Search

Computational studies on artificial neural networks under changing structural design and data distributions are discussed. Researchers investigate a two-group classification problem.

We built a multi-agent information system (MAIS) called Sherpa using distributed artificial intelligence architecture. The system integrates distributed knowledge sources and information to help the Wisconsin Division of Narcotics Enforcement [...]

Labor cost is one of the major contributors of software development cost. Among the variables that affects labor cost is the software development team size. There are very few methods available in literature to determine software development team size because team size selection happens during early stages of software development, and varies throughout systems development cycle. Under such circumstances, only methods that can be used to predict team size are Bayesian and analytical methods. In this paper, we use both Bayesian and analytical methods to predict team size. Specifically, we use a hybrid Bayesian network and simulation methodology for estimating posterior distributions of the team size using a real-world software engineering dataset, and a Cobb-Douglas function to estimate optimal team size. Using the leave-one-out sampling, we test our Bayesian approach and find that our approach predicts appropriate team size category with over 90% accuracy. However, our tests with optimal team size indicate that less than 20% of real-world software projects use optimal team size. [ABSTRACT FROM AUTHOR]

Abstract: We propose a bisection method for varying classification threshold value for cost sensitive neural network learning. Using simulated data and different misclassification cost asymmetries, we test the proposed threshold varying bisection method and compare it with the traditional fixed-threshold method based neural network and a probabilistic neural network. The results of our experiments illustrate that the proposed threshold varying bisection method performs better than the traditional fixed-threshold method based neural network. However, when compared to probabilistic neural network, the proposed method works well only when the misclassification cost asymmetries are low. [Copyright &y& Elsevier]

We compare log maximum likelihood gradient ascent, root-mean-square error minimizing gradient descent and genetic-algorithm-based artificial neural network procedures for a binary classification problem. We use simulated data and real-world data sets, and four different performance metrics of correct classification, sensitivity, specificity and reliability for our comparisons. Our experiments indicate that a genetic-algorithm-based artificial neural network that maximizes the total number of correct classifications generally fares well for the binary classification problem. However, if the training data set contains inconsistent decisions or noise then the log maximum likelihood maximizing gradient ascent may be the best classification approach to use. The root-mean-square minimizing gradient descent approach appears to overfit training data and has the lowest reliability among the approaches considered for our research. At the end of the paper, we provide a few guidelines, including computational complexity, for selection of an appropriate technique for a given binary classification problem. [ABSTRACT FROM AUTHOR]

In this paper, we describe an information technology capital budgeting (ITCB) problem, show that the ITCB problem can be modeled as a 0–1 knapsack optimization problem, and propose two different simulated annealing (SA) heuristic solution procedures to solve the ITCB problem. Using several simulations, we empirically compare the performance of two SA heuristic procedures with the performance of two well-known ranking methods for capital budgeting. Our results indicate that the information technology (IT) investments selected using the SA heuristics have higher after-tax profits than the IT investments selected using the two ranking methods. [ABSTRACT FROM AUTHOR]

In this paper, we describe an approach for building a hybrid Bayesian network-based multi-agent system for drug crime knowledge management. We use distributed artificial intelligence architecture to create a multi-agent information system that integrates distributed knowledge sources and information to aid decision-making. Our comparison of the hybrid system with a previously developed stand-alone expert system Sherpa, which was in use at a large drug crime investigation facility, shows that the current system compares similar to the existing system in terms of efficiency and effectiveness of knowledge management. We illustrate how the proposed hybrid bayesian network-based can be implemented in the distributed computing network environment. [ABSTRACT FROM AUTHOR]

This paper provides a comparative study of machine learning techniques for two-group discrimination. Simulated data is used to examine how the different learning techniques perform with respect to certain data distribution characteristics. Both linear and nonlinear discrimination methods are considered. The data has been previously used in the comparative evaluation of a number of techniques and helps relate our findings across a range of discrimination techniques. [ABSTRACT FROM AUTHOR]