Course in Artificial Neural Networks: Applications to Finance and Strategy

At IIM-C we firmly believe that information is power. In the years to come the need for increased sharing of information will govern the changes in organizational structure. A basic understanding of information systems is thus mandatory. We endeavor to ensure that you drive the e-commerce revolution. These courses not only equip you with software tools but also impart an understanding of the hardware which will help you set design your own database. It's time you started your own dotcom at IIMC.

Introduction:

Artificial Neural Network (ANN) is currently probably the most recognized Soft Computing / Computational Intelligence tool which is widely used in the area of problem modeling, particularly in the case where there is no known relationship that exists between the input and the output, or there is adequate knowledge regarding the relationship among the variables involved, or even there is any good understanding about the patterns (such as clustering structure, topological or causal map) that exist in the data space. So far the success rate of ANN in such problem solving in the areas of Finance and Strategy is quite commendable, and more and more ANN is either replacing the classical methods such as Statistical tools (e.g. Time Series methods), or enriching these by supplementing such efforts. In the process ANN itself is going through a lot of evolutionary processes (e.g. development of higher order networks, or recurrent networks - networks with loops) to tackle newer problems particularly those arising in highly uncertain markets in a dynamic or even chaotic situation (often confused with randomness) in the presence of strongly globalized economy and also for achieving more efficiency and generalizeability.

Probability and Statistics among other mathematical tools provide a strong mathematical basis for ANN, which is why ANN is such a robust technique. Statistics continues to contribute to the development of ANN, particularly for solving data rich problems (yet avoiding data over-fit tendencies). But ANN also enriches Statistics, e.g. Support Vector Regression. ANN can be an efficient tool for performing linear or non-linear as well as multiple regressions. ANN also is powerful for computing Principal Components. More and more ANN is coming out of the infamous 'Black Box' syndrome; people are now better able to conceptualize the ANN models to map their understanding about the complex nature of the problem. There are also now a lot of hybrid methods developed around ANN. For examples, a) tune an ANN model using approaches such as Genetic Algorithm (for optimizing the architecture, for instance), b) develop the black-box ANN model using fuzzy expert rule base which are easier for the modeler to comprehend, and c) use ANN to optimize popular Fuzzy Inference Systems such as Takagi-Sugeno model, as in ANFIS. There are also ensemble methods such as expert selection (by voting) from among different ANN architectures where the domain knowledge is poor. With the development of Computational Learning Theory (COLT) (also referred as Statistical Learning Theory - SLT) it is not only possible to give often practical confidence bounds for an ANN model provided the sample size is adequately large even if we do not know the underlying probability distributions, it is also possible to compare the generalization capabilities between different ANN models, or compare ANN models vs. other models such as SVM (Support Vector Machines).
Course Objectives

The objectives of this application oriented course are:

1. To bring the student closer to the heart of the application problem of her individual choice or interest in the areas of finance and strategy, to expose her to existing solutions, and finally to enable the student to make an attempt to solve the problem better from a practitioner's point of view.
2. To bring home the points to the students regarding general issues of problem solving in a data modeling situation.
3. To make the student aware of the problems of finance and strategy at large that are up for grab for solving using ANN.
4. To expose the student to the power and pitfalls of ANN vis-à-vis financial modeling.



The course will introduce the concept of ANN and its applications to Finance and Strategy.

Course Content:

a) Exposure to relevant concepts, architectures and tools of ANN and also to the general issues regarding the 'learning from data' paradigm.
b) Exposure to special areas of finance and strategy depending on students' interests.
c) Review of ANN applications for modeling selected problems in finance and strategy.
Pedagogy



Project on live data, Case Discussions and Presentation, ANN Demo cum Problem Solving, and Lectures.
Session Plan

1. ANN and general modeling concepts: 6 sessions
2. Special topics in Finance and Special Demo Sessions on Finance applications (Finance Research and Trading Lab): 4 sessions
3. Case Discussions by Faculty: 4 sessions
4. Presentation of the Finance topic based on the project by students: 2 sessions
5. Case presentation by students: 2 sessions
6. Presentation / Demo of project by students: 2 sessions



The number of sessions can vary depending on the number of students and projects etc.
Evaluation

Project and Case Discussion-70%
End-term-30%
Sample References:

* Neural networks in the capital markets, Ed. by Apostolos-Refenes, John Wiley & Sons,1995.
* Neural networks in time series forecasting of financial markets, by E Michael Azoff, John Wiley & Sons, 1995.
* Neural networks in finance and investing, Ed. by Robert R Trippi and Efraim Turban, Heinemann Asia, 1995.
* Neural Networks in Finance: Gaining Predictive Edge in the Market, P. D. McNeils, Elsevier Academic Press, 2005.
* Soft Computing Approach to Pattern Recognition and Image Processing, A Ghosh and S K Pal, World Scientific, 2002.
* Modelling the market, by Focardi, Jonas C., Frank Fabozzi & Associates, 1997.
* Trading on The Edge, Guido J. Deboeck (Ed), John Wiley, 1994
* Financial optimization, by Zenios S, 1994.
* Tools for Computaional finance (Third Edition), R. Seydel. Springer, 2006 (Universitext).
* Applied Computational Economics and Finance, M.J. Miranda and P.L. Fackler. MIT Press, 2002 (Pearson Education, Low Price Edition, 2004.).