Dinesh R. Pai

Software Effort Estimation Using a Neural Network Ensemble

Accurate software effort estimation is crucial for software consulting organizations to stay competitive in their software development costs and retain customers. Artificial Neural Network (ANN) is an effective tool to obtain accurate effort estimates. In this paper, software effort estimation models using Artificial Neural Network (ANN) ensembles and regression analysis are developed based on data collected from 163 software development projects. The main emphasis of the paper is in developing an effective experimental design to achieve superior effort estimation results. In addition, we compare the software effort estimation of ANNs and multiple regression analysis. We found two interesting results. First, variables other than size (function points) are not especially helpful in predicting software development effort. Second, a properly designed ANN ensemble significantly outperforms estimation using regression analysis and can achieve better effort estimate predictions.

Experimental comparison of parametric, non-parametric, and hybrid multigroup classification

Highlights? The proposed hybrid method dominates almost all the other methods on classification performance. ? Logistic regression and neural network provides worst relative performance under most scenarios. ? This shows that the data complexities have adverse impact on the multinomial logistic regression. ? The results indicate that all classification methods are adversely affected by the nonstatic data. ? This study demonstrates the effectiveness of the hybrid method in improving classification accuracy. This study evaluates the relative performance of some well-known classification techniques, as well as a proposed hybrid method. The proposed hybrid method is a combination of k-nearest neighbor (kNN) and linear programming (LP) method for four group classification. Computational experiments are conducted to evaluate the performances of these classification techniques. Monte Carlo simulation is used to generate dataset with varying characteristics such as multicollinearity, nonlinearity, etc. for the experiments. The experimental results indicate that LP approaches, in general, and the proposed hybrid method, in particular, consistently have lower misclassification rates for most data characteristics. Furthermore, the hybrid method utilizes the strengths of both methods - k-NN and linear programming - resulting in considerable improvement in the classification accuracy. The results of this study can aid in the design of various hybrid techniques that combine the strengths of different methods to improve classification accuracy and reliability.

The US motor carrier industry: estimation of operational efficiency using DEA

In this paper, a performance analysis of the US motor carriers is carried out from the perspectives of a shipper and a motor carrier. Various efficiency measures are estimated using Data Envelopment Analyses (DEAs) with publicly available financial data on a representative sample of 21 publicly listed motor carriers. The analysis reveals that the average efficiencies of US motor carriers have gradually increased during the study period; however, a majority of the carriers are scale-inefficient, which demonstrates potential savings through benchmarking input targets.

Forecasting new adoptions: A comparative evaluation of three techniques of parameter estimation

Forecasting sales for an innovation before the products introduction is a necessary but difficult task. Forecasting is a crucial analytic tool when assessing the business case for internal or external investments in new technologies. For early stage investments or internal business cases for new products, it is essential to have some understanding of the likely diffusion of the technology. Diffusion of innovation models are important tools for effectively assessing the merits of investing in technologies that are new or novel and do not have prima facie, predictable patterns of user uptake. Most new product forecasting models require the estimation of parameters for use in the models. In this chapter, we evaluate three techniques to determine the parameters of the Bass diffusion model by using an example of a new movie.

Short-Term Predictions of the Total Medical Costs of California Counties

The aim of this research is to develop a model to forecast short-term health cost changes. The motivation for producing such a model is to provide local decision makers with a tool to predict short-term health-care costs in their localities. In order to achieve this objective, we collected data on total health-care expenditures and demographic data for California counties from 2000 to 2007. We then used various statistical methods to better understand the data and developed a regression model. Each years prediction model was then used to forecast the following years total health-care expenditure. The model developed adequately predicted health-care costs for the years on which the model was developed (2000–2006), and adequately forecast health-care costs for the holdout year, 2007. The average adjusted R2 value was 0.57, with an average mean absolute deviation score of 34. The best predictors of total health-care expenditures were county population, the number of county health-care facilities, and county per capita personal income. The practical implications of the model are that it will provide public and private decision makers with a useful tool for forecasting short-term demand for health-care services, enabling better planning for health-care manpower, facility planning, and financial planning needs. The contribution of this paper contrasts with the earlier work in that it supports short-term operational, not strategic, planning needs. The papers limitation is that it relies on data from one state. It should be tested in other, dissimilar, areas of the United States.

Analyzing the balancing of error rates for multi-group classification

This paper reports the relative performance of an experimental comparison of some well-known classification techniques such as classical statistical, artificial intelligence, mathematical programming (MP), and hybrid approaches. In particular, we examine the four-group, three-variable problem and the associated error rates for the four groups when each of the models is applied to various sets of simulated data. The data had varying characteristics such as multicollinearity, nonlinearity, sample proportions, etc. We concentrate on individual error rates for the four groups, i.e., we count the number of group 1 values classified into group 2, group 3, and group 4 and vice versa. The results indicate that in general not only are MP, k-NN, and hybrid approaches relatively better at overall classification but they also provide a much better balance between error rates for the top customer groups. The results also indicate that the MP and hybrid approaches provide relatively higher and stable classification accuracy under all the data characteristics.

Bankruptcy Prediction in Retail Industry Using Logistic Regression

In view of the failure of high-profile companies such as Circuit City and Linens n Things, Financial distress or bankruptcy prediction of retail and other firms has generated much interest recently. Recent economic conditions have led to predictions of a wave of retail bankruptcies (e.g., McCracken and O’Connell, 2009). This research develops and tests a model for the prediction of bankruptcy of retail firms. We use accounting variables such as inventories, liabilities, receivables, net income (loss), and revenue. Some guiding discriminate rule is given, and a few factors were identified as measures of a profitable company.

Productivity in the US Telecommunications Industry: A DEA Approach

Abstract This chapter develops a productivity analysis of the US telecommunications industry using a data envelopment analysis (DEA) approach. The study concerns itself with eight telecommunications companies. Output variables used are market price, return on equity, and debt equity ratio. The input variables are sales to profit, return on equity, and debt ratio to capital.

An Examination of Determinants of Software Testing and Project Management Effort

ABSTRACT Software estimation research has primarily focused on software effort involved in direct software development. As more and more organizations buy instead of building software, more effort is spent on software testing and project management. In this empirical study, the effect of program duration, computer platform, and software development tool (SDT) on program testing effort and project management effort is studied. The study results point to program duration and software tool as significant determinants of testing and management effort. Computer platform, however, does not have an effect on testing and management effort. Furthermore, the mean testing effort for third generation (3G) development environment was significantly higher than the mean testing effort for fourth generation (4G) environments that used IDE. In addition, the management effort for 4G environment projects without the use of IDE was lower than nonprogramming report generation projects.

Does efficiency and quality of care affect hospital closures

Abstract In recent decades, a large number of hospitals in Pennsylvania and across the United States have been forced to close entirely, or to transform their beds for alternative uses including outpatient care. Hospital closures have severe repercussions for the stakeholders. A better understanding of hospital closures could help take corrective measures to alleviate the adverse impact closures have on communities. Using Pennsylvania Department of Health data compiled from various sources, we address the following questions: Are less efficient hospitals less likely to survive in the long run? What are the effects of quality of care on hospital closures? Does teaching status and location (urban or rural) have any impact on the probability of hospital closure? The result demonstrates several factors of varying significance affect hospital closures/survivals. Hospitals with higher ratio of registered nurses per bed, higher operating margin, lower percentage of revenues from Medicare and Medicaid, and lower competition were less likely to close. Efficiency measures such as DEA efficiency, cost per patient day, and cost per discharge were not found to have a significant impact on hospital closures. The results suggest that hospital administrators may focus more on quality of care and less on cost reduction and efficiency.