Machine Learning in Investment Funds Industry

Introduction

Technological advancement has continually changed the operations of the various industries in the market. Some of the areas that have progressively been impacted are such as fund investments industries, transport sector, manufacturing, among others. Due to the surge in demand for more goods and services, it’s becoming paramount to heighten the offers which are already presented in the market. Machine learning is defined by Nastase & Szpakowicz (2006), as the technique that gives the computer the ability to compute sophisticated statistical tasks, to improve the productivity. The definition is extrapolated by Hu & Liu (2004), who affirms that the machines are programmed with certain algorithms, which enables them to perform tasks beyond human capability. Investment funds industry has embraced the use of computers in lowering the operations costs, and other peripheral expenses that are associated with human errors.

Among the benefits that machine learning has brought to the economy is reduction of the errors in making judgments. Research conducted by Bjornsdotter (2010) illustrated that most of the well-performing companies in the world such as PWC, Deloitte, and others, are revolutionizing their operations, through replacing the routine jobs with robots. The move has enabled the said organizations to register positive growth for the past five years, hence sustaining the competitive edge (Hu & Liu, 2004). Industries have faced tough times due to the demise of the traditionally recognized methods of predicting the future performance. Some time back, the management relied wholly on the accounting records to make decisions (Zhi-Chen, 2013). However, the trend has shifted, due to the advent of new technology. The customer’s perspectives concerning a particular brand affect the placement of the product in the market. Therefore, there is need to leverage on big data, which is the major proposition of the machine learning. Zhi-Chen (2013) propounds that big data serves the purposes that physical human interaction cannot. In his argument, the feedbacks collected from the various clients have profound importance, in establishing weaknesses and strengths of the business. Thus, investment funds industry is one area that braced the changes in technology. According to Beygelzimer et al.(2009), the cyclicality of the economy can be easily interpreted through the use of automated tools. The method is devoid of human intervention, hence more accurate.

To achieve the major objective of this paper, the work will be organized from the larger theme, narrowing down to the finite ideas. The paper gives an introduction to the topic, to illustrate the development of the machine learning from the traditional point of view, to the current form. Again, the background information will allow the reader to acquaint with the problem statement being discussed. Moreover, identification of the gap follows, to elaborate on what has been done and not yet achieved by the previous authors. This is followed by theoretical foundation, and then a review of secondary sources. The research concludes by reaffirming the point of concern identified in the literature review.

Progressive review of the past literature and case studies were consulted, to get the stands of the various researchers. According to Zhi-Chem (2013) secondary sources allow access to information from different authors. It has an impact on improving the reliability of the results, hence reducing the bias. Furthermore, varied ideas improve the reliability. The approach that was adopted aimed to derive as much literature as possible, with minimal errors of retrieval. The papers were chosen depending on the year of publication, and much consideration was taken on the language used. In this case, English was considered as the formal dialect of conducting the research, thus peer-reviewed journals were selected based on these criteria. The following table shows the phrases that were used and the corresponding number of appearance in the search databases.

Search terms Results
Machine learning… 595
Importance of machine learning…. 80
Machine learning in funds industry…. 20
Machine learning in capital market… 8
Investment funds industry… 165
Factors affecting investment funds industry… 19
Machine learning algorithms… 220
Financial analytics… 169
Investment funds….. 342
Financial indicators … 155

Table 1: Search terms

Background

Disruptive technologies have emerged in the current economy more than before. Use of big data, artificial intelligence, virtual reality, among others, was not embraced in the past in some of the countries. Beygelzimer et al. (2009) notes that developing nations have lapsed in embracing the new technologies in the various sectors of production. That said, research conducted by  Rapach et al.(2018) affirm that well-developed countries integrated the modern technologies in their operations long before the third world states could do so. Due to this effect, machine learning is taken as a new thing in Africa, where the transition takes long (Beygelzimer et al., 2009). However, the major concern is the rate at which various industries can use the same, to lower the operating costs. While some areas such as banking, manufacturing, and others can leverage on this model to make more profits, management lacks the personal initiative to advocate for the same (Rapach et al., 2018).

Furthermore, the costs associated with the use of machine learning in the industries frustrate the efforts of the managers to introduce the new system (Chun & Tang, 2007).Modern technology comes along with peripheral expenses. Some of them which were mentioned are the cost of employing the analysts and installation of machines that can run sophisticated computations (Boselli et al., 2017). Although there is little use of machine learning in the investments funds industry, experts argue that this is the only way to shun the anomalies of the unpredictable economic cycles. Rapach et al. (2018) further adds to the sentiments by affirming that analysis of the elements that affect the fund’s industry can be accurately tracked through use of an algorithm.

In the same vein, Chun & Tang (2007) opines that financial indicators reflect the performance of the investment funds industry. In their argument, the world markets vary depending on the trading commodities. Several factors such as political instabilities, suctions, change of trading policies, among others have direct impacts on the performance of the industry. The nature of the economy has transformed from a narrowed coverage to inclusion of pertinent elements that were ignored before (Boselli et al., 2017). For example, countries are much inclined on the intangible characteristics such as the political will to engage in any form of agreement with another.  Also, companies are much focused on how they fair in the capital market. Lindgren (2004) affirms that both large and small corporations are not immune to disruptive technology. Their response dictates how well they list in the stock market.

Machine learning is perceived to impact the fund’s industry positively. Rapach et al., (2018) is of the idea that it reduces the human intervention that is linked to errors. However, the research conducted by Beygelzimer et al. (2018) contends that industries have not embraced this technology fully. To be specific, investment funds requires accurate predictions by analysis of data and deriving meaningful information from them. They have not integrated the IT infrastructure to assist in the process of assessing ups and downs of the market (Boselli et al., 2017).

Identification of the Gap

The past literatures have discussed the topic of machine learning in investments finance industry from different angles. Some authors propose that the integration of the technology has not been fully utilized, citing the cases of developing countries to have slow adoption (Grenton, 2001). Furthermore, Furnkranz (2001) opines that more focuses have been directed to artificial intelligence, while little efforts beings employed on the issue of big data. According to the definition placed across by Furnkranz & Flach (2005) the ability of the computer to analyze without human intervention is leveraged to improve the production. He cites its use in a survey of the feedbacks from the clients, and understanding of the seasonal changes of the market. The study conducted by Genton (2001) illustrated that top executive shy off from using machine learning due to the costs associated with it.

Application of machine learning has evolved from the past, due to the increased demand for more accurate information to make informed decisions (Genton, 2001). The most significant and modern method that has taken precedence in most of the organizations is the data mining. It’s defined by Furnkranz (2001) as the process of consolidating raw data and making meaningful information that can be utilized to project the future performance. About ten years ago, businesses and large corporations did not find the importance of utilizing the advantages of the huge data (Zhi-Chen, 2013). This was partially contributed by lack of stiff competitions from the market, and little knowledge concerning what algorithms could impact their businesses. In this regard, the stereotyping of complete reliance on marketing DNAs has completely changed (Lindgren, 2004). Instead of focusing on the aspects that would forge the business forward only, other playing factors such as inputs of the customers are taken into considerations (Rapach et al., 2018). Analysis of intangible elements can only be analyzed by machine learning. Some of the tools used to collect pertinent information are social media and the organization’s websites (Hu & Liu, 2004). Compounding the responses have seen some of the organizations grow from one level to another. However, the major challenge is the sustainability of the system. There exist a huge disconnect between what the management believes, and the actual output achieved by use of machine learning. Research conducted by Lindgren (2004) among 100 countries spread across twenty developing countries indicated that about 20%, slated use of big data in marketing and improving the performance. He notes that much effort is needed to increase the awareness and champion for the use of more recent technology in various sectors of production.

Genton (2001) notes that machine learning applications have been utilized in various areas. In his proposition, data mining is one of the most recent advancement. The results are supported by Boselli et al.(2017) who affirm that data mining is the process of extracting the useful information, which is used to make decisions regarding the direction a certain industry should take. He cites mining of cryptocurrencies, as one area that has emerged in the present time. Rapach et al. (2018) feel that the slow adoption of machine learning is contributed by lack of good will by respective countries. The political class restricts some new advancement as a way to protect their sovereignty. For example, mining of digital currency has faced impediments in some of the nations and has been declining since the year 2017(Rapach et al., 2018).

While Markov et al. (2007) view that use of machine learning in investments fund industry has not achieved full appreciation, Nastase & Szpakowicz (2006) opine on the need for personal initiatives to improve the technology. In their argument, it’s the responsibility of the management to come up with a way of improving the efficiency and output. Genton (2001) classifies machine learning into three broad areas: one is supervised learning. It is defined as the process of generating an output based on the solutions of the previous problems. The computer stores the outputs of the previous computations, thus makes it easy to solve a similar problem when it appears.  This classification is also recognized as known labels (Furnkranz, 2001). Additionally, the second one is unsupervised, which compels the machine to compute for new answers every time a new challenge appears (Markov et al., 2007). Alongside, reinforcement discovers new ways that suit the challenge at hand.

Understanding the performance of the capital markets has been a long-standing debate among the scholars. Some authors such as (Rapach et al., 2018) hold the assertion that fund size affects the direction in which the economy takes. Based on their argument, a huge pool of fund is a challenge to manage, if the portfolio managers lack information about the future performance. They affirm that such funds erode the actual growth, which should be experienced. Chun & Tang (2007) argue that irrespective of the amount of the funds available for investment, overhead and research costs remain stable. Thus, huge funds have no impact on the performance. In the same vein, Boselli et al. (2017) contends that proper analysis of the markets has to be conducted, to make right and accurate decisions.  Investment funds industry is one of the most volatile sectors and requires informed plans on how to tap profit from the complicated external environment. In this regard, resources have to be directed towards upgrading the machine learning to foster growth in funds industry.

Theoretical Foundation/Conceptual Framework

Literature consolidates some theories and models that relate machine learning to the topic of investments and funds industry. Often quoted Rapach et al. (2018) describes probability approximately correct (PMC) model to have a huge contribution, like output that the machine learning achieves in a given period. Based on this model, two parameters are central in making the right judgment. The first one is accuracy, which is defined by Lindgren (2004) as the degree of approximation from the optimal value being studied. In this case, if an individual has a large set of data, the statistical analysis should not deviate much from the actual expected solution. The dispersion towards against or towards the correct figures should be minimal, to guarantee the dependability of the information. The second one is confidence parameters. The latter measures the likelihood of meeting the accuracy of all the requirements (Genton, 2001). While analyzing any form of data, it’s difficult to eliminate the detailed factors, since they have a direct implication on the decisions that analysts make regarding the future performance of the market. In the investments and funds industry, understanding the indicators from different perspectives places the investor at an upper hand to control the anomalies.

Information theory was proposed by Claude, in the year 1948(Markov et al., 2007). Based on his proposition, this theory studies the transmission and extraction of information over a noisy channel.  The research conducted by Claude was aimed to investigate the obstructions that cause errors in the final output that the receiver on the other end gets. The results confirm that the number of disruptions has fundamental impacts on the processing of data. To achieve the objective, this theorist describes entropy as the level of uncertainty involved in various variables in the communication and analysis process. Abstractly, the aim of this theory justifies that the receiver has to reconstruct the final information, with a low probability of error. The idea is supported by Lindgren (2004), who stresses the need of having a less noisy study of the market. In his argument, machine learning can filter all unnecessary data, and focus on what is important. This is more helpful to investors in the fund’s industry. They can make accurate decisions on when to buy shares with prospects of making profits. Additionally, the portfolio managers can direct the pool of funds to the right channels, hence mitigating the challenges of decreasing the dividends.

Moreover, portfolio analysis theory (Hu & Liu, 2004) aims to analyze the individual securities; in order to choose the best personality to suit the role. The task is performed by accurately performing the scrutiny of a broad array of the investors, and their financial contribution to the organization. The prediction of the securities allows choosing the person who could be the best performing, in regards to the securities. Elements such as risks are considered, to make the right choice of the suitable candidate (Furnkranz, 2001). Some of the roles of the portfolio analyst are: translating technical predictions, to a more light and understandable manner. It allows comparison of the performance of various individuals and the profitability of their securities. Second, he or she selects the most efficient individual. The listed roles can only be achieved with accurate information. According to Genton(2001), machine learning provides the accuracy and derives helpful indicators that direct the analyst on the different movements of the economic phases. Having the right management team improves the acquisition of machine learning leading to resourceful market knowledge.

Review of Literature

This section will discuss extant case studies and researches conducted by other authors based on the topic under discussion. The purpose of selecting the past literatures is to give the current research more evidence, hence increasing the reliability. Again, the wide coverage elaborates more on the theme, thus avoiding the errors. Rapach et al. (2018) affirms that comparing the papers and critiquing the opinions of the authors enables the reader, to identify the gap. This is important in proposing the necessary recommendations and possible solutions to the problem statement. To have a deep understanding of the topic, it will be split into two major sections: Machine learning and investment funds industry. Furthermore, the sub-topics will be segmented into smaller sections.

Machine learning

Machine learning is a term that refers to the detecting of patterns in a given set of huge data (Lindgren, 2004). Most of the activities human beings engage in are surrounded by machine learning. Some of the areas where it’s heavily used are in digital cameras for face recognition. According to Chan & Tang (2007) this is a trend that is gaining momentum in the current era. Instead of using passwords to access personal information, prompting the machine to register the facial appearance beefs up the security. Again, the same technology is employed in antispam software. It’s used to filter the emails to avoid retaining nonessential ones. Most importantly, the use of this system in enhancing the security of credit cards has received appreciation from different nations. Genton (2001) propounds that the wake of digital insecurity lowers the prospects of digital transactions. With machine learning, the financial sectors have stepped up the security of the cards to detect any form of fraud. Due to the complexity of the tasks that need to be extracted and analyzed, the human programmer cannot be able to provide specific and accurate solutions (Furnkranz & Flach, 2005). The following paragraphs will discuss the various factors that influence machine learning.

Machine learning algorithms

This is an application that allows the system to learn and improve the outputs based on the previous experiences (Hu & Liu, 2004). For example, if a computer solves complex problems on data analysis, the procedure that was adopted is stored for future use. In case a similar problem appears again, the memory of the machine searches for all possible solutions, without approaching it from initial stages. Bjornsdotter (2010) affirms that learning involves identification of statistical patterns that guide in solving a particular challenge. In the same vein, he notes that algorithms barely resemble how human beings approach a task. They explicitly give more insight on how to arrive at the answers to the most sophisticated task.

One type of algorithm is supervised learning. According to Genton (2001), it’s defined by feeding the machine with a classification system that is already created by programmers. The algorithm generates the functions which later deconstruct the major task, into smaller problems that can be easily interpreted(Markov et al.,2007). In this case, human intervention is required, to give the initial data. Later, the machine is tasked with the role of interpreting the metadata at different steps, to have n output that is accurate and more reliable. Genton (2001) is of the idea that when the input is incomplete, the machine can infer to the past recordings, to search for the appropriate information to fill out the space. Furnkranz & Flach (2005) contends that one of the striking characteristics of supervised learning is the classification of the processes. In his assertion, Complex data is split into smaller elements, whose outputs are consolidated in the final solution. The figure below shows the different phases of classification (Lindgren, 2004).

Figure 1: Stages of classifications (Lindgren, 2004)

The second type of algorithm is unsupervised. This is the opposite of the supervised, in the sense that the computer is required to learn how to execute a certain task (Markov et al., 2007). In this case, no classifications are fed in the input. According to Markov et al.(2007) this type of algorithm is useful in incidences where the output is already known, hence the procedure to achieve the final results lacks. Genton (2001) is of the ideas that this technology is used in data clustering, which is essential in the investment fund industry. The money market is filled with a lot of uncertainties. If the analyst has the results of the present performance of the economy, unsupervised learning can be employed to find out the variables that would influence the future performance (Chun &Tang, 2007).

A study conducted by Lindgren (2004) among the top performing companies in the world indicated that the use of algorithms has continually gained momentum, in performing some of the routine jobs. Some of the works that were shown to have been highly replaced by machine are clerical jobs and data analysis. In his findings, he noted that organizations registered in the U.S embraced the new technology more than any other country. The responses from the top executives indicated that most of the businesses failed to integrate the algorithms due to the high initial and maintenance costs. In the same vein, Beygelzimer et al. (2009) opine that machine learning algorithms are faced with challenges of communication latency. Though Furnkranz (2001) stresses that the process enables interpretation of complex data, the accuracy is still not achieved fully. Boselli et al.(2017) note that if the system could solve the uncertainties in the world economy, then more stable trading could be witnessed, which is currently being swayed by unseen factors.

Financial analytics

The complexity of the economies presents the traders with the tasks of debunking the right method, which they can use to understand the stocks and performance of the forex trades. According to Baumgartner & Serpen (2010), observing the market requires an individual to have some background information on what to expect, to make correct judgments.  Financial analytics is defined by Kalousis et al. (2004) as the process of interpreting the raw data, that is collected from the market and making meaningful information from them. Baumgartner & Serpen (2010) is of the idea that the sophistication of the market fuels individuals and companies to engage modern technologies to assist in statistical analysis. The research conducted by Joseph (2015) among one hundred companies listed in Canada’s stock market illustrated that various top executives are shunning away from the involvement of human interpretations. The proposition placed across by the respondents showed a clear indication of the transition from traditional reliance on human experts, to embracing what machines can achieve.

Barga et al. (2015) opines that machine learning has continually revolutionized the way businesses operate as opposed to human beings. It has the capability of synthesizing huge data, which have similar characteristics to each other. In this regard, the computer can track all the activities that are performed in the owner’s account and make an adjustment based on the changes of the habit. It could be spending trait of the customers and also the company itself. Joseph (2015) contends that huge data is a lead in directing the management in identifying the weaknesses and strengths of the organization. According to his arguments, large corporations such as Amazon, Google among others, have been able to retain the competitive edge due to a greater understanding of the market. Machine learning can console all the finances; hence the managers make decisions from an informed point of view. The argument is supported by Zhang (2017), who is of the idea that most of the businesses fail due to inability to project their financial stand in the coming futures. Factors such as customers spending cannot be analyzed by using human efforts. An algorithm can easily compare large data, collected from different places all over the world (Kalousis et al., 2004).

Optimization of machine learning in the finance prediction is subtle in the present market (Joseph, 2015). Studies conducted by Wang et al.(2010)and Joseph(2015)illustrated that banking sectors could leverage this process, to backtest the loan portfolio. It is an important tool to determine the interest rates and the premiums that an individual is expected to pay after being awarded the loan. Barga et al. (2015) finds that logistic regression is utilized to derive the meaning of figures recorded in the books of the account. This allows mitigation of unforeseen risks, such as default from the loanees. Zhang (2017) also affirms that decision tree algorithm is useful when dealing with complicated data. According to his argument, the mechanism works by segmenting tasks into respective nodes, each possessing a particular value of information. Such, the output is more organized and happens within a short period.

Figure 2: Tree algorithm (Zhang, 2017)

Winning strategies in trade involve a proper understanding of the performances of the finance of the various companies (Zhang, 2017). Traders are keen on where they place their money, hence takes their time to make decisions on whether to buy or sell. Such a decision is not arrived at by plainly studying the candlesticks and direction of the market. According to Aggarwal (2017), making the entry and exit strategies in a given market is arrived at after invoking various tools of the trade. Some of those are moving averages, supports, and resistances, among others. All these are designed to assist the trader, to have a clear understanding of the market.

Risk management

In the previous five years, the amount of financial information has increased significantly (Aggarwal, 2017). The scenario is contributed by the increase in number of transactions that are performed on daily basis, by the funds industry.  In the wake of digitalization, the circulation of money has also changed (Hryshko &Downs,n.d). Online payments methods such as PayPal, Skrill among others, have taken precedence, increasing the chances of risks. Furthermore, the financial institutions are mandated by the regulatory bodies to retain compliance, while handling customers’ funds. The financial crisis of the year 2008-2009 necessitated various governments to increase the number of elements that every organization was expected to report on the books of accounts (Akansu et al., 2016). Based on the survey conducted by Zhang (2017), he noted that the corporation was supposed to report more than 2000 variables. Moreover, additional policies were introduced, which increased the stored data. Working on such information requires machines that guarantee accuracy.

One of the areas where the machine learning can be applied in risk management is the credit modeling. Akansu et al. (2016) applied neural networks, to consolidate and interpret the information from one hundred SMEs. In his finding, the output of the data relating to the financial performance reflected their actual position. The results indicated a strong relationship of accuracy and the method that was invoked to make judgments. Furthermore, Eilon (2010) used regression tree to find out the solvency of the companies. The author aimed to find out the ability of the organization to offer credit facilities over a given period. The output of the research confirmed that variables such as the current assets and liabilities had a strong influence on the capability of the institution to allow credits. Moreover, the algorithm could identify the strength of the customer’s loyalty, which is a strong factor in deciding the growth path of the business (Caley, 2000).

Eilon (2010) affirms that fraudulent activities are major risks for the fund industry. The increase of online payments and use of credit cards extrapolates the cyber-aggressors.  One area that has been heavily regulated due to cyber insecurity is the banks. The cards are designed with workflow engines that can detect any form of manipulation (Wang et al., 2010). The system works by previous experiences, to determine whether the payments activities executed by the traders are genuine. Traditionally, banks used the manual measures to track the cases of insecurity, which was inefficient. Again, the method was time-consuming and in most cases, the fraudsters were not apprehended(Joseph,2015). Machine learning has eased the way of doing things because a credit card can be used to trace all past transactions. A study conducted by Caley(2000) involving the top banks in the U.S and those in Africa contrasted their integration of machine learning in their operations. The result confirmed that the banks in the U.S embraced more of the technology compared to those in developing states. Based on the results, the banks located in Africa recorded high cases of fraud, with an estimated loss of close to $1billion through the cybercrimes (Aggarwal, 2017).

Far and above, risk management can be linked to the surveillance of the market. In trading of stocks, the investors are required to remain vigilant of their activities (Eilon, 2010). According to Caley (2000), machine learning can be used to tame rogue traders, who manipulate the information for their gain.  The system provides the behavior of every member, hence mitigating executing changes that might affect other trades in the same circle. However, the challenge of machine learning is lack of labeled data. As it was eluded earlier, this technology relies on previous data (Caley, 2000). Companies might fail to disclose financial information, as a way to shun the competitors from understanding their operations. Again, it’s difficult to explain correlations of huge data (McCarthy, 2014). Since the computer provides the output based on what was achieved in the prior problems, human beings are unable to explain the results. The disconnect can necessitate the management to commit more errors. There is need of human intervention in machine learning to improve efficiency. Although it’s considered accurate and more efficient, professionals are needed to add insight. The computer can ignore some pertinent information that is not recorded in the books of accounts. Further research is required to ascertain the best way to explain the output of machine learning.

Algorithmic trading

Trading is a risky business that involves understanding the real performance of the markets. In instances where the various trades are affected by factors such as political instability and other fundamental analysis elements, the possibility of the trader to make loses are too high(Karan & Kumar,2016). Eilon (2010) notes that the accuracy of any entry or exit of a particular market set is determined by correct analysis of prevailing economic environment. According to Caley (2000), the sophistication of the economy goes beyond human judgment. In some instances, looking through the economic graphs might fail to make actual meaning, when human intellect is used without any supporting material. In the same vein Joseph (2015) notes that some of the technical interpretations are made using machines. For example, the tools of trading such as fibonacci retracements, fractals, and others, are essentially utilized to enable the person to minimize the chances of making losses, thus improve the profit (Wang et al., 2010).

A study conducted by Zhang (2017) illustrated that machines use some prior information about the market, to make a proper and accurate prediction regarding the possible movements in the following days. Trading has been there for quite some time and is ideally one of the oldest forms of exchange. One of the elements that the machine uses to guide the trader is the news. For example, whenever a certain country registers political instability, a downtrend is experienced forcing the investors to pull out of the market significantly (Caley, 2000). Subsequently, anything that favors the markets results in an uptrend. The machine can interpret such information since the market repeats itself. It’s all based on some pre-installed information that is compared to estimate a near accuracy answer.

Barga et al. (2015) details some of the advantages of algorithmic trading in his research. According to the proposition placed across this paper, he notes that the method is more accurate, and precise as opposed to the human brain. This is supported by Joseph (2015) who noted that the algorithm searches through a huge database, and filtering unnecessary materials. The ability to utilize huge data increases the chances of accuracy.  Another advantage is that it is fast to use. Hryshko & Downs (n.d) propounds that machine gives clear indicators of the real-time performance of the market. Use of human intellects to interpret the graphical representation of different markets is encompassed by major errors. The consequence of this could result in making huge loses, which might be difficult to recover. Thus, risk management is the optimal benefit of algorithmic trading.

However, despite the massive growth of machine learning in the current times, Caley (2000) holds the idea that human intervention cannot be eliminated completely. The perfection of the computer-guided trading can mislead the investor in some instances. One of the problems that are explained in the study conducted by Eilon (2010) is the latency of the indicators. McCarthy (2014) interviewed four hundred avid traders on the issue of the use of machine learning in trading. 20% of the responses indicated that following the interpretations of the computer without invoking more theoretical knowledge increases the chances of loses. There is need to incorporate the aspect of human judgment, to ameliorate the latency of time when conducting the analysis.

Investment funds industry

Eilon (2010) defined investment funds as the supply of capital, which is owned by numerous people. The pool of money is consolidated on a common point and used to make investments. The investors benefit in the form of dividends. This strategy is more helpful compared to when an individual makes own investment plans, subject to errors associated with human weakness(Baumgartner & Serpen,2010). Under this classification, mutual funds and money markets are included. According to Zhang (2017) making correct decisions on the nature of the market is guided by how well investors can understand the prevailing environment. In this regard, machine learning has contributed enormously to the understanding of various economic players. 

Mutual funds

According to Wang et al. (2010) mutual funds are those that are contributed by the shareholders, and managed by a technical team to make profits. The portfolio of the manager selected to oversee the funds is essential in minimizing the levels of loses that investors can make. Currently, more than $1trillion dollars are invested in the mutual funds, with anticipation of making profits by diversifying the investment (Joseph, 2015). If for example the fees and expenses of managing the funds average at 1%, it would generate a loss of $10 billion in one year. Authors such as Eilon (2010) contend that the most probable cause of such an incident is lack of proper selection of stock. In his argument, the managers and investors fail to integrate machine interpretation, hence rely on human judgment. The minor errors performed by the investors aggregate to major risks. According to Caley (2000) the analysis should be asymmetrical to the elements that control the nature of the market.

Several academic studies affirm that mutual funds do not have a significant impact on stock-picking (McCarthy, 2014). The research concludes that the amount of money that is pooled together for investment cannot influence the nature of the market if the right portfolio is responsible for its management. The assertion is supported by Eilon (2010), who noted in his study that the characteristics of investors and the professionals managing the funds have a direct impact on performance. In another research conducted by Karan &Kumar (2016), he noted that managers who outperform the currents standards of the market heighten the productivity.

Joseph (2015) opines that one way to understand the mutual funds is through evaluating the actual return. Based on his assertion, the mutual funds are huge and sometime might impede the entrusted people to make the right decision while planning where to channel the money. In the same vein, Hryshko & Downs (n.d) extrapolates the idea by suggesting that designing benchmarks would assist in increasing the possibility of making correct judgments.  In his argument, managers should benchmark by setting hypothetical achievements. The only way to be sure of what to expect is by invoking the machine learning. The overall advantage of computerized projection is the ability to utilize an array of data, collected from previous similar incidences.

Financial Markets

Financial markets are huge and have continually been expanding. Due to the increase of demand and supply, trading in intangible asset has become central in the world’s economies. According to McCarthy (2014), some of the classifications of financial markets involve securities. Understanding the shifts in the markets involves the employment of high level of creativity and analytical skills. Research conducted by Aggarwal (2017) illustrated that most of the companies in the stock market operate by interpreting huge numbers. They register enormous funds that cannot make any meaning if the proper and correct analysis is not conducted. Similarly, financial institutions understand that instructing the computers to do certain tasks is not helpful compared to training them on how to write without human intervention. Along, Akansu et al. (2016) argue that the modern financial industries are moving away from the traditional paperwork. Instead, works that were initially done by people have been consolidated and are presently executed by a single machine. The strategy has increased the accuracy of studying the market, hence eliminating conflicting information.

Caley (2000) notes that one area that machine learning has brought revolution is on customer care services. Traditionally, customers could travel to the actual premise to get the assistance with minor issues. It was time-consuming, both to the client and the company. Instead, machine learning has eliminated this latency of communication, by providing all the service at one common point. The company’s website integrates the chat system that allows feedbacks from customers spread across the world. Furthermore, the computerized customer care directs the client to the right department; hence the problems are solved by the appropriate expert. One advantage of resorting to this kind of system is promptness of response. Barga et al. (2015) is of the idea that handling customers manually is time-consuming, and increases congestion at the place of work. For example, in financial markets, the companies involved in the business have a wide base of investors from different states. To serve them well, use of modern technology is the sure way to respond to their concerns (Zhang, 2017). Again, the managers can inform them of the performance of securities and possible periods on when to buy or sell. In return, the customer enjoys expeditiousness of the recent machine learning.

Akansu et al. (2016) argue that large financial institutions have replaced human beings with machines in handling the customers. In his assertions, the computer is trained to study the responses of the customers based on the previous solutions. Future queries related to the prior information are solved instantly, without the need of human intervention. Again, if the system is unable to solve the problem, the client is referred to an expert, who gives more details concerning the problem (Karan & Kumar, 2016).

Companies save a lot regarding operational cost because a huge number of customers are served per a given time. Similarly, the human labor is significantly reduced. However, Joseph (2015) is of the idea that machine learning in customer care does not fulfill the entirety of human curiosity. Some customers feel more comfortable to exchange conversation with people, as opposed to computers. In such a scenario, the financial institutions might lose potential investor, who would want to learn more about the culture and financial standing of the company.

Far and above, considerations of network security in financial institutions form the debate of the investors. According to Eilon (2010), any transaction that involves money is a risky business that can necessitate downfall. The advent of cyber attackers has increased the challenges of managing the huge data, thus placing the management of various organizations at the task of heightening the security. Caley (2000) notes that even the large corporations seem defenseless on this issue. Mammoths such as Oracle, LinkedIn have faced external penetration by aggressors.  Machine learning can be used to study suspicious trends in the market.  Such, organizations should rethink their approach on how to use machine learning, to avoid losing funds to the enemies.

 

 

 

 

 

Summary

Machine learning has found huge use in the investment funds industry, both in managing the huge data and deriving meaningful information from the same. The transition is slowly getting the form in developing nations such as Africa, while the developed states have fully incorporated machine learning in financial systems. Based on the literatures that were reviewed, it was observed that adopting new technology in conducting various activities in the funds industry has enormous advantages. Some of them are increased output, hence profits. Aggarwal (2017) noted that corporation that studies the patterns of the market benefit from making correct predictions. Also, the time latency is reduced. McCarthy (2014) stresses on the idea of bridging the gap of communication between the investors and the respective companies. With machine learning, it’s easy to communicate the same message to people located in different geographic regions promptly.

The reviewed papers identify that organizations have not embraced the idea of machine learning completely. This is due to the costs of installing the infrastructure and training the employees on data analytics. However, some of the large organizations leverage on the accuracy of the system. A gap exists in coupling the intelligence of the machine with human judgment. Although the computers can show accurate output, deriving the meaning of the information becomes challenging.  To achieve the objective of this research paper, secondary sources were widely consulted. Themes from varied authors were compared and critiqued, which justifies the methodology used.

Moreover, theoretical reviews identified probability approach model, as one element which affects the output of machine learning. The model stands on the ground that computers are able to give more accurate data, compared to human beings. This is more helpful in investment funds industry, where the figures of transactions are high. Furthermore, the paper discusses information theory that has illustrated the importance of extracting correct information in a given market. Far and above, portfolio analysis enables close study of individual securities, which helps in selecting the right candidate to oversee the fund investments.

The literature that was reviewed identified common themes regarding the topic under discussion. The paper indicated that some of the factors influencing machine learning are: machine learning algorithms, financial analytics, risk management and algorithmic trading. Furthermore, mutual funds and financial markets were discussed in details, to answer the research question. Conclusively, machine learning is beneficial to all organizations. Investment funds industry should increase efforts in integrating the system, so as to reduce the risks of fraud and making unprecedented loses. 

Reference

Aggarwal, S. (2017). Comparative Analysis of Hedge Funds in Financial Markets using Machine Learning Models. International Journal of Computer Applications163(3), 25-29.

Akansu, A. N., Kulkarni, S. R., & Malioutov, D. (2016). Overview. Financial Signal Processing and Machine Learning, 1-10.

Barga, R., Fontama, V., & Tok, W. H. (2015). Introduction to Statistical and Machine Learning Algorithms. Predictive Analytics with Microsoft Azure Machine Learning, 133-148.

Baumgartner, D., & Serpen, G. (2010). Fast Preliminary Evaluation of New Machine Learning Algorithms for Feasibility. 2010 Second International Conference on Machine Learning and Computing.

Beygelzimer, A., Dasgupta, S., & Langford, J. (2009). Importance weighted active learning. Proceedings of the 26th Annual International Conference on Machine Learning – ICML ’09.

Bjornsdotter, M. (2010). Machine Learning for Functional Brain Mapping. Application of Machine Learning.

Boselli, R., Cesarini, M., Mercorio, F., & Mezzanzanica, M. (2017). Using Machine Learning for Labour Market Intelligence. Machine Learning and Knowledge Discovery in Databases, 330-342.

Caley, J. (2000). A Survey of Systems for Predicting Stock Market Movements, Combining Market Indicators and Machine Learning Classifiers.

Chun, Q., & Tang, J. C. (2007). Foreign Direct Investment in Chinese Industries: A Genetic Algorithm Approach. 2007 International Conference on Machine Learning and Cybernetics.

Eilon, V. (2010). Learning about the learning-Brain: mutual learning of brain and Machine in adaptive BMI settings. Frontiers in Neuroscience4.

Furnkranz, J. (2001). Round Robin Rule Learning. In Proceedings of the 18th International Conference on Machine Learning (ICML-01), 146- 153.

Furnkranz, J., Flach, P. (2005), ROC ‘n’ Rule Learning—Towards a Better Understanding of Covering Algorithms, Machine Learning, Volume 58 (1), pp. 39 – 77.

Genton, M. (2001). Classes of Kernels for Machine Learning: A Statistics Perspective. Journal of Machine Learning Research 2: 299-312.

Hryshko, A., & Downs, T. (n.d.). Development of Machine Learning Software for High-Frequency Trading in Financial Markets. Business Applications and Computational Intelligence.

Hu, M., & Liu, B. (2004). Mining and summarizing customer reviews. In Proc 10th ACM SIGKDD International Conf on Knowledge Discovery and Data Mining KDD’04, 168–177.

Joseph, H. R. (2015). Poster: Software Development Risk Management: Using Machine Learning for Generating Risk Prompts. 2015 IEEE/ACM 37th IEEE International Conference on Software Engineering.

Karan, A., & Kumar, P. (2016). Predicting Bankruptcy Using Machine Learning Algorithms. International Journal on Cybernetics & Informatics5(1), 91-105.

Kalousis, A., Gama, J., & Hilario, M. (2004). On Data and Algorithms: Understanding Inductive Performance. Machine Learning54(3), 275-312.

Lindgren, T. (2004), Methods for Rule Conflict Resolution, Lecture Notes in Computer Science, Volume 3201, Pages 262 – 273.

Markov, M., Muchnik, I., Mottl, V., & Krasotkina, O. (2007). Machine-Learning for Dynamic Reverse Engineering of Hedge Funds. 2007 International Conference on Machine Learning and Cybernetics.

McCarthy, D. F. (2014). Hedge Funds versus Mutual Funds (2):An Examination of Multialternative Mutual Funds. The Journal of Alternative Investments.

Nastase, V., & Szpakowicz, S. (2006). Group Decision and Negotiations, Special Issue on Formal and Informal Information Exchange in E-negotiations, volume 15(2).

Rapach, D., Strauss, J., Tu, J., & Zhou, G. (2018). Dynamic Return Dependencies Across Industries: A Machine Learning Approach. SSRN Electronic Journal.

Wang, Y., Jia, J., & Qu, Y. (2010). The “Earth-Moon” model on software project risk management. 2010 International Conference on Machine Learning and Cybernetics.

Zhang, W. (2017). Machine Learning Approaches to Predicting Company Bankruptcy. Journal of Financial Risk Management06(04), 364-374.

Zhi-Chen T., & Zan Zhang. (2013). Data and fuzzy partition-based importance measure and its application in a comprehensive evaluation. 2013 International Conference on Machine Learning and Cybernetics.

Also look at:

Share this Post