univerge site banner
Original Article | Open Access | Int. J. Manag. Account. 2021; 3(4), 83-90. | doi: 10.34104/ijma.021.083090

Portfolio Construction Using Financial Ratio Indicators and Classification through Machine Learning

Hasan M Sami* Mail Img

Abstract

Financial Ratios have been a major indicator for financial asset selection. Its seen that the decision taken to construct a portfolio based on financial ratio indicators has been able to make better returns than the random asset allocation process in the portfolio. This research will show multiple classifications based on unsupervised machine learning processes to satisfactorily determine investable assets or securities for portfolio contribution. Our suggested portfolio would then be compared with a random portfolio for a specific time frame in order to determine portfolio return, Sharpe ratio, and portfolio performance. 

INTRODUCTION

Financial performance optimization has been a major concern for most investors and market analysts. Its seen that investment management process is highly dependent on investment policy, market timing and security selection (Brinson et al., 1986). For the pur-pose of solving the research perspective our concern of management would associate with security selection. Brinson et al. (1986) also mentions that an effective investment policy would be dependent on these three factors, but in process of understanding the quail-fication of well-judged financial ratio indicators for assets, our comparison with randomized port-folios would be compared with portfolios created by assets suggested by financial ratio indicators classified upon Machine Learning processes (Azam et al., 2020). 

Primary concern of portfolio construction is risk minimization through diversification of assets in a portfolio (Nanda et al., 2010). This research would in-volve itself with selection of assets based on sugges-tive financial ratio indicators. Through the suggestive assets our target would be construct a portfolio and check its return options for a given time period. Nanda et al. (2010) have found that, through K Means Clus-tering we have been able to identify more specific asset classes rather than SOM or Fuzzy C-Means Clus-tering process. This research would also compare the portfolio risk measurement with market benchmark index. 

Its seen that sustainable economic growth and development of social market plays an important role for financial markets and asset performance (Khan et al., 2020). Khan et al. (2020) suggested that through studying various key features such as skewness, kur-tosis, symmetry and fat tail risk portfolio optimization could be estimated through weighted allocation of assets.  In this research process, the key feature also in-cludes minimum investment contribution for sustain-able growth, but our comparison needs to be justified with smaller investment and educated asset choices rather than random asset choices and indefinite contri-bution for assets. Its observed that rather than insti-tutional investors in case of smaller investors, the con-sidered standard deviation does not go beyond a pres-cribed risk level (Baumann et al., 2012). 

As Machine Learning undertakes an approach to classify assets on the basis of defined educated criteria in reference to the considered financial ratios, its highly imperative that the classification methods need to be unbiased and solution centric based on situational factors (Li et al., 2014). As Barnes suggests that in order to understand the key situational constraints of a firm, all the key financial ratios needs to be under-stood (Barnes, 1987) hence asset allocation decision making needs to be dependent on financial ratios of assets. Its also seen that financial ratio indicators remain as key factors to specifically identify assets or securities to prohibit bad investments (Lin et al., 2011). Hence the primary research concern that gets developed through this process is: 

a) Are Machine Learning approaches useful for selection of assets based on classification through financial ratios? 

b) Can Portfolio Optimizations be achieved on the basis of evaluations of selected assets? 

Literature Review 

Its been observed that various financial ratios play different roles in order to exhibit financial conditions of an organization in different conditions. Chen & Shimerda (Chen et al., 1981) explains the importance of financial ratios as whole in order to determine overall condition of a company. Jordan et al, (2007) have evidenced that EPS is by far the most important financial ratio that determines the importance of selection of an asset for financial decision making. Through various research prospects its also seen that EPS plays a vital role for companys future prospect (Rashidi et al., 2007). Its observed that firms liqui-dity capability is a strong indicator in terms financial standing and further financial investments. In terms considering important constituents for portfolio cons-truction, it found that investment is highly based on ratio analysis processes (Brinson et al., 1986). Sustain-ability is a method of suggestion generated by portfolio construction and risk management process for the highest possible return (Khan et al., 2020). In cases of online portfolio formation Baumann et al. (2012); Li et al. (2014) as the investors are mostly making smaller investment hence the risk allocation is different in comparison to institutional investors.  

Current Ratio is considered as another important ratio that submits the liquidity of any organization. Its found that in operational level (industry level) a firm considers to measure importance in relation to profit (Eljeily, 2004) with cash conversion and availability. Its observed that although the relationship between investments in current assets and liability are a major indicator for firms financial standing (Qudah et al., 2015), its observed that a moderate policy of following a good current ratio always makes a favorable decision. 

As Nanda et al. (2010) have successfully suggested the importance of clustering as a financial benchmark for segregating different class of assets for performance analysis, Momeni et al. (2015) have proved how the decision of selecting a good class of assets can make ultimate benefits. Abhishek & Samidha, (2017) found beneficial result through clustering method for segre-gating stocks in terms of performance measures for Shanghai Stock Exchange. Hence, in our research process we are considering K Means Clustering an important and derivative technique for Stock selection and portfolio formation. 

Alkhatib et al. (2018) have found that in case of pre-diction-based systems, K Nearest Neighbor algorithm has the lowest error margin and most rational results with highest similarity with predicted results. Bauman & Miller, (1997) have suggested that continuously favorable EPS value plays an important role for a companys financial future. In order to make our res-earch process successful enough we would initiate the suggestion of best EPS and Current ratio value sug-gested by market index review for financial decision-making processes. 

As Martin Suggested that (Robert, 2021) through pro-gramming language-based portfolio optimizer PyPort-folioOpt uses all sectors of risk return parity meas-urements in a favorable account through mean vari-ance allocation, Black-Litterman allocation (Idjorek, 2007) and machine learning inspired Hierarchical Risk Parity Algorithm (Lopez, 2016), hence the results obtained through PyPortfolio optimization processes could be beneficial for research and portfolio perfor-mance evaluation purposes. 

In accordance to Saito in Investor Business (Saito, 2021), a company with more than 25% EPS value in comparison to previous years data is found to be in strong demand. Its also seen that the higher the EPS value the better is the companys indication of overall stock price increase is there in future (Lopez, 2016). Similarly, its seen that a company with having a better than one 1 current ratio is supposed to have more liquidities in comparison to companies which have less than 1 current ratio (Saito, 2021). In suggestion it found that any companies which have current ratio in bet-ween 1.2 to 2 perform right at its best in terms of liquidation (Saito, 2021).  

Finally in reference to portfolio return and risk mea-sure a company needs to evaluate. The overall measure of Sharpe ratio Bodnar et al. (2016); Geltner, (2003) and MFoxCNBC, (2020) has led benefits for overall portfolio risk and return measure in a profitable man-ner. Furthermore, its seen selection of investment in a specific time period is of paramount importance in order to understand portfolio quality (Zhou, 2000). 

Theoretical Framework 

Importance of Financial Ratios - Chen & Shimerda, (1981) have found that key financial ratios are the most important determinant of organizational health. It found that – 

a) They are used to evaluate key warning signs for any operating cycle.

b) They are highly industry specific indicators of organizational performance measurements. 

c) Its a measurement tool that compares one com-pany with the industry and other companies of industry. 

EPS - As Jordan et al. (2007) have suggested that EPS is one of the key indicators of an organization in terms of its performance measures. In accordance to CFI, how much each share is earns in respect of any situ-ational constraint is represented through EPS hence under general rule, higher value of EPS the better the company is performing and lowers EPS shows a lower performance.

Current Ratio - Eljelly suggested (Eljeily, 2004) that current ratio basically shows how much current liabi-lities a company own in comparison to how much cur-rent assets the company possess under any given period of time. In general, the higher the ratio of cur-rent assets divided by current liabilities the better the performance of the assets. 

Importance of Portfolio Optimization - Portfolio management is the art of selecting the right investment tools in the right proportion to generate optimum re-turns with a balance of risk from the investment made in accordance to Bauman et al. (1997). In order to successfully implement portfolio optimization through PyPortfolio Optimizer its seen that an optimum port-folio can reduce the diversifiable risk till zero (López, 2016).

As the Portfolio consists of Assets of various weights with each having different return capability, hence the overall portfolio will have the capability of return in comparison to defined weight allocations to different assets which will be assigned in terms of their expec-ted returns, the overall return of all these assets will be the complete return of the portfolio. 

Sharpe Ratio - Sharpe Ratio: The Sharpe Proportion may bea budgetary metric frequently utilized by specu-lators when surveying the execution of speculation ad-ministration items and experts. It comprises of taking the abundance return of the portfolio, relative to the risk-free rate, and separating it by the standard devi-ation of the portfolios abundance returns (Bodnar et al., 2016). A great Sharpe proportion will have values more noteworthy than or rise to 1.00 which is by and large considered “good”, as this would propose that the portfolio is advertising abundance returns relative to its instability (Dowd, 2000). Financial specialists will regularly compare (Dowd, 2000) the Sharpe Pro-portion of a portfolio relative to its peers. A portfolio with a Sharpe Proportion of 1.00 may well be consi-dered lacking on the off chance that the competitors in its peer gather have a normal Sharpe Proportion over 1.00 (Dowd, 2000). Geltner, (2003) moreover notices that in connection to assessing Sharpe proportion of a portfolio there ought to moreover be IRR related with portfolio execution degree. In its peer group have an average Sharpe Ratio above 1.00 MFoxCNBC, (2020); Geltner, (2003) also mentions that in relation to eva-luating Sharpe ratio of a portfolio there should also be IRR associated with portfolio performance measure.

Importance of Classifying Assets for Selection – Security selection is the process of identifying indi-vidual securities within a certain asset class that will make up the portfolio (Nanda et al., 2010; Momeni et al., 2010). The process of making effective selections is guided through the process of making effective selections for assets based on their financial perfor-mance measures (Khan et al., 2020; Qudah et al., 2000). Our theoretical frame work thus addresses to confirm following activities in order to address the solution process effectively.

Process 1: 

1) Random selection of assets from various online sources like -

a) www.fknol.com

b) www.fools.com

c) www.investors.com

d) www.yahoo.finance.com

2) Selecting a specific time period for those sele-cted assets. 

3) Collect the asset prices for those selected dates 

4) Apply those assets in a PyPortfolioOpt Opti-mizer and get a result in terms of Sharpe Ratio & Portfolio Return. 

Process 2: 

1) Collect EPS & Current Ratio for those ran-domly selected assets 

2) Apply K Means Clustering in order to segre-gate the selected assets into the desirable asset class 

Feature Settings: 

EPS: Greater Than 1

Current Ratio: Greater than 1 and less than 2

3) Applying the Selected asset classes through Py-PortfolioOpt for Portfolio Optimization process. 

4) Evaluating the result of Newly Selected Port-folio with Randomized Portfolio in-terms port-folio performance through Sharpe ratio and Port-folio return. 

METHODOLOGY

In order to understand the true potential of portfolio management, the selected time period for performance is considered as January & February 2020 (MFox-CNBC, 2020). In the process of application, K Means Clustering the primary task would be associated with feature selections by which an asset could be classified as good investment property from another asset which is not. Feature selection properties have been demon-strated throughout as 

EPS >>> As High as Possible

1.0 < Current Ratio <= 2.0

In the position of Given set of observations between xi and yi, where the observations are d-dimensional real vector, the target of K Means clustering is to segregate the n observations in K sets. Formally the main target is to generate: 

The minimum point of difference from the existing true reference point of i (1, 2, …., N) C where 1, 2, ……. , N are all the observations for variables. 

Here, Xi is representing EPS 

         Yi is representing Current Ratio

Hence K Means Clustering would truly find the actual difference of points with best performing EPS & Cur-rent Ratio values in order to segregate the best stocks from the worst stocks for portfolio selection.

Arg-Min (SQRT|| xi – Ux||^2 + || yi – Uy|| ^ 2)=

Minimum Point of Difference 

Ux are the average points for EPS 

Uy are the average points for Current Ratio

Process 2: Secondary Check: 

Using K Nearest Neighbor Clustering, this paper would initiate the task to initially select the effective EPS and Current ratio value and select the assets only with the minimum difference attribute. 

Hence, the best asset choice will be -

Min (ABS (K (Xi, Yi) – K (Mod X, Mod Y)))

ABS refers to absolute value

K (xi, yi) refers the respective weighted attribute of EPS and Current ratio of Assets

K (Mod X, Mod Y) determines the ultimate quality attribute of best EPS and Current Ratio. 

K refers to the values of variables 

Assets will be the list of all assets; those are randomly selected for portfolio with their quality attributes of EPS and Current ratio 

Hence the Algorithm would be as such 

Final Step: Apply the selected group of assets into the portfolio. After making the selection we would apply them using the portfolio return formula: 

Expected Return = WAi×Rai+…….+WAnXRAn

Where;

WAi = Weight of security Ai

RAi = Expected return of security Ai

WAn = Weight of security An

RAn = Expected return of security 

Here i refers to all the assets which are present in the portfolio from 1, 2, …. n

RESULTS AND FINDINGS

In order to make the portfolio optimization process effective, the Stock suggestion taken using theoretical framework (Nanda et al, 2010; Momeni et al., 2015) we have taken several stocks as suggestions without the boundary of their effective EPS and Current Ratio. 

Under the effective suggestion we have considered two months ratio price as the benchmarks for portfolio opti- mization test process. 

As PyPortfolioOpt mentions everything under the perspective of yearly gains hence the overall returns are forecasted as annualized returns. But this research paper has brought how pre-evaluated assets can make tremendous developments in Portfolio gains as the benchmark is yearly return. In accordance to the suggestions of various online medium its found that various assets like popular ones are only suggested as the best contributor of the portfolio which as a whole are good contributors for the portfolio but for short term gains various different assets should be suggested in accordance to their financial ratios.  

The assets which were selected for random methods are seen to have major problems with Current ratio and some have extremely high EPS while some have ex-tremely low EPS. The Overall return and risk variance for the randomized portfolio is found out through PyPortfolioOpt in order to understand the Portfolios return.

In the process of making selection of assets where the relevant features are only selected are demonstrated for the preferential portfolio. The selected assets are des-cribed with their specific EPS and Current Ratio here: 

Here the entire assets were selected using K-Nearest Neighbor algorithm and later on tested through K Means Clustering Method for selection of absolutely same as-sets for portfolio optimization. Using the PyPortfolio-Opt its seen that the above-mentioned assets make the overall return and risk variance:  

The portfolio optimization method also has suggested which two assets were the biggest contributor of financial portfolio. The discrete allocation process has made the total numbers of assets contribution process much more effective and efficient through outPyPortfolioOpt process. 

Its seen that the portfolio analysis process was highly capable of making accurate suggestions regarding the asset selection but failed to make exact predictions regarding number of asset allocations or weight distributions. 

Its found that the overall portfolio created through effective selection of assets are highly capable of per-forming and they are nearly 89% similar in reference to actual or best performing portfolio with an effective gain of 45% in an average with an effective contribu-tion margin of $1000 as the initial investment amount.

CONCLUSION

In our research process we have identified, how stocks in NYSE could be used effectively through classifi-cation-based Machine Learning processes. Also, its seen that return and risk measure calculations are al-ways much better under biased and favorable circum-stances rather than randomized circumstances. Its seen that effective selection of assets always bring better profits as random portfolio made a annualized return of 23% whereas assets with favorable financial ratios made 75% overall annualized return. The overall risk measure of the randomized portfolio was less than the return whereas for selected asset portfolio its seen to have a very low risk in comparison to its return. Finally, the annualized variance of high return portfolio is lower than randomized portfolio at 4% and 6% respectively. Our overall portfolio management process didnt make use of any initiation of time series forecasting method that can show if the selected assets would be able to make a positive and negative gain for the considered period. Although the assets were referred through good suggestion by classification process suggested through financial principles, but if the time series forecasting were not applied then all these assets could have a negative ROI which was not discussed in this paper.  This paper was highly focusing towards elaborating the benefit of portfolio construction process and classi-fication-based selection process; hence the major effectivity of time series forecasting was not mentio-ned in this paper. The secondary data sources were also not highly reliable as there were various other important financial ratios which were not found in the open data sources. 

ACKNOWLEDGEMENT

This research paper is a contribution effort in order to understand the benefit of financial ratios as a bench-mark for classifying important assets. Hence all the related researches are of paramount importance for this paper. Also, various theoretical aspects such as EPS ratio and Current Ratio and their suggestions as stan-dard point is referred as benefits understanding for asset selection for portfolio; so the theoretical aspects need to acknowledge as well for the research paper. 

CONFLICTS OF INTEREST

The author declared there is no potential conflict of interest to publish it.

Article References:

  1. Azam MS, Rahman A, and Ahmed MT. (2020). Prediction of liver diseases by using few machine learning based approaches, Aust. J. Eng. Innov. Technol., 2(5), 85-90. https://doi.org/10.34104/ajeit.020.085090 
  2. Barnes, (1987). The Analysis and Use of Finan- cial Ratios: A Review. Journal of Business Finance & Accounting, 14(4); 0306-686X. 
  3. Baumann, P., & Trautmann, N. (2012). Port-folio-optimization models for small investors. Mathematical Methods of Operations Research, 77(3), 345–356. https://doi.org/10.1007/s00186-012-0408-3  
  4. Bauman, W Scott; Miller, Robert E. (1997). Investor expectations and the performance of value stocks versus growth stocks. Journal of Portfolio Management; London, 23(3): 57-68. 
  5. Bodnar, T., & Zabolotskyy, T. (2016). How risky is the optimal portfolio which maximizes the Sharpe ratio? AStA Advances in Statistical Analysis, 101(1), 1–28. https://doi.org/10.1007/s10182-016-0270-3  
  6. Brinson, G. P., Hood, L. R., & Beebower, G. L. (1986). Determinants of Portfolio Perfor-mance. Financial Analysts Journal, 42(4), 39–44. https://doi.org/10.2469/faj.v42.n4.39   
  7. Chen, K. H., & Shimerda, T. A. (1981). An Empirical Analysis of Useful Financial Ratios. Financial Management, 10(1), 51. https://doi.org/10.2307/3665113   
  8. Dowd, K. (2000). Adjusting for risk: Inter-national Review of Economics & Finance, 9(3), 209–222. https://doi.org/10.1016/s1059-0560(00)00063-0   
  9. Eljelly, A. M. A. (2004). Liquidity ‐ profit-ability tradeoff: An empirical investigation in an emerging market. International Journal of Commerce and Management, 14(2), 48–61. https://doi.org/10.1108/10569210480000179    
  10. Fresh Books, (2020, July 20). What Is a Good Liquidity Ratio? Fresh Books. https://www.freshbooks.com/hub/accounting/good-liquidity-ratio#:%7E:text=Current%20Ratio,-The%20current%20ratio&text=The%20current%20liabilities%20refer%20to,liabilities%20to%20covers%20its%20debts 
  11. Geltner, D. (2003). IRR-Based Property-Level Performance Attribution. The Journal of Port-folio Management, 29(5), 138–151. https://doi.org/10.3905/jpm.2003.319915 
  12. Idzorek, T. (2007). A step-by-step guide to the Black-Litterman model. Forecasting Expected Returns in the Financial Markets, 17–38. https://doi.org/10.1016/b978-075068321-0.50003-0 
  13. Jordan, C. E., Clark, S. J., & Smith, W. R. (2007). Should Earnings per Share (EPS) Be taught as a Means of Comparing Intercom-pany Performance? Journal of Education for Business, 82(6), 343–348. https://doi.org/10.3200/joeb.82.6.343-348   
  14. Khan, K. I., Naqvi, S. M. W. A., Ghafoor, M. M., & Akash, R. S. I. (2020). Sustainable Port-folio Optimization with Higher-Order Moments of Risk. Sustainability, 12(5), 2006. https://doi.org/10.3390/su12052006   
  15. Li, B., & Hoi, S. C. H. (2014). Online port-folio selection. ACM Computing Surveys, 46(3), 1–36. https://doi.org/10.1145/2512962 
  16. Lin, F., Liang, D., & Chen, E. (2011). Finan-cial ratio selection for business crisis pre-diction. Expert Systems with Applications, 38(12), 15094–15102. https://doi.org/10.1016/j.eswa.2011.05.035   
  17. López de Prado, M. (2016). Building Diversi-fied Portfolios that Outperform out of Sample. The J. of Portfolio Management, 42(4), 59–69. https://doi.org/10.3905/jpm.2016.42.4.059   
  18. Mamun AMA, Islam T, Siam MMS, and Kabir ME. (2020). Development of smart librarian with the virtual assistant (PRIMO), Aust. J. Eng. Innov. Technol., 2(4), 54-65. https://doi.org/10.34104/ajeit.020.054065 
  19. MFoxCNBC, (2020). Coronavirus crisis is caus-ing financial stress for nearly 9 in 10 Ameri-cans, CNBC. Available online at -https://www.cnbc.com/2020/04/16/coronavirus-crisis-causing-financial-stress-for-near-9-in-10-americans.html 
  20. Momeni M, Mohseni M, Soofi M. (2015). Clustering Stock Market Companies via K- Means Algorithm Kuwait Chapter of Arabian, Journal of Business and Management Review, 4(5); January. 2015.
  21. Nanda, S. R., Mahanty, B., & Tiwari, M. K. (2010). Clustering Indian stock market data for portfolio management. Expert Systems with Applications, 37(12), 8793–8798. https://doi.org/10.1016/j.eswa.2010.06.026   
  22. Qudah A.A.A & Al-Afeel M.A.M. (2015). The Relationship between the Investment in Current Assets and Profitability & Liquidity, Journal of Finance and Investment Analysis, 4(4), 11-22.
  23. Rashidi, P., and Analoui, M. (2007). Modified k-means algorithm for clustering stock market companies, 1st Iran Data Mining Conference 201-21, Tehran: Amir Kabir University
  24. Robert A. M. (2021), PyPortfolioOpt: portfolio optimization in Python, The Journal of Open Source Software, Available online at - https://doi.org/10.21105/joss.03066  
  25. Saito-Chung, D. (2021). Why The EPS Rating Is One Key To Picking The Best Stocks. Investors Business Daily, online at -https://www.investors.com/how-to-invest/investors-corner/eps-rating-is-key-to-picking-great-stocks/   
  26. Zhou, X. Y. (2000). Continuous-Time Mean-Variance Portfolio Selection: A Stochastic LQ Framework. Applied Mathematics and Optimi-zation, 42(1), 19–33. https://doi.org/10.1007/s002450010003 

Article Info:

Academic Editor

Dr. Doaa Wafik Nada, Associate Professor, School of Business and Economics, Badr University in Cairo (BUC), Cairo, Egypt.

Received

May 27, 2021

Accepted

July 1, 2021

Published

July 8, 2021

Article DOI: 10.34104/ijma.021.083090

Corresponding author

Hasan M Sami*

School of Business, Canadian University Bangladesh, Dhaka, Bangladesh. 

Cite this article

Sami HM. (2021). Portfolio construction using financial ratio indicators and classification through machine learning, Int. J. Manag. Account. 3(4), 83-90. https://doi.org/10.34104/ijma.021.083090

Views
281
Download
655
Citations
Badge Img
Share