Timothy D. Hill

Artificial neural network models for forecasting and decision making

Abstract Some authors advocate artificial neural networks as a replacement for statistical forecasting and decision models; other authors are concerned that artificial neural networks might be oversold or just a fad. In this paper we review the literature comparing artificial neural networks and statistical models, particularly in regression-based forecasting, time series forecasting, and decision making. Our intention is to give a balanced assessment of the potential of artificial neural networks for forecasting and decision making models. We survey the literature and summarize several studies we have performed. Overall, the empirical studies find artificial neural networks comparable with their statistical counterparts. We note the need to consider the many mathematical proofs underlying artificial neural networks to determine the best conditions for their use in forecasting and decision making.

Time series forecasting using neural networks: should the data be deseasonalized first?

This research investigates whether prior statistical deseasonalization of data is necessary to produce more accurate neural network forecasts. Neural networks trained with deseasonalized data from Hill et al. (1996) were compared with neural networks estimated without prior deseasonalization. Both sets of neural networks produced forecasts for the 68 monthly time series from the M-competition (Makridakis et al., 1982). Results indicate that when there was seasonality in the time series, forecasts from neural networks estimated on deseasonalized data were significantly more accurate than the forecasts produced by neural networks that were estimated using data which were not deseasonalized. The mixed results from past studies may be due to inconsistent handling of seasonality. Our findings give guidance to both practitioners and researchers developing neural networks. Copyright

Heparin inhibits inositol trisphosphate-induced calcium release from permeabilized rat liver cells

Neoplastic rat liver epithelial (261B) cells made permeable by electroporation released 0.2-0.3 microM Ca2+ from intracellular stores in response to 0.5 microM Ins(1,4,5)P3 stimulation. This Ca2+ release response was found to be inhibited by heparin in a dose-dependent manner (Ki of 15 micrograms/ml). Two other glycosaminoglycans, chondroitin sulfate and hyaluronic acid, showed no inhibitory effect at doses as high as 0.2 mg/ml. Passive Ca2+ release, and sequestration of Ca2+ into intracellular storage sites by the action of Ca2+-ATPase were unaffected by heparin treatment. We conclude that the inhibitory action of heparin treatment on Ca2+ mobilization in permeabilized 261B cells is mediated through its interaction at the Ins(1,4,5)P3 receptor binding site.

Neural network models as an alternative to regression

Neural networks can provide several advantages over conventional regression models. They are claimed to possess the property to learn from a set of data without the need for a full specification of the decision model; they are believed to automatically provide any needed data transformations. They are also claimed to be able to see through noise and distortion. An empirical study evaluating the performance of neural network models on data generated from three known regression models is presented. The results of this study indicate that neural network models perform best under conditions of high noise and low sample size. With less noise or larger sample sizes, they become less competitive. However, in two of the three cases, the neural network models were able to maintain mean absolute percentage errors (MAPE) within 2% of those of the true model.<<ETX>>

Neural network models for forecast: a review

Neural networks are advocated as a replacement for statistical forecasting methods. The authors review the literature comparing neural networks and classical forecasting methods, particularly in causal forecasting, time series forecasting, and judgmental forecasting. They provide not only an overview and evaluation of the literature but also summarize several studies performed which address the typical criticisms of work in this area. Overall, the empirical studies find neural networks at least as good as their classical counterparts.<<ETX>>

The influence of glutathione depleting agents on human platelet function

Human platelets were dose- and time-dependently depleted of intracellular glutathione (GSH) by treatment with the chemical oxidizing agents diamide and N-ethylmaleimide (NEM), and by formation of chemical conjugates with 1-chloro-2,4-dinitrobenzene (CDNB) catalyzed by GSH-S-transferase. In addition to effects upon GSH, these agents also inhibited platelet GSH-peroxidase activity. The inhibitory effect of CDNB was selective for GSH-peroxidase, while diamide and NEM treatment caused inhibition of several other cytosolic enzymes tested. Arachidonic acid (AA) induced aggregation and secretion responses measured in platelets depleted of GSH by diamide and NEM were attenuated. In contrast, these platelet functions remained identical to control following GSH depletion by CDNB treatment, suggesting that GSH is not required for normal platelet aggregation or secretion. Effects of diamide and NEM apart from their action on GSH may account for the platelet dysfunction induced by these compounds.

Function approximation using backpropagation and general regression neural networks

The approximation capabilities of backpropagation (BP) neural networks and D. Spechts (1991) general regression neural network (GRNN) are compared using data generated from 14 functions under three levels of random noise. The results show that the BP approach provides significantly more accurate estimates than the GRNN approach, especially when the level of random noise in the data is low.<<ETX>>

Extracellular ATP mobilizes intracellular Ca2+ in T51B rat liver epithelial cells: A study involving single cell measurements

T51B rat liver epithelial cells were stimulated with extracellular ATP. Changes in cytoplasmic free Ca2+ concentration [( Ca2+]i) were measured by fura-2 both in a large population of cells on coverslips in a cuvette and in single cells in a microscopic system. Extracellular ATP evoked a prompt increase in [Ca2+]i in both the presence and absence of extracellular Ca2+, although the effect was less pronounced in the latter case. These findings indicate that at least part of the [Ca2+]i increase is due to mobilization of intracellularly bound calcium. Stimulation with ATP did not mobilize the total pool of intracellular releasable Ca2+, as evidenced from experiments where subsequent addition of ionomycin evoked a pronounced increase in [Ca2+]i in the absence of extracellular Ca2+. The effect of ATP was maintained at room temperature but was markedly impaired in the absence of continuous stirring of the buffer solution. In the absence of stirring, ATP had to be increased to the millimolar range in order to evoke a pronounced effect. Single cell measurements revealed a heterogenous Ca2+ response to ATP, with some cells failing to respond with a detectable increase in [Ca2+]i. The actual increase in [Ca2+]i was not uniform throughout the cytoplasm, but seemed to start in one part of the cell. Even if part of the [Ca2+]i increase might be accounted for by ATP promoting the hydrolysis of phosphatidylinositol 4,5-bisphosphate and thereby a generation of InsP3 and diacylglycerol, there was no initiation of DNA synthesis under the present conditions. Hence, extracellular growth factors exert either a quantitative difference in second messenger production or additional stimulatory effects by activating intracellular signal pathways beyond these represented by [Ca2+]i and protein kinase C.

Optimal telephone line allocation of voice ads system

Find-it is a computer driven, interactive voice response (IVR) system. A caller can dial into Find-It to retrieve information from a database or record his own advertisement. With Find-1~ a caller ean avoid the delay of putting advertisement in a newspaper. In additioq a caller can mod@ or withdraw her/his advertisement at anytime. A simulation study was eondueted based on the assumption that the calling distribution during the day will be similar to those of ASK 2000, a telephone inquiry system of the State of Hawaii. DMerent calling volumes were generated based on the calling volume of ASK 2000. A simulation model was developed for Find-It allowing us to study different calling volumes from 10’%o(60 calls per day) to 200’XO (1200 calls per day) of that of ASK 2000. Line configurations were allowed to change from 4 lines to 20 lines. The simulation model was implemented with the INSIGHT simulation package.

Platelet hypersensitivity induced by 1-chloro-2,4-dinitrobenzene, hydroperoxides and inhibition of lipoxygenase

Platelets administered 1-chloro-2,4-dinitrobenzene to deplete intracellular glutathione (GSH) and inhibit GSH-peroxidase responded with irreversible aggregation to low doses of arachidonic acid (AA) more rapidly than control cells. This increase in sensitivity was correlated to inhibition of GSH-peroxidase, and not with the depletion of GSH. Addition of hydrogen peroxide, 15-hydroperoxyeicosatetraenoic acid, or inhibition of the lipoxygenase metabolic pathway by 4,7,10,13-eicosatetraynoic acid also induced a hypersensitive aggregation response to AA. These results suggest that the three modes of treatment share a common mechanism of increasing AA metabolism to biologically active prostaglandins and thromboxane A2 through alterations in cyclooxygenase kinetics and available enzyme substrate.