Electrical computers: arithmetic processing and calculating – Electrical digital calculating computer – Particular function performed
Reexamination Certificate
2001-04-12
2004-08-03
Malzahn, David H. (Department: 2124)
Electrical computers: arithmetic processing and calculating
Electrical digital calculating computer
Particular function performed
C714S746000
Reexamination Certificate
active
06772185
ABSTRACT:
TECHNICAL FIELD
The present invention relates to a time-series prediction method and apparatus utilizing wavelet coefficient series.
BACKGROUND ART
Conventionally, time-series prediction has been performed in solely the time domain or in solely the frequency domain. Time-domain prediction is performed as shown in
FIG. 1
, and many models, such as a linear regression model, pertain to time-domain prediction. Frequency-domain prediction is performed as shown in
FIG. 2
, and a model for inferring the shape of a power spectrum pertains to frequency-domain prediction. These traditional methods have been studied widely and applied to actual time-series prediction.
DISCLOSURE OF THE INVENTION
However, in the above-described conventional methods, prediction accuracy decreases for a time series having a certain time configuration in solely a specific frequency band or for a composite time series having a different time configuration in each of different frequency bands. This is because prediction in solely the time domain or in solely the frequency domain cannot simultaneously handle information in the time domain and information in the frequency domain.
Incidentally, when a time series is subjected to wavelet expansion, the time series is represented in the form of the sum of frequency components localized in the time domain.
In view of the foregoing, an object of the present invention is to provide a time-series prediction method and apparatus utilizing wavelet coefficient series which can obtain a prediction value of an original time series through an operation of wavelet-transforming the time series to thereby decompose the time series into a plurality of time series which are band-restricted in the frequency domain, and then inversely transforming the individual prediction values of the plurality of time series of the respective frequency components.
To achieve the above object, the present invention provides the following.
[1] A time-series prediction method utilizing wavelet coefficient series comprising: (a) wavelet-transforming a time series by use of a wavelet transformation unit in order to decompose the time series into a plurality of time series which are band-restricted in the frequency domain; (b) predicting values of frequency components obtained as a result of decomposition, by use of corresponding prediction units; and (c) reconstructing the prediction values of the respective frequency components by use of an inverse wavelet transformation unit to thereby obtain a prediction value of the original time series.
[2] A time-series prediction method utilizing wavelet coefficient series as described in [1] above, characterized in that ordinary wavelet transformation is used in order to wavelet-transform the time series to thereby univocally decompose the time series into a plurality of time series of the respective frequency components.
[3] A time-series prediction method utilizing wavelet coefficient series as described in [1] above, characterized in that stationary wavelet transformation is used for decomposition of the time series in order to observe the frequency components of the time series while aligning their observation times.
[4] A time-series prediction apparatus utilizing wavelet coefficient series, comprising: (a) a wavelet transformation unit for decomposing a time series into a plurality of coefficient series of different frequency components; (b) prediction units for predicting values of the respective frequency components; and (c) an inverse wavelet transformation unit for reconstructing the prediction values of the respective frequency components to thereby obtain a prediction value of the original time series.
REFERENCES:
patent: 5420872 (1995-05-01), Hyodo et al.
patent: 6295624 (2001-09-01), Gillard
Aihara Kazuyuki
Masuda Naoki
Japan Science and Technology Corporation
Lorusso, Loud & Kelly
Malzahn David H.
LandOfFree
Time-series predicting method using wavelet number series... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Time-series predicting method using wavelet number series..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Time-series predicting method using wavelet number series... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3355131