May 3-7, 2005
Thesis Presentations
721 Broadway
at Waverly Place
4th Floor
South Elevators
New York, NY 10003
For additional
information email
itp.inquiries@nyu.edu
or call 212.998.1880
| Dynamic Information Visualization | ||
|---|---|---|
| Author(s): | Wontae Kim | |
| Instructor: | Hechinger, Nancy | |
| Class: | Final Project Seminar (Wed.) | |
| URL: | http://home.nyu.edu/~wk292/Thesis.dcr | |
| Keywords: | Information Visualization,Technical Analysis,Stock Market | |
| Exploration of Information Visualization systems that utilize dynamic information sources. | ||
| The ability to access and analyze dynamic information quickly is one of the most important factors to finding good investment opportunities in the financial markets. My thesis project attempts to examine new ways to visualize information pertaining to stock market data. My goal is to provide an innovative tool for investors that will allow them to understand market movements more easily and quickly than via traditional means such as charts and tables. The product is targeted primarily for professional investors, such as portfolio managers, and traders. | ||
| Personal Statement: | The ability to access and analyze information in a timely manner is one of the most important factors to finding good investment opportunities in the financial markets. For technical analysts, many of their decisions are based upon certain proprietary models, as well as standard historical graphs that display all a wide variety of data including, price, volume, momentum, stochastics, etc. However, due to the nature of bar and line graphs, analysis is usually limited to one company, or one market index at a time. While the standard bar graph is a useful tool for analyzing one company, it is difficult to see how that company is performing relative to its competitors, as well as the rest of the market. In order to obtain information on a number of different companies at a time, one must resort to tables that list the current data for the information required. In this case, however, it is difficult to see any trends that may be occurring.
Over the past few years, a few products have been developed allowing users to view the market from a top-down approach. These products allow users to see how industries, sectors, as well as individual stocks are performing in relation to the market, other companies, and even its own industry. While this takes a large step forward in helping technical analysts discover trends, it lacks the ability for users to determine buy/sell triggers because there is no reference to how the company is performing in relation to its own past performance. I believe that it is possible to create a new charting system that allows a user to quickly determine current and historical information for all companies within a market or index. This system would give users a top-down view of the market, including both historical and current data for all of the companies that comprise it. Would allow one to see what the trends are for specific companies, industries, as well as the entire market as a whole simultaneously. As a result, I believe that it would be easier to find good stock picks. Moreover, I hope to develop a new indicator that automatically factors in its own past performance and highlights potential buy/sell triggers. My thesis project attempts to examine the new ways of visualizing information pertaining to stock market data. My goal is to provide an innovative tool for investors that will allow them to understand market movements more easily and quickly than via traditional means such as charts and tables, as well as providing a useful product that will help users make buy/sell decisions. The product is targeted primarily for professional investors, such as portfolio managers, and traders. The core features includes a web based interface showing the stock market. Individual stocks are grouped into industries so that users are able to look at which particular industries are performing well. There is a list of indicators, so that the user can choose which indicator to view the market by, and a historical playback function that allows the user to see the movement of the market over a specified time frame. From a technical analysis standpoint, I have developed a new indicator that collects the values of certain existing indicators and performs statistical analysis on these values, highlighting values that are out-of-the-ordinary as potential buy/sell triggers. | |
| Methodology: | The development of the prototype for my product is done with Macromedia Director. However, since the product is designed to be web based, or incorporated with existing technical analysis products, the final version of the product should be done in Java.
Data Sources The price and stochastic data used in my thesis project were taken from eSignal, a technical analysis tool. For my prototype, I decided to use the 30 companies that comprise Dow Jones Industrial Average (DJIA). Since the two indicators I would be using were price and stochastic data, I downloaded daily price and stochastic information for the past 300 days for each of the companies that comprise the DJIA. The date, opening price, closing price, and stochastic value for each company were stored in a separate array for each company to be used for all functions of my prototype. Organization of Data Following the organizational structure of stocks by the Standard Industrial Classification (SIC) code, I grouped the companies by sector. This resulted in nine sectors including basic materials, capital goods, conglomerates, consumer, energy, financial, healthcare, services and technology. Ticker Sector AA Basic Materials DD Basic Materials BA Capital Goods CAT Capital Goods HON Capital Goods GE Conglomerates MMM Conglomerates UTX Conglomerates GM Consumer Cyclical KO Consumer Non-Cyclical MO Consumer Non-Cyclical PG Consumer Non-Cyclical XOM Energy AIG Financial AXP Financial C Financial JPM Financial JNJ Healthcare MRK Healthcare PFE Healthcare DIS Services HD Services MCD Services SBC Services VZ Services WMT Services HPQ Technology IBM Technology INTC Technology MSFT Technology Statistical Analysis of Stochastic Buy/Sell Triggers One of the most important features of my thesis project is the Buy/Sell trigger that alerts investors when a stock may be ready to move. This is done by determining whether or not the value of the stochastic indicator is abnormal. In order to do this, the program will run a statistical analysis of the stochastic values to determine what value the stochastic indicator must be in order to be two standard deviations away from the mean value. Before talking about the statistical analysis of the stochastic indicator, I would like to give an overview of what a stochastic value is, and how it can be applicable when making buy/sell decisions. The formulas for determining stochastic values are as follows: http://www.stockcharts.com/education/IndicatorAnalysis/indic_stochasticOscillator.html The stochastic indicator was developed by George Lane, and measures the relationship between the stock’s closing price, and the price range of the stock over a period of time. The theory behind the importance of the stochastic is that the closing price of a stock will end near the high end of the trading range when the trend for the price movement is positive, and vice versa for when prices are falling. The stochastic value of the stock will oscillate between 0 and 100. When the value of the stochastic is between 80 – 100, the stock is considered overbought, and prices are generally expected to fall. When the value is between 0 – 20, the stock is considered to be oversold, and prices are expected to rise. Calculation of Values at Overbought and Oversold Levels I believe that looking for abnormally high or low stochastic values will indicate to technical analysts that a stock’s price is ready to move. Therefore, my first attempt was simply to create a distribution curve with all stochastic values over the past 100 days. The results were not favorable, because the values between 21 – 79, which are usually discarded when making buy/sell decisions, were throwing off the mean and standard deviation that determined the buy/sell trigger. I determined that the overbought and oversold signals should be calculated separately. Therefore, I decided to run two distributions, one for abnormally high overbought values, and one for abnormally low oversold values. In order to determine overbought values, I discarded all values that were below 80. Then I ran a statistical analysis to determine the mean and standard deviation for all value between 80 and 100. In order to calculate the extremely overbought value, I added the mean value plus two standard deviations. This means that only about 5% of the stochastic values will be higher than this number. Similarly, to calculate the abnormally low values I ran the distribution of all values between 0 and 20. These two values were calculated and stored for each company. Calculation of Peaks and Valleys at Overbought and Oversold Levels After completing the first set of analysis on all 30 companies, I wondered if attempting to determine the abnormally high peaks and valleys of a company’s stochastic indicator would be a more useful gauge of overbought and oversold levels. The reasoning behind this, is that simply calculating a abnormally high overbought or oversold stochastic level did not necessarily indicate when a stock’s stochastic indicator would reverse direction. In order to calculate these values, I had to find all values above 80 or below 20 that changed direction. In order to compute peaks, I took all values above 80 and compared them to the previous value as well as the next value. If the current value was higher than the previous value and lower than the next value, then I stored this value in an array as a peak. I then ran a distribution of all the values in the new array to determine the mean and standard deviation of the peaks. I used the same methodology to determine valleys. Graphical User Interface The design of the graphical user interface (GUI) is inspired by Tree Maps, developed by the Human-Computer Interaction Lab (HCIL) at the University of Maryland. Tree Map The GUI has similar functionality of the Tree Map and Market Carpet with the added ability to use filters that determine abnormally high or low values of specific indicators. Different views of the market can be selected by using a drop down menu at the top left corner, allowing users to switch between a price view and a stochastic view. The triggers can be selected through the use of checkboxes. When checked, any stock that fits the requirements of each trigger will display a colored box at the top right corner of the stock. At the bottom, there is a scroll bar which allows users to scroll back and forth through time. Price View Thesis Project Price View As you can see, each square represents a company that comprises the DJIA. Additionally, the companies are grouped by industry code. The color of each square represents the price change for that particular stock when compared to the previous days’ closing price. Companies that are bright green have gained 5% or more while bright red represents companies that have lost 5% or more of their value. There is no need to display company’s name or any other information in this view because this is purely technical analysis. However, when a particular square is clicked, then a link providing further information for that specific stock will open in a new window for further company specific analysis. When using the price view, the price change of each company was calculated by determining the current price divided by the previous days’ closing price minus 1. Price change = (Current price / Previous Closing price) – 1 Stochastic View Thesis Stochastic View The stochastic view requires no calculation at all. It simply shows the stochastic value of each stock in a color that represents its value. Stochastic values of 0 are shown in white. As the stochastic value increases to 100, the color becomes black. Historical Scroll Bar The historical scroll bar, located at the bottom, allows users to scroll back and forth through time. This simply passes the values stored in the initial array and displays the change in price for that particular day. | |
| Sources: | Investopedia – http://www.investopedia.com/ - Provided definitions and background information for stock market terminology and technical analysis.
StockCharts.com - http://stockcharts.com/education/Overview/techAnalysis1.html - Provided information relating to technical analysis and indicators. StockTA.com – http://www.stockta.com – Technical Analysis Charts. Andrews, Keith – Information Visualization Tutorial Notes, http://www.iicm.edu/ivis/, July 2002 Fry, Benjamin – Organic Information Design, Massachusetts Institute of Technology, May 1997 Human-Computer Interaction Lab, University of Maryland – Tree Map - http://www.cs.umd.edu/hcil/treemap/index.shtml eSignal.com – http://www.esignal.com – Source for stock market data Yahoo Quotes – http://quote.yahoo.com – Source for the companies that make up the DJIA and the SIC for each company. Map of the Market - http://www.smartmoney.com/marketmap Market Carpet - http://www.stockcharts.com/carpet.html | |
| Conclusions: | Overall, I believe that my thesis project was a success. While the use of Treemaps for displaying stock data is already in use, my thesis project’s stock specific analysis tied with the use of Treemaps is unique and provides a valuable tool for investors.
Graphical User Interface As far as the development of the user interface goes, I was disappointed to find so many tools that were similar to what I have developed. I believe that the use of Treemaps provides users with a very quick and simple way to determine the general status of the market, as well as specific industries and companies. As previously mentioned, a technical analyst begins searching for new ideas by first analyzing the market, then determining the strongest and weakest industries, and finally attempting to analyze the strongest and weakest stocks within those industries. The Treemap view that my GUI utilizes helps a potential technicial analyst to do this rapidly and easily in one simple, yet effective screen. Buy/Sell Triggers The real power behind my thesis project lies in the analysis of individual companies tied into the Treemap type interface. My research has indicated that the existing products do not analyze stocks at a company specific level, which is a main requirement for technical analysis. The stock trigger that I have developed performs analysis of each company, and displays them on top of the market view. As far as the specific trigger that I have developed goes, further research and backtesting must be done to show its efficacy as either a trigger to buy or sell, or simply to show whether the stock’s price is ready to move. To date, my analysis does show that a stocks price will generally decrease when the overbought trigger has been reached and vice versa for the oversold trigger. However, much more research must be done to analyze the risks and rewards associated with trading utilizing these triggers. Strengths The strengths of the project include the ability to quickly gather a lot of information about the market, including leading and lagging industries, as well as potential investment ideas. Weaknesses As far as an analysis tool is concerned, it is still a very high level analysis tool. Professional investors will need to do company specific research when making final buy or sell determinations. Future Work There is much more work to be done with this project. I am happy with the use of Treemaps for the user interface. I would like to incorporate additional buy/sell triggers that are already in existence as well as continue to improve upon the stochastic buy/sell trigger that I have developed. I believe that this product would be most effective when incorporated into an existing technical analysis product such as eSignal. Because of its main strength as an idea generator, tying my project into an existing system would allow users to quickly generate new ideas using my project, and then allow them to perform in-depth research using the tools provided by existing systems. Conclusion The use of new interfaces for the technical analysis of the stock market remains to be a new and rapidly developing area. My thesis project incorporates the strengths of many features that are already on the market and provides new functionality where existing products have failed. | |