Ralf Östermark

Solving Irregular Econometric and Mathematical Optimization Problems with a Genetic Hybrid Algorithm

In the present paper we apply a new Genetic Hybrid Algorithm (GHA) to globally minimize a representative set of ill-conditioned econometric/mathematical functions. The genetic algorithm was specifically designed for nonconvex mixed integer nonlinear programming problems and it can be successfully applied to both global and constrained optimization. In previous studies, we have demonstrated the efficiency of the GHA in solving complicated NLP, INLP and MINLP problems. The present study is a continuation of this research, now focusing on a set of highly irregular optimization problems. In this paper we discuss the genetic hybrid algorithm, the nonlinear problems to be solved and present the results of the empirical tests.

A fuzzy control model (FCM) for dynamic portfolio management

Abstract In the present study we continue previous work by recasting the portfolio management problem into a control theoretic form, still retaining a recursive programming structure through time. The uncertainty of future prices and risk levels is recognized by fuzzy numbers. Our system has interesting directions for future research: improving performance through expert systems support for forecasting, using the experience from the model in efforts to theory design, solving the fuzzy control model (FCM) when the state objectives, return predictions and the level of economic friction are fuzzy.

Vector forecasting and dynamic portfolio selection: Empirical efficiency of recursive multiperiod strategies

Abstract In the paper we present a dynamic portfolio selection system. The system combines recent results from the domains of time series analysis, portfolio theory and multiperiod linear programming under uncertainty. The performance of the decision support system is tested empirically against a dynamic super criterion defined in the study. We demonstrate that suitable combination of ex ante and ex post information with initial balances for the planning horizon yields better portfolio efficiency through time than a strategy relying purely on ex ante information. The system was programmed on Abo Academys vax 8800 computer, mainly in fortran , with access to the imsl library and ifps/optimum .

Gestalt system of holistic graphics: new management support view of MCDM

We present a holistic system of unaggregated multidimensional “profiles”, avoiding the information-destroying reduction of multicriterion dimensionality to a single “number”. Visualization of system information is proposed, transforming numerical data into graphical images. This respects and enhances human ability (and preference) to reason directly on the basis of a graphical Gestalt.

Growth and Profitability in Small Privately Held Biotech Firms: Preliminary Findings

This paper reports on preliminary findings on a study of the relationship of growth and profitability among small privately held Finnish Life Science firms. Previous research results concerning growth and profitability are mixed, ranging from strongly positive to a negative relationship. The conventional wisdom states that growth is a prerequisite for profitability. Our results suggest that the reverse is the case. A high profitability-low growth biotech firm is more probably to make the transition to high profitability-high growth than a firm that starts off with low profitability and high growth.

Fuzzy linear constraints in the capital asset pricing model

Abstract In this paper the topic of portfolio management is tackled by a fuzzy mathematical programming approach. It is demonstrated that managerial imprecision may be explicitly incorporated in the policy constraints augmented coefficient matrix of the quadratic portfolio problem. By organizing the first-order conditions the quadratic problem is linearized and solvable by matrix inversion. Through fuzzification, the policy constraints augmented problem is solvable by parametric methods of fuzzy linear programming. The augmented portfolio program is directly amenable to the position vector method develop by the author.

A hybrid genetic fuzzy neural network algorithm designed for classification problems involving several groups

We propose a multigroup classification algorithm based on a hybrid genetic fuzzy neural net (GFNN) framework. Recent results on evolutionary computation and fuzzy neural network methodology are combined to effectively adapt the membership functions of the fuzzifier and the defuzzifier to the data set. Separate membership functions are defined for each dimension in the fuzzifier and for each fuzzy output group in the defuzzifier. The signal inherent in the fuzzifier is aggregated by a suitable T-norm and transmitted to the defuzzifier. The defuzzifier aggregates the response, i.e., the predicted group membership, by a suitable conorm. If misclassifications occur during training, the membership functions of both the fuzzifier and the defuzzifier are adapted by a systematic, robust procedure. The algorithm is successfully tested with real economic data. In total, the GFNN performs as good as the best of the competing methods in our test. The results suggest economically meaningful interpretations.

A fuzzy neural network algorithm for multigroup classification

We propose a multigroup classification algorithm based on a hybrid fuzzy neural net framework. A key feature of the approach is the adaptation of membership functions to new data. In this way, learning is reflected in the shape of the membership functions. By defining separate membership functions for each fuzzy output class, we allow dynamic adjustment of the functions during training. The algorithm is successfully tested with real economic data. The results suggest economically meaningful interpretations.

VARMAX-modelling of blast furnace process variables

Abstract In this study we present some preliminary evidence on Vector Autoregressive Moving Average (VARMAX) modelling of the hot metal silicon content and temperature of pig iron using key process data from a blast furnace. In many empirical estimation problems, a priori knowledge of the dynamic structure of the observed processes is lacking. In the blast furnace case, the dynamics of pig iron production is roughly known, thus alleviating the search for a proper lag structure. In the paper we present fairly adequate VARMAX-models for the silicon content and temperature. The approach of the present paper thus should provide a useful framework for practical modelling and control of the production process.

Temporal interdependence in fuzzy MCDM problems

Abstract Contemporaneous MCDM methodology is based on the simplifying assumption of independent objectives. This restriction was partly relaxed through the concept of (static) interdependent objectives introduced by C. Carlsson and R. Fuller. Most real world managerial decision problems involve interdependent objectives, yet in a temporal setting. In the paper we generalize the static concept to temporal fuzzy multiobjective programming problem. We introduce the concepts temporal support and temporal conflict in the objective set within both infinite and finite planning horizons. We also formulate dynamic versions of membership functions for interdependent objectives in crisp and fuzzy multiobjective programming problems. The new concepts are used to describe and model temporal goal conflicts in numerical illustrations.