Mary Malliaris

A Neural Network Model for Estimating Option Prices

A neural network model that processes financial input data is developed to estimate the market price of options at closing. The networks ability to estimate closing prices is compared to the Black-Scholes model, the most widely used model for the pricing of options. Comparisons reveal that the mean squared error for the neural network is less than that of the Black-Scholes model in about half of the cases examined. The differences and similarities in the two modeling approaches are discussed. The neural network, which uses the same financial data as the Black-Scholes model, requires no distribution assumptions and learns the relationships between the financial input data and the option price from the historical data. The option-valuation equilibrium model of Black-Scholes determines option prices under the assumptions that prices follow a continuous time path and that the instantaneous volatility is nonstochastic.

Using neural networks to forecast the S&P 100 implied volatility

The implied volatility, calculated using the Black-Scholes model, is currently the most popular method of estimating volatility and is considered by traders to be a significant factor in signalling price movements in the underlying market. Thus, the ability to develop accurate forecasts of future volatility allows a trader to establish the proper strategic position in anticipation of changes in market trends. A neural network has been developed to forecast future volatility, using past volatilities and other options market factors. The performance of this network demonstrates its value as a predictive tool.

Beating the best: A neural network challenges the Black-Scholes formula

A neural network model which processes financial input data is presented to estimate the market price of options. The networks ability to estimate option prices is compared to estimates generated by the Black-Scholes model, a traditional financial model. Comparisons reveal that the neural network outperforms the Black-Scholes model in about half of the cases examined. While the two modeling approaches differ fundamentally in their methodology for determining option prices, some common results emerge. While the neural network performs better than Black-Scholes on prices out-of-the-money, estimations near the expiration data are accurate for both.<<ETX>>

Forecasting energy product prices

Five inter-related energy products are forecasted one month into the future using both linear and nonlinear techniques. Both spot prices and data derived from those prices are used as input data in the models. The models are validated by running data from the following year. Results show that, even though all products are highly correlated, the prediction results are asymmetric. In forecasts for crude oil, heating oil, gasoline and natural gas, the nonlinear forecasts were best while for propane, the nonlinear model had the largest average absolute error.

Time series and neural networks comparison on gold, oil and the euro

Gold, oil, and the euro are three very important economic markets that have been studied individually by numerous authors. But certain basic questions about their inter-relationships since the year 2000 remain unaddressed. Gold has been an important commodity for several centuries. Oils importance grew during the 20th century, and the euro has become important during the 21st. Standard economic analysis allows us to hypothesize about a specific relationship and test for it, and a neural network gives us the ability to identify important variables in a forecast without forming a prior hypothesis about the relationship of each variable to the target. This paper analyzes the inter-relationships among the price behavior of gold, oil and the euro using a standard time series methodology then employs neural networks to build a forecast for each of the three variables. We then compare the results of the neural network to those implied by the time series tests. The statistical evidence of time series analysis demonstrates that both short-term and long-term relationships exist between the three variables. Both the time series and neural network results indicate that the series move together though they identify slightly different relationships. The time series results imply that oil adjusts to gold, the euro and oil have equal affects on each other, and the weakest relationship is between gold and the euro. The neural network indicates that oil impacts gold more than gold impacts oil, oils affect on the euro is greater than the euros effect on oil and last, golds impact on the euro is greater and faster than the euros impact on gold.

Modeling the behavior of the S&P 500 index: a neural network approach

The October 1987 stock market crash challenged the prevailing financial models of a random walk and led to the emergence of a new and competing model of stock price time series. This new approach supports a nonrandom underlying structure and is labeled chaotic dynamics. If a neural network can be constructed which determines market prices better than the random walk model, it would support those who claim that they have found statistical evidence that a chaotic dynamics structure underlies the market. This paper constructs a neural network which lends support to the deterministic paradigm.<<ETX>>

Online Analytical Processing (OLAP) for Decision Support

Online analytical processing (OLAP) refers to the general activity of querying and presenting text and number data from data warehouses and/or data marts for analytical purposes. This chapter gives an overview of OLAP and explains how it is used for decision support. Before the specific OLAP functions and platforms are presented, the connection between the OLAP systems and analytical data repositories is covered. Then, an overview of functionalities that are common for all OLAP tools is presented.

Do-ahead replaces run-time: a neural network forecasts options volatility

Compares three methods of estimating the volatility of daily S&P 100 Index stock market options. The implied volatility, calculated via the Black-Scholes model, is currently the most popular method of estimating volatility and is used by traders in the pricing of options. Historical volatility has been used to predict the implied volatility, but the estimates are poor predictors. A neural network for predicting volatility is shown to be far superior to the historical method.<<ETX>>

Are Foreign Currency Markets Interdependent? Evidence from Data Mining Technologies

This study uses two data mining methodologies: Classification and Regression Trees (C&RT) and Generalized Rule Induction (GRI) to uncover patterns among daily cash closing prices of eight currency markets. Data from 2000 through 2009 is used, with the last year held out to test the robustness of the rules found in the previous nine years. Results from the two methodologies are contrasted. A number of rules which perform well in both the training and testing years are discussed as empirical evidence of interdependence among foreign currency markets. The mechanical rules identified in this paper can usefully supplement other types of financial modeling of foreign currencies.

Modeling Federal Funds rates: a comparison of four methodologies

Monthly Federal Fund interest rate values, set by the Federal Open Market Committee, have been the subject of much speculation prior to the announcement of their new values each period. In this study we use four competing methodologies to model and forecast the behavior of these short term Federal Fund interest rates. These methodologies are: time series, Taylor, econometric and neural network. The time series forecasts use only past values of Federal Funds rates. The celebrated Taylor rule methodology theorizes that the Federal Fund rate values are influenced solely by deviations from a desired level of inflation and from potential output. The econometric and neural network models have inputs used by both the time series and Taylor rule. Using monthly data from 1958 to the end of 2005 we distinguish between sample and out-of-sample sets to train, evaluate, and compare the models’ effectiveness. Our results indicate that the econometric modeling performs better than the other approaches when the data are divided into two sets of pre-Greenspan and Greenspan periods. However, when the data sample is divided into three groups of low, medium and high Federal Funds, the neural network approach does best.