to come in the
trading automation.
28 Automated Trading Software - Design and Integration in Business Intelligence Systems
References
[1] V. Barbour, Capitalism in Amsterdam
in the Seventeenth Century, University of
Michigan Press, 1963, ISBN 978-
0404613136, pp. 17.
[2] L. Petram, The World’s First Stock
Exchange, Columbia University Press
2014, ISBN 978-0231163781, pp. 9.
[3] Wikipedia - Electronic trading -
Internet:https://en.wikipedia.org/wiki/Stoc
k_market, January 29, 2018 [January 31,
2018].
[4] Wikipedia - Open outcry - Internet:
https://en.wikipedia.org/wiki/Open_outcry,
January 9, 2018 [January 28, 2018].
[5] Wikipedia - Electronic trading -
Internet:https://en.wikipedia.org/wiki/Elect
ronic_trading, December 1, 2017 [January
28, 2018].
[6] P. Gomber, B. Arndt, M. Lutat, T.
Uhle, High-Frequency Trading, Business
& Information Systems Engineering
Journal, April 2013, pp. 6-15.
[7] S. Ouf, M. Nasr, The Cloud
Computing: The Future of BI in the Cloud,
International Journal of Computer Theory
and Engineering, Vol. 3, No. 6, 2011, pp.
750.
[8] P. Treleaven, M. Galas, V. Lalchand,
Algorithmic Trading Review,
Communication of the ACM, 2013, vol 56,
no. 11, pp. 78.
[9] D.L. Olson, D. Wu, Predictive Data
Mining Models, Springer Science Business
Media Singapore, 2017, ISBN 978-981-10-
2543-3, pp. 5-6.
[10] J.O. Katz, D.L. McCormick, The
encyclopedia of trading strategies,
McGraw-Hill, New York, USA, 2000
ISBN 978-007-0580990-2, pp. 22.
Cristian PĂUNA graduated the Faculty of Cybernetics, Statistics and
Economic Informatics of the Academy of Economic Studies in 1999 and
also he is a graduate of the Faculty of Aircrafts of the Polytechnic
University of Bucharest in 1995. He got the title of Master of Science in
Special Aerospace Engineering in 1996. In the last decades he had a
sustained activity in software development industry, especially applied in
the financial trading domain. Based on several original mathematical
algorithms he is the author of more automated trading software. At
present he is the Principal Software Developer of Algo Trading Service
Ltd. and he is involved as PhD student in Economic Informatics Doctoral
School of the Academy of Economic Studies.