WO2019028931A1 - 一种确定事件周期值的方法及装置 - Google Patents

一种确定事件周期值的方法及装置 Download PDF

Info

Publication number
WO2019028931A1
WO2019028931A1 PCT/CN2017/097772 CN2017097772W WO2019028931A1 WO 2019028931 A1 WO2019028931 A1 WO 2019028931A1 CN 2017097772 W CN2017097772 W CN 2017097772W WO 2019028931 A1 WO2019028931 A1 WO 2019028931A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
candidate period
determining
candidate
peak
Prior art date
Application number
PCT/CN2017/097772
Other languages
English (en)
French (fr)
Inventor
陈迅
Original Assignee
网宿科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 网宿科技股份有限公司 filed Critical 网宿科技股份有限公司
Priority to US16/061,802 priority Critical patent/US20210165850A1/en
Priority to EP17897206.3A priority patent/EP3460714A4/en
Publication of WO2019028931A1 publication Critical patent/WO2019028931A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G06F17/141—Discrete Fourier transforms
    • G06F17/142—Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B9/00—Measuring instruments characterised by the use of optical techniques
    • G01B9/02—Interferometers
    • G01B9/02083—Interferometers characterised by particular signal processing and presentation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B9/00—Measuring instruments characterised by the use of optical techniques
    • G01B9/02—Interferometers
    • G01B9/02083—Interferometers characterised by particular signal processing and presentation
    • G01B9/02084—Processing in the Fourier or frequency domain when not imaged in the frequency domain
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G06F17/141—Discrete Fourier transforms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/15—Correlation function computation including computation of convolution operations
    • G06F17/156—Correlation function computation including computation of convolution operations using a domain transform, e.g. Fourier transform, polynomial transform, number theoretic transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks

Definitions

  • the present invention relates to the field of data processing technologies, and in particular, to a method and apparatus for determining an event period value.
  • the period value at which an event occurs when determining the period value at which an event occurs, it can usually be implemented based on a Fourier transform periodogram method or by an autocorrelation algorithm.
  • the Fourier transform-based periodic graph method first performs a Fourier transform on the processed time series, and then squares the Fourier coefficients in the transformed result, and the maximum value in the obtained result corresponds to the period of the original time series. value.
  • the shortcoming of this method is that the detection accuracy of the large time span period is not good.
  • the autocorrelation algorithm finds periodicity by calculating the cross-correlation of the time series with itself at different points in time. The distance between the most significant set of peaks in the algorithm results is the periodicity of the original time series.
  • the limitation of the autocorrelation algorithm is that because there may be more than one set of peaks corresponding to the same period, it is difficult to use the computer to automatically determine important peaks in the output of the algorithm, so it is often necessary for the user to manually determine the most important in the resulting image. The set of peaks.
  • the purpose of the present application is to provide a method and apparatus for determining an event period value, which can improve the accuracy of determining the event period value when the event period value is automatically determined.
  • an aspect of the present application provides a method for determining an event period value, the method comprising: acquiring a time series of a target event, wherein the time series includes a preset number of sequence values, and the sequence value is used for Characterizing the number of times the target event occurs in a unit time; calculating an autocorrelation sequence of the time series, and determining a first candidate period corresponding to the autocorrelation sequence based on peak values and trough values in the autocorrelation sequence a set; wherein, the candidate period value in the first candidate period set is associated with a confidence; calculating a Fourier transform result of the time series, and based on an amplitude value of a frequency point in the Fourier transform result, Determining a second candidate period set corresponding to the Fourier transform result; wherein, the candidate period value in the second candidate period set is associated with a confidence; taking the first candidate period set and the second candidate a union of periods, determining a total value of confidence corresponding to each candidate period value in the
  • determining the first candidate period set corresponding to the autocorrelation sequence comprises: calculating a peak height corresponding to each wave peak in the autocorrelation sequence, and extracting from a peak value of the autocorrelation sequence based on the peak height Marking the peak value of the wave; calculating the time span between the peaks of any two adjacent marker waves, and counting the number of repetitions of each of the time spans; using the time span in which the number of repetitions meets the specified condition as the candidate period value of the autocorrelation sequence And adding the candidate period value to the first candidate period set.
  • calculating a peak height corresponding to each wave peak in the autocorrelation sequence includes: calculating, for the target wave peak in the autocorrelation sequence, a first between the target wave peak and the adjacent two wave valley values a difference value and a second difference value, and determining a smaller one of the first difference value and the second difference value as a peak height of the target wave peak value.
  • extracting the mark wave peak from the peak value of the autocorrelation sequence based on the peak height includes: comparing a peak height corresponding to each wave peak with a specified peak height threshold, and greater than or equal to the specified peak height The peak value corresponding to the peak height of the threshold is determined as the peak value of the marker wave.
  • the specified peak height threshold is subordinate to the threshold set; the threshold set includes at least two different specified peak height thresholds; wherein each of the specified peak height thresholds corresponds to a respective extracted marker wave peak .
  • the number of repetitions satisfying the specified condition includes: the number of repetitions in the calculated time span The proportion of the total number is greater than or equal to the specified ratio threshold.
  • a confidence level associated with the candidate period value in the first candidate period set is determined according to the ratio of the number of repetitions of the candidate period value to the total number of calculated time spans The confidence as the candidate period value is associated.
  • determining the second candidate period set corresponding to the Fourier transform result includes: randomly scrambling the order of the sequence values in the time series N times, obtaining N out-of-order time series, and calculating the out-of-order a time series corresponding to the out-of-order Fourier transform result; wherein N is an integer greater than or equal to 1; a maximum amplitude value is determined from amplitude values of respective frequency points in the out-of-order Fourier transform result; A target frequency point whose amplitude value is greater than or equal to the maximum amplitude value is determined in a Fourier transform result of the sequence, and a reciprocal of the target frequency point is used as a candidate period value in the second candidate period set.
  • a confidence level associated with a candidate period value in the second candidate period set is determined in a manner that a ratio of an amplitude value of the target frequency point to the maximum amplitude value is used as the target frequency The confidence associated with the candidate period value corresponding to the point.
  • determining the period value of the target event comprises determining one or more candidate period values whose total confidence value is greater than or equal to the specified threshold as the set of period values of the target event.
  • the method further includes: if the union is an empty set, determining that the target event does not have periodicity.
  • an aspect of the present application further provides an apparatus for determining an event period value, the apparatus comprising a processor and a memory, wherein the memory stores a computer program, when the computer program is executed by the processor And performing the following steps: acquiring a time series of the target event, where the time series includes a preset number of sequence values, the sequence value is used to represent the number of times the target event occurs in a unit time; and calculating the time series And determining, according to the peak value and the trough value in the autocorrelation sequence, a first candidate period set corresponding to the autocorrelation sequence; wherein, the candidate period value and the confidence in the first candidate period set Calculating a Fourier transform result of the time series, and determining a second candidate period set corresponding to the Fourier transform result based on the amplitude value of the frequency point in the Fourier transform result; wherein The candidate period value in the second candidate period set is associated with a confidence; the first candidate period set and the second candidate are taken And set period, and
  • the following steps are further implemented: calculating a peak height corresponding to each wave peak in the autocorrelation sequence, and based on the peak height, a wave from the autocorrelation sequence Extracting the peak value of the marker wave; calculating the time span between the peaks of any two adjacent marker waves, and counting the number of repetitions of each of the time spans; taking the time span of the number of repetitions satisfying the specified condition as the autocorrelation sequence A candidate period value is added to the first candidate period set.
  • the following steps are further implemented: randomly arranging the order of the sequence values in the time series N times, obtaining N out-of-order time series, and calculating the chaos An out-of-order Fourier transform result corresponding to the sequence time sequence; wherein N is an integer greater than or equal to 1; a maximum amplitude value is determined from amplitude values of respective frequency points in the out-of-order Fourier transform result; A target frequency point whose amplitude value is greater than or equal to the maximum amplitude value is determined in a Fourier transform result of the time series, and a reciprocal of the target frequency point is used as a candidate period value in the second candidate period set.
  • the present application combines the autocorrelation algorithm and the Fourier algorithm to determine the first candidate period set corresponding to the autocorrelation sequence by using the peak value and the trough value in the autocorrelation sequence, and further, through the Fourier transform.
  • the amplitude value of the frequency point in the result may determine a second candidate period set corresponding to the Fourier transform result.
  • each candidate period value may be included, and each candidate period value may be associated with a confidence level.
  • the union of the two sets can be taken. In this union, the total confidence value for each candidate period value can be calculated.
  • the total confidence value may be the sum of the confidences of the same candidate period value in the set of two candidate periods.
  • the period value of the target event can be determined based on the total confidence value. It can be seen from the above that by combining the two algorithms, the accuracy of the period value can be ensured to be higher, and the period value of the target event can be automatically determined by determining the total value of the confidence.
  • Embodiment 1 is a flowchart of a method for determining an event period value in Embodiment 1 of the present invention
  • FIG. 2 is a schematic diagram of determining a threshold set in the first embodiment of the present invention
  • Embodiment 3 is a functional block diagram of a device in Embodiment 2 of the present invention.
  • the present application provides a method of determining an event period value. Referring to FIG. 1, the method includes the following steps.
  • S1 Obtain a time series of target events, where the time series includes a preset number of sequence values, and the sequence values are used to represent the number of times the target event occurs in a unit time.
  • the number of times the target event to be analyzed occurs in a unit time can be counted.
  • the unit time can be flexibly adjusted according to actual application scenarios.
  • the unit time can be 1 hour, so that the number of times the target event occurs within an hour can be counted.
  • the time series of the target event can be generated.
  • the time series may include a preset number of sequence values, the preset number of sequence values may be used to characterize the number of times the target event occurs in a unit time, and each of the sequence values may be related to the unit Time corresponds.
  • the time sequence of the target event may be ⁇ 2, 3, 6, 8, 9 ⁇ , then each number in the time series may represent the number of times the target event occurred within a unit time. If the unit time is 1 hour, the statistical duration of each digit in the time series may be 1 hour.
  • S3 calculating an autocorrelation sequence of the time series, and determining, according to a peak value and a trough value in the autocorrelation sequence, a first candidate period set corresponding to the autocorrelation sequence; wherein the first candidate period set The candidate period value in is associated with the confidence.
  • the corresponding autocorrelation sequence can be calculated by an autocorrelation algorithm.
  • the formula for calculating the autocorrelation sequence is a technical means commonly used in the art, and will not be described here.
  • the autocorrelation sequence may be expressed as ⁇ R i ⁇ , where i has a value of 1 to m, and m is the total number of values in the autocorrelation sequence.
  • the values in the autocorrelation sequence may be in a state of up and down fluctuations, which As such, the values in the autocorrelation sequence may have peaks and troughs. Wave peaks and trough values may be extracted sequentially from the autocorrelation sequence, wherein the wave peaks are greater than two values adjacent thereto, the trough values being less than two values adjacent thereto.
  • the peak height corresponding to each wave peak in the autocorrelation sequence can be calculated.
  • the peak height may be determined by calculating a first difference between the target wave peak and the adjacent two valley values for the target wave peak in the autocorrelation sequence. And a second difference, and determining a smaller one of the first difference and the second difference as a peak height of the target wave peak. For example, if the peak value is 12 and the two valley values adjacent thereto are 3 and 7, respectively, then the first difference and the second difference are 9 and 5, respectively, then 5 can be used as a wave. Peak height of peak 12.
  • the marker wave peak value can be extracted from the peak value of the autocorrelation sequence based on the peak height. Specifically, the peak height corresponding to each wave peak may be compared with a specified peak height threshold, and the peak value corresponding to the peak height greater than or equal to the specified peak height threshold may be determined as the marker wave peak. For example, if the specified peak height threshold is 5, then the peak value of the peak height greater than or equal to 5 can be used as the marker wave peak.
  • the time span between the peaks of two adjacent marker waves may be the period value corresponding to the time series. Therefore, the time span between the peaks of any two adjacent marker waves can be calculated, and the number of repetitions of each of the time spans can be counted. The more the number of repetitions, the more likely the time span is to be the period value of the time series.
  • the calculated time span between any two adjacent marker wave peaks may be 3 hours, 4 hours, 3 hours, 3 hours, 3 hours, 4 hours, 3 hours, The 10 time spans of 6 hours, 4 hours, and 3 hours, so that the number of repetitions corresponding to the time span of 3 hours can be statistically calculated, and the number of repetitions corresponding to the time span of 3, 4 hours is 3, and the other time spans correspond.
  • the number of repetitions is 1.
  • a time span in which the number of repetitions satisfies a specified condition may be used as a candidate period value of the autocorrelation sequence, and the candidate period value may be added to the first candidate period set.
  • the fact that the number of repetitions meets the specified condition may mean that the proportion of the number of repetitions in the total number of calculated time spans is greater than or equal to the specified ratio threshold.
  • the specified ratio threshold is 50%
  • the ratio of the number of repetitions corresponding to the 3-hour time span to the total number is 60%, exceeding the specified ratio threshold, and other time spans.
  • the proportion of the corresponding number of repetitions in the total number does not exceed 50%, so 3 hours can be added as the candidate period value to the first In the candidate cycle set.
  • the number of candidate period values obtained by using only one specified peak height threshold may be relatively small.
  • a threshold set H may be provided in advance, in which the threshold set is At least two specified peak height thresholds can be included.
  • the marker peaks can be obtained in the manner described above.
  • each of the specified peak height thresholds corresponds to the respective extracted marker wave peak.
  • the candidate period value may be determined according to the peak value of the extracted marker wave, so that the candidate period value in the first candidate period set can be relatively complete.
  • the specified peak height threshold in the threshold set may have a value interval, and according to the value interval, each specified peak height threshold may be obtained according to a fixed step size.
  • the threshold set corresponds to a value interval of 0, and the fixed step size may be 2, so that three specified peak height thresholds of 2, 4, and 6 can be generated.
  • the associated confidence level may also be set for each candidate period value.
  • the confidence can be used to indicate the likelihood of a candidate period value as a true period value.
  • the proportion of the number of repetitions of the candidate period value in the total number of calculated time spans may be used as the confidence level of the candidate period value association.
  • the confidence level associated with the candidate period value of 3 hours is 0.6.
  • S5 calculating a Fourier transform result of the time series, and determining, according to the amplitude value of the frequency point in the Fourier transform result, a second candidate period set corresponding to the Fourier transform result; wherein The candidate period values in the second set of candidate periods are associated with a confidence level.
  • the second candidate period set may also be determined based on the method of Fourier transform.
  • the time series may be subjected to discrete Fourier transform to obtain a Fourier transform result corresponding to the time series.
  • the amplitude value corresponding to each frequency point can be displayed by means of a periodic diagram.
  • the larger the amplitude value the higher the confidence that the frequency point corresponding to the amplitude value can be used as the frequency value of the time series (that is, the reciprocal of the period value).
  • the order of the sequence values in the time series may be randomly scrambled in advance to obtain an out-of-order time series.
  • the random scrambling step may be performed N times, thereby obtaining N out-of-order time series, where N is an integer greater than or equal to 1.
  • each of the out-of-order time can be calculated by using a discrete Fourier transform formula.
  • the sequence corresponds to the result of the out-of-order Fourier transform, so that N out-of-order Fourier transform results can be obtained.
  • the maximum amplitude value can be determined.
  • the frequency point corresponding to the maximum amplitude value may represent a period value that the out-of-order time sequence may have after randomly scrambling the original time series. If the original time series has significant periodicity, the amplitude value of the frequency point corresponding to the periodicity should be greater than the maximum amplitude value obtained according to the out-of-order time series.
  • a target frequency point whose amplitude value is greater than or equal to the maximum amplitude value may be determined in the Fourier transform result of the time series, and the reciprocal of the target frequency point is used as a candidate in the second candidate period set.
  • Period value For example, if the maximum amplitude value determined in the result of the out-of-order Fourier transform is 15, then the frequency point of the Fourier transform result of the normal time series may be extracted from the frequency point with the amplitude value greater than or equal to 15, and The reciprocal of each extracted frequency point is calculated in turn as a candidate period value in the second candidate period set.
  • the associated confidence levels can be set for each candidate period value.
  • the ratio of the amplitude value of the target frequency point to the maximum amplitude value may be used as a confidence level associated with the candidate period value corresponding to the target frequency point. For example, if the maximum amplitude value is 15, and the amplitude value of one target frequency point is 30, then the confidence value associated with the candidate period value corresponding to the target frequency point may be 2.
  • the union of the two sets may be taken, thereby merging the candidate period values in the two sets.
  • some candidate period values may correspond to only one confidence level, and some candidate period values may correspond to two confidence levels.
  • the candidate period value corresponding to the two confidences is present in both sets, and the probability of being the true period value of the time series is higher. Therefore, the total value of the confidence corresponding to each candidate period value in the merged set may be determined, and the period value of the target event is determined based on the total value of the confidence.
  • the total confidence value may be the sum of the confidence levels in the two sets.
  • one or more candidate period values having a total confidence value greater than or equal to the specified threshold may be determined as the set of period values of the target event. It should be noted that the number of determined period values may be more than one, which indicates that there may be two different period values in the original time series, which is also common in actual scenarios.
  • the union is an empty set, it indicates that the first candidate period set and the second candidate period set are both empty sets, that is, there is no candidate period value, so that the target event may be determined not to be With periodicity.
  • the present application further provides an apparatus for determining an event period value, the apparatus comprising a processor 100 and a memory 200, wherein the memory 200 stores therein a computer program, when the computer program is used by the processor 100 When executed, implement the following steps:
  • Obtaining a time series of the target event where the time series includes a preset number of sequence values, and the sequence value is used to represent the number of times the target event occurs in a unit time;
  • a time span in which the number of repetitions satisfies the specified condition is taken as a candidate period value of the autocorrelation sequence, and the candidate period value is added to the first candidate period set.
  • N is an integer greater than or equal to 1;
  • a target frequency point whose amplitude value is greater than or equal to the maximum amplitude value is determined in the Fourier transform result of the time series, and a reciprocal of the target frequency point is used as a candidate period value in the second candidate period set.
  • the application can be described in the general context of computer-executable instructions executed by a computer, such as a program module.
  • program modules include routines, programs, objects, components, data structures, and the like that perform particular tasks or implement particular abstract data types.
  • the present application can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are connected through a communication network.
  • program modules can be located in both local and remote computer storage media including storage devices.
  • the present application combines the autocorrelation algorithm and the Fourier algorithm to determine the first candidate period set corresponding to the autocorrelation sequence by using the peak value and the trough value in the autocorrelation sequence, and further, through the Fourier transform.
  • the amplitude value of the frequency point in the result may determine a second candidate period set corresponding to the Fourier transform result.
  • each candidate period value may be included, and each candidate period value may be associated with a confidence level.
  • the union of the two sets can be taken. In this union, the total confidence value for each candidate period value can be calculated.
  • the total confidence value may be the sum of the confidences of the same candidate period value in the set of two candidate periods.
  • the period value of the target event can be determined based on the total confidence value. It can be seen from the above that by combining the two algorithms, the accuracy of the period value can be ensured to be higher, and the period value of the target event can be automatically determined by determining the total value of the confidence.
  • the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, ie may be located A place, or it can be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment. Those of ordinary skill in the art can understand and implement without deliberate labor.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Signal Processing (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Surgery (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Discrete Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Auxiliary Devices For Music (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本发明公开了一种确定事件周期值的方法及装置，其中，所述方法包括：获取目标事件的时间序列，所述时间序列中包括预设数量的序列值；计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。本申请提供的技术方案，能够在自动确定事件周期值的情况下，提高事件周期值的确定精度。

Description

一种确定事件周期值的方法及装置 技术领域
本发明涉及数据处理技术领域，特别涉及一种确定事件周期值的方法及装置。
背景技术
在自然社会中，许多事情都在遵循一定的周期发生。例如，部分地区的季节按照一年四季交替变化，周而复始。又例如，声波是物体进行周期性振动所产生的结果。此外，在人文社会中，许多人类从事的行业中，事件也往往遵循一定的周期发生。例如，银行会定期向储户支付存款利息。又例如，雇主会定期向员工发放薪水。
在实际生活中，周期性的事件通常会伴随着同类随机事件的发生，随机事件作为噪音，往往会影响对事件周期性的识别。因此，需要一种有效的算法，能够在存在噪音的情况下正确地识别出事件的周期性。
当前，在确定事件发生的周期值时，通常可以基于傅里叶变换的周期图方法或者通过自相关算法来实现。其中，基于傅里叶变换的周期图方法先对处理的时间序列进行傅里叶变换，再对变换结果中的傅里叶系数求平方，所得结果中的最大值就对应于原时间序列的周期值。这一方法的不足之处在于：对大时间跨度周期的检测准确度欠佳。
自相关算法通过计算时间序列与其自身在不同时间点的互相关来发现周期性。算法结果中最显著的一组波峰之间的距离即指示原时间序列的周期性。自相关算法的局限性在于：因为对应于同一周期的波峰可能不只一组，在算法的输出结果中使用计算机自动地确定重要的波峰难度较大，故常常需要用户在结果图像中手动确定最重要的那组波峰。
由上可见，当前亟需一种精度高并且能够自动确定事件周期值的方法。
发明内容
本申请的目的在于提供一种确定事件周期值的方法及装置，能够在自动确定事件周期值的情况下，提高事件周期值的确定精度。
为实现上述目的，本申请一方面提供一种确定事件周期值的方法，所述方法包括：获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数；计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联；计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联；取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
进一步地，确定所述自相关序列对应的第一候选周期集合包括：计算所述自相关序列中各个波峰值对应的峰高，并基于所述峰高从所述自相关序列的波峰值中提取标记波峰值；计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数；将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。
进一步地，计算所述自相关序列中各个波峰值对应的峰高包括：针对所述自相关序列中的目标波峰值，分别计算所述目标波峰值与相邻两个波谷值之间的第一差值和第二差值，并将所述第一差值和所述第二差值中的较小值确定为所述目标波峰值的峰高。
进一步地，基于所述峰高从所述自相关序列的波峰值中提取标记波峰值包括：将各个波峰值对应的峰高与指定峰高阈值进行对比，并将大于或者等于所述指定峰高阈值的峰高对应的波峰值确定为所述标记波峰值。
进一步地，所述指定峰高阈值从属于阈值集合中；所述阈值集合中包括至少两个不同的指定峰高阈值；其中，各个所述指定峰高阈值均与各自提取的标记波峰值相对应。
进一步地，重复次数满足指定条件包括：重复次数在计算出的时间跨度的 总数量中所占的比例大于或者等于指定比例阈值。
进一步地，按照下述方式确定与所述第一候选周期集合中的候选周期值相关联的置信度：将所述候选周期值的重复次数在计算出的时间跨度的总数量中所占的比例作为所述候选周期值关联的置信度。
进一步地，确定所述傅里叶变换结果对应的第二候选周期集合包括：随机打乱所述时间序列中序列值的排列顺序N次，得到N个乱序时间序列，并计算所述乱序时间序列对应的乱序傅里叶变换结果；其中，N为大于或者等于1的整数；从所述乱序傅里叶变换结果中各个频率点的振幅值中确定最大振幅值；在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。
进一步地，按照下述方式确定与所述第二候选周期集合中的候选周期值相关联的置信度：将所述目标频率点的振幅值与所述最大振幅值的比值作为与所述目标频率点对应的候选周期值相关联的置信度。
进一步地，确定所述目标事件的周期值包括：将置信度总值大于或者等于指定阈值的一个或多个候选周期值确定为所述目标事件的周期值集合。
进一步地，所述方法还包括：若所述并集为空集，判定所述目标事件不具备周期性。
为实现上述目的，本申请另一方面还提供一种确定事件周期值的装置，所述装置包括处理器和存储器，所述存储器中存储有计算机程序，当所述计算机程序被所述处理器执行时，实现以下步骤：获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数；计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联；计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联；取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
进一步地，当所述计算机程序被所述处理器执行时，还实现以下步骤：计算所述自相关序列中各个波峰值对应的峰高，并基于所述峰高从所述自相关序列的波峰值中提取标记波峰值；计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数；将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。
进一步地，当所述计算机程序被所述处理器执行时，还实现以下步骤：随机打乱所述时间序列中序列值的排列顺序N次，得到N个乱序时间序列，并计算所述乱序时间序列对应的乱序傅里叶变换结果；其中，N为大于或者等于1的整数；从所述乱序傅里叶变换结果中各个频率点的振幅值中确定最大振幅值；在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。
由上可见，本申请结合自相关算法和傅里叶算法，通过自相关序列中的波峰值和波谷值，可以确定所述自相关序列对应的第一候选周期集合，此外，通过傅里叶变换结果中频率点的振幅值，可以确定所述傅里叶变换结果对应的第二候选周期集合。在两个候选周期集合中，分别可以包括各个候选周期值，并且每个候选周期值可以与置信度相关联。这样，在最终确定事件的周期值时，可以取两个集合的并集。在该并集中，可以计算每个候选周期值的置信度总值。该置信度总值可以是同一个候选周期值在两个候选周期集合中置信度之和。这样，基于所述置信度总值便可以确定所述目标事件的周期值。由上可见，通过结合两种算法，可以保证周期值的精度更高，并且通过置信度总值进行判定，能够自动确定出目标事件的周期值。
附图说明
为了更清楚地说明本发明实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本发明实施例一中确定事件周期值的方法流程图；
图2是本发明实施例一中阈值集合的判定示意图；
图3是本发明实施例二中装置的功能模块图。
具体实施方式
为使本发明的目的、技术方案和优点更加清楚，下面将结合附图对本发明实施方式作进一步地详细描述。
实施例一
本申请提供一种确定事件周期值的方法，请参阅图1，所述方法包括以下步骤。
S1：获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数。
在本实施方式中，可以统计待分析的目标事件在单位时间内发生的次数。所述单位时间可以根据实际应用场景进行灵活调整。例如，所述单位时间可以是1小时，这样，便可以统计所述目标事件在每小时内发生的次数。
在本实施方式中，在统计了目标事件在单位时间内发生的次数后，便可以生成所述目标事件的时间序列。所述时间序列中可以包括预设数量的序列值，所述预设数量的序列值可以用于表征所述目标事件在单位时间内发生的次数，并且每个所述序列值可以与所述单位时间相对应。例如，所述目标事件的时间序列可以是{2，3，6，8，9}，那么该时间序列中的各个数字可以表示在单位时间内目标事件发生的次数。若所述单位时间为1小时，那么该时间序列中的每个数字的统计时长便可以是1小时。
S3：计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联。
在本实施方式中，针对所述时间序列，可以通过自相关算法计算得到其对应的自相关序列。计算自相关序列的公式是本领域常用的技术手段，这里便不再赘述。所述自相关序列可以表示为{Ri}，其中i的取值为1至m，m为所述自相关序列中各个数值的总数。
在本实施方式中，所述自相关序列中的数值可以呈现上下波动的状态，这 样，所述自相关序列中的数值便可以存在波峰值和波谷值。可以从所述自相关序列中按序提取出波峰值和波谷值，其中，所述波峰值大于与其相邻的两个数值，所述波谷值小于与其相邻的两个数值。
在提取出波峰值和波谷值之后，可以计算所述自相关序列中各个波峰值对应的峰高。在本实施方式中，所述峰高可以按照下述方式确定：针对所述自相关序列中的目标波峰值，分别计算所述目标波峰值与相邻两个波谷值之间的第一差值和第二差值，并将所述第一差值和所述第二差值中的较小值确定为所述目标波峰值的峰高。例如，所述波峰值为12，与其相邻的两个波谷值分别为3和7，那么所述第一差值和所述第二差值分别为9和5，那么便可以将5作为波峰值12的峰高。
在本实施方式中，确定出各个波峰值的峰高后，可以基于所述峰高从所述自相关序列的波峰值中提取标记波峰值。具体地，可以将各个波峰值对应的峰高与指定峰高阈值进行对比，并将大于或者等于所述指定峰高阈值的峰高对应的波峰值确定为所述标记波峰值。例如，所述指定峰高阈值为5，那么便可以将峰高大于或者等于5的波峰值作为标记波峰值。
在本实施方式中，确定出所述标记波峰值后，相邻两个标记波峰值之间的时间跨度便可能是所述时间序列对应的周期值。因此，可以计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数。其中，重复次数越多，表明该时间跨度越可能作为所述时间序列的周期值。例如，当前共计有11个标记波峰值，计算出的任意两个相邻标记波峰值之间的时间跨度可以是3小时、4小时、3小时、3小时、3小时、4小时、3小时、6小时、4小时、3小时这10个时间跨度，那么由此可以统计得到3小时的时间跨度对应的重复次数为6，4小时的时间跨度对应的重复次数为3，其它的时间跨度对应的重复次数为1。在本实施方式中，可以将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。所述重复次数满足指定条件可以指：重复次数在计算出的时间跨度的总数量中所占的比例大于或者等于指定比例阈值。假设所述指定比例阈值为50％，按照上述的例子可见，3小时的时间跨度对应的重复次数在总数量中所占的比例为60％，超过了所述指定比例阈值，而其它的时间跨度对应的重复次数在总数量中所占的比例均未超过50％，因此，可以将3小时作为候选周期值加入第一 候选周期集合中。
在实际应用场景中，仅通过一个指定峰高阈值得到的候选周期值的数量可能比较少，为了不遗漏时间序列中可能存在的周期值，可以预先提供阈值集合H，在所述阈值集合中，可以包括至少两个指定峰高阈值。请参阅图2，针对每个指定峰高阈值，均可以通过上述的方式得到标记波峰值。这样，各个所述指定峰高阈值均与各自提取的标记波峰值相对应。后续则可以根据每次提取出的标记波峰值确定出候选周期值，从而可以保证第一候选周期集合中的候选周期值能够比较完备。在本实施方式中，所述阈值集合中的指定峰高阈值可以具备取值区间，基于该取值区间，按照固定的步长，便可以得到各个指定峰高阈值。例如，所述阈值集合对应的取值区间为0值6，所述固定的步长可以是2，那么便可以生成2、4、6这三个指定峰高阈值。
在本实施方式中，在确定了第一候选周期集合中的各个候选周期值之后，还可以为每个候选周期值设定关联的置信度。所述置信度可以用于表明候选周期值作为真实周期值的可能性。具体地，可以将所述候选周期值的重复次数在计算出的时间跨度的总数量中所占的比例作为所述候选周期值关联的置信度。例如，上述的例子中，3小时这个候选周期值关联的置信度便为0.6。
S5：计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联。
在本实施方式中，还可以基于傅里叶变换的方法来确定第二候选周期集合。具体地，可以将所述时间序列进行离散傅里叶变换，从而得到该时间序列对应的傅里叶变换结果。在傅里叶变换结果中，通过周期图的方式可以显示出各个频率点对应的振幅值。其中，振幅值越大，表明该振幅值对应的频率点可以作为时间序列的频率值(也就是周期值的倒数)的置信度越高。基于此，在本实施方式中，为了判断所述时间序列是否存在周期值，可以预先随机打乱所述时间序列中序列值的排列顺序，得到乱序时间序列。如果原先的时间序列具备一定的周期性，那么在随机打乱之后得到的乱序时间序列中可能不具备明显的周期性。在实际应用场景中，为了保证随机打乱的完备性，可以执行随机打乱的步骤N次，从而得到N个乱序时间序列，其中，N为大于或者等于1的整数。
在本实施方式中，可以利用离散傅里叶变换公式，计算各个所述乱序时间 序列分别对应的乱序傅里叶变换结果，从而可以得到N个乱序傅里叶变换结果。从这N个结果中各个频率点的振幅值中，可以确定出最大振幅值。该最大振幅值对应的频率点可以表示在随机打乱原先的时间序列之后，乱序时间序列可能具备的周期值。而如果原先的时间序列具备明显的周期性，该周期性对应的频率点的振幅值应当大于根据乱序时间序列得到的所述最大振幅值。基于此，可以在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。例如，根据乱序傅里叶变换结果中确定出的最大振幅值为15，那么便可以将正常的所述时间序列的傅里叶变换结果中振幅值大于或者等于15的频率点提取出来，并依次计算每个提取出的频率点的倒数，从而作为第二候选周期集合中的候选周期值。
同样地，在确定出第二候选周期集合中的候选周期值后，可以为各个候选周期值设定关联的置信度。在本实施方式中，可以将所述目标频率点的振幅值与所述最大振幅值的比值作为与所述目标频率点对应的候选周期值相关联的置信度。例如，所述最大振幅值为15，其中一个目标频率点的振幅值为30，那么该目标频率点对应的候选周期值相关联的置信度便可以为2。
S7：取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
在本实施方式中，在确定出第一候选周期集合以第二候选周期集合后，可以取这两个集合的并集，从而将这两个集合中的候选周期值合并。在合并之后，有的候选周期值可能只对应一个置信度，而有的候选周期值可能对应两个置信度。对应两个置信度的候选周期值说明同时存在于这两个集合中，其作为所述时间序列的真实周期值的可能性就越高。因此，可以确定合并后的集合中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。具体地，所述置信度总值可以是两个集合中的置信度之和。在确定所述目标事件的周期值时，可以将置信度总值大于或者等于指定阈值的一个或多个候选周期值确定为所述目标事件的周期值集合。需要说明的是，确定出的周期值的数量可能不止一个，这就表明在原先的时间序列中，可能存在两个不同的周期值，这在实际场景中也是比较常见的。
当然，若所述并集为空集，则表明所述第一候选周期集合以及所述第二候选周期集合均为空集，也就是不存在候选周期值，这样则可以判定所述目标事件不具备周期性。
实施例二
请参阅图3，本申请还提供一种确定事件周期值的装置，所述装置包括处理器100和存储器200，所述存储器200中存储有计算机程序，当所述计算机程序被所述处理器100执行时，实现以下步骤：
获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数；
计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联；
计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联；
取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
在本实施方式中，当所述计算机程序被所述处理器执行时，还实现以下步骤：
计算所述自相关序列中各个波峰值对应的峰高，并基于所述峰高从所述自相关序列的波峰值中提取标记波峰值；
计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数；
将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。
在本实施方式中，当所述计算机程序被所述处理器执行时，还实现以下步骤：
随机打乱所述时间序列中序列值的排列顺序N次，得到N个乱序时间序列， 并计算所述乱序时间序列对应的乱序傅里叶变换结果；其中，N为大于或者等于1的整数；
从所述乱序傅里叶变换结果中各个频率点的振幅值中确定最大振幅值；
在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。
本说明书中的各个实施方式均采用递进的方式描述，各个实施方式之间相同相似的部分互相参见即可，每个实施方式重点说明的都是与其他实施方式的不同之处。尤其，针对装置的实施方式来说，均可以参照前述方法的实施方式的介绍对照解释。
本申请可以在由计算机执行的计算机可执行指令的一般上下文中描述，例如程序模块。一般地，程序模块包括执行特定任务或实现特定抽象数据类型的例程、程序、对象、组件、数据结构等等。也可以在分布式计算环境中实践本申请，在这些分布式计算环境中，由通过通信网络而被连接的远程处理设备来执行任务。在分布式计算环境中，程序模块可以位于包括存储设备在内的本地和远程计算机存储介质中。
由上可见，本申请结合自相关算法和傅里叶算法，通过自相关序列中的波峰值和波谷值，可以确定所述自相关序列对应的第一候选周期集合，此外，通过傅里叶变换结果中频率点的振幅值，可以确定所述傅里叶变换结果对应的第二候选周期集合。在两个候选周期集合中，分别可以包括各个候选周期值，并且每个候选周期值可以与置信度相关联。这样，在最终确定事件的周期值时，可以取两个集合的并集。在该并集中，可以计算每个候选周期值的置信度总值。该置信度总值可以是同一个候选周期值在两个候选周期集合中置信度之和。这样，基于所述置信度总值便可以确定所述目标事件的周期值。由上可见，通过结合两种算法，可以保证周期值的精度更高，并且通过置信度总值进行判定，能够自动确定出目标事件的周期值。
上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。
以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。 可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下，即可以理解并实施。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如ROM/RAM、磁碟、光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
以上所述仅为本发明的较佳实施例，并不用以限制本发明，凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (14)

  1. 一种确定事件周期值的方法，其特征在于，所述方法包括：
    获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数；
    计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联；
    计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联；
    取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
  2. 根据权利要求1所述的方法，其特征在于，确定所述自相关序列对应的第一候选周期集合包括：
    计算所述自相关序列中各个波峰值对应的峰高，并基于所述峰高从所述自相关序列的波峰值中提取标记波峰值；
    计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数；
    将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。
  3. 根据权利要求2所述的方法，其特征在于，计算所述自相关序列中各个波峰值对应的峰高包括：
    针对所述自相关序列中的目标波峰值，分别计算所述目标波峰值与相邻两个波谷值之间的第一差值和第二差值，并将所述第一差值和所述第二差值中的较小值确定为所述目标波峰值的峰高。
  4. 根据权利要求2或3所述的方法，其特征在于，基于所述峰高从所述自相关序列的波峰值中提取标记波峰值包括：
    将各个波峰值对应的峰高与指定峰高阈值进行对比，并将大于或者等于所述指定峰高阈值的峰高对应的波峰值确定为所述标记波峰值。
  5. 根据权利要求4所述的方法，其特征在于，所述指定峰高阈值从属于阈值集合中；所述阈值集合中包括至少两个不同的指定峰高阈值；其中，各个所述指定峰高阈值均与各自提取的标记波峰值相对应。
  6. 根据权利要求2所述的方法，其特征在于，重复次数满足指定条件包括：
    重复次数在计算出的时间跨度的总数量中所占的比例大于或者等于指定比例阈值。
  7. 根据权利要求2所述的方法，其特征在于，按照下述方式确定与所述第一候选周期集合中的候选周期值相关联的置信度：
    将所述候选周期值的重复次数在计算出的时间跨度的总数量中所占的比例作为所述候选周期值关联的置信度。
  8. 根据权利要求1所述的方法，其特征在于，确定所述傅里叶变换结果对应的第二候选周期集合包括：
    随机打乱所述时间序列中序列值的排列顺序N次，得到N个乱序时间序列，并计算所述乱序时间序列对应的乱序傅里叶变换结果；其中，N为大于或者等于1的整数；
    从所述乱序傅里叶变换结果中各个频率点的振幅值中确定最大振幅值；
    在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。
  9. 根据权利要求8所述的方法，其特征在于，按照下述方式确定与所述第二候选周期集合中的候选周期值相关联的置信度：
    将所述目标频率点的振幅值与所述最大振幅值的比值作为与所述目标频率点对应的候选周期值相关联的置信度。
  10. 根据权利要求1所述的方法，其特征在于，确定所述目标事件的周期值包括：
    将置信度总值大于或者等于指定阈值的一个或多个候选周期值确定为所述目标事件的周期值集合。
  11. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    若所述并集为空集，判定所述目标事件不具备周期性。
  12. 一种确定事件周期值的装置，其特征在于，所述装置包括处理器和存储器，所述存储器中存储有计算机程序，当所述计算机程序被所述处理器执行时，实现以下步骤：
    获取目标事件的时间序列，所述时间序列中包括预设数量的序列值，所述序列值用于表征所述目标事件在单位时间内发生的次数；
    计算所述时间序列的自相关序列，并基于所述自相关序列中的波峰值和波谷值，确定所述自相关序列对应的第一候选周期集合；其中，所述第一候选周期集合中的候选周期值与置信度相关联；
    计算所述时间序列的傅里叶变换结果，并基于所述傅里叶变换结果中频率点的振幅值，确定所述傅里叶变换结果对应的第二候选周期集合；其中，所述第二候选周期集合中的候选周期值与置信度相关联；
    取所述第一候选周期集合以及所述第二候选周期的并集，确定所述并集中各个候选周期值对应的置信度总值，并基于所述置信度总值，确定所述目标事件的周期值。
  13. 根据权利要求12所述的装置，其特征在于，当所述计算机程序被所述处理器执行时，还实现以下步骤：
    计算所述自相关序列中各个波峰值对应的峰高，并基于所述峰高从所述自相关序列的波峰值中提取标记波峰值；
    计算任意两个相邻标记波峰值之间的时间跨度，并统计各个所述时间跨度出现的重复次数；
    将重复次数满足指定条件的时间跨度作为所述自相关序列的候选周期值，并将所述候选周期值加入第一候选周期集合中。
  14. 根据权利要求12所述的装置，其特征在于，当所述计算机程序被所述处理器执行时，还实现以下步骤：
    随机打乱所述时间序列中序列值的排列顺序N次，得到N个乱序时间序列，并计算所述乱序时间序列对应的乱序傅里叶变换结果；其中，N为大于或者等于1的整数；
    从所述乱序傅里叶变换结果中各个频率点的振幅值中确定最大振幅值；
    在所述时间序列的傅里叶变换结果中确定振幅值大于或者等于所述最大振幅值的目标频率点，并将所述目标频率点的倒数作为第二候选周期集合中的候选周期值。
PCT/CN2017/097772 2017-08-08 2017-08-17 一种确定事件周期值的方法及装置 WO2019028931A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US16/061,802 US20210165850A1 (en) 2017-08-08 2017-08-17 Method and device for determining event periodic value
EP17897206.3A EP3460714A4 (en) 2017-08-08 2017-08-17 METHOD AND DEVICE FOR DETERMINING VALUE OF EVENT PERIOD

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710669625.0 2017-08-08
CN201710669625.0A CN107463904B (zh) 2017-08-08 2017-08-08 一种确定事件周期值的方法及装置

Publications (1)

Publication Number Publication Date
WO2019028931A1 true WO2019028931A1 (zh) 2019-02-14

Family

ID=60548491

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/097772 WO2019028931A1 (zh) 2017-08-08 2017-08-17 一种确定事件周期值的方法及装置

Country Status (4)

Country Link
US (1) US20210165850A1 (zh)
EP (1) EP3460714A4 (zh)
CN (1) CN107463904B (zh)
WO (1) WO2019028931A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111358454A (zh) * 2020-03-17 2020-07-03 乐普（北京）医疗器械股份有限公司 一种基于差异波占比检测信号质量的方法和装置

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11635345B2 (en) 2013-07-02 2023-04-25 Computational Systems, Inc. System for separating periodic frequency of interest peaks from non-periodic peaks in machine vibration data
CN110912860B (zh) * 2018-09-18 2022-02-18 北京数安鑫云信息技术有限公司 一种检测伪周期性访问行为的方法及装置
CN109918272B (zh) * 2018-12-18 2022-06-21 创新先进技术有限公司 报警消息的周期性确定方法、装置及设备
DE102019127211A1 (de) * 2019-03-05 2020-09-10 Computational Systems, Inc. System zum Trennen von periodischen Amplitudenspitzen von nicht-periodischen Amplitudenspitzen in Maschinenschwingungsdaten
CN110390594A (zh) * 2019-07-16 2019-10-29 华泰证券股份有限公司 经济金融数据周期变化规律的可视化方法、装置及介质
CN113811908B (zh) * 2019-09-02 2023-01-31 西门子（中国）有限公司 用于确定生产设备的生产周期的方法和装置
CN110611684A (zh) * 2019-09-27 2019-12-24 国网电力科学研究院有限公司 一种周期性Web访问行为的检测方法、系统及存储介质
CN111768287A (zh) * 2020-05-22 2020-10-13 支付宝(杭州)信息技术有限公司 周期识别方法、装置、服务器及可读存储介质
CN111967421B (zh) * 2020-08-27 2023-07-28 广东工业大学 周期性功率跃迁与常规设备启动事件的分类方法和装置
US11624831B2 (en) 2021-06-09 2023-04-11 Suteng Innovation Technology Co., Ltd. Obstacle detection method and apparatus and storage medium
CN113253299B (zh) * 2021-06-09 2022-02-01 深圳市速腾聚创科技有限公司 障碍物检测方法、装置及存储介质
CN114579554A (zh) * 2022-03-08 2022-06-03 京东科技信息技术有限公司 运维时间序列数据处理方法及装置、存储介质及电子设备
CN115168727B (zh) * 2022-07-20 2023-08-08 北京百度网讯科技有限公司 用户习惯的挖掘方法、装置及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060242214A1 (en) * 2005-03-15 2006-10-26 Daisuke Tominaga Periodicity judgement apparatus, periodicity judgement method and periodicity judgement program
CN101572089A (zh) * 2009-05-21 2009-11-04 华为技术有限公司 信号周期检测方法及其装置
CN101853240A (zh) * 2009-03-31 2010-10-06 华为技术有限公司 一种信号周期的估计方法和装置
US20100305944A1 (en) * 2009-05-28 2010-12-02 Cambridge Silicon Radio Limited Pitch Or Periodicity Estimation
CN106842161A (zh) * 2017-01-04 2017-06-13 北京环境特性研究所 基于置信度评价的微动周期提取方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6718317B1 (en) * 2000-06-02 2004-04-06 International Business Machines Corporation Methods for identifying partial periodic patterns and corresponding event subsequences in an event sequence
US7369961B2 (en) * 2005-03-31 2008-05-06 International Business Machines Corporation Systems and methods for structural clustering of time sequences
JP5530812B2 (ja) * 2010-06-04 2014-06-25 ニュアンス コミュニケーションズ，インコーポレイテッド 音声特徴量を出力するための音声信号処理システム、音声信号処理方法、及び音声信号処理プログラム
CN102346745B (zh) * 2010-08-02 2014-04-02 阿里巴巴集团控股有限公司 单词的用户行为数的预测方法和装置
CN104378219A (zh) * 2013-08-12 2015-02-25 中兴通讯股份有限公司 路由器流量数据智能分析的方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060242214A1 (en) * 2005-03-15 2006-10-26 Daisuke Tominaga Periodicity judgement apparatus, periodicity judgement method and periodicity judgement program
CN101853240A (zh) * 2009-03-31 2010-10-06 华为技术有限公司 一种信号周期的估计方法和装置
CN101572089A (zh) * 2009-05-21 2009-11-04 华为技术有限公司 信号周期检测方法及其装置
US20100305944A1 (en) * 2009-05-28 2010-12-02 Cambridge Silicon Radio Limited Pitch Or Periodicity Estimation
CN106842161A (zh) * 2017-01-04 2017-06-13 北京环境特性研究所 基于置信度评价的微动周期提取方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3460714A4 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111358454A (zh) * 2020-03-17 2020-07-03 乐普（北京）医疗器械股份有限公司 一种基于差异波占比检测信号质量的方法和装置
CN111358454B (zh) * 2020-03-17 2022-07-29 乐普（北京）医疗器械股份有限公司 一种基于差异波占比检测信号质量的方法和装置

Also Published As

Publication number Publication date
EP3460714A1 (en) 2019-03-27
CN107463904B (zh) 2021-05-25
US20210165850A1 (en) 2021-06-03
CN107463904A (zh) 2017-12-12
EP3460714A4 (en) 2019-04-24

Similar Documents

Publication Publication Date Title
WO2019028931A1 (zh) 一种确定事件周期值的方法及装置
WO2019214248A1 (zh) 一种风险评估方法、装置、终端设备及存储介质
Borges et al. Predicting the popularity of github repositories
CN112989266A (zh) 时间序列中的周期性检测和周期长度估计
WO2019200702A1 (zh) 去网纹系统训练方法、去网纹方法、装置、设备及介质
Zhu et al. Conditional mean spectrum of aftershocks
WO2011008198A1 (en) Determining a seasonal effect in temporal data
Kugiumtzis Surrogate data test on time series
WO2017063516A1 (zh) 噪音信号确定方法、语音去噪方法及装置
CN106327230B (zh) 一种异常用户检测方法及设备
CN110598959A (zh) 一种资产风险评估方法、装置、电子设备及存储介质
CN109410198B (zh) 一种时序动作检测方法、装置及设备
CN110737650A (zh) 数据质量检测方法及装置
CN107203720B (zh) 风险值计算方法及装置
CN105761119B (zh) 在线人数分布计算方法及装置
US11792197B1 (en) Detecting malicious user accounts of an online service using major-key-shared-based correlation
CN114726623A (zh) 一种高级威胁攻击评估方法、装置、电子设备及存储介质
Wu et al. Data dependency on measurement uncertainties in speaker recognition evaluation
CN114020996A (zh) 一种识别异常观看用户的方法、装置、介质及计算机设备
Feng et al. An acoustic method for condition classification of a water-filled underground siphon
CN110399399B (zh) 用户分析的方法、装置、电子设备和存储介质
CN113076451A (zh) 异常行为识别和风险模型库的建立方法、装置及电子设备
Mackrell et al. Discovering anomalous patterns in network traffic data during crisis events
Yeşilova et al. AN APPLICATION OF HOT DECK IMPUTATION AND SUBSTITUTION METHODS IN THE ESTIMATION OF MISSING DATA
CN111708813A (zh) 一种用户日常行为异常检测方法和装置

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2017897206

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2017897206

Country of ref document: EP

Effective date: 20180827

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17897206

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE