Catálogo de publicaciones - libros

This chapter provides an overview of the book.

It has been twenty-five years since DeWit(1982) first applied fuzzy logic (FL) to insurance. That article sought to quantify the fuzziness in underwriting. Since then, the universe of discourse has expanded considerably and now also includes FL applications involving classification, projected liabilities, future and present values, pricing, asset allocations and cash flows, and investments. This article presents an overview of these studies. The two specific purposes of the article are to document the FL technologies have been employed in insurance-related areas and to review the FL applications so as to document the unique characteristics of insurance as an application area.

Traditionally, autoregressive integrated moving average (ARIMA) models have been one of the most widely used linear models in time series forecasting. However, ARIMA models can not easily capture nonlinear patterns. In the last two decades artificial neural networks (ANNs) have been proposed as an alternative to traditional linear models, particularly in the presence of nonlinear data patterns. Recent research suggests that a hybrid approach combining both ARIMA models and ANNs can lead to further improvements in the forecasting accuracy compared with pure models. In this paper, a hybrid model that combines a seasonal ARIMA model and an Elman neural network (ENN) is used to forecast agricultural commodity prices. Different approaches for specifying the ANNs are investigated among others, and a genetic algorithm (GA) is employed to determine the optimal architecture of the ANNs. It turns out that the out-of-sample prediction can be improved slightly with the hybrid model.

To improve the management of the Employment Time Guarantee Fund (Fundo de Garantia do Tempo de Serviço - FGTS), a study in Brazil is conducted to analyze past data and anticipate the future trends of this fund. In this paper, Nonlinear Principal Component Analysis (NLPCA) - with the Artificial Neural Network architecture and Back-Propagation algorithm - is used to reduce the data dimension in describing various causes of withdrawals from the FGTS. With the analysis of the properties of these withdrawals, the paper discusses the correlation between the policy of free treatment of AIDS patients and their withdrawal from the plan. Nonlinear time series corresponding to each cause of withdrawal over 75 months - from 1994 to 2000 - are collected from the administrator of the FGTS. Using NLPCA, 17 small quantity time series (Group 1) are combined into one variable and then combined with other 7 middle quantity series (Group 2) to form another variable. Finally, four combined time series (Group 3) are formed which can well represent features of the total of 27 kinds of withdrawals with respect to their different causes. As a criterion for dimension reducing, the coefficient of correlation between the output of Group 1 and the sum of 17 is 0.8486 and that between Group 2 and sum of 8 is 0.9765.

Female Labor Force Participation (FLFP) is perhaps one of the most relevant theoretical issues within the scope of studies of both labor and behavioral economics. Many statistical models have been used for evaluating the relevance of explanatory variables. However, the decision to participate in the labor market can also be modeled as a binary classification problem. For this reason, in this paper, we compare four techniques to estimate the Female Labor Force Participation. Two of them, Probit and Logit, are from the statistical area, while Support Vector Machines (SVM) and Hamming Clustering (HC) are from the machine learning paradigm. The comparison, performed using data from the Venezuelan Household Survey for the second semester 1999, shows the advantages and disadvantages of the two methodological paradigms that could provide a basic motivation for combining the best of both approaches.

The DEA model provides scores regarding firms’ efficiency, but it does not obtain an overall map about each unit’s position, in order to identify competitive clusters and improve the design of complete benchmarking policies. This lack makes the interpretation of DEA results difficult, together with its real applications for the management of firms.

This paper outlines a data mining approach to the analysis and prediction of the trend of stock prices. The approach consists of three steps, namely, partitioning, analysis and prediction. A commonly used -means clustering algorithm is used to partition stock price time series data. After data partition, linear regression is used to analyse the trend within each cluster. The results of the linear regression are then used for trend prediction for windowed time series data. Using our trend prediction methodology, we propose a trading strategy TTP (Trading based on Trend Prediction). Some results of applying TTP to stock trading are reported. The trading performance is compared with some practical trading strategies and other machine learning methods. Given the volatility nature of stock prices the methodology achieved limited success for a few countries and time periods. Further analysis of the results may lead to further improvement in the methodology. Although the proposed approach is designed for stock trading, it can be applied to the trend analysis of any time series, such as the time series of economic indicators.

This paper is a practical guide to the application of instance-based machine learning techniques to the solution of a financial problem. A broad class of instance-based families is considered for classification using the WEKA software package. The problem selected for analysis is a common one in financial and econometric work: the use of publicly available economic data to forecast future changes in a stock market index. This paper examines various stages in the analysis of this problem including: identification of the problem, considerations in obtaining and preprocessing data, model and parameter selection, and interpretation of results. Finally, the paper offers suggestions of areas of future study for applying instance-based machine learning in the setting of solving financial problems.

Instance based learning is a class of data mining learning paradigms that applies specific cases or experiences to new situations by matching known cases and experiences with new cases. This paper presents an application of the instance-based learning algorithm for predicting daily stock index price changes of the S&P 500 stock index between October 1995 and September 2000, given the daily changes in the exchange rate of the Canadian Dollar, the Pound Sterling, the French Franc, the Deutsche Mark and the Yen, the monthly changes in the consumer price index, GDP, and the changes in the monthly rates of certificates of deposit. The algorithm is used to predict an increase, decrease or no change in the S&P 500 stock index between a business day and the previous business day. The predictions are carried out using the IB3 variant of the IBL algorithms. The objective is to determine the feasibility of stock price prediction using the IB3 variant of the IBL algorithms. Various testing proportions and normalization methods are experimented with to obtain good predictions.

It is well known that index fund optimization is important when hedge trading in a stock market. By “optimization” is meant the optimization of the proportion of funds in the index fund. Index funds consisting of a small number of listed companies are constructed in this paper by means of a genetic algorithm method based on the coefficient of determination between the return rate of the fund price and the changing rate of the market index. The method is examined with numerical experiments applied to the Tokyo Stock Exchange. The results show that the index funds work well in forecasting over a future period when a market index has followed a downward or a flat trend. In addition, we reveal problems arising from this optimization in that the coefficient of determination depends on the characteristics of the scatter diagram between the index fund price and the market index.