Economic phenomenon predicting and analyzing system using neural network
First Claim
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.
1 Assignment
0 Petitions
Accused Products
Abstract
An economic phenomenon predicting and/or analyzing system using a neural network. In the disclosed system, time series data indicating economic phenomena are input to preparation modules, and moving-average values and their differences are generated. One of the preparation modules performs a predetermined process over the time series data indicating an economic phenomenon, i.e. the change of TOPIX, to remove trends. A pattern sorter sorts the trend-free data into a certain number of groups. Average values of various time series data, their differences and the result of pattern sorting are input to input layer neurons of the network. The network is provided in advance with learning information of the change of TOPIX in the past. The output of the output layer neurons will be a value of prediction of the change of TOPIX. For the output of hidden layer neurons, principal components are obtained by principal analysis modules. A correlation analysis module obtains a distribution of frequency of principal component rankings and analyzes the correlation between the explanation variants and the output of the neural network based on the obtained distribution of frequency.
-
Citations
34 Claims
-
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 Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11)
-
-
12. An economic phenomenon predicting and analyzing system comprising:
-
(a) a neural network for predicting economic phenomenon, 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 having respective 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 having a predetermined number of output layer neurons, each of said hidden layer neurons making a synaptic combinations with arbitrary ones of said output layer neurons, each synaptic combination having a weight, and each of said output layer neurons providing an output signal, wherein the weights of the synoptic combinations between said input layer neurons and hidden layer neurons and the weights of the synoptic 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 phenomenon; (b) moving-average-value arithmetic means for inputting time series data indicating various principal economic indices 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 a first to an n-th order difference between said moving-average values, said difference arithmetic means outputting 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) principal component arithmetic means for obtaining principal components of the output of said hidden layer neurons by principal analysis; and (e) correlation analyzing means for analyzing a correlation between variation of the economic phenomenon to be predicted and variation of the output of said output layer neurons by analyzing the obtained principal components. - View Dependent Claims (13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30)
-
-
31. An economic phenomenon predicting and analyzing method comprising the steps of:
-
(a) obtaining moving-average values of time series data indicating various principal economic indices for a plurality of recent predetermined periods, each of said various principal economic indices including an economic phenomenon to be predicted; (b) obtaining differences between said moving-average values, which are obtained in an individual common period, for at least a first to an n-th order difference; (c) removing trends from time series data indicating the economic phenomenon to be predicted by subtracting from time series data indicating the economic phenomenon to be predicted an individual moving-average value of the economic phenomenon for any of the predetermined periods; (d) sorting said time series data indicating the economic phenomenon to be predicted, after removing trends, into patterns; and (e) predicting the economic phenomenon, based on said moving-average values, said differences and said sorting, using a neural network including (e1) an input layer comprising a plurality of input layer neurons for inputting said moving-average values and differences, and for inputting data indicating the variations of various principal economic indices including the economic phenomenon to be predicted and said patterns as data indicating a variation patterns of the economic phenomenon, (e2) a predetermined number of hidden layers comprising a plurality 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 (e3) 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 generating an output signal representing a result of prediction of the economic phenomenon, (e4) wherein weights of the synaptic combinations between said input layer neurons and hidden layer neurons and weights of the synaptic combinations between said hidden layer neurons and said output layer neurons are organized by learning in such a manner that when a plurality of data indicating variations of the various principal economic indices and a plurality of data indicating the variation patterns of the economic phenomenon are input to said input layer neurons, said signals output from said output layer neurons represent the result of prediction of the economic phenomenon.
-
-
32. An economic phenomenon predicting and analyzing method comprising the steps of:
-
(a) obtaining moving-average values of time series data indicating various principal economic indices for a plurality of recent predetermined periods, each of said various principal economic indices including an economic phenomenon to be predicted; (b) obtaining differences between said moving-average values, which are obtained in the individual common period, for at least a first to an n-th order difference; (c) predicting the economic phenomenon, based on said moving-average values and differences, using a neural network including (c1) an input layer having a plurality of input layer neurons for inputting said moving-average values and differences, data indicating the variations of various principal economic indices including the economic phenomenon to be predicted, (c2) a predetermined number of hidden layers having a plurality of hidden layer neurons, respectively, each of said input neurons making synaptic combination with arbitrary ones of said hidden layer neurons, each synaptic combination having a weight, and (c3) an output layer having a predetermined number of output layer neurons, each of said hidden layer neurons making a synaptic combinations with arbitrary one of said output layer neurons, each of said output layer neurons generating an output signal representing the result of prediction of the economic phenomenon, and each synaptic combination having a weight, (c4) wherein weights of the synaptic combinations between said input layer neurons and hidden layer neurons and weights of the synaptic combinations between said hidden layer neurons and said output layer neurons are organized by learning in such a manner that when a plurality of data indicating the variations of the various principal economic indices and a plurality of data indicating the variation pattern of the economic phenomenon are input to said input layer neurons, said signals output from said output layer neurons represent a result of prediction of the economic phenomenon; (d) obtaining a plurality of principal components by a principal component analysis for the output of said hidden layer neurons; and (e) analyzing a correlation between the variation pattern of economic phenomenon and the variation of output of said output layer neurons by analyzing the obtained principal components.
-
-
33. An economic phenomenon predicting system comprising:
-
(a) a predetermined number of neural networks organized to output a result of prediction of an economic phenomenon when data indicating variations of various principal economic indices including the economic phenomenon and data indicating a variation pattern of the economic phenomenon are provided as input; (b) moving-average-value arithmetic means for obtaining moving-average values of time series data indicating various principal economic indices 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 networks; (c) difference arithmetic means for obtaining a plurality of differences between said moving-average values obtained in an individual common period, for at least from a first to an n-th difference, 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, to said networks; (d) trend removing means for removing a plurality of trends from the time series data indicating the economic phenomenon to be predicted by subtracting from said time series data an individual moving-average value of the economic phenomenon for any of the predetermined periods; and (e) pattern-sorting means for sorting said time series data indicating the economic phenomenon to be predicted into patterns, said pattern-sorting means generating output patterns, which are obtained from the sorting, as data indicating variation patterns of the economic phenomenon to said networks.
-
-
34. An economic phenomenon predicting and analyzing system comprising:
-
(a) a predetermined numbers of neural networks that generates a number of hidden layer outputs according to both data indicating variations of various principal indices, including an economic phenomenon to be predicted, and data indicating a variation pattern of the economic phenomenon, and so as to output signals indicating the economic phenomenon by combining said hidden layer outputs; (b) moving-average-value arithmetic means for obtaining moving-average values of inputted time series data indicating various principal economic indices 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 networks; (c) difference arithmetic means for obtaining a difference between said moving-average values, Which are obtained in an individual common period, for at least from a first to an n-th order difference, said difference arithmetic means including means for 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, to said networks; (d) principal component arithmetic means for obtaining principal components of the hidden layer outputs by principal analysis; and (e) correlation analyzing means for analyzing a correlation between variation of the economic phenomenon to be predicted and variation of the output of said output layer neurons by analyzing the obtained principal components.
-
Specification