Dennis Ridley

Moving-window spectral neural-network feedforward process control

Unlike reactive feedback control, feedforward control is a proactive method by which information about a measurable disturbance is fed, ahead of time, to the manipulated inputs of a process, the output of which is to be controlled, so as to counteract the effect of the disturbance. Discretized observations on the process variable are indexed to form a time series. A time-series model is fitted to the series. The ultrahigh signal-to-noise ratio fitted values are examined by a neural network, for patterns which detect when the future process is expected to become out of control. The neural-network diagnosis forms the basis for corrective action, prior to the process becoming out of control. In principle, this goes beyond SPC to achieve a process which is never actually out of control.

The univariate moving window spectral method

A major difficulty encountered in time series analysis is the bias in model parameter estimates, resulting in multiple-period lead time forecast error divergence. An approach, which mitigates the effect of this bias, is described. The spectral approach offers the potential for better estimation of cyclical components in time series. When recombined by the moving window spectral (MWS) paradigm, better long range forecast, are possible. Illustration is by comparisons to 24 other models, applied to complex non-linear multiple component time series, and 111 empirical time series. The MWS method requires the least user expertise, it explains, and it forecasts the time series the best. It is applicable to a broad range of time series associated with the physical, economic, and social sciences.

The predictive ability of accounting operating cash flows: a moving window spectral analysis

In this paper, evidence is provided on the predictive ability of quarterly operating Cash Flows (CFs). The inability of creditors and investors to anticipate future CFs based on historical CFs, with any degree of accuracy, may suggest that historical forecasting models are underspecified. Indeed, the discontinuities, variability, seasonality and trend in CF data may require additional, and as of yet, undisclosed variables, to enhance the predictability of extant forecasting models. In this study, Moving Window Spectral (MWS) analysis, a frequency domain approach, is applied to accounting time series data for the first time in an effort to assess the predictability of aggregate operating CFs. This method is adopted due to its ability to capture trend and multiple cyclical components in the data. Our results show that CFs can be reliably predicted using aggregate data on a firm-by-firm basis. In addition, our results outperform the results previously reported in the accounting literature. This research provides insight into the properties of accounting time series data not possible from a strictly time domain analysis. The implications of this and other findings for accounting and auditing are discussed.

Complementary antithetic weights for lognormal time-series forecasting

Abstract A simplified antithetic time-series model is presented. Two independent combining weights are replaced by a single weight and its complement. The number of computations and the time required to perform them are thereby reduced. Scope and purpose The fitting of a time-series model to serially correlated data often results in model parameters that are biased. The fitted values obtained from such a model are in turn biased. This is quite often unavoidable. Antithetic time-series modeling is a process for removing bias from the fitted values. The process may be retrofitted to any time-series model.