begin title three inches from top of paper

49
Innovation Brief Portfolio Selection of the NASDAQ and NYSE Energy Sectors using Genetic Optimization By Evan, L, Hjelmstad University of Advancing Technology

Upload: butest

Post on 07-Nov-2014

323 views

Category:

Documents


1 download

DESCRIPTION

 

TRANSCRIPT

Page 1: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Innovation Brief

Portfolio Selection of the NASDAQ and NYSE Energy Sectors using Genetic

Optimization

By Evan, L, Hjelmstad

University of Advancing Technology

Page 2: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

ABSTRACT

With the currently unstable economy and the upcoming transfer of our dependency on oil

to renewable and cleaner energy, it is often hard to create a portfolio of energy stocks

with solid financials, especially if the investor is inexperienced. A program that utilizes

genetic algorithms to analyze and compare stocks in the NASDAQ and NYSE energy

sectors could possibly aid an investor in creating such a portfolio. It is the intent of this

project to create such a program that utilizes a genetic algorithm to quickly and

efficiently compile a portfolio that is the best possible for the energy sector. It will

determine this by taking into account the fundamental analysis and dividends of its

individual stocks and attempting to find a balance between them.

Page 3: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

TABLE OF CONTENTS

1. ABOUT THE INNOVATION………………………………………

Innovation

Today’s Situation about the Innovation

Innovation Timeline

Innovation Inquiry

2. REVIEW OF RELATED MATERIALS……………………………..

3. LEARNING PROCESS………………………………………………

4. RESULTS…………………………………………………………….

5. REFERENCES……………………………………………………….

6. APPENDIX A. Portfolio Analysis Results Excel Insert……………...

7. APPENDIX B. Adjusted Scoring Ranges…..………………………...

2

2

3

4

5

6

11

18

22

---

25

1

Page 4: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

ABOUT THE INNOVATION

Innovation

Due to the unstable US economy, reduction of dependence on foreign oil and

transferring that dependence to renewable energy, it is more difficult than ever for an

individual to make a balanced portfolio of energy stocks. A computer program that

utilizes genetic algorithms can help to aid an individual investor’s decision by giving

unbiased stock picks that are based on solid financials. The purpose of this project is to

see if the use of a genetic algorithm to determine and sort through the most financially

healthy energy stocks and then compiling these into a portfolio will be more profitable

than a randomly compiled energy portfolio. For the scope of this project it will only look

at the NASDAQ and NYSE, specifically their energy sectors. However, if using a

genetic algorithm can assist an individual on the NASDAQ and NYSE energy sectors it is

reasonable to assume that it would work for the energy sectors of other exchanges with

minimal changes.

This project will be done in several stages. The first stage will select the most

secure small, medium and large cap stocks off the NASDAQ and NYSE energy indexes

using the fundamental analysis methods explained later. The next stage will generate a

random population of portfolios using the previously selected small, medium and large

cap stocks as its population. The last stage will use a genetic algorithm which would

recombine and evolve the population of portfolios until it found the final portfolio that

best fit the criteria.

2

Page 5: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Today’s Situation about the Innovation

The NYSE was created in 1792 and remains one of the most respected stock

exchanges in the world. However, it also remains a purely broker to broker floor traded

system where individual investors cannot execute trades themselves. When the NASDAQ

was established in 1971 it was the first implementation of electronic stock trading and

ever since then it has been easy for individuals to manage their own investments. One

thing that individual investors learn early on is how easy it is to get emotionally involved

while trading stocks. Remaining emotionless during trades is a critical aspect of

investing. (Paulos, 2003). This is where automated investment and analysis programs

come into play.

Automated investment and analysis computer programs have existed since the

1980’s but until recently were only used by large banks and investment companies as

these were the only entities that could afford to have such programs made. (Khan &

Sharma, 2008). Recently however, the amount of individual investors has increased, thus

increasing the appeal for the mass market of investor aiding programs. These programs

analyze stocks from a purely statistical and analytical point of view, only reacting to

changes in data. They are not susceptible to outside influences such as emotions, rumors

or stress. According to The Korea Times, “A third of all EU and U.S. stock trades in

2006 were driven by such automatic programs, or algorithms, and the figure will reach 50

percent by 2010, according to Boston-based consulting firm Aite Group.” (Jin-seo, 2008,

koreatimes.co.kr). Taking this into account, it is easy to see how the market for programs

that aid investor decisions will only increase over the years.

3

Page 6: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Innovation Timeline

This project is broken into several key sections. The first is acquiring the skills necessary

to complete the project as well as doing the initial research and designs. The second is

creating the program by implementing the designs. Finally the third and last section is

testing the genetic portfolio against a multitude of random portfolios and then analyzing

the results.

The following is a week by week breakdown of what will be accomplished.

Weeks 1 – 12 Learn basics of C#, experiment with Genetic Algorithms

Weeks 12 – 15 Design graphical user interface

Weeks 15 – 18 Code program to gather needed information off of the web

Weeks 18 – 20 Code program to take user input

Weeks 20 – 26 Code Genetic Algorithm

Weeks 26 – 30 Test, debug and determine most efficient criteria for genetic

algorithm

Week 30 Run program to create genetic portfolio

Week 30 Compile random portfolios

Weeks 30 – 34 Track portfolios progress

Weeks 34 – 40 Analyze and document results

4

Page 7: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Innovation Inquiry

The purpose of this project is to see if the use of a genetic algorithm to determine and sort

through the most financially healthy energy stocks and then compiling these into a

portfolio will be more profitable than a randomly compiled energy portfolio. For the

scope of this project it will only look at the NASDAQ and NYSE, specifically their

energy sectors. At the conclusion of this project, through the analysis methods described

below, it will be determined if using a genetic algorithm is a viable solution to aiding

investors in compiling a balanced and solid portfolio of stocks. This will be determined

during the analysis stage of this project and will involve competing one genetic energy

portfolio that the program generates against 100 randomly compiled energy portfolios.

The analysis will take place over four stages and if the genetic portfolio outperforms over

half of the random portfolios in over half of the time periods it will be considered

successful. The time periods will be as follows, one week, two weeks, one month, and

one year (using 11 months of prior data if available).

5

Page 8: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

REVIEW OF RELATED MATERIAL

Introduction

The idea of being able to outsmart the stock market or create a strategy that

predicts which direction it will move has been the fantasy of professional and private

investors alike. This is shown through the creation of such fictional movies like Pi

(1998) and many books such as The Five Greatest Stock Traders of All Time that depict

stories of people that do manage to outsmart the market for at least short while and the

strategies they used. Regardless of the success that a strategy may meet, the drive for

more efficient strategies that use increasingly complex techniques will never stop.

Techniques to Analyze a Stock

In the beginning the techniques used to analyze a stock were simple. Many times

investors would look at the fundamentals of a company such as what industry they were

in, what direction they were taking, who ran the company and what their near term goals

were. Then the techniques grew more complicated as the stocks began to be analyzed

quantitatively and numerous ratios were used to judge the safety and value of a stock.

Collectively these techniques are known as fundamental analysis. After fundamental

analysis came technical analysis which involves using a series of techniques to analyze a

company’s stock chart. From this an investor can attempt to determine a variety of things

including which direction the stock will move. (Motley Fool Staff, 2008).

6

Page 9: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Genetic Algorithms

While the above methods help an investor analyze a stock, it does not provide any

conclusive evidence as to what separates a successful stock from an unsuccessful one.

With ever increasing amounts of processing power and more advanced algorithms being

produced and improved every day, the idea of being able rationalize the movement of

certain stocks to specific variables and causes is becoming more of a reality. One way

that this is being achieved is through the use of genetic algorithms.

Genetic Algorithms are a subset of evolutionary algorithms that are based on

inheritance, mutation, selection and crossover. They work by essentially sorting through

a population using a variety of parameters until they have a new population. Just as the

theory of evolution works in real life, in each generation the most fit genes are kept and

the rest are thrown out. (Shapcott, 1992). They then repeat this process narrowing down

the results more and more each time until they have the correct or most correct answer.

This can be helpful in creating an efficient and balanced portfolio by sorting through a

selection of stocks until it has the very best (defined by its parameters) of each industry or

sector. These can then be compiled into a balanced portfolio. Each stock is assigned a

score based on different criteria and then randomly assigned to a portfolio. This is then

repeated until a specified number of random portfolios are created. The genetic

algorithm would then proceed through a specified number of generations, keeping the

most fit portfolios and randomly mutating these with a selected list of stocks until a prime

portfolio was discovered. (Thatcher, 2004; Orito & Yamazaki, 2001).

7

Page 10: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Other Variations involving Genetic Algorithms

From the base idea of using genetic algorithms to create a portfolio of stocks,

many variations have been created. One idea consists of implementing a neural network

into the genetic optimization. Neural networks are closely related to genetic algorithms

in that they use a biological process to analyze and sort data to solve a specific problem.

Neural networks are structured similar to a human brain in that they are made up of

neurons which essentially implement the basic structure of how the human brain makes

connections. Thus a web of connections is created throughout a data source that can be

explored for a solution. By implementing this functionality into a genetic algorithm it

would allow the program to actually learn as it creates portfolios and theoretically

improve its methods each time it executes. This offers additional ideas for expanding the

project upon its completion. (Khan & Sharma, 2008).

Another variation improves upon the initial idea by using two separate genetic

algorithms in a two step process to create the portfolio. The first step consists of

selecting only the most fit stocks to begin with and compiling these into a list. This list is

then used in the creation of the random portfolios. This is more efficient as any

particularly weak stocks have already been eliminated. Essentially, the random portfolios

consist of only the best, resulting in the final portfolio theoretically containing the best of

the best. It is an adaptation of this method which I plan to employ to analyze the energy

sector of the NASDAQ and NYSE, but rather than using a two stage genetic algorithm, I

will be using a more direct analysis and sorting algorithm before I deploy the genetic

algorithm. (Keung, Yu, Wang, & Zhou, 2006).

8

Page 11: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

In addition to genetic algorithms being utilized to compile portfolio of stocks,

they have been used with much success in comparing and contrasting trading strategies to

determine the best combination of strategies and the optimum situations in which to

deploy them. This was accomplished by using the behavior of groups of stock traders

that was determined by variables such as amount of daily trades, the volume of these

trades, current price of the specific stocks and whether they took a profit or loss on these

trades and then using this information as the data set. The population for the algorithm

was compiled of objects that represented these traders via rule sets which live stock data

was fed into each day. The rule sets determined if that specific trader bought or sold a

specific stock and recorded if they made a profit. What was discovered was that the rule

sets that were output by this algorithm far exceeded any individual rule sets profitability.

While it is not the purpose of this project to create such an algorithm it does offer many

ideas for expanding this project upon its successful completion. (Eiben and Smith 2007).

Problems Faced

One problem faced with the ever increasing complexity of algorithms in the

investment arena is that as one algorithm reaches the efficiency of another algorithm the

competition between the two decreases their profits. Thus the developers for the

algorithms are stuck in a never ending battle for creativity, complexity and efficiency.

(Duhigg, 2006).

Another problem that is often faced is the dataset for the genetic algorithm is

incomplete. The algorithm relies entirely on the data for generating great results and

9

Page 12: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

without a complete dataset, which is made difficult by the massive amount of companies

that are listed on various exchanges, it cannot give optimum output.

Conclusion

Based on the research done on the application of genetic algorithms in regards to

stock portfolio selection, an opportunity arises to see if this method can be successfully

applied to a new and especially volatile sector such as that of the NASDAQ and NYSE

energy sectors. If this is proven to be true then more research is merited to find out in

what other volatile sectors genetic algorithms may prove useful and in what other ways

they can be applied to stock market analysis.

10

Page 13: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

LEARNING PROCESS

In order to prove that genetic optimization can indeed create a more profitable

portfolio of energy stocks than one that is randomly selected, the program must be

designed, created and tested. The program will be created in the C# programming

language and will contain the following characteristics.

Upon running the user will be required to input the number of stocks

desired in the portfolio.

The overall portfolio score will be out of 300 and will be comprised of

several individual scores including how close the portfolio is to the

dividend and fundamental analysis.

The user will not input a desired fundamental analysis score as it will

always be at a desired score of 100. The fundamental analysis score of an

entire portfolio will be the average of all the fundamental analysis scores

of its individual stocks.

The user will also not be required to input a dividend score as the best

possible balance between dividend and fundamental analysis will be

sought. The default dividend will be at $1.00 per share but the total

dividends can exceed this.

If an internet connection is available, it will navigate to NASDAQ.com and will

read information about all of the companies that are listed under the energy sector. This

information includes stocks on both the NASDAQ and NYSE. If an internet connection is

not available, it will use the most recent data contained in a database. From this it will

11

Page 14: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

assign points to each company based on the fundamental analysis of its stock and will

determine if the stock is a small, medium or large cap stock. For each of these categories

the most financially secure stocks will be selected based on their fundamental analysis

score. The collection of stocks that make up all three portfolios will make up a new

population. From this new population the genetic algorithm will randomly create a

population of portfolios which will be slightly and randomly adjusted (mutated) each

generation until a portfolio is created that meets all of the user’s criteria.

Fundamental analysis involves an analysis of a company’s fundamentals such as

its true value, assets owned, growth over the past quarters and debt status. The

fundamental analysis score will be comprised of ten sub scores which will add up to its

total score of 100. This score will then be doubled to be out of 200 to give fundamental

analysis double the weight of the dividend score which will be out of 100. Together, the

fundamental analysis and dividend scores will be out of 300 which will provide direct

comparisons between portfolios. The individual methods of analysis which are described

below were chosen as they represent some of the most common methods for determining

the health of a stock. If the stock receives great results across all of the categories it is a

huge indicator that this stock is likely to do better in the long term than other stocks with

less impressive results. Please note however that while the scales of each method are

described as being between 1 and 10 there are exceptions where they can receive a

negative score for abnormally bad results or extra points for abnormally great results in a

specific category, with the max score at either end being -5 and 15. The 1 through 10

score is used to total to 100 and most scores fall within this range. The exact process was

determined through a series of trials which can be found in Appendix B.

12

Page 15: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Trailing P/E – The trailing P/E of a stock will be scored on a scale of 1 – 10. The price to

earnings of a company is often considered the king of fundamental analysis measures.

Since this is the trailing price to earnings it compares the current price of a stock with its

past earnings. A P/E ratio gives the buyer an idea of how much earnings power they are

buying and is a standard way to compare two companies’ earnings that have different

stock prices. However, more than the P/E ratio of a company must be taken into account,

as in the past companies have been able to manipulate their earnings in such a way that

they appear much more solvent than they really are. Two historical examples of this are

Enron and WorldCom. More recent examples include Lehman Brothers and Merrill

Lynch. (Kelly, 2003).

P/B – The P/B ratio of a stock will be scored on a scale of 1 – 10. The price to book ratio

compares a stock’s price with the total value of the company. This helps an investor

judge if the stock they are buying is under or overvalued. If price to book ratio is one

then the stock is selling for exactly what it is worth. If it sells for more than it is worth

than its ratio will be above one. However, being above one is not always a concern.

Some companies have intrinsic value that is not reflected in their book value. (Kelly,

2003).

P/S - The P/S ratio of a stock will be scored on a scale of 1-10. The price to sales value

of a stock compares its stock price with its total sales for the last four quarters. It is

advantageous to include a price to sales ratio in your analysis as while companies can

13

Page 16: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

manipulate earnings to their liking, it is practically impossible to manipulate your sales.

(Kelly, 2003).

Current Ratio – The current ratio of a company will be scored on a scale of 1 – 10. The

current ratio gives you an idea of how solvent a company is. It is simply an assets

divided by liabilities ratio. Ideally you would like to see a company have at least a 2 to 1

current ratio. (Kelly, 2003).

Quick Ratio –The quick ratio of a company will be scored on a scale of 1 – 10. The

quick ratio is similar to the current ratio but gives a more accurate reading of how well a

company can deal with unseen expenses or opportunities as it is only the cash on hand

divided by the current liabilities. Ideally you would like to see a quick ratio of at least .5.

(Kelly, 2003).

Net Profit Margin – The net profit margin of a company will be scored on a scale of 1 –

10. The net profit margin of a company is found by dividing the money it has left after

paying expenses by the money it had before paying expenses. This gives you an idea of

how much money a company keeps from its revenue. This is also a great way to compare

companies within the same industry. If two companies are of similar size, create similar

products, have similar stock prices and other similar fundamental analysis scores it may

be difficult to tell which company is healthiest. However, this will be clear when you

look at the net profit margin as the company with the higher margin has found how to

14

Page 17: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

squeeze more profit out of their sales. This means if business becomes challenging they

will be more likely to survive and adapt. (Kelly, 2003).

Cash Flow per Share – The cash flow per share will be scored on a scale of 1 – 10. The

cash flow per share is simply the company’s total cash flow divided by the total number

of shares. A company’s cash flow tells you how well that company manages its money

and how efficiently it reinvests it. By dividing it by the total number of shares you can

see how much you must pay for a share of the company’s cash flow. (Kelly, 2003).

Beta – The beta will be scored on a scale of 1 – 10. The beta of a company compares

how volatile its stock is with the rest of the market. The NASDAQ is measured by the

NASDAQ 500 which is comprised of the five hundred most influential stocks on the

exchange. The NASDAQ 500 has a beta of one. Every single stocks beta in the

NASDAQ is in relationship to the NASDAQ 500’s beta. The NYSE is measured by the

S&P 500 which again is comprised of the five hundred most influential stocks on the

exchange. The S&P 500 also has a beta of one, thus no changes need to be made when

analyzing the beta of a stock on either exchange. If a stock has a beta higher than one it

can either mean that the stock has been more successful than the market, or it has not

been as successful. If its beta is less than one that could mean that the market has been

successful but the stock has generally stayed the same. It could also mean that the market

is doing worse and that stock is not going down but again staying the same. Either way a

good solid stock should have a beta slightly greater than one. (Kelly, 2003).

15

Page 18: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

ROE – The ROE of a stock will be scored on a scale of 1 – 10. The return on equity of a

company is what many people believe to be the greatest measure of a stock’s success.

This tells how the stock has done in the past by dividing the net income by the total

shareholders’ equity. This is essentially determining how much money a company has

made from the investments it made with shareholders money. (Kelly, 2003).

EPS – The EPS of a company will be scored on a scale of 1 – 10. The earnings per share

of a company is a standard way to measure the growth of a company. However, like with

the P/E ratio it can be easily manipulated. The earnings per share is simply the total

earnings for that quarter or year divided by the total amount of shares outstanding.

Ideally you want this number to be as large as possible. (Kelly, 2003).

A dividend is a way that a company shares profit with its shareholders and is a specific

amount that is paid each quarter for each share of stock an investor owns of that

company. Many people wish to incorporate dividend paying stocks into their portfolio as

it offers a guaranteed income even if the current markets are volatile. The target amount

is $1.00 of dividends per share. It is assumed that it is alright if the portfolio exceeds that

amount of dividends but it is not alright if it falls short and thus for every one cent that

the total dividend amount falls short of the requested amount, one point will be subtracted

from the total score of 100. The portfolio dividend score will be an average of all the

dividends of its stock resulting in the dividends per share of that portfolio.

16

Page 19: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Upon completion of the program running, the test phase will begin and the price of each

stock will be noted. For the sake of testing, an even amount of virtual money will be

invested in each stock. Then, another program will randomly create fifty portfolios of the

same number and ratio of small, medium and large cap stocks but drawing from the entire

energy sector rather than the genetically optimized one. In addition, five people will each

be asked to create ten random portfolios again based on the entire energy sector but again

keeping the same ratio of small, medium and large cap stocks in each one. This will

ensure that neither one method is biased in its randomization and will result in exactly

one hundred randomly created portfolios and one genetically created portfolio. Again,

the price of each stock will be noted and an even amount of virtual money invested. Due

to the constant fluctuation of the market, the portfolios will be compared and ranked after

one week, two weeks, one month, and one year (using prior data if available). If the

genetically created portfolio has outperformed a majority of the randomly created

portfolios in at least two of these time frames then the project will be deemed successful.

If not, an attempt will be made to isolate the variables that led to it being outperformed

for future correction.

17

Page 20: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

RESULTS

Thus far the project has turned out better than expected as it exceeded each of the

criteria’s for success by outperforming a majority of the random portfolios in three of the

time slots and tying in one. Development of the program was completed in late May but

many additional features have been added to it since. On June 8th, 2009, the analysis

stage began which involved competing the genetic portfolio the program created against

100 random portfolios. As stated earlier, the analysis was performed in four stages across

different time periods: one week, two weeks, one month and using 11 months of prior

data (if available), one year. The following is a breakdown of how the genetic portfolio

did during each stage and how it compared to the random portfolios. While the random

portfolios exact percentage gains and losses are too numerous to state in this report, you

can view all the data gathered during the analysis stage in Appendix A.

Analysis Period One (One Week)

2.35% overall genetic portfolio gain

50 random portfolios outperformed / 50 random portfolios outperformed by

Analysis Period Two (Two Weeks)

-5.55% overall genetic portfolio loss

93 random portfolios outperformed / 7 random portfolios outperformed by

Analysis Period Three (One Month)

-9.51% overall genetic portfolio loss

18

Page 21: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

74 random portfolios outperformed / 26 random portfolios outperformed by

Analysis Period Four (One Year, using 11 Months of Prior Data)

-43.85% overall genetic portfolio loss

71 random portfolios outperformed / 28 random portfolios outperformed by / 1 tied

Although at first glance it may seem as if the algorithm was unsuccessful as there was

only one time period when the portfolio was profitable, the goal of this project was to

create a program that would consistently outperform a random portfolio and thus aid an

investor in being profitable. Anyone who has invested in the stock market before

understands that you cannot always be profitable as there are times when all stocks go

down. During times like these it is your goal to lose as little money as possible which is

exactly what the genetic portfolio accomplished. Even while the genetic portfolio lost

value, when compared to the random portfolios it was doing extremely well. In fact, the

genetic portfolio only started excelling against the random portfolios when the energy

sectors of the NASDAQ and NYSE took a turn for the worse. This indicates that this

portfolio is not nearly as volatile and is much more financially stable than the random

portfolios. During the first analysis period there were exactly fifty portfolios that

outperformed the genetic portfolio. However, these portfolios quickly lost their value as

soon as bad news hit the sector, whereas the genetic portfolio was not hit nearly as hard.

During the next two analysis periods the genetic portfolio outperformed an incredible 93

and 74 random portfolios respectively. Even when using 11 months of prior data

combined with the one month of current data during the fourth analysis stage, the genetic

19

Page 22: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

portfolio outperformed an amazing 71 random portfolios. Although the portfolio posted

a loss during this time period of 43.85%, this is due to the horrendous market conditions

caused by the failures of many prominent banks and auto companies during the last

months of 2008. It is believed that it will be several years before the markets fully

recover from such a hard hit. The data stated above is more easily seen in the graph

below.

Thus this project can be deemed successful by the criteria stated earlier which stated that

the genetic portfolio must outperform at least half of the random portfolios in at least half

of the analysis periods. The genetic portfolio beat a majority of the portfolios in three of

the time periods and tied in one.

As it has been proven that the current fundamental analysis methods when combined with

a genetic algorithm can produce an above average portfolio in the energy sector, there is

still much to be explored. These same techniques can be applied to other sectors and

even exchanges as well. Furthermore, the fundamental analysis methods can be used

20

Page 23: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

separate from the genetic algorithm to rank and display the top stocks from each sector

and exchange, further helping an investor make educated decisions. Regardless of the

direction this project runs, the first successful stepping stone is in place.

21

Page 24: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

REFERENCES

Eiben, A. E., Smith, J. E. (2007). Introduction to Evolutionary Computing. New York:

Springer Berlin Heidelberg.

Jin-seo, C. (06/29/2008). The Korea Times: Algorithm Stock Trading Spreading.

Retrieved September 24th, 2008, from

http://www.koreatimes.co.kr/www/news/biz/2008/06/123_26677.html.

Kelly, J. (2003). The Neatest Little Guide to Stock Market Investing. New York: Plume.

Keung Lai, K., Yu, L., Wang, S., Zhou, Z. A Double Stage Genetic Optimization

Algorithm for Portfolio Selection. Retrieved from

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.87.9105.

Khan, A. & Bandopadhyaya, T. K., & Sharma, S. (2008). Comparisons of Stock Rates

Prediction Accuracy using Different Technical Indicators with Back propagation

Neural Networks and Genetic Algorithm Based Back propagation Neural

Networks. Proceedings of the First International Conference on Emerging Trends

in Engineering and Technology. Retrieved from

http://ieeexplore.ieee.org/servlet/opac?punumber=4579839.

Lin, L., & Cao, L., & Wang, J., & Zhang, C. (2000). The Applications of Genetic

Algorithms in Stock Market Data Mining. Retrieved from http://www-

staff.it.uts.edu.au/~lbcao/publication/DM2004.pdf.

22

Page 25: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Motley Fool Staff. (2008). Investing Strategies: Your First Stock. Retrieved October 11th,

2008 from http://www.fool.com/investing/beginning/investing-strategies-your-

first-stock.aspx.

Orito, Y. & Yamazaki, G. (2001). Index Fund Portfolio Selection Using GA.

Proceedings of the Fourth International Conference on Computational Intelligence

and Multimedia Applications. Retrieved from

http://ieeexplore.ieee.org/servlet/opac?punumber=7656.

Paulos, J. (2003). A Mathematician Plays the Stock Market. New York: Basic Books.

Shapcott, J. Index Tracking: Genetic Algorithms for Investment Portfolio Selection.

Retrieved from http://citeseerx.ist.psu.edu/viewdoc/summary?

doi=10.1.1.56.7737.

Thatcher, G. (2004). Using Genetic Algorithms to Analyze Stock Portfolios. Retrieved

September 24th, 2008, from

http://www.gregthatcher.com/Papers/AI/GeneticAlgorithms.aspx.

Wang, J. (1998). Evolutionary Stock Trading Decision Support System Using Sliding

Window. Proceedings of the 1998 International Conference on Evolutionary

Computation. Retrieved from http://ieeexplore.ieee.org/servlet/opac?

punumber=5621.

Wang, Y. & Tang, W. (2005). Research on Intelligent Algorithm for Portfolio Selection

with Credibility Criterion. Proceedings of the Fourth International on Machine

23

Page 26: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

Learning and Cybernetics. Retrieved from http://ieeexplore.ieee.org/servlet/opac?

punumber=10231.

24

Page 27: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

APPENDIX B. Adjusted Scoring Ranges

Upon completion of the initial rating and sorting system a series of test were

begun to optimize and tweak this process before feeding the resulting data into the

genetic algorithm. When the top companies were looked over by hand it was discovered

that the previously determined rating system was inefficient at properly sorting great

companies out from the phenomenal. As the energy sector tends to be abnormally

volatile when compared to other sectors and as small cap stocks are generally more

volatile than medium and large cap stocks, small cap stocks were used as an indicator of

how balanced the rating system was. Below are the top five companies from the small

cap subcategory and the corresponding data for each company.

1: 91.847 GulfMark Offshore, Inc.2: 90.261 EV Energy Partners, L.P.3: 87.664 Contango Oil & Gas Company4: 86.793 Goodrich Petroleum Corporation5: 84.799 Clayton Williams Energy, Inc.

1: GulfMark Offshore, Inc.Total FA: 91.847Score: 1.43 Trailing P/E: 3.43Score: 10 P/B: 0.73Score: 1.54 P/S: 1.54Score: 8.843 Current Ratio: 8.843Score: 10 Quick Ratio: 2Score: 12 Net Profit Margin: 44.64Score: 9.97 Cash Flow Per Share: 3.97Score: 10 Beta: 1.25Score: 9.604 ROE: 24.01Score: 10 EPS: 7.56

25

Page 28: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

2: EV Energy Partners, L.P.Total FA: 90.261Score: 0 Trailing P/E: 1.5Score: 10 P/B: 0.59Score: 1.31 P/S: 1.31Score: 10 Current Ratio: 10Score: 10 Quick Ratio: 2Score: 12 Net Profit Margin: 109.61Score: 8.581 Cash Flow Per Share: 2.581Score: 7.68 Beta: 0.96Score: 12 ROE: 60.9Score: 10 EPS: 11.14

3: Contango Oil & Gas CompanyTotal FA: 87.664Score: 1.74 Trailing P/E: 3.74Score: 10 P/B: 1.84Score: 3.17 P/S: 3.17Score: 5.024 Current Ratio: 5.024Score: 9.4 Quick Ratio: 1.2Score: 12 Net Profit Margin: 90.34Score: 10 Cash Flow Per Share: 4.722Score: 9.84 Beta: 1.27Score: 12 ROE: 42.13Score: 10 EPS: 11

4: Goodrich Petroleum CorporationTotal FA: 86.793Score: 4.31 Trailing P/E: 6.31Score: 10 P/B: 1.18Score: 3.58 P/S: 3.58Score: 7.708 Current Ratio: 7.708Score: 9.6 Quick Ratio: 1.3Score: 12 Net Profit Margin: 63.06Score: 9.921 Cash Flow Per Share: 3.921Score: 6.08 Beta: 0.76Score: 10.854 ROE: 29.27Score: 9.48 EPS: 3.48

26

Page 29: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

5: Clayton Williams Energy, Inc.Total FA: 84.799Score: 0.76 Trailing P/E: 2.76Score: 10 P/B: 1.19Score: 0.72 P/S: 0.72Score: 4.084 Current Ratio: 4.084Score: 8 Quick Ratio: 0.5Score: 12 Net Profit Margin: 26.97Score: 9.395 Cash Flow Per Share: 3.395Score: 8.56 Beta: 1.43Score: 12 ROE: 59.11Score: 10 EPS: 11.67

The next trial used a completely unrestricted rating system where there were no minimum

or maximum caps. While this did increase the gaps between the stocks, it resulted in

unrealistic differences that would only serve to skew the output of the genetic algorithm.

Below are the top five small cap stocks and the corresponding data for each company.

1: 3952.39 North European Oil Royality Trust2: 144.18 New Concept Energy, Inc3: 122.73 Permian Basin Royalty Trust4: 102.986 EV Energy Partners, L.P.5: 99.373 Contango Oil & Gas Company

1: North European Oil Royality TrustTotal FA: 3952.39Score: 4.96 Trailing P/E: 6.96Score: 10 P/B: 1Score: 6.66 P/S: 6.66Score: 4.068 Current Ratio: 4.068Score: 0 Quick Ratio: 0Score: 18.213 Net Profit Margin: 97.13Score: 4.312 Cash Flow Per Share: 1.078Score: 1.84 Beta: 0.23Score: 3898.342 ROE: 38908.42Score: 9.975 EPS: 3.975

27

Page 30: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

2: New Concept Energy, IncTotal FA: 144.18Score: -1.65 Trailing P/E: 0.35Score: 10 P/B: 0.34Score: 2.31 P/S: 2.31Score: 8.34 Current Ratio: 8.34Score: 8.6 Quick Ratio: 3.5Score: 61.272 Net Profit Margin: 527.72Score: 1.4 Cash Flow Per Share: 0.35Score: 8.64 Beta: 1.42Score: 21.94 ROE: 144.4Score: 16.258 EPS: 10.258

3: Permian Basin Royalty TrustTotal FA: 122.73Score: 2.41 Trailing P/E: 4.41Score: -832.4 P/B: 423.2Score: 4.39 P/S: 4.39Score: 4 Current Ratio: 4Score: 0 Quick Ratio: 0Score: 18.421 Net Profit Margin: 99.21Score: 0.44 Cash Flow Per Share: 0.11Score: 4.32 Beta: 0.54Score: 911.928 ROE: 9044.28Score: 8.391 EPS: 2.391

4: EV Energy Partners, L.P.Total FA: 102.986Score: -0.5 Trailing P/E: 1.5Score: 10 P/B: 0.59Score: 1.31 P/S: 1.31Score: 8.734 Current Ratio: 8.734Score: 8.3 Quick Ratio: 2Score: 19.461 Net Profit Margin: 109.61Score: 8.581 Cash Flow Per Share: 2.581Score: 7.68 Beta: 0.96Score: 13.59 ROE: 60.9Score: 17.14 EPS: 11.14

28

Page 31: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

5: Contango Oil & Gas CompanyTotal FA: 99.373Score: 1.74 Trailing P/E: 3.74Score: 10 P/B: 1.84Score: 3.17 P/S: 3.17Score: 5.024 Current Ratio: 5.024Score: 8.14 Quick Ratio: 1.2Score: 17.534 Net Profit Margin: 90.34Score: 10.722 Cash Flow Per Share: 4.722Score: 9.84 Beta: 1.27Score: 11.713 ROE: 42.13Score: 17 EPS: 11

The next and final trial involved using an adjusted rating system similar to the one

above but with minimum and maximum restrictions. After a company’s financials

surpass the 1 through 10 rating scale on either end, a new scale kicks in which adds

points to the rating scale much less willingly and subtracts points slightly less willingly.

This properly allowed the exceeding companies to stand out and helped to separate the

companies that had abnormally bad ratings but were not distinguished before as their

financials capped the rating scale. Overall, this creates a rating scale where it is much

easier to lose points then it is to gain them, again aiding in selecting only the most

financially secure stocks. Below are the top five small cap stocks and the corresponding

data for each company.

1: 96.385 EV Energy Partners, L.P.2: 94.839 Contango Oil & Gas Company3: 93.997 GulfMark Offshore, Inc.4: 90.407 Clayton Williams Energy, Inc.5: 89.71 New Concept Energy, Inc

29

Page 32: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

1: EV Energy Partners, L.P.Total FA: 96.385Score: -0.5 Trailing P/E: 1.5Score: 10 P/B: 0.59Score: 1.31 P/S: 1.31Score: 8.734 Current Ratio: 8.734Score: 8.3 Quick Ratio: 2Score: 15 Net Profit Margin: 109.61Score: 8.581 Cash Flow Per Share: 2.581Score: 7.68 Beta: 0.96Score: 13.59 ROE: 60.9Score: 15 EPS: 11.14

2: Contango Oil & Gas CompanyTotal FA: 94.839Score: 1.74 Trailing P/E: 3.74Score: 10 P/B: 1.84Score: 3.17 P/S: 3.17Score: 5.024 Current Ratio: 5.024Score: 8.14 Quick Ratio: 1.2Score: 15 Net Profit Margin: 90.34Score: 10.722 Cash Flow Per Share: 4.722Score: 9.84 Beta: 1.27Score: 11.713 ROE: 42.13Score: 15 EPS: 11

3: GulfMark Offshore, Inc.Total FA: 93.997Score: 1.43 Trailing P/E: 3.43Score: 10 P/B: 0.73Score: 1.54 P/S: 1.54Score: 8.169 Current Ratio: 8.169Score: 8.3 Quick Ratio: 2Score: 12.964 Net Profit Margin: 44.64Score: 9.97 Cash Flow Per Share: 3.97Score: 10 Beta: 1.25Score: 9.604 ROE: 24.01Score: 13.56 EPS: 7.56

30

Page 33: BEGIN TITLE THREE INCHES FROM TOP OF PAPER

4: Clayton Williams Energy, Inc.Total FA: 90.407Score: 0.76 Trailing P/E: 2.76Score: 10 P/B: 1.19Score: 0.72 P/S: 0.72Score: 4.084 Current Ratio: 4.084Score: 8 Quick Ratio: 0.5Score: 11.197 Net Profit Margin: 26.97Score: 9.395 Cash Flow Per Share: 3.395Score: 8.56 Beta: 1.43Score: 13.411 ROE: 59.11Score: 15 EPS: 11.67

5: New Concept Energy, IncTotal FA: 89.71Score: -1.65 Trailing P/E: 0.35Score: 10 P/B: 0.34Score: 2.31 P/S: 2.31Score: 8.34 Current Ratio: 8.34Score: 8.6 Quick Ratio: 3.5Score: 15 Net Profit Margin: 527.72Score: 1.4 Cash Flow Per Share: 0.35Score: 8.64 Beta: 1.42Score: 15 ROE: 144.4Score: 15 EPS: 10.258

31