CN104915846A - Electronic commerce time sequence data anomaly detection method and system - Google Patents
Electronic commerce time sequence data anomaly detection method and system Download PDFInfo
- Publication number
- CN104915846A CN104915846A CN201510342240.4A CN201510342240A CN104915846A CN 104915846 A CN104915846 A CN 104915846A CN 201510342240 A CN201510342240 A CN 201510342240A CN 104915846 A CN104915846 A CN 104915846A
- Authority
- CN
- China
- Prior art keywords
- data
- electronic commerce
- abnormal
- normal value
- statistics
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Landscapes
- Complex Calculations (AREA)
Abstract
The invention discloses an electronic commerce time sequence data anomaly detection method and a system. The method comprises steps: electronic commerce data based on a time sequence are acquired; an Nth phase electronic commerce data adjacent to to-be-detetced data are selected as window statistical data, quantile statistics is carried out on the window statistical data, a normal value upper boundary and a normal value lower boundary in the window statistical data are thus determined, and data out of a normal value range determined by the normal value upper boundary and the normal value lower boundary in the window statistical data are abnormal data; and the abnormal data serve as an application interface for being called by a demand side. Through benchmark detection, time sequence fluctuation identification based on a robust statistics method is realized, and the method and the system of the invention are applied to various distribution conditions. According to different electronic commerce service scenes and different data distribution forms, data anomaly can be automatically found out.
Description
Technical field
The present invention relates to ecommerce correlative technology field, particularly a kind of method for detecting abnormality of ecommerce time series data and system.
Background technology
Time series is the ordered set of in chronological sequence each observational record tactic.In electronic commerce affair, As time goes on, time series comprises a large amount of data usually, for seasonal effect in time series analysis, can disclose the inherent law of electronic commerce affair motion, change and progress, especially for data exception, often contain more more important information and knowledge, therefore, how fast and effeciently to detect these abnormal be a significant job, such as sometimes order data is large singularly, may mean the huge market opportunity behind; The abnormal growth of profit data, may mean and have the place reducing cost of products or promote profit have to be positioned and excavate; And the exception of number of users reduces, the generation of certain market risk or the low inferior problem of efficiency of operation may be meaned, etc. abnormal data need to find in time and localized reason in electronic commerce affair operation, and the features such as the polytrope of electronic commerce affair, complicacy and big data quantity, the detection for abnormal data brings no small challenge.
For the abnormality detection of time series data, existing technical scheme utilizes the statistical model of service logic and strong assumption usually, roughly adopts two kinds of methods:
(1) the subjective threshold method of service logic
Adopt moving average or chain rate, on year-on-year basis as reference value, waiting to judge that the rate of change of data field reference value calculates, then comparing with threshold value, exceeding threshold value and being then judged to be exception, this threshold value is that subjectivity is determined usually.Be made with two shortcomings like this, one is threshold value is that business personnel's subjectivity is determined, the degree of understanding of different people to business is different may produce different threshold values, and interpretation is poor; When passing through chain rate, on year-on-year basis data as reference value on the other hand, when multiple abnormal data occurs simultaneously time, due to the excessive or too small meeting of abnormal data above make the chain rate of abnormal data below or on year-on-year basis data tend to be steady, thus continuous print abnormal data below cannot be found, just as by " shielding ".As shown in Figure 1:
Such as, before and after June 18, the data of three days become large all extremely, but when calculating by chain rate data, two days next due to first day data large especially, the chain rate change calculated will be very little, so just can only find that the data point of first day is abnormal, after the abnormal data of two days will be fallen by first day data " shielding " and can not detect out.
(2) strong assumption statistical model diagnostic method
Statistical model often has the assumed condition that data meet certain specific distribution; conventional is meet normal distribution; under normal distribution hypothesis; normal value interval range is: average n times of standard deviation; time n gets 1-3; the probability dropping on this region is respectively 68.29%, 95.45% and 99.73%, and the situation of data outside normal value interval belongs to small probability event, is defined as exceptional value.
In the method for current above-mentioned discovery time sequence data exception, there is the multiple condition such as threshold definitions subjectivity, existence " shielding " effect in the subjective threshold method of the first service logic, detection interpretation and precision all exist larger problem; There is hypothesis and rely on too strong problem in another kind of strong assumption statistical model judging rules, actual electronic commerce data is very complicated, seldom have the situation meeting certain specific distribution, this is restricted with regard to making the usable range of statistical model, and model effect also weakens greatly.
Summary of the invention
Based on this, be necessary the abnormal data can not checking out electronic commerce data for prior art well, a kind of method for detecting abnormality and system of ecommerce time series data are provided.
A method for detecting abnormality for ecommerce time series data, comprising:
Data acquisition step, comprising: obtain based on seasonal effect in time series electronic commerce data, and perform benchmaring step to each data in described electronic commerce data, the data performing benchmaring step are data to be tested;
Benchmaring step, comprise: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then execute exception invocation step, wherein, described N is the default natural number being greater than 1;
Exception call step, comprising: abnormal data is supplied to party in request as application interface and calls.
An abnormality detection system for ecommerce time series data, comprising:
Data acquisition module, for: obtain based on seasonal effect in time series electronic commerce data, perform benchmaring module to each data in described electronic commerce data, the data performing benchmaring module are data to be tested;
Benchmaring module, for: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then execute exception calling module, wherein, described N is the default natural number being greater than 1;
Exception call module, for: abnormal data is supplied to party in request as application interface and calls.
The present invention, by benchmaring, realizes sequential fluctuation based on robust statistics methods and identifies, be applicable to various distribution situation.The present invention for ecommerce different business scene, different pieces of information distribution form, can find data exception automatically.
Accompanying drawing explanation
Fig. 1 is prior art chain rate index shielding effect schematic diagram;
Fig. 2 is the workflow diagram of the method for detecting abnormality of a kind of ecommerce time series data of the present invention;
Fig. 3 is fractile abnormity point schematic diagram;
Fig. 4 is normal distribution and fractile comparison diagram;
Fig. 5 is the system construction drawing of most preferred embodiment of the present invention;
Fig. 6 is the workflow diagram of preferred embodiment;
Fig. 7 is the construction module figure of the abnormality detection system of a kind of ecommerce time series data of the present invention.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be further described in detail.
Be illustrated in figure 2 the workflow diagram of the method for detecting abnormality of a kind of ecommerce time series data of the present invention, comprise:
Step S201, comprising: obtain based on seasonal effect in time series electronic commerce data, and perform step S202 to each data in described electronic commerce data, the data performing step S202 are data to be tested;
Step S202, comprise: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then perform step S203, wherein, described N is the default natural number being greater than 1;
Step S203, comprising: abnormal data is supplied to party in request as application interface and calls.
The electronic commerce data that step S201 obtains is based on seasonal effect in time series data, and in general, electronic commerce data can not meet existing statistical distribution pattern completely, thus causes the method for existing statistic mixed-state abnormal data not use.The present invention is in step S202, have employed the statistical method of fractile, the method of fractile is for any distribution, the statistical computation even comprising improper value or multiple abnormal data is all very sane, and ultimate principle calculates data position shared in all data of this statistical window thus determines range of normal value.
Step S203, specifically, abnormal data result is processed into the structural data of standard, the forms such as such as hdfs file, hbase file, xml or text, stored in MySQL or distributed data base, directly called by Database Systems, API Calls or the Internet data transmission agreement by standard, for downstream demand side
Abnormal data is supplied to party in request by step S203 and calls, such as, carry out the subsequent operations such as abnormal alarm after determining abnormal data by the statistical method of employing fractile by the present invention.
The present invention is according to the feature of electronic commerce affair complex data, do not rely on subjective judgement and strong assumption condition, carry out general sane anomaly data detection, overcome the shortcoming of prior art, thus greatly improve the scope of application and the Detection results of overall abnormality detection scheme.
Wherein in an embodiment, also comprise: if the data of described electronic commerce data continuous N phase perform step S202 do not detect abnormal data, then using the data of described electronic commerce data continuous N phase as combine detection data, perform combine detection step, wherein M is the default natural number being greater than 1, described combine detection step, sequential trend analysis or sequential causal inference are comprised to described combine detection data, and the abnormal data that sequential trend analysis or sequential causal inference obtain is carried out backtracking breakpoint analysis.
Other Model Fusion of universality universal model and specific transactions scene get up by the present embodiment, substantially increase effect and the precision of anomaly data detection, make the more intelligent and high efficiency of the abnormality detection of electronic commerce affair time series data.
Wherein in an embodiment,
Described sequential trend analysis comprises: be linear growth trend, rapid growth trend, periodically rising tendency by described combine detection data based on Time Series, choose from described electronic commerce data do not meet described linear growth trend, rapid growth trend, periodically rising tendency data as abnormal data;
Described sequential causal inference comprises: from described combine detection data, select the first data group and the second data group, described first data group and the second data group have the probability distribution of identical type, the data variation scope calculating the first data group, as normal data variation range, will exceed the data of described normal data variation range as abnormal data in the second data group.
Embodiment adds sequential trend analysis and sequential causal inference, wherein:
Sequential trend analysis: the superposition of time series various factors often, the residing environment of such as order volume and whole company, overall growth speed, periodically, the factor such as seasonal is relevant, the thinking of trend analysis is exactly that the overall data of complexity is resolved into the single factor that simply can quantize to explain, such as conceptual data=linear growth trend+rapid growth trend+smooth growth+periodicity growth waits quantizing factor, convenient analysis.
Described sequential trend analysis is linear growth trend, rapid growth trend, periodically rising tendency based on Time Series, represents for Y with mathematical function
t=S
t+ T
t+ C
t, wherein
Y
tfor combine detection data;
S
tfor linear growth trend, obey shape such as the once linear of y=a+bx and distribute;
T
tfor rapid growth trend, obey shape as y=a+bx
nthe high order nonlinear Distribution of (wherein n>1);
C
tfor periodicity rising tendency, obey shape as the periodic distribution of y=a+b cos (x);
One by one several trend after decomposition are detected from described combine detection data, choose do not meet described linear growth trend, rapid growth trend or periodically rising tendency data as abnormal data, the mode not meeting trend can adopt existing various trend analysis mode to judge, such as the regression algorithm of various function judges, as linear regression and non-linear regression mode judge, or according to linear regression or non-linear regression mode, corresponding regression function is calculated respectively to combine detection data, and judge whether regression function meets Y
t=S
t+ T
t+ C
t.
Sequential causal inference be by advance known two distribution same or similar two groups of data group y1 and y2, at certain time point, if only have wherein one group of data group, such as y1 changes, and data group y2 does not change, then can the data group y1 to having changed be gone to calculate and judge with the data group y2 do not changed, owing to can calculate the scope of normal data variation according to same distributed data, therefore, if in the data group y1 changed, the data that amplitude of fluctuation exceedes the normal data mobility scale of the data group y2 that employing does not change are judged as abnormal data.
Such as, for the data of normal distribution, its average and variance can be calculated, its variance is calculated for the data group y2 do not changed, the data of data group y2 are exceeded as abnormal data for amplitude of fluctuation in data group y1.The selection of data group by pre-determining, such as, can have two groups of data to be buy different commodity respectively in same time section for purchase, if the classification of commodity is comparatively similar, then can think that these two groups of data have identical probability distribution.Specifically how selecting, can be pre-determined by user and write in configuration file, in computation process, determining by reading configuration file.
Wherein in an embodiment, also comprise:
Described backtracking breakpoint analysis comprises: using time point corresponding for the abnormal data of described sequential trend analysis or described sequential causal inference as current point in time t
nowforward trace, between the front average D1 of the combine detection data at every turn between more each time point proparea in range1 and time point back zone, in range2, the time interval of rear average D2, range1 and the range2 of combine detection data is identical, if time point t
forefront average and the change of rear average exceed predetermined threshold value, then think time point t
foreproparea between range1 and back zone the data of range2 have exception, from time point t
foreto t
nowcombine detection data in time period are as abnormal data.
Embodiment adds breaking point detection (Breakout Detection): for some time series datas, short-term may not have large fluctuation, but the change by also becoming, form another trend scope, breaking point detection is exactly change by the trend in the long-term scope of monitoring period sequence indicator the existence noted abnormalities a little, average or accumulated change amount mainly by calculating a period of time before and after each data point contrast, if the Change in Mean before and after this data point or accumulated change amount larger, then be likely a breakpoint, data trend before and after this data point changes.
Wherein in an embodiment, described fractile statistics, specifically comprises:
To window statistics by size of data sequence, the median of calculation window statistics, upper quartile, lower quartile, described median is the data being in all data centre positions after the sequence of window statistics, described upper quartile is the data being in all data 1/4th positions after the sequence of window statistics, described lower quartile is the data for all data 3/4ths positions after the sequence of window statistics, calculate the absolute value of the difference of lower quartile and upper quartile as interquartile-range IQR, determine that described normal value upper boundary values is that median deducts k times of interquartile-range IQR, determine that described normal value lower border value is that median adds k times of interquartile-range IQR, wherein, described k be greater than 1 natural number.
The sequence of window statistics can be sort by order from small to large, then interquartile-range IQR=lower quartile-upper quartile, and the data being less than normal value upper boundary values are abnormal data, and the data being greater than normal value lower border value are abnormal data.
The sequence of window statistics also can be sort by order from big to small, then interquartile-range IQR=upper quartile-lower quartile, and the data being greater than normal value upper boundary values are abnormal data, and the data being less than normal value lower border value are abnormal data.
Preferably, k is 1,2 or 3.
Be illustrated in figure 3 fractile abnormity point schematic diagram.Fractile detection has any distribution all applicable, and its test effect is also fine, and when data Normal Distribution, can obtain detection effect suitable with mean-standard deviation.As shown in Figure 4, when data fit normal distribution, the fractile statistics of the present embodiment is suitable with the power of test of normal distribution, and median is equal with the average μ of normal distribution, and interquartile-range IQR approximates the standard deviation sigma of 1.4 times of normal distributions.
Be illustrated in figure 5 the system construction drawing of most preferred embodiment of the present invention, comprise: data preparation module 501, benchmaring module 502, combine detection module 503, result Fusion Module 504, detection application module 505 5 part composition, wherein.
Data preparation module 501: the major function of data preparation module carries out the data sample pre-service of training pattern, it does not mainly meet the data of time series or service logic by rejecting improper value, null value etc., and removes the impact of the means exclusive PCR factors such as extreme value according to service needed.
Benchmaring module 502: the major function of benchmaring module is to provide the basic detection of abnormality detection is adopt the fractile abnormal point detecting method comprising N phase unusual fluctuation window data here.N is a parameter, and can specify according to service needed, in general N value is larger, and the time cycle of consideration is longer, and the variation of exceptional value bounds is less, is applicable to the data that fluctuation situation is little, such as buys user's change conditions; Otherwise the variation of exceptional value border is larger, is applicable to the data that fluctuation situation is larger, such as user browses quantity, buys the amount of money etc.Fractile outlier detection is a kind of method of robust statistics, what is called is steadily and surely that data have when changing that even data make mistakes, the adaptability problem of statistical method, the result difference that the method that robustness is strong obtains when in the face of different pieces of information situation is little, the method of fractile is for any distribution, the statistical computation even comprising improper value or multiple abnormal data is all very sane, ultimate principle calculates data position shared in all data of this statistical window, as shown in Figure 3, after median just refers to and all data is sorted by size, be in the number in all data centre positions, upper quartile is the data being in all data 1/4th (25%) positions, lower quartile is the data for all data 3/4ths (75%) positions, the same as the statistic of metric data dispersion degree with standard deviation in the middle of normal distribution, in fractile statistics, definition interquartile-range IQR is as the tolerance of data discrete degree, specifically be calculated as: lower quartile deducts upper quartile.So just can define the interval range of a normal data, be generally: median ± n times interquartile-range IQR, n generally goes to get between 1 to 3, average in the corresponding normal distribution of median in such fractile and interquartile-range IQR and standard deviation, just achieve all general rejecting outliers of all Data distribution8.Point outside above-mentioned normal value up-and-down boundary is defined as outlier.Fractile detection has any distribution all applicable, and its test effect is also fine, and when data Normal Distribution, can obtain detection effect suitable with mean-standard deviation.As shown in Figure 4, when data fit normal distribution, both power of tests are suitable, and median is equal with average, and interquartile-range IQR approximates 1.4 times of standard deviations.
Combine detection module 503: the major function of combine detection module is except benchmaring module, also provides the detection method of several specific transactions scene, to improve overall plan Detection results; Wherein sequential Trend Decomposition is that each type detects respectively and processes various dissimilar trend such as linear for complex time series data decomposition rising tendency, rapid growth trend, smooth growths; Sequential causal inference is for fine-grained data, and the concrete impact of the abnormal data caused is impacted in the quantitative judge activity in some special advertising campaigns of such as some departments, category data; Backtracking property breakpoint analysis is the complementary module detecting for benchmaring data instantaneity and increase, need accumulation a period of time just can find out for there being some abnormal datas, what breakpoint analysis solved is exactly after abnormal generation a period of time, do interim backtracking, calculate by the accumulation of abnormal data thus judge the different abnormal stages, finally finding the trend in long-term scope to change.
Result Fusion Module 504: the major function of result Fusion Module is result by service logic and degree of confidence set of weights altogether, exports the result of the higher and better effects if of final confidence.
Detect application module 505: the major function detecting application module is to provide an abnormality detection data transmission interface, according to the abnormal point numerical certificate calculated, be transferred to each business application module or warning module, abnormality detection information be transmitted to fast and use in staff.
Be illustrated in figure 6 the workflow diagram of preferred embodiment, comprise:
Step S601, data preparation module carries out data integration and data prediction;
Step S602, through the data input reference detection module of the integrated process of data preparation module, benchmaring module detects substantially, if detect abnormal data, then be sent to result Fusion Module, perform step S604, if continuous many phases all do not detect abnormal data, then data are sent to combine detection module and perform step S603;
Step S603, carries out sequential Trend Decomposition and sequential causal inference to data, when detecting that abnormal data then carries out breaking point detection, abnormal data is sent to result Fusion Module, performs step S604;
Step S604, carries out model result fusion to the abnormal data received from combine detection module, is sent to detection application model after obtaining final abnormality detection result, performs step S605;
Step S605, exports data to detection application interface.
Be illustrated in figure 7 the construction module figure of the abnormality detection system of a kind of ecommerce time series data of the present invention, comprise:
Data acquisition module 701, for: obtain based on seasonal effect in time series electronic commerce data, perform benchmaring module to each data in described electronic commerce data, the data performing benchmaring module are data to be tested;
Benchmaring module 702, for: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then execute exception calling module, wherein, described N is the default natural number being greater than 1;
Exception call module 703, for: abnormal data is supplied to party in request as application interface and calls.
Wherein in an embodiment, also comprise: if the data of described electronic commerce data continuous N phase perform benchmaring module do not detect abnormal data, then using the data of described electronic commerce data continuous N phase as combine detection data, perform combine detection module, wherein M is the default natural number being greater than 1, described combine detection module, sequential trend analysis or sequential causal inference are comprised to described combine detection data, and the abnormal data that sequential trend analysis or sequential causal inference obtain is carried out backtracking breakpoint analysis.
Wherein in an embodiment,
Described sequential trend analysis comprises: be linear growth trend, rapid growth trend, periodically rising tendency by described combine detection data based on Time Series, choose from described electronic commerce data do not meet described linear growth trend, rapid growth trend, periodically rising tendency data as abnormal data;
Described sequential causal inference comprises: from described combine detection data, select the first data group and the second data group, described first data group and the second data group have the probability distribution of identical type, the data variation scope calculating the first data group, as normal data variation range, will exceed the data of described normal data variation range as abnormal data in the second data group.
Wherein in an embodiment, described backtracking breakpoint analysis comprises: using time point corresponding for the abnormal data of described sequential trend analysis or described sequential causal inference as current point in time t
nowforward trace, between the front average D1 of the combine detection data at every turn between more each time point proparea in range1 and time point back zone, in range2, the time interval of rear average D2, range1 and the range2 of combine detection data is identical, if time point t
forefront average and the change of rear average exceed predetermined threshold value, then think time point t
foreproparea between range1 and back zone the data of range2 have exception, from time point t
foreto t
nowcombine detection data in time period are as abnormal data.
Wherein in an embodiment, described fractile statistics, specifically comprises:
To window statistics by size of data sequence, the median of calculation window statistics, upper quartile, lower quartile, described median is the data being in all data centre positions after the sequence of window statistics, described upper quartile is the data being in all data 1/4th positions after the sequence of window statistics, described lower quartile is the data for all data 3/4ths positions after the sequence of window statistics, calculate the absolute value of the difference of lower quartile and upper quartile as interquartile-range IQR, determine that described normal value upper boundary values is that median deducts k times of interquartile-range IQR, determine that described normal value lower border value is that median adds k times of interquartile-range IQR, wherein, described k be greater than 1 natural number.
The above embodiment only have expressed several embodiment of the present invention, and it describes comparatively concrete and detailed, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that for the person of ordinary skill of the art, without departing from the inventive concept of the premise, can also make some distortion and improvement, these all belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.
Claims (10)
1. a method for detecting abnormality for ecommerce time series data, is characterized in that, comprising:
Data acquisition step, comprising: obtain based on seasonal effect in time series electronic commerce data, and perform benchmaring step to each data in described electronic commerce data, the data performing benchmaring step are data to be tested;
Benchmaring step, comprise: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then execute exception invocation step, wherein, described N is the default natural number being greater than 1;
Exception call step, comprising: abnormal data is supplied to party in request as application interface and calls.
2. the method for detecting abnormality of ecommerce time series data according to claim 1, it is characterized in that, also comprise: if the data of described electronic commerce data continuous N phase perform benchmaring step do not detect abnormal data, then using the data of described electronic commerce data continuous N phase as combine detection data, perform combine detection step, wherein M is the default natural number being greater than 1, described combine detection step, sequential trend analysis or sequential causal inference are comprised to described combine detection data, and the abnormal data that sequential trend analysis or sequential causal inference obtain is carried out backtracking breakpoint analysis.
3. the method for detecting abnormality of ecommerce time series data according to claim 2, is characterized in that,
Described sequential trend analysis comprises: be linear growth trend, rapid growth trend, periodically rising tendency by described combine detection data based on Time Series, choose from described electronic commerce data do not meet described linear growth trend, rapid growth trend, periodically rising tendency data as abnormal data;
Described sequential causal inference comprises: from described combine detection data, select the first data group and the second data group, described first data group and the second data group have the probability distribution of identical type, the data variation scope calculating the first data group, as normal data variation range, will exceed the data of described normal data variation range as abnormal data in the second data group.
4. the method for detecting abnormality of ecommerce time series data according to claim 2, it is characterized in that, described backtracking breakpoint analysis comprises: using time point corresponding for the abnormal data of described sequential trend analysis or described sequential causal inference as current point in time t
nowforward trace, between the front average D1 of the combine detection data at every turn between more each time point proparea in range1 and time point back zone, in range2, the time interval of rear average D2, range1 and the range2 of combine detection data is identical, if time point t
forefront average and the change of rear average exceed predetermined threshold value, then think time point t
foreproparea between range1 and back zone the data of range2 have exception, from time point t
foreto t
nowcombine detection data in time period are as abnormal data.
5. the method for detecting abnormality of ecommerce time series data according to claim 1, is characterized in that, described fractile statistics, specifically comprises:
To window statistics by size of data sequence, the median of calculation window statistics, upper quartile, lower quartile, described median is the data being in all data centre positions after the sequence of window statistics, described upper quartile is the data being in all data 1/4th positions after the sequence of window statistics, described lower quartile is the data for all data 3/4ths positions after the sequence of window statistics, calculate the absolute value of the difference of lower quartile and upper quartile as interquartile-range IQR, determine that described normal value upper boundary values is that median deducts k times of interquartile-range IQR, determine that described normal value lower border value is that median adds k times of interquartile-range IQR, wherein, described k be greater than 1 natural number.
6. an abnormality detection system for ecommerce time series data, is characterized in that, comprising:
Data acquisition module, for: obtain based on seasonal effect in time series electronic commerce data, perform benchmaring module to each data in described electronic commerce data, the data performing benchmaring module are data to be tested;
Benchmaring module, for: select the N phase electronic commerce data of contiguous data to be tested as window statistics, fractile statistics is carried out to described window statistics, thus determine the normal value coboundary in described window statistics and normal value lower boundary, the data be in described window statistics beyond described normal value coboundary and the determined range of normal value of described normal value lower boundary are abnormal data, if there is abnormal data, then execute exception calling module, wherein, described N is the default natural number being greater than 1;
Exception call module, for: abnormal data is supplied to party in request as application interface and calls.
7. the abnormality detection system of ecommerce time series data according to claim 6, it is characterized in that, also comprise: if the data of described electronic commerce data continuous N phase perform benchmaring module do not detect abnormal data, then using the data of described electronic commerce data continuous N phase as combine detection data, perform combine detection module, wherein M is the default natural number being greater than 1, described combine detection module, sequential trend analysis or sequential causal inference are comprised to described combine detection data, and the abnormal data that sequential trend analysis or sequential causal inference obtain is carried out backtracking breakpoint analysis.
8. the abnormality detection system of ecommerce time series data according to claim 7, is characterized in that,
Described sequential trend analysis comprises: be linear growth trend, rapid growth trend, periodically rising tendency by described combine detection data based on Time Series, choose from described electronic commerce data do not meet described linear growth trend, rapid growth trend, periodically rising tendency data as abnormal data;
Described sequential causal inference comprises: from described combine detection data, select the first data group and the second data group, described first data group and the second data group have the probability distribution of identical type, the data variation scope calculating the first data group, as normal data variation range, will exceed the data of described normal data variation range as abnormal data in the second data group.
9. the abnormality detection system of ecommerce time series data according to claim 7, it is characterized in that, described backtracking breakpoint analysis comprises: using time point corresponding for the abnormal data of described sequential trend analysis or described sequential causal inference as current point in time t
nowforward trace, between the front average D1 of the combine detection data at every turn between more each time point proparea in range1 and time point back zone, in range2, the time interval of rear average D2, range1 and the range2 of combine detection data is identical, if time point t
forefront average and the change of rear average exceed predetermined threshold value, then think time point t
foreproparea between range1 and back zone the data of range2 have exception, from time point t
foreto t
nowcombine detection data in time period are as abnormal data.
10. the abnormality detection system of ecommerce time series data according to claim 6, is characterized in that, described fractile statistics, specifically comprises:
To window statistics by size of data sequence, the median of calculation window statistics, upper quartile, lower quartile, described median is the data being in all data centre positions after the sequence of window statistics, described upper quartile is the data being in all data 1/4th positions after the sequence of window statistics, described lower quartile is the data for all data 3/4ths positions after the sequence of window statistics, calculate the absolute value of the difference of lower quartile and upper quartile as interquartile-range IQR, determine that described normal value upper boundary values is that median deducts k times of interquartile-range IQR, determine that described normal value lower border value is that median adds k times of interquartile-range IQR, wherein, described k be greater than 1 natural number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510342240.4A CN104915846A (en) | 2015-06-18 | 2015-06-18 | Electronic commerce time sequence data anomaly detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510342240.4A CN104915846A (en) | 2015-06-18 | 2015-06-18 | Electronic commerce time sequence data anomaly detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104915846A true CN104915846A (en) | 2015-09-16 |
Family
ID=54084892
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510342240.4A Pending CN104915846A (en) | 2015-06-18 | 2015-06-18 | Electronic commerce time sequence data anomaly detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104915846A (en) |
Cited By (48)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105653129A (en) * | 2015-12-29 | 2016-06-08 | 江苏飞尚安全监测咨询有限公司 | Classic algorithm based real-time signal discrimination and correction method |
| CN105808368A (en) * | 2016-03-15 | 2016-07-27 | 南京联成科技发展有限公司 | Information security abnormity detection method and system based on random probability distribution |
| CN106209496A (en) * | 2016-06-23 | 2016-12-07 | 珠海格力电器股份有限公司 | Communication abnormity detection device and method |
| CN106227640A (en) * | 2016-07-27 | 2016-12-14 | 北京市环境保护监测中心 | A kind of abnormal deviation data examination method based on automatic monitor and system |
| CN106339316A (en) * | 2016-08-24 | 2017-01-18 | 上海爱企网络科技有限公司 | Method and device for diagnosing code segment in e-commerce platform |
| CN106548035A (en) * | 2016-11-24 | 2017-03-29 | 腾讯科技（深圳）有限公司 | A kind of diagnostic method and device of data exception |
| CN106815452A (en) * | 2015-11-27 | 2017-06-09 | 苏宁云商集团股份有限公司 | A kind of cheat detection method and device |
| CN106933691A (en) * | 2017-03-09 | 2017-07-07 | 深圳市彬讯科技有限公司 | A kind of recognition methods and system for abnormity point in seasonal time series data sequence |
| CN107204875A (en) * | 2017-05-11 | 2017-09-26 | 腾讯科技（深圳）有限公司 | Data reporting links monitoring method, device, electronic equipment and storage medium |
| CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
| CN107729293A (en) * | 2017-09-27 | 2018-02-23 | 中南大学 | A kind of geographical space method for detecting abnormal based on Multivariate adaptive regression splines |
| CN108154342A (en) * | 2017-12-25 | 2018-06-12 | 苏州大学 | Intelligent bus data collaborative method and its system based on cloud storage |
| CN108320810A (en) * | 2018-04-11 | 2018-07-24 | 平安科技（深圳）有限公司 | Disease abnormal deviation data examination method and device, computer installation and storage medium |
| CN108536777A (en) * | 2018-03-28 | 2018-09-14 | 联想(北京)有限公司 | A kind of data processing method, server cluster and data processing equipment |
| CN108804703A (en) * | 2018-06-19 | 2018-11-13 | 北京焦点新干线信息技术有限公司 | A kind of data exception detection method and device |
| CN108900478A (en) * | 2018-06-11 | 2018-11-27 | 阿里巴巴集团控股有限公司 | The detection method and device of unusual fluctuation attack, safety protection equipment |
| CN109189827A (en) * | 2018-08-16 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Time Series Processing method and apparatus, electronic equipment |
| CN109861857A (en) * | 2019-01-28 | 2019-06-07 | 网联清算有限公司 | Fault detection method and device |
| CN110071913A (en) * | 2019-03-26 | 2019-07-30 | 同济大学 | A kind of time series method for detecting abnormality based on unsupervised learning |
| CN110083638A (en) * | 2019-04-24 | 2019-08-02 | 广东联合电子服务股份有限公司 | A kind of regular base construction method of delay and data retention analysis method |
| CN110163457A (en) * | 2018-02-14 | 2019-08-23 | 北京京东尚科信息技术有限公司 | A kind of abnormal localization method and device of operational indicator |
| CN110334849A (en) * | 2019-05-30 | 2019-10-15 | 深圳壹账通智能科技有限公司 | Resource method for early warning, device and computer equipment based on data monitoring |
| CN110348718A (en) * | 2019-06-28 | 2019-10-18 | 北京淇瑀信息科技有限公司 | Financial business index monitoring method, device and electronic equipment |
| CN110399903A (en) * | 2019-06-28 | 2019-11-01 | 上海上湖信息技术有限公司 | The detection method and device of abnormal data, computer readable storage medium |
| CN110399400A (en) * | 2018-10-31 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Detect method, apparatus, equipment and the medium of abnormal data |
| CN110474862A (en) * | 2018-05-10 | 2019-11-19 | 中移(苏州)软件技术有限公司 | A kind of network flow abnormal detecting method and device |
| CN110751371A (en) * | 2019-09-20 | 2020-02-04 | 苏宁云计算有限公司 | Commodity inventory risk early warning method and system based on statistical four-bit distance and computer readable storage medium |
| CN111241158A (en) * | 2020-01-07 | 2020-06-05 | 清华大学 | Anomaly detection method and device for aircraft telemetry data |
| CN111382944A (en) * | 2020-03-11 | 2020-07-07 | 江苏苏宁物流有限公司 | Job behavior risk identification method and device, computer equipment and storage medium |
| CN111382919A (en) * | 2018-12-28 | 2020-07-07 | 顺丰科技有限公司 | Operation index automatic monitoring method and device |
| US10769658B2 (en) | 2016-09-30 | 2020-09-08 | International Business Machines Corporation | Automatic detection of anomalies in electronic communications |
| CN111650346A (en) * | 2020-07-14 | 2020-09-11 | 中科三清科技有限公司 | Automatic checking method and device for atmospheric pollution monitoring data and electronic equipment |
| CN111755125A (en) * | 2020-07-07 | 2020-10-09 | 医渡云（北京）技术有限公司 | Method, device, medium and electronic device for analyzing patient measurement index |
| CN111783804A (en) * | 2019-04-04 | 2020-10-16 | 中国移动通信集团上海有限公司 | Abnormal call bill determining method, device, equipment and storage medium |
| CN111984930A (en) * | 2020-08-19 | 2020-11-24 | 河海大学 | Method and system for identifying abnormal value of underground water level monitoring data |
| CN112116014A (en) * | 2020-09-24 | 2020-12-22 | 贵州电网有限责任公司 | Test data outlier detection method for distribution automation equipment |
| CN112651539A (en) * | 2019-10-12 | 2021-04-13 | 富士通株式会社 | Information processing apparatus, information processing method, and computer program |
| CN112926906A (en) * | 2019-12-06 | 2021-06-08 | 顺丰科技有限公司 | Method, device, storage medium and equipment for distributing transfer shifts of cargo item receiving and dispatching points |
| CN113076349A (en) * | 2020-01-06 | 2021-07-06 | 阿里巴巴集团控股有限公司 | Data anomaly detection method, device and system and electronic equipment |
| CN113159597A (en) * | 2021-04-29 | 2021-07-23 | 浙江陀曼云计算有限公司 | Machine tool production interval calculation method and system based on time sequence data fluctuation |
| CN113434823A (en) * | 2021-08-26 | 2021-09-24 | 企查查科技有限公司 | Data acquisition task abnormity early warning method and device, computer equipment and medium |
| CN113742118A (en) * | 2020-05-29 | 2021-12-03 | Ovh公司 | Method and system for detecting anomalies in a data pipeline |
| CN114493230A (en) * | 2022-01-20 | 2022-05-13 | 清华大学 | Abnormal service data processing method and device and storage medium |
| CN114780400A (en) * | 2022-04-18 | 2022-07-22 | 南京安元科技有限公司 | Method for blocking cyclic calling among services based on periodic data balance statistics |
| CN115168159A (en) * | 2022-09-06 | 2022-10-11 | 北京达佳互联信息技术有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and storage medium |
| CN117093917A (en) * | 2023-10-18 | 2023-11-21 | 北京天泽智云科技有限公司 | Data anomaly detection method and system based on fusion cumulative amount |
| CN117648232A (en) * | 2023-12-11 | 2024-03-05 | 武汉天宝莱信息技术有限公司 | Application program data monitoring method, device and storage medium |
| CN118624840A (en) * | 2024-08-12 | 2024-09-10 | 浙江华消科技有限公司 | Calibration method and device of gas detection equipment, storage medium and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100199285A1 (en) * | 2009-02-05 | 2010-08-05 | Vmware, Inc. | Virtual machine utility computing method and system |
| CN102945320A (en) * | 2012-10-29 | 2013-02-27 | 河海大学 | Time series data abnormity detection method and device |
| CN103561418A (en) * | 2013-11-07 | 2014-02-05 | 东南大学 | Anomaly detection method based on time series |
| CN103631681A (en) * | 2013-12-10 | 2014-03-12 | 国家电网公司 | Method for online restoring abnormal data of wind power plant |
| CN104350471A (en) * | 2012-06-28 | 2015-02-11 | 国际商业机器公司 | Detecting anomalies in real-time in multiple time series data with automated thresholding |
-
2015
- 2015-06-18 CN CN201510342240.4A patent/CN104915846A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100199285A1 (en) * | 2009-02-05 | 2010-08-05 | Vmware, Inc. | Virtual machine utility computing method and system |
| CN104350471A (en) * | 2012-06-28 | 2015-02-11 | 国际商业机器公司 | Detecting anomalies in real-time in multiple time series data with automated thresholding |
| CN102945320A (en) * | 2012-10-29 | 2013-02-27 | 河海大学 | Time series data abnormity detection method and device |
| CN103561418A (en) * | 2013-11-07 | 2014-02-05 | 东南大学 | Anomaly detection method based on time series |
| CN103631681A (en) * | 2013-12-10 | 2014-03-12 | 国家电网公司 | Method for online restoring abnormal data of wind power plant |
Non-Patent Citations (3)
| Title |
|---|
| 吴林: "一种基于趋势分析的网络性能异常检测方法", 《科学技术与工程》 * |
| 金中淦: "四分位数稳健统计和迭代稳健统计在室间质量评价结果分析中的比较", 《临床检验杂志》 * |
| 陈运文: "基于异常检测的时间序列研究", 《计算机技术与发展》 * |
Cited By (72)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815452A (en) * | 2015-11-27 | 2017-06-09 | 苏宁云商集团股份有限公司 | A kind of cheat detection method and device |
| CN105653129A (en) * | 2015-12-29 | 2016-06-08 | 江苏飞尚安全监测咨询有限公司 | Classic algorithm based real-time signal discrimination and correction method |
| CN105808368A (en) * | 2016-03-15 | 2016-07-27 | 南京联成科技发展有限公司 | Information security abnormity detection method and system based on random probability distribution |
| CN105808368B (en) * | 2016-03-15 | 2019-04-30 | 南京联成科技发展股份有限公司 | A kind of method and system of the information security abnormality detection based on random probability distribution |
| CN106209496A (en) * | 2016-06-23 | 2016-12-07 | 珠海格力电器股份有限公司 | Communication abnormity detection device and method |
| CN106227640A (en) * | 2016-07-27 | 2016-12-14 | 北京市环境保护监测中心 | A kind of abnormal deviation data examination method based on automatic monitor and system |
| CN106227640B (en) * | 2016-07-27 | 2019-03-19 | 北京市环境保护监测中心 | A kind of abnormal deviation data examination method and system based on automatic monitor |
| CN106339316A (en) * | 2016-08-24 | 2017-01-18 | 上海爱企网络科技有限公司 | Method and device for diagnosing code segment in e-commerce platform |
| CN106339316B (en) * | 2016-08-24 | 2019-01-11 | 上海爱企网络科技有限公司 | A kind of method and device that code segment is diagnosed in e-commerce platform |
| US10769658B2 (en) | 2016-09-30 | 2020-09-08 | International Business Machines Corporation | Automatic detection of anomalies in electronic communications |
| CN106548035A (en) * | 2016-11-24 | 2017-03-29 | 腾讯科技（深圳）有限公司 | A kind of diagnostic method and device of data exception |
| CN106548035B (en) * | 2016-11-24 | 2019-08-06 | 腾讯科技（深圳）有限公司 | A kind of diagnostic method and device of data exception |
| CN106933691B (en) * | 2017-03-09 | 2020-11-10 | 深圳市彬讯科技有限公司 | Identification method and system for abnormal points in seasonal time series data sequence |
| CN106933691A (en) * | 2017-03-09 | 2017-07-07 | 深圳市彬讯科技有限公司 | A kind of recognition methods and system for abnormity point in seasonal time series data sequence |
| CN107204875A (en) * | 2017-05-11 | 2017-09-26 | 腾讯科技（深圳）有限公司 | Data reporting links monitoring method, device, electronic equipment and storage medium |
| CN107729293B (en) * | 2017-09-27 | 2019-11-12 | 中南大学 | A kind of geographical space method for detecting abnormal based on Multivariate adaptive regression splines |
| CN107729293A (en) * | 2017-09-27 | 2018-02-23 | 中南大学 | A kind of geographical space method for detecting abnormal based on Multivariate adaptive regression splines |
| CN107657288B (en) * | 2017-10-26 | 2020-07-03 | 国网冀北电力有限公司 | Power dispatching flow data anomaly detection method based on isolated forest algorithm |
| CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
| CN108154342A (en) * | 2017-12-25 | 2018-06-12 | 苏州大学 | Intelligent bus data collaborative method and its system based on cloud storage |
| CN110163457A (en) * | 2018-02-14 | 2019-08-23 | 北京京东尚科信息技术有限公司 | A kind of abnormal localization method and device of operational indicator |
| CN108536777A (en) * | 2018-03-28 | 2018-09-14 | 联想(北京)有限公司 | A kind of data processing method, server cluster and data processing equipment |
| CN108320810A (en) * | 2018-04-11 | 2018-07-24 | 平安科技（深圳）有限公司 | Disease abnormal deviation data examination method and device, computer installation and storage medium |
| CN110474862B (en) * | 2018-05-10 | 2021-08-13 | 中移(苏州)软件技术有限公司 | Network traffic anomaly detection method and device |
| CN110474862A (en) * | 2018-05-10 | 2019-11-19 | 中移(苏州)软件技术有限公司 | A kind of network flow abnormal detecting method and device |
| CN108900478A (en) * | 2018-06-11 | 2018-11-27 | 阿里巴巴集团控股有限公司 | The detection method and device of unusual fluctuation attack, safety protection equipment |
| CN108804703B (en) * | 2018-06-19 | 2021-09-17 | 北京焦点新干线信息技术有限公司 | Data anomaly detection method and device |
| CN108804703A (en) * | 2018-06-19 | 2018-11-13 | 北京焦点新干线信息技术有限公司 | A kind of data exception detection method and device |
| CN109189827B (en) * | 2018-08-16 | 2022-04-15 | 创新先进技术有限公司 | Time sequence processing method and device and electronic equipment |
| CN109189827A (en) * | 2018-08-16 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Time Series Processing method and apparatus, electronic equipment |
| CN110399400B (en) * | 2018-10-31 | 2023-08-15 | 腾讯科技（深圳）有限公司 | Method, device, equipment and medium for detecting abnormal data |
| CN110399400A (en) * | 2018-10-31 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Detect method, apparatus, equipment and the medium of abnormal data |
| CN111382919A (en) * | 2018-12-28 | 2020-07-07 | 顺丰科技有限公司 | Operation index automatic monitoring method and device |
| CN109861857A (en) * | 2019-01-28 | 2019-06-07 | 网联清算有限公司 | Fault detection method and device |
| CN110071913A (en) * | 2019-03-26 | 2019-07-30 | 同济大学 | A kind of time series method for detecting abnormality based on unsupervised learning |
| CN111783804B (en) * | 2019-04-04 | 2023-11-24 | 中国移动通信集团上海有限公司 | Abnormal ticket determining method, device, equipment and storage medium |
| CN111783804A (en) * | 2019-04-04 | 2020-10-16 | 中国移动通信集团上海有限公司 | Abnormal call bill determining method, device, equipment and storage medium |
| CN110083638A (en) * | 2019-04-24 | 2019-08-02 | 广东联合电子服务股份有限公司 | A kind of regular base construction method of delay and data retention analysis method |
| CN110334849A (en) * | 2019-05-30 | 2019-10-15 | 深圳壹账通智能科技有限公司 | Resource method for early warning, device and computer equipment based on data monitoring |
| CN110348718B (en) * | 2019-06-28 | 2023-11-14 | 北京淇瑀信息科技有限公司 | Service index monitoring method and device and electronic equipment |
| CN110399903A (en) * | 2019-06-28 | 2019-11-01 | 上海上湖信息技术有限公司 | The detection method and device of abnormal data, computer readable storage medium |
| CN110348718A (en) * | 2019-06-28 | 2019-10-18 | 北京淇瑀信息科技有限公司 | Financial business index monitoring method, device and electronic equipment |
| CN110399903B (en) * | 2019-06-28 | 2021-07-13 | 上海上湖信息技术有限公司 | Abnormal data detection method and device and computer readable storage medium |
| CN110751371A (en) * | 2019-09-20 | 2020-02-04 | 苏宁云计算有限公司 | Commodity inventory risk early warning method and system based on statistical four-bit distance and computer readable storage medium |
| WO2021052031A1 (en) * | 2019-09-20 | 2021-03-25 | 苏宁云计算有限公司 | Statistical interquartile range-based commodity inventory risk early warning method and system, and computer readable storage medium |
| CN112651539A (en) * | 2019-10-12 | 2021-04-13 | 富士通株式会社 | Information processing apparatus, information processing method, and computer program |
| CN112926906B (en) * | 2019-12-06 | 2023-06-13 | 顺丰科技有限公司 | Method, device, storage medium and equipment for allocating connection shifts of goods receiving and dispatching points |
| CN112926906A (en) * | 2019-12-06 | 2021-06-08 | 顺丰科技有限公司 | Method, device, storage medium and equipment for distributing transfer shifts of cargo item receiving and dispatching points |
| CN113076349A (en) * | 2020-01-06 | 2021-07-06 | 阿里巴巴集团控股有限公司 | Data anomaly detection method, device and system and electronic equipment |
| CN113076349B (en) * | 2020-01-06 | 2024-06-11 | 阿里巴巴集团控股有限公司 | Data anomaly detection method, device and system and electronic equipment |
| CN111241158A (en) * | 2020-01-07 | 2020-06-05 | 清华大学 | Anomaly detection method and device for aircraft telemetry data |
| CN111241158B (en) * | 2020-01-07 | 2020-09-18 | 清华大学 | Anomaly detection method and device for aircraft telemetry data |
| CN111382944A (en) * | 2020-03-11 | 2020-07-07 | 江苏苏宁物流有限公司 | Job behavior risk identification method and device, computer equipment and storage medium |
| CN113742118B (en) * | 2020-05-29 | 2024-04-26 | Ovh公司 | Method and system for detecting anomalies in data pipes |
| CN113742118A (en) * | 2020-05-29 | 2021-12-03 | Ovh公司 | Method and system for detecting anomalies in a data pipeline |
| CN111755125A (en) * | 2020-07-07 | 2020-10-09 | 医渡云（北京）技术有限公司 | Method, device, medium and electronic device for analyzing patient measurement index |
| CN111755125B (en) * | 2020-07-07 | 2024-04-23 | 医渡云（北京）技术有限公司 | Method, device, medium and electronic equipment for analyzing patient measurement index |
| CN111650346A (en) * | 2020-07-14 | 2020-09-11 | 中科三清科技有限公司 | Automatic checking method and device for atmospheric pollution monitoring data and electronic equipment |
| CN111984930A (en) * | 2020-08-19 | 2020-11-24 | 河海大学 | Method and system for identifying abnormal value of underground water level monitoring data |
| CN112116014A (en) * | 2020-09-24 | 2020-12-22 | 贵州电网有限责任公司 | Test data outlier detection method for distribution automation equipment |
| CN113159597B (en) * | 2021-04-29 | 2022-08-16 | 浙江陀曼云计算有限公司 | Machine tool production interval calculation method and system based on time sequence data fluctuation |
| CN113159597A (en) * | 2021-04-29 | 2021-07-23 | 浙江陀曼云计算有限公司 | Machine tool production interval calculation method and system based on time sequence data fluctuation |
| CN113434823B (en) * | 2021-08-26 | 2021-11-16 | 企查查科技有限公司 | Data acquisition task abnormity early warning method and device, computer equipment and medium |
| CN113434823A (en) * | 2021-08-26 | 2021-09-24 | 企查查科技有限公司 | Data acquisition task abnormity early warning method and device, computer equipment and medium |
| CN114493230A (en) * | 2022-01-20 | 2022-05-13 | 清华大学 | Abnormal service data processing method and device and storage medium |
| CN114780400B (en) * | 2022-04-18 | 2023-06-27 | 南京安元科技有限公司 | Method for blocking cyclic call between services based on periodic data balance statistics |
| CN114780400A (en) * | 2022-04-18 | 2022-07-22 | 南京安元科技有限公司 | Method for blocking cyclic calling among services based on periodic data balance statistics |
| CN115168159A (en) * | 2022-09-06 | 2022-10-11 | 北京达佳互联信息技术有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and storage medium |
| CN117093917A (en) * | 2023-10-18 | 2023-11-21 | 北京天泽智云科技有限公司 | Data anomaly detection method and system based on fusion cumulative amount |
| CN117648232A (en) * | 2023-12-11 | 2024-03-05 | 武汉天宝莱信息技术有限公司 | Application program data monitoring method, device and storage medium |
| CN117648232B (en) * | 2023-12-11 | 2024-05-24 | 武汉天宝莱信息技术有限公司 | Application program data monitoring method, device and storage medium |
| CN118624840A (en) * | 2024-08-12 | 2024-09-10 | 浙江华消科技有限公司 | Calibration method and device of gas detection equipment, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104915846A (en) | Electronic commerce time sequence data anomaly detection method and system | |
| WO2021052031A1 (en) | Statistical interquartile range-based commodity inventory risk early warning method and system, and computer readable storage medium | |
| US11182394B2 (en) | Performing database file management using statistics maintenance and column similarity | |
| CN107086944A (en) | A kind of method for detecting abnormality and device | |
| US20060167704A1 (en) | Computer system and method for business data processing | |
| CN108846691A (en) | Regional grain and oil market price monitoring analysing and predicting system and monitoring method | |
| US20130085801A1 (en) | Supply Chain Performance Management Tool Having Predictive Capabilities | |
| CN111324639A (en) | Data monitoring method and device and computer readable storage medium | |
| Kusrini | Grouping of retail items by using K-means clustering | |
| US20090327049A1 (en) | Forecasting discovery costs based on complex and incomplete facts | |
| Yang et al. | Optimal acquisition policy in remanufacturing under general core quality distributions | |
| US20160132798A1 (en) | Service-level agreement analysis | |
| KR101886943B1 (en) | A method for establishing power supply and demand early warning model based on dynamic clustering | |
| Salamai et al. | Risk identification-based association rule mining for supply chain big data | |
| CN115330531B (en) | Enterprise risk prediction method based on electricity consumption fluctuation period | |
| Ruifeng et al. | Increasing production rate in Kanban controlled assembly lines through preventive maintenance | |
| CN112541635A (en) | Service data statistical prediction method and device, computer equipment and storage medium | |
| CN117397204A (en) | Machine learning time series anomaly detection | |
| CN116126642A (en) | Information processing method, device, equipment and storage medium | |
| CN105005575A (en) | Quick developing interface method for enterprise intelligent prediction | |
| CN110874640B (en) | Distribution selection and simulation of intermittent data using machine learning (ML) | |
| CN117252624A (en) | Intelligent cost accounting management system and method | |
| Asghari Torkamani et al. | Estimating the change point of correlated Poisson count processes | |
| Li et al. | iMiner: mining inventory data for intelligent management | |
| Castiñeira et al. | A new approach for fast evaluations of large portfolios of oil and gas fields |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150916 |
|
| RJ01 | Rejection of invention patent application after publication |