×

Economic phenomenon predicting and analyzing system using neural network

  • US 5,444,819 A
  • Filed: 04/27/1993
  • Issued: 08/22/1995
  • Est. Priority Date: 06/08/1992
  • Status: Expired due to Fees
First Claim
Patent Images

1. An economic phenomenon predicting system comprising:

  • (a) a predicting neural network comprising (a1) an input layer having a plurality of input layer neurons for inputting signals indicating respective economic phenomena, (a2) a predetermined number of hidden layers comprising a plurality of hidden layer neurons, respectively, each of said input neurons making synaptic combinations with arbitrary ones of said hidden layer neurons, each synaptic combination having a weight, and (a3) an output layer comprising a predetermined number of output layer neurons, each of said hidden layer neurons making synaptic combinations with arbitrary ones of said output layer neurons, each synaptic combination having a weight, and each of said output layer neurons outputting an output signal, and (a4) wherein the weights of the synaptic combinations between said input layer neurons and hidden layer neurons and the weights of the synaptic combinations between said hidden layer neurons and said output layer neurons are organized by learning so that when data indicating variations of various principal economic indices including an economic phenomenon to be predicted and data indicating a variation pattern of the economic phenomenon to be predicted are input to said input layer neurons, said signals output from said output layer neurons represent a result of prediction of the economic phenomena;

    (b) moving-average-value arithmetic means for inputting time series data indicating various principal economic indices including the economic phenomenon to be predicted and for obtaining moving-average values for a plurality of recent predetermined periods, said moving-average-value arithmetic means supplying said obtained moving-average values, as part of said data indicating the variations of the various principal economic indices including the economic phenomenon to be predicted, to said input layer neurons;

    (c) difference arithmetic means for obtaining difference at least a first to an n-th order between said moving-average values, said difference arithmetic means supplying the obtained differences, as part of said data indicating the variations of said various principal economic indices including the economic phenomenon to be predicted;

    (d) trend removing means for removing trends from the time series data indicating the economic phenomenon to be predicted by subtracting from said time series data indicting the economic phenomenon to be predicted the individual moving-average value of the economic phenomenon to be predicted for any of the plurality of predetermined periods; and

    (e) pattern-sorting means for sorting said time series data indicating the economic phenomenon to be predicted after removing the trends into patterns, said pattern-sorting means outputting patterns, said patterns being obtained from the sorting, as data indicating a variation pattern of the economic phenomenon to be predicted to said input layer neurons.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×