Seamus Hill

Fuzzy, Distributed, Instance Counting, And Default Artmap Neural Networks For Financial Diagnosis

This paper shows the potential of neural networks based on the Adaptive Resonance Theory as tools that generate warning signals when bankruptcy of a company is expected (bankruptcy prediction problem). Using that class of neural networks is still unexplored to date. We examined four of the most popular networks of the class — fuzzy, distributed, instance counting, and default ARTMAP. In order to illustrate their performance and to compare with other techniques, we used data, financial ratios, and experimental conditions identical to those published in previous studies. Our experiments show that two financial ratios provide highest discriminatory power of the model and ensure best prediction accuracy. We examined performance and validated results by exhaustive search of input variables, cross-validation, receiver operating characteristic analysis, and area under curve metric. We also did application-specific cost analysis. Our results show that distributed ARTMAP outperforms the other three models in general, but the fuzzy model is best performer for certain vigilance values and in the application-specific context. We also found that ARTMAP outperforms the most popular neural networks — multi-layer perceptrons and other statistical techniques applied to the same data.

Analysing the effects of combining fitness scaling and inversion in genetic algorithms

Genetic Algorithms in their original form as presented by Holland [10] included four operators selection, reproduction, mutation and inversion. Today most attention is given to selection, crossover and mutation, whereas inversion is rarely used. We compare the effectiveness of an inversion operator in a basic GA, and in a GA using fitness scaling. Results indicate that at higher levels of epistasis inversion is more useful in a basic GA than a GA with fitness scaling.

An analysis of multi-chromosome GAs in deceptive problems

This paper discusses a new approach to using GAs to solve deceptive fitness landscapes by incorporating mechanisms to control the convergence direction instead of simply increasing the population diversity. In order to overcome some of the difficulties that GAs face when searching deceptive landscapes, we introduce two new multi-chromosome genetic algorithms. These multi-chromosome genetic algorithms have been designed to accelerate the GAs search speed in more complicated deceptive problems by looking for a balance between diversity and convergence. Five different problems are used in testing to illustrate the usefulness of our proposed approaches. The results show that the lack of diversity is not the only reason that normal GAs have difficulty in solving deceptive problems but that convergence direction is also important.

Analysis of a triploid genetic algorithm over deceptive and epistatic landscapes

This paper examines the performance of a canonical genetic algorithm (CGA) against that of the triploid genetic algorithm (TGA) introduced in [14], over a number of well known deceptive landscapes and a series of NK landscapes in order to increase our understanding of the the TGAs ability to control convergence. The TGA incorporates a mechanism to control the convergence direction instead of simply increasing the population diversity. Results indicate that the TGA appears to have the highest level of difficulty in solving problems with a disordered pattern. While these problems seem to improve the CGAs performance, it has a negative affect on the performance of the TGA. However, the results illustrate that the TGA performs better on NK-like problems (i.e. the overlapped problems) and NK problems with higher levels of epistasis.

Insolvency Prediction of Irish Companies Using Backpropagation and Fuzzy ARTMAP Neural Networks

This study explores experimentally the potential of BPNNs and Fuzzy ARTMAP neural networks to predict insolvency of Irish firms. We used financial information for Irish companies for a period of six years, preprocessed properly in order to be used with neural networks. Prediction results show that with certain network parameters the Fuzzy ARTMAP model outperforms BPNN. It outperforms also self-organising feature maps as reported by other studies that use the same dataset. Accuracy of predictions was validated by ROC analysis, AUC metrics, and leave-one-out cross-validation.

Comparing Small Population Genetic Algorithms over Changing Landscapes.

This paper examines the performance and adaptability of a number of small population Genetic Algorithms (GAs) over a selection of dynamic landscapes. Much of the research in this area tends to focus on GA with relatively large populations for problem optimisation. However there is research, which suggests that GAs with smaller populations can also be effective over changing landscapes. This research compares the performance and adaptability of a number of these small population GA over changing landscapes. With small population GAs, convergence can occur quickly, which in turn affects the adaptability of a GA over dynamic landscapes. In this paper five GA variants using small population sizes are run over well-known unimodal and multimodal problems, which were tailored to produce dynamic landscapes. Adaptability within the population is considered a desirable feature for a GA to optimise a changing landscape and different methods are used to maintain a level of diversity within a population to avoid the problem of premature convergence, thereby allowing the GA population adapt to the dynamic nature of the search space. Initial results indicate that small population GAs can perform well in searching changing landscapes, with GAs which possess the ability to maintain diversity within the population, outperforming those that do not.

Diversifying Techniques & Neutrality in Genetic Algorithms

This paper examines the implicit maintenance of diversity w ithin a population through the inclusion of a layered genotype-phenotype map (GP-map) in a Genetic Algor ithm (GA), based on the principal of Neutral theory (Kimura, 1968). The paper compares a simple GA (SGA), incorporating a variety of diversifying techniques, to the multi-layered GA (MGA) as proposed by the auth ors. The MGA creates a neutral representation by including a layered GP-map based on the biological concep ts ofTranscriptionandTranslation. In standard GAs, each phenotype is represented by a distinct genotype. H owever by allowing a higher number of alleles to encode phenotypic information on the genotype, one can crea te a situation where a number of genotypes may represent the same phenotype. Through this process one can i ntroduce the idea of redundancy or neutrality into the representation. This representation allows for ad aptive mutation (hot spots) and silent mutation (cold spots). This combination enables the level of diversity to d ynamically adjust during the search, and directs the search towards closely related neutral sets. Previous work has shown that introducing this type of representation can be beneficial; in this paper we show how this represen tatio is useful at introducing and maintaining diversity. Here we compare the performance of the MGA agains t traditional diversifying techniques used in conjunction with a SGA over a fully deceptive changing lands cape.

Examining the impact of Neutral theory on Genetic Algorithm population evolution

This paper examines the introduction of neutrality as proposed by Kimura (Kimura, 1968) into the genotype-phenotype mapping of a Genetic Algorithm (GA). The paper looks at the evolution of both a simple GA (SGA) and a multi-layered GA (MGA) incorporating a layered genotype-phenotype mapping based on the biological concepts of Transcription and Translation. Previous research in comparing GAs often use performance statistics; in this paper an analysis of population dynamics is used for comparison. Results illustrate that the MGA populations evolution trajectory is quite different to that of the SGA population over dynamic landscapes and that the introduction of neutrality implicitly maintains genetic diversity within the population primarily through genetic drift in association with selection.

Altering the Granularity of Neutrality in a Multi-layered Genetic Algorithm

By adopting a basic interpretation of the biological processes of transcription and translation, the multilayered GA (MGA) introduces a genotype-phenotype mapping for a haploid genotype, which allows the granularity of the representation to be tuned. The paper examines the impact of altering the level of neutrality through changes in the granularity of the representation and compares the performance of a standard GA (SGA) to that of a number of multi-layered GAs, each with a different level of neutrality, over both static and changing environments. Initial results indicate that it appears advantageous to include a multi-layered, biologically motivated genotype-phenotype encoding over more difficult landscapes. The paper also introduces an interpretation of missense mutation, which operates within the genotype-phenotype map (GP-map). Results also suggest that this mutation strategy can assist in tracking the optimum over various landscapes.

Analysing the impact of dimensionality on diversity in a multi-layered Genotype-Phenotype mapped genetic algorithm

This paper examines the impact of changes in dimensionality on a multi-layered genotype-phenotype mapped GA. To gain an understanding of the impact we carry out a series of experiments on a number of well understood problems and compare the performance of a simple GA (SGA) to that of a multi-layered GA (MGA) to demonstrate their ability to search landscapes with varying degrees of difficulty due to changes in the dimensionality of each function. The paper also examines the impact of diversity maintenance in assisting the search and identifies the natural increase in diversity as the level of problem difficulty increases, as a result of the layered Genotype-Phenotype mapping. Initial results indicate that it may be advantageous to include a multi-layered genotype-phenotype mapping under certain circumstances.