Development of a recurrent Sigma-Pi neural network rainfall forecasting system in Hong Kong

At the moment, weather forecasting is still an art — the experience and intuition of forecasters play a significant role in determining the quality of forecasting. This paper describes the development of a new approach to rainfall forecasting using neural networks. It deals with the extraction of information from radar images and an evaluation of past rain gauge records to provide shortterm rainfall forecasting. All of the meteorological data were provided by the Royal Observatory of Hong Kong (ROHK). Preprocessing procedures were essential for this neural network rainfall forecasting. The forecast of the rainfall was performed every half an hour so that a storm warning signal can be delivered to the public in advance. The network architecture is based on a recurrent Sigma-Pi network. The results are very promising, and this neural-based rainfall forecasting system is capable of providing a rain storm warning signal to the Hong Kong public one hour ahead.     

CMV colitis masquerading as colon cancer--an unusual presentation of acquired immunodeficiency syndrome

To determine whether there is any correlation between sudden decrease in barometric pressure and onset of labor, a non-experimental, retrospective study at a 948-bed tertiary care hospital was done. Pregnant patients of 36 weeks gestation or more who presented with spontaneous onset of labor during the 48 hours surrounding the 12 occurrences of significant drop in barometric pressure in 1992 were included in the study. Significantly more occurrences of onset of labor were identified in the 24 hours after a drop in barometric pressure than were identified in the 24 hours prior to the drop in barometric pressure (P < 0.05). Therefore, the overall number of labor onsets increased in the 24 hours following a significant drop in barometric pressure.     

Power Distribution Fault Cause Identification With Imbalanced Data Using the Data Mining-Based Fuzzy Classification E-Algorithm

Power distribution systems have been significantly affected by many outage-causing events. Good fault cause identification can help expedite the restoration procedure and improve the system reliability. However, the data imbalance issue in many real-world data sets often degrades the fault cause identification performance. In this paper, the E-algorithm, which is extended from the fuzzy classification algorithm by Ishibuchi to alleviate the effect of imbalanced data constitution, is applied to Duke Energy outage data for distribution fault cause identification. Three major outage causes (tree, animal, and lightning) are used as prototypes. The performance of E-algorithm on real-world imbalanced data is compared with artificial neural network. The results show that the E-algorithm can greatly improve the performance when the data are imbalanced     

A new method to estimate parameters of linear compartmental models using artificial neural networks

At present, the preferred tool for parameter estimation in compartmental analysis is an iterative procedure; weighted nonlinear regression. For a large number of applications, observed data can be fitted to sums of exponentials whose parameters are directly related to the rate constants/coefficients of the compartmental models. Since weighted nonlinear regression often has to be repeated for many different data sets, the process of fitting data from compartmental systems can be very time consuming. Furthermore the minimization routine often converges to a local (as opposed to global) minimum. In this paper, we examine the possibility of using artificial neural networks instead of weighted nonlinear regression in order to estimate model parameters. We train simple feed-forward neural networks to produce as outputs the parameter values of a given model when kinetic data are fed to the networks' input layer. The artificial neural networks produce unbiased estimates and are orders of magnitude faster than regression algorithms. At noise levels typical of many real applications, the neural networks are found to produce lower variance estimates than weighted nonlinear regression in the estimation of parameters from mono- and biexponential models. These results are primarily due to the inability of weighted nonlinear regression to converge. These results establish that artificial neural networks are powerful tools for estimating parameters for simple compartmental models.     

A behavioral medicine intervention in persons with HIV

This study used a pretest-posttest preexperimental design to examine the effect of a 10-week behavioral medicine support group intervention in a sample of persons with HIV. Using Solomon's psychoneuroimmunologic framework, the 10-week behavioral medicine program focused on the mind/body interaction, the relaxation response, coping with illness, hardiness, and nutrition. Pearson correlation coefficients and t tests were performed on the pre- and postintervention measures of hardiness, social support, immune function, and perceived health status. Results of the study indicated that hardiness (preintervention) and CD4 counts (pre- and postintervention) were significantly correlated with health status; however, CD4 counts decreased over the course of the behavioral medicine program. Implications for nursing and recommendations for further research are discussed.