Analyzing Divisia Rules Extracted from a Feedforward Neural Network

ship exists between the quantity of money and the general level of prices. Confidence This paper introduces a mechanism for in this relationship, expressed in terms of generating a series of rules that characterize long-run rates of money growth and inflation, the money-price relationship, defined as the along with an accumulation of evidence suprelationship between the rate of growth of porting a seemingly stable linear demand for the money supply and inflation. Divisia broad money aggregates, led the major cencomponent data is used to train a selection of tral banks of the world to accept monetary candidate feedforward neural networks. The targeting as the means of controlling inflaselected network is mined for rules, expressed tion. A specific measure of the rate of growth in human-readable and machine-executable of money stock, known as a monetary aggreform. The rule and network accuracy are gate, is derived from the various constituent compared, and expert commentary is made liquid liabilities of commercial and savings on the readability and reliability of the banks. For monetarists the ultimate policy extracted rule set. The ultimate goal of this goal of low inflation is achieved by keeping research is to produce rules that meaningfully the growth of the chosen aggregate within a and accurately describe inflation in terms of target range. the Divisia component dataset. Clearly, the foundations of the construction of monetary aggregates are well rooted in monetary aggregation theory and require