Integrating Neural Networks and Semi-Markov Processes for Automated Knowledge Acquisition: An Application to Real-time Scheduling*

Recently, artificial neural networks (ANN) have gained attention as a promising modeling tool for building intelligent systems. A number of applications have been reported in areas varying from pattern recognition to bankruptcy prediction. In this paper, we present a creative methodology that integrates computer simulation, semi-Markov optimization, and ANN techniques for automated knowledge acquisition in real-time scheduling. The integrated approach focuses on the synergy between operations research and ANN in eliciting human knowledge, filtering inconsistent data, and building competent models capable of performing at the expert level. The new approach includes three main components. First, computer simulation is used to collect expert decisions. This step allows expert knowledge to be obtained in a non-intrusive way and minimizes the difficulties involved in interviewing experts, constructing repertory grids, or using other similar structures required for manual knowledge acquisition. The data collected from computer simulation are then optimized using a semi-Markov decision model to remove data redundancies, inconsistencies, and errors. Finally, the optimized data are used to build ANN-based expert systems. The integrated approach is evaluated by comparing it with the human expert and using ANN alone in the domain of real-time scheduling. The results indicate that ANN-based systems perform worse than human experts from whom the data were collected, but the integrated approach outperforms human experts and ANN models alone.     

Assessment of Cigarette Smoking and Smoking Susceptibility among Youth: Telephone Computer-Assisted Self-Interviews versus Computer-Assisted Telephone Interviews

This experimental study assesses the effect of two survey methods,telephone audio computer-assisted self-interviewing (T-ACASI)and computer-assisted telephone interviewing (CATI), on self-reportsof smoking behavior and smoking susceptibility among adolescents12–17 years of age in California. In T-ACASI, participantslisten to prerecorded, computer-controlled questions and respondby pressing the keypad on a touch-tone telephone. In CATI, interviewersadminister the questions and enter responses into a computer.Prior research suggests that youth may be more likely to reportsensitive behaviors in a self-administered survey like T-ACASIcompared to an interviewer-administered survey like CATI, dueto greater perceived confidentiality. Logistic regression analyseswere conducted on unweighted data, controlling for demographicdifferences. Adjusted estimates of current smoking (past 30days) were significantly greater in T-ACASI (8.3 percent) thanCATI (4.5 percent). Smoking susceptibility (i.e., lack of afirm commitment not to smoke among those who have never smoked)was also greater in T-ACASI (45.0 percent) than CATI (34.9 percent).In both surveys, social desirability response bias was negativelyassociated with smoking, which suggests that response bias wasproblematic for both modes. Many respondents reported that aparent was present during the interview (59.4 percent in CATI;42.0 percent in T-ACASI). In both surveys, parental presencewas negatively associated with smoking, which suggests thatthis factor could also contribute to underreporting. Applicationof sample weights to the data eliminated the survey mode effects;however, the CATI current smoking estimate (9.3 percent) fromthis study was significantly less than an estimate (14.2 percent)obtained from a self-administered, school-based survey conductedthe same year on California adolescents.     

An Analysis of Damage to U.S. Soybean Yields from 1982 Acid Deposition Levels

Experimental data were compiled for seven soybean varieties studied at four different locations in the U.S. for a period of one to three years. Three dose–response functions (linear, quadratic, and Weibull) were fitted, as appropriate, to the experimental data. Effects of acid deposition on soybean yields were estimated by combining variety-specific dose–response and production data. These calculations suggest that U.S. soybean production (2.2 billion bushels in 1982) would increase by 3.2 million bushels with a 10% reduction in 1982 hydrogen ion concentration levels in all counties where soybeans were grown, and by almost 70 million bushels with uniform improvement to pH 5.2. These estimates are based on many simplifying assumptions and are subject to several important sources of uncertainty explored in this study. The calculated changes are small in comparison with effects from other natural and anthropogenic stresses.