Michael Y. Hu

Neural network forecasting of the British pound/US dollar exchange rate

Neural networks have successfully been used for exchange rate forecasting. However, due to a large number of parameters to be estimated empirically, it is not a simple task to select the appropriate neural network architecture for an exchange rate forecasting problem. Researchers often overlook the effect of neural network parameters on the performance of neural network forecasting. This paper examines the effects of the number of input and hidden nodes as well as the size of the training sample on the in-sample and out-of-sample performance. The British pound/US dollar is used for detailed examinations. It is found that neural networks outperform linear models, particularly when the forecast horizon is short. In addition, the number of input nodes has a greater impact on performance than the number of hidden nodes, while a larger number of observations do reduce forecast errors.

An empirical analysis of factors explaining foreign joint venture performance in China

Abstract Little empirical evidence exists to explain the success factors in international joint ventures in developing countries. This article fills a void in the literature by studying factors underlying the performance of foreign joint ventures in China. An intersection of 2,442 joint ventures registered with the Ministry of Foreign Economics Relations and Trade from 1979 to 1990 and 383 joint ventures honored by the China Association of Enterprises with Foreign Investment constitutes the sample of this study. Level of partner commitment, the number of joint venture partners, sociocultural distance among partners, product/industry characteristics, foreign control, and joint venture location in China are factors used to explain performance. Some of the prescribed relationships are supported by the empirical findings. With the exception of product/industry, control, and location, the factors used in this study are found to be significantly related to performance.

Foreign ownership in Chinese joint ventures: A transaction cost analysis

The transaction cost framework is used to identify the factors that influence the percent ownership of foreign partners in joint ventures with China. The ownership literature is extended and new empirical evidence of the theoretical predictions is presepted. Sociocultural distance, economic risks and industry-related factors are identified as potential factors determining the percent ownership. The findings support the prescribed relationships of the first two factors and not the third, probably due to the fact that old technologies are being transferred to China in a joint venture partnership.

Effect of data standardization on neural network training

Data transformation is a popular option in training neural networks. This study evaluates the effectiveness of two well-known transformation methods: linear transformation and statistical standardization. These two are referred to as data standardization. A carefully designed experiment is used in which data from two-group classification problems were trained by feedforward networks. Different kinds of classification problems, from relatively simple to hard, were generated. Other experimental factors include network architecture, sample size, and sample proportion of group 1 members. Three performance measurements for the effect of data standardization are employed. The results suggest that networks trained on standardized data yield better results in general, but the advantage diminishes as network and sample size become large. In other words, neural networks exhibit a self-scaling capability. In addition, impact of data standardization on the performance of training algorithm in terms of computation time and number of iterations is evaluated. The results indicate that, overall, data standardization slows down training. Finally, these results are illustrated with a data set obtained from the American Telephone and Telegraph Company.

An agent-based diffusion model with consumer and brand agents

Market members interact within a complex, adaptive system to effect adoption decisions and the resulting diffusion of innovations. Agent-based modeling (ABM) is a methodology well suited for simulating this system. It complements and extends econometric approaches by incorporating interactions among system members, and adaptation in the system, revealing emergent results. Since ABM allows study at the individual unit level, heterogeneity among system members is reflected and modeling at the brand level is possible. Here an ABM with consumer and brand agents is described. The brand and product diffusion curve output allows study of diffusion at micro and macro levels, respectively.

Estimation of posterior probabilities of consumer situational choices with neural network classifiers

Abstract This study shows how neural networks can be used to estimate the posterior probabilities in a consumer choice situation. We provide the theoretical basis for its use and illustrate the entire neural network modeling procedure with a situational choice data set from AT&T. Our findings supported the appropriateness of this application and clearly illustrate the nonlinear modeling capability of neural networks. The posterior probability estimates clearly add to the usefulness of the technique for marketing research.

Estimating breast cancer risks using neural networks

Breast cancer is one of the most important medical problems. In this paper, we report the results of using neural networks for breast cancer diagnosis. The theoretical advantage is that posterior probabilities of malignancy can be estimated directly, and coupled with resampling techniques such as the bootstrap, distributions of the probabilities can also be obtained. These allow a researcher much more insight into the variability of estimated probabilities. Another contribution is that we present an integrative approach to building neural network models. The issues of model selection, feature selection, and function approximation are discussed in some detail and illustrated with the application to breast cancer diagnosis.

Natural mortality and participation fatigue as potential biases in diary panels: Impact of some demographic factors and behavioral characteristics on systematic attrition

Abstract Although many researchers have studied factors contributing to attrition bias and sample nonrepresentativeness, there has been no attempt to identify the sources of attrition. We used AT&T data collected using the before-after with control group experimental design to identify, isolate, and measure panel attrition due to two sources: natural mortality and participation fatigue. We then conducted statistical tests and ran logistic regressions to identify some demographic factors and behavioral characteristics that significantly contributed to systematic attrition biases due to natural mortality and participation fatigue. Marital status and past mobility seem to matter in the case of natural mortality, whereas age and sex were found to selectively affect attrition due to participation fatigue. We also found that systematic attrition significantly affected sample representativeness, especially when participation fatigue is mostly involved. We then show that our results are largely consistent with previous findings. Finally, we conclude by outlining some of the operational implications of our findings and the steps that can be taken to overcome attrition bias and sample depletion.

Service The Key to Frequent-Guest Programs

Abstract Discriminant analyses indicate that members of hotel frequent-guest programs do, indeed, consider those programs in their hotel selection—but top-notch service is also essential

Modeling consumer situational choice of long distance communication with neural networks

This study shows how artificial neural networks can be used to model consumer choice. Our study focuses on two key issues in neural network modeling - model building and feature selection. Using the cross-validation approach, we address these two issues together and specifically examine the effectiveness of a backward feature selection algorithm for consumer situational choices of communication modes. Results indicate that the proposed heuristic for feature selection is robust with respect to validation sample variation. In fact, the feature selection approach produces the same best subset of features as the all-possible-subset approach.