電気工学および電子技術ジャーナル

Neuro-Fuzzy Approach Towards Technoeconomic Forecasting

Dolores De Groff, Mohammad Dabbas and Perambur Neelakanta

Proposed in this paper is a fuzzy inference engine (FIE) intended to ascertain ex ante forecast details on a dependent variable y, based on a set of ex post information gathered on y in technoeconomic contexts. The FIE constructed thereof conforms to an artificial neural network (ANN), and, the ANN outcome deduced yields the forecasting on the temporal evolution of y(t) in the ex ante time-frame (t) vis-à-vis a set of ex post data availed. The ex post data available is however, sparse and inadequate for robust forecasting. Therefore, its cardinality is first improved and sufficient number of such sets is obtained as pseudoreplicates via statistical bootstrapping. The test ANN then uses these pseudoreplicates as training inputs toward robust prediction/forecast schedules. Further, the pseudoreplicated sets are considered as overlapping and hence, fuzzy. Therefore, the test ANN adopted is relevant to a FIE realization. Real-world technoeconomic data set on ADSL sales-cum-facility details at a wire-center in a telecommunication company (telco) is used to test the efficacy of the FIE proposed and validate the forecasting method described.