Neural networks have been advocated as an alternative to traditional statistical forecasting methods. In the present experiment, time series forecasts produced by neural networks are compared with forecasts from six statistical time series methods generated in a major forecasting competition (Makridakis et al. [Makridakis, S., A. Anderson, R. Carbone, R. Fildes, M. Hibon, R. Lewandowski, J. Newton, E. Parzen, R. Winkler. 1982. The accuracy of extrapolation (time series) methods: Results of a forecasting competition. J. Forecasting 1 111--153.]); the traditional method forecasts were estimated by experts in the particular technique. The neural networks were estimated using the same ground rules as the competition. Across monthly and quarterly time series, the neural networks did significantly better than traditional methods. As suggested by theory, the neural networks were particularly effective for discontinuous time series.neural networks, time series, back propagation

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 to their statistical counterparts.We note the need for using the many mathematical proofs underlying artificial neural networks to determine the best conditions for using artificial neural networks in the forecasting and decision making.Key Words: Artificial Neural Networks, Regression, Forecasting, Decision Making, Time Series. ARTIFICIAL NEURAL NETWORK MODELS FOR FORECASTING AND DECISION MAKINGINTRODUCTION Over the last few decades, there has been much research directed at predicting the future and making better decisions. This research has led to many developments in forecasting methods. Most of these methodological advances have been based on statistical techniques. Currently, there is a new challenger for these methodologies -artificial neural networks.Artificial neural networks (ANN) have been widely touted as solving many forecasting and decision modeling problems (e.g., Hiew and Green, 1992). For example, they are argued to be able to model easily any type of parametric or non-parametric process and automatically and optimally transform the input data. These sorts of claims have led to much interest in artificial neural networks. On the other hand, Chatfield (1993) has queried whether artificial neural networks have been oversold or are just a fad.In this paper, we will attempt to give a balanced review of the literature comparing artificial neural networks and statistical techniques.Our review will be segmented into three different application areas: time series forecasting, regressionbased forecasting, and regression-based decision models. Additionally, we will note the literature comparing artificial neural networks and other models such as discriminant analysis. But before that review, we will first examine the general claims made for artificial neural networks that are relevant to forecasting and decision making.THE POTENTIAL OF ARTIFICIAL NEURAL NETWORKS Artificial neural networks are mathematical models inspired by the organization and functioning of biological neurons. There are numerous artificial neural network variations that are related to the nature of the task assigned to the network. There are also numerous variations in how the neuron is modeled. In some cases, these models correspond closely to biological neurons (e.g., Gluck and Bower, 1988;Granger et al., 1989) and in other cases the models depart from bio...

Abstract. The project upon which this paper is based is a qualitative study of the supervision [thesis advising] 1 of research students [graduate students] in departments of education and psychology in three British universities. Two models are apparent in the literature of supervision. The technical rationality model gives priority to issues of procedure or technique, while the negotiated order model conceptualizes supervision as a process open to negotiation and change. We look at supervisory style, reporting findings on the nature of tutorials [meetings] between supervisor and student, the extent of direction given by the supervisor to the project, and the nature of the interpersonal relationship between the parties. We also consider student strategies. Our findings suggest that although the technical rationality model has much to recommend it, a negotiated order model is a better description of what happens in practice.The project upon which this article is based is one of several linked qualitative studies on the research [graduate] ~ student experience in the social sciences. 2 Together these studies covered a range of subjects, regions, and institutions, although each had a somewhat different focus from the others. Our particular interest was in the supervision [advising] of research students in education and psychology. Thesis supervision has been little studied, despite numerous testimonies to its critical importance and exceptional difficulty, e.g. 'the most complex and subtle form of teaching in which we engage ' (Brown and Atkins 1988, p. 115).In the UK, doctoral students and many master's degree students in the social sciences typically obtain their degrees 'by research' and do not follow coursework programs on the North American model, although there now tends to be some systematic research training provided in the first year) Graduate education remains relatively marginal in the British system (Becher 1993) and arts and social science students are unlikely to have either a large cohort of fellow students or a thesis committee. Consequently the student's relationship with a supervisor is a key element in his or her progress.A common concern in graduate education across countries is the number of students who fail to complete their dissertations (Blume 1986). In Britain, the Economic and Social Research Council (ESRC) brought about a dramatic improvement by prohibiting social science students with grants from studying in departments with poor completion rates. There are a number of reasons why such a policy might be successful. Improved supervision might be one of these reasons.

No abstract