CN112308414A - Income abnormity detection method and device, electronic equipment and storage medium - Google Patents
Income abnormity detection method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112308414A CN112308414A CN202011192328.XA CN202011192328A CN112308414A CN 112308414 A CN112308414 A CN 112308414A CN 202011192328 A CN202011192328 A CN 202011192328A CN 112308414 A CN112308414 A CN 112308414A
- Authority
- CN
- China
- Prior art keywords
- income
- time sequence
- historical
- revenue
- sequence
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 63
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 74
- 238000000034 method Methods 0.000 claims abstract description 44
- 230000002159 abnormal effect Effects 0.000 claims abstract description 35
- 230000000737 periodic effect Effects 0.000 claims description 42
- 238000012549 training Methods 0.000 claims description 26
- 238000004590 computer program Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000004891 communication Methods 0.000 abstract description 6
- 235000019580 granularity Nutrition 0.000 description 10
- 230000006870 function Effects 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- IWIANZLCJVYEFX-RYUDHWBXSA-N Pro-Phe Chemical compound C([C@@H](C(=O)O)NC(=O)[C@H]1NCCC1)C1=CC=CC=C1 IWIANZLCJVYEFX-RYUDHWBXSA-N 0.000 description 5
- 150000001875 compounds Chemical class 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 229920006395 saturated elastomer Polymers 0.000 description 3
- 230000003442 weekly effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 229960001948 caffeine Drugs 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000037211 monthly cycles Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- RYYVLZVUVIJVGH-UHFFFAOYSA-N trimethylxanthine Natural products CN1C(=O)N(C)C(=O)C2=C1N=CN2C RYYVLZVUVIJVGH-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Strategic Management (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Operations Research (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Educational Administration (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Probability & Statistics with Applications (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention relates to the field of big data of communication services, and discloses a method and a device for detecting income abnormity, electronic equipment and a storage medium. The method for detecting the income abnormity comprises the following steps: according to a preset time sequence decomposition model and a historical time period before the current day, obtaining a decomposition result of a historical income time sequence in the historical time period, wherein the decomposition result comprises: historical trend item time sequence, historical period item time sequence and historical holiday item time sequence; acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result; acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model; and detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and acquiring a detection result. By adopting the embodiment, the income abnormal condition at any time interval can be accurately detected, the detection accuracy is high, and the speed is high.
Description
Technical Field
The embodiment of the invention relates to the field of big data of communication services, in particular to a method and a device for detecting income abnormity, electronic equipment and a storage medium.
Background
As the mobile communication market is increasingly competitive, how to realize revenue growth is an important issue for each large operator. Meanwhile, the income of an enterprise has great influence on network construction investment, business development planning and the like, so the income is used as a core key index of an operator, scientific and effective evaluation is realized on the current state of the operator, the future development condition of the operator is accurately predicted, and in-time warning or early warning is carried out on abnormal fluctuation, thereby having great significance on operation decision, business operation and risk management.
The inventors found that at least the following problems exist in the related art: at present, an income abnormity detection model is established in a machine learning mode or a deep learning mode to detect abnormal income conditions, a large amount of historical data is needed to be used as training data in both the machine learning mode and the deep learning mode, and otherwise, the income abnormity detection model is low in accuracy. However, in some cases, training is not performed without a large amount of revenue history data, which results in low accuracy of the anomaly detection model.
Disclosure of Invention
The embodiment of the invention aims to provide a method, a device, electronic equipment and a storage medium for detecting income abnormity, which can accurately detect the income abnormity condition in any time period, and have high detection accuracy and high speed.
To solve the above technical problem, an embodiment of the present invention provides a method for detecting an abnormal income, including: according to a preset time sequence decomposition model and a historical time period before the current day, obtaining a decomposition result of a historical income time sequence in the historical time period, wherein the decomposition result comprises: the time sequence decomposition model is obtained by training a Prophet model based on a historical income time sequence; acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result; acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model; and detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and acquiring a detection result.
An embodiment of the present invention further provides an apparatus for detecting an abnormal income, including: the device comprises a first acquisition module, a second acquisition module, a third acquisition module and a detection module; the first obtaining module is used for obtaining a decomposition result of a historical income time sequence in a historical time period according to a preset time sequence decomposition model and the historical time period before the current day, and the decomposition result comprises: the time sequence decomposition model is obtained by training a Prophet model based on a historical income time sequence; the second acquisition module is used for acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result; the third acquisition module is used for acquiring a revenue fluctuation interval of the time period to be detected according to a revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model; the detection module is used for detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and obtaining a detection result.
An embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of revenue anomaly detection described above.
Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which when executed by a processor implements the method of revenue anomaly detection described above.
In the embodiment of the application, the preset time sequence decomposition model is obtained by training the Prophet model according to a historical income time sequence, and the Prophet model can decompose the historical income time sequence into 3 parts, namely a historical trend item time sequence based on the overall development of the historical income sequence, a historical period item time sequence used for reflecting the influence of a periodic rule on the historical income and a holiday item time sequence used for reflecting the influence of holiday factors on the historical income; fully analyzing the rule in the historical income time sequence, further acquiring an income fluctuation threshold according to the decomposition result, and acquiring an income fluctuation interval of a time period to be measured according to the income fluctuation threshold; the income fluctuation threshold is related to the decomposition result, so that the determined income fluctuation threshold is more reasonable, and the income fluctuation area of the period to be measured is related to the period to be measured and the income fluctuation threshold instead of a fixed income fluctuation interval, so that the determined income fluctuation interval based on the period to be measured is more accurate and reasonable; in addition, the Prophet model has low requirements on training data sets for training, the quantity of the training data is greatly less than that of a model for detecting the income abnormity based on deep learning, the threshold for using the income abnormity detection is lowered, the application scene of the income abnormity detection is improved, and the accuracy of the income detection of a communication operator can be improved.
Additionally, obtaining a revenue fluctuation threshold based on the historical revenue sequence and the decomposition result includes: acquiring a first average value time sequence of the historical periodic item, wherein the first average value time sequence is used for representing a time sequence of income average values corresponding to the periodic regularity in the historical period; taking the sum of the first average time sequence, the historical trend item time sequence and the historical holiday item time sequence as the historical income average time sequence in the historical time period; acquiring the historical income mean value time sequence and a difference sequence between the historical income time sequences; and acquiring the income fluctuation threshold according to the difference sequence. The income has a periodic rule, so that the income of the same month and date can be different in different years, and by acquiring a first average value sequence of the historical period items, the first average value sequence is used for representing the time sequence of the income average value corresponding to the periodic rule in the historical period; the influence of periodic regularity can be reflected more accurately, and the historical income mean value time sequence can be predicted more accurately by taking the sum of the first mean value time sequence, the historical trend item time sequence and the historical holiday item time sequence as the historical income mean value time sequence in the historical time period; and acquiring the historical income mean value time sequence and a difference sequence between the historical income time sequences, reflecting the difference between a predicted value and an actual historical income time sequence, and further determining an accurate income fluctuation threshold.
Additionally, the revenue fluctuation threshold includes: an upper threshold and a lower threshold; the obtaining the revenue fluctuation threshold value according to the difference sequence comprises: obtaining the statistical distribution of the difference sequences; obtaining a value corresponding to a first preset digit from the statistical distribution as an upper threshold; and acquiring a value corresponding to a second preset digit from the statistical distribution as a lower threshold, wherein the second preset digit is smaller than the first preset digit. The first preset digit and the second preset digit can be set as required, and then a flexible income fluctuation threshold value can be obtained.
In addition, the history cycle entry timing sequence includes: m periodic sequences, m being an integer greater than 0; the obtaining of the first mean time sequence of the history cycle items includes: obtaining respective average period mean value sequences of the m period sequences; and taking the sum of the average periodic mean sequences of the m periodic sequences as the first mean sequence. The cycle duration of each cycle sequence is different, and the different cycle durations cause the phenomenon that a plurality of cycle sequences are overlapped, the respective average cycle mean value sequence of each cycle sequence is determined, and the first mean value sequence can be accurately obtained.
In addition, obtaining an average cycle mean sequence of each of the m cycle sequences includes: dividing the periodic sequence into n subsequences according to the period length of the periodic sequence, wherein n is an integer greater than 0; acquiring a revenue average value corresponding to the same sampling time in each period length as an average value subsequence of each subsequence; and combining the n permutation subsequences of the mean value as an average periodic mean value sequence of the periodic sequence. The permutation and combination of the n average value subsequences make the sequence length of the permutation period consistent with the length of the historical period item sequence, and further improve the accuracy of obtaining the average period average value sequence of the m period sequences.
In addition, acquiring the income fluctuation interval of the time period to be detected according to the income fluctuation threshold, the time period to be detected and the time sequence decomposition model, and the method comprises the following steps: obtaining a trend item time sequence, a cycle item time sequence and a holiday item time sequence in the time period to be detected according to the time sequence decomposition model and the time period to be detected; obtaining the sum of the trend item time sequence, the cycle item time sequence, the holiday item time sequence and the upper limit threshold value in the period to be measured to obtain a fluctuating upper limit interval; and taking the sum of the trend item time sequence, the cycle item time sequence, the holiday item time sequence and the lower limit threshold value in the period to be measured as a lower limit interval of income fluctuation. And by adding the difference sequence, the income fluctuation interval can be accurately acquired.
In addition, whether the income in the time period to be detected is abnormal or not is detected according to the income fluctuation interval in the time period to be detected, and a detection result is obtained, wherein the method comprises the following steps: inputting the time period to be measured into the time sequence decomposition model, and acquiring a income time sequence corresponding to the time period to be measured; judging whether the income time sequence corresponding to the time period to be detected exceeds the income fluctuation interval or not; if so, the detection result indicates that the income time sequence corresponding to the time period to be detected is abnormal, otherwise, the detection result indicates that the income time sequence corresponding to the time period to be detected is normal. The income time sequence in any time period to be detected can be predicted through the time sequence decomposition model, and based on the acquired income time sequence in the time period to be detected, the abnormal detection can be carried out on the historical income, the income in the future time period after the current time period can be predicted, whether the income in the future time period is abnormal or not can be detected, no additional model needs to be trained, and the detection cost is reduced.
Drawings
One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.
Fig. 1 is a flow chart of a method of revenue anomaly detection provided in accordance with a first embodiment of the present invention;
FIG. 2 is a flow chart of a method of revenue anomaly detection provided in accordance with a second embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating the effect of an abnormal income detection method on an abnormal income detection during a period of time to be measured according to a second embodiment of the present invention;
FIG. 4 is a schematic diagram of obtaining a first mean sequence in a method for revenue anomaly detection according to a third embodiment of the present invention;
fig. 5 is a block diagram illustrating an apparatus for detecting an abnormal income traffic according to a fourth embodiment of the present invention;
fig. 6 is a block diagram of an electronic device according to a fifth embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments.
The following embodiments are divided for convenience of description, and should not constitute any limitation to the specific implementation manner of the present invention, and the embodiments may be mutually incorporated and referred to without contradiction.
A first embodiment of the present invention relates to a method of revenue anomaly detection. The flow is shown in figure 1:
step 101: according to a preset time sequence decomposition model and a historical time period before the current day, obtaining a decomposition result of a historical income time sequence in the historical time period, wherein the decomposition result comprises: historical trend item time sequence, historical cycle item time sequence and historical holiday item time sequence, the time sequence decomposition model is obtained by training the Prophet model based on the historical income time sequence.
Step 102: and acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result.
Step 103: and acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model.
Step 104: and detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and acquiring a detection result.
In the embodiment of the application, the preset time sequence decomposition model is obtained by training the Prophet model according to a historical income time sequence, and the Prophet model can decompose the historical income time sequence into 3 parts, namely a historical trend item time sequence based on the overall development of the historical income sequence, a historical period item time sequence used for reflecting the influence of a periodic rule on the historical income and a holiday item time sequence used for reflecting the influence of holiday factors on the historical income; fully analyzing the rule in the historical income time sequence, further acquiring an income fluctuation threshold according to the decomposition result, and acquiring an income fluctuation interval of a time period to be measured according to the income fluctuation threshold; the income fluctuation threshold is related to the decomposition result, so that the determined income fluctuation threshold is more reasonable, and the income fluctuation area of the period to be measured is related to the period to be measured and the income fluctuation threshold instead of a fixed income fluctuation interval, so that the determined income fluctuation interval based on the period to be measured is more accurate and reasonable; in addition, the Prophet model has low requirements on training data sets for training, the quantity of the training data is greatly less than that of a model for detecting the income abnormity based on deep learning, the threshold for using the income abnormity detection is lowered, the application scene of the income abnormity detection is improved, and the accuracy of the income detection of a communication operator can be improved.
A second embodiment of the present invention relates to a method of revenue anomaly detection. A second embodiment is a detailed description of the first embodiment, and the method for detecting income anomaly can be applied to an electronic device, and the flow is shown in fig. 2.
Step 201: according to a preset time sequence decomposition model and a historical time period before the current day, obtaining a decomposition result of a historical income time sequence in the historical time period, wherein the decomposition result comprises: historical trend item time sequence, historical cycle item time sequence and historical holiday item time sequence, the time sequence decomposition model is obtained by training the Prophet model based on the historical income time sequence.
The revenue anomaly detection method may be used to detect revenue conditions of an enterprise, such as revenue of a communications operator, revenue of the travel industry, and the like. The income time sequence has complex characteristics including long-term development trend and periodic regularity, wherein the time period comprises a year period, a month period and a week period, and the three periods are mutually overlapped; the revenue timing is also affected by random, random factors, such as major events, emergencies, national policies, and the like. The income indexes comprise a plurality of time granularities such as daily granularity, monthly granularity and the like, and the time granularity is the minimum time unit in the time sequence. While the nature and richness of the data varies at different time granularities. The modeling analysis is carried out by adopting a traditional time sequence model or a statistical model, the income data characteristic and rule description capacity is poor, and the accuracy is low. The complex deep learning method needs a large amount of historical data to be accumulated, has low applicability on monthly granularity income data, is easy to overfit, is used as a black box model, has poor interpretability of a result on business, cannot effectively assist a user to know the abnormal symptom of income, and cannot effectively solve the problem of abnormal income.
In this example, before step 201 is executed, a time sequence decomposition model obtained by training a Prophet model according to a historical revenue time sequence may be trained in advance. The training process of the time-series decomposition model is described in detail below.
The Prophet algorithm is an open-source time series prediction tool. The Prophet model is composed of three parts as a whole: data growth trend, cycle item trend and holiday influencing factors, and t is date. The form of the Prophe model is shown in formula (1):
y(t)＝g(t)+s(t)+h(t)+εtformula (1);
wherein g (t) representsA trend growth function used for fitting aperiodic changes of income prediction values in the time series and divided into saturated linear growth and piecewise linear growth; s (t) is used to represent the effect of periodic changes, say weekly, in each year, on revenue prediction; h (t) represents the influence of holidays with non-fixed periods in the time series on the revenue prediction value, εtExpressed as a noise term, representing a random factor that is not predicted by the Prophe model, which in this example may be assumed to be gaussian.
According to the Prophe model, the income time sequence can be regularly decomposed, different types of subsequences can be obtained, and g (t) in the Prophe model can be used for fitting the overall development trend of the income sequence to be an ascending period, a stationary period or a declining period and the overall change speed. s (t) can represent the superposition of a plurality of composite periods by a Fourier series modeling method, and the form of s (t) is shown in formula (2):
wherein, P represents a period expected in the income sequence, such as 365.25 days of annual period, 30.5 days of monthly period and 7 days of weekly period in the sequence with day granularity; and defining whether high-frequency change is considered in the Prophe model by using the Fourier order N, wherein the higher the value of the high-frequency change is, the higher the fitting precision is, and the excessive N value can generate overfitting. In this example, when the period is the income time sequence of the year, N may take a value of 10, if the period is the income time sequence of the month, the corresponding N takes a value of 5, and if the period is the income time sequence of the week, the corresponding N takes a value of 3. When the value of N is determined, the parameter β can be estimated as a set of parameters, i.e., β ═ a1,b1,…,aN,bN]T. Due to the characteristic of the various periodic overlaps in the revenue timing, in this example, s (t) of the various periodic overlaps can be expressed as in equation (3).
s(t)＝syear(t)+smonth(t)+sweek(t) formula (3);
h (t) represents a holiday factor, and includes its own factors such as temporary adjustment of income settlement means, in addition to external situations such as marketing activities, emergency policies, and spring festival.
The historical revenue timing over the historical period is denoted as yhistory(t) of (d). The parameters of g (t), s (t) and h (t) in the Prophet model can be solved by an L-BFGS optimization method.
In this example, the historical period refers to a period before the current day, and the revenue timing within the historical period may be collected; the historical time sequence can be day granularity, month granularity or year granularity, and the occurrence time and the influence time range of the holiday factors are recorded at the same time, wherein the holiday factors comprise events such as marketing activities, holiday factors and policies. The revenue schedule of the collection may be pre-processed.
The pretreatment process comprises the following steps: and judging whether the ratio of the missing time in the collected income time sequence to the whole time interval is smaller than a preset threshold value, if so, filling the default time sequence by adopting a difference method, and if the default ratio is larger than the preset threshold value, determining that the quality of the currently collected income time sequence is poor, and outputting prompt information for indicating to obtain the income time sequence again. Wherein the preset threshold may be set to 30%. And taking the income time sequence after preprocessing as a historical income time sequence.
In the Prophet model, the division of g (t) is saturated linear growth and piecewise linear growth. For stable revenue data, e.g., revenue for stable traffic, total revenue, etc.; if the revenue data is stable, a saturated linear growth function can be selected; the selection of the saturation linear growth function can be shown as equation (4):
the maximum capacity c (t) may be set as a constant according to the actual maximum income of the service, or may be set as a variation value with respect to the time t. For revenue data in a fast-developing period and a predicted future time period is short, a piecewise linear growth function may be selected, which may be shown in equation (5);
g(t)＝(k+a(t)Tδ)t+(m+a(t)Tγ) formula (5);
where k denotes a growth rate, δ and a both denote rate changes, m denotes a compensation parameter, and γ denotes a smoothing optimization parameter.
s (t) completing the parameters for different periodicities; and according to different periodic characteristics of different income indexes, the periodic characteristics are as follows: selecting a single period, two compound periods or a plurality of compound periods, and selecting corresponding periodic functions to be superposed, wherein the functions are expressed as annual compound periods, monthly compound periods and weekly compound periods as shown in a formula (6), a formula (7) and a formula (8);
for example, if s (t) includes a year cycle and a month cycle, s (t) is syear(t)+smonth(t)。
In the Prophet model, for events such as holiday factors, marketing activities, holiday factors, policies and the like, the time of occurrence of the event and the time windows of front and back influences can be set. If the twenty-one activity occurs on 11 months and 11 days, the influence range is the first 7 days and the last 3 days according to business experience.
According to the historical income time sequence and the selected function, the Prophet algorithm of python or R can be used for opening a source packet to complete the solving of the Prophet model parameters, and the solving can also be realized by programming by an L-BFGS optimization method; and obtaining a time sequence decomposition model.
Inputting the historical time interval into the time sequence decomposition model to obtain the historical income time sequence yhistory(t) decomposition results:historical trend item time sequence ghistory(t), historical period item sequence shistory(t) and historical holiday term sequence hhistory(t)。
It should be noted that the time sequence decomposition model can also be used for predicting the income in the time period after the current day, for example, the income time sequence in the future time period can be obtained by inputting the future time period t into the time sequence decomposition model, so as to realize the prediction of the income in the future time period.
It is worth mentioning that in the training process of training the Prophet model, the data amount of the collected income time sequence is small, for example, only income data of one to two months can be collected, the obtained time sequence decomposition model is accurate, the model training needs to be carried out by the income data at least taking years as units as a deep learning training mode, and the difficulty of the model training is greatly reduced.
Step 202: and acquiring a first average value time sequence of the historical periodic item, wherein the first average value time sequence is used for representing a time sequence of the income average value corresponding to the periodic regularity in the historical period.
Specifically, in this example, taking the history period item as a sequence of the period, the history period item time sequence is expressed as formula (9);
shistory(t)＝sweek(t) formula (9);
solving for the first sequence of means of the history period term, i.e.
The first mean sequence solving method may be: the time length of the historical income time sequence is L, and the period duration of the period item is 7 days, then the period time sequence can be divided into n-L/7 subsequences, where n is an integer. Calculating the average value of the same sampling time in each period as the average period subsequence
The average periodic subsequence can be used as in equation (10):
each one of which is
The length of (2) is 7, and if the length is the same as the length of L, the length can be
Filling the time sequence to the length of L to obtain
Taking the sequence of weeks as an example, 7 days are one cycle, one cycle from monday to sunday. And if the period sequence s comprises n periods, the average fluctuation condition of the monday is the average value of n mondays in the period sequence s, the average fluctuation condition of the tuesday is the average value of n tuesdays in the period sequence s, and so on, and the average fluctuation sequence of the monday is repeated to the length L. It should be noted that, for the unstable cycle length of month and year, the month cycle is 30 days and the year cycle is 365 days for convenience of calculation. The length can be eliminated or supplemented according to the characteristics of income data. For example, for a series of monthly cycles, the first 10 days of the month and the last 10 days of the month are retained, and the average of the income for the remaining intermediate days of the month is solved. When the remaining days in the middle of the month are 11 days, removing 1 day with the largest difference value with the average value, and supplementing the remaining days to 10 days by using the average value when the remaining days are less than 10 days; and removing 1 day with the largest difference value between the income average values of the 2 middle-of-month and the 2 middle-of-month from the income data of leap years.
Step 203: and taking the sum of the first average time sequence, the historical trend item time sequence and the historical holiday item time sequence as the historical income average time sequence in the historical period.
Specifically, the sum of the first average time sequence, the historical trend item time sequence and the historical holiday item time sequence is taken as the historical income average time sequence in the historical period, which can be expressed by formula (11):
the
Represented as a historical revenue mean time sequence,
is a first mean sequence.
Step 204: obtaining a historical revenue mean timing sequence and a sequence of differences between the historical revenue timing sequences.
Specifically, it can be shown as formula (12):
wherein e ishistory(t) represents the difference sequence.
Step 205: and acquiring a revenue fluctuation threshold according to the difference sequence.
In one example, a statistical distribution of the difference sequences is obtained; acquiring a first difference value between a first value corresponding to a first preset digit acquired in the statistical distribution and a value corresponding to a second preset digit in the statistical distribution, wherein the second preset digit is smaller than the first preset digit; taking the sum of the first value and the first difference as an upper threshold; a second difference between the second value and the first difference is taken as a lower threshold.
Specifically, the first preset number of bits is 3/4, and the second preset number of bits is 1/4; then the first predetermined number of bits corresponds to a value e3/4The value corresponding to the second predetermined digit is e1/4(ii) a The upper and lower threshold values may be obtained according to, for example, equation (13).
The upper threshold is expressed as ΔupperThe lower threshold is expressed as Δlower。
Step 206: and acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model.
In one example, a trend item time sequence and a holiday item time sequence in a time period to be detected are obtained according to the time sequence decomposition model and the time period to be detected; obtaining the sum of the trend item time sequence, the first mean value time sequence, the holiday item time sequence and the upper limit threshold value in the period to be measured to obtain a fluctuating upper limit interval; and taking the sum of the trend item time sequence, the first mean value time sequence, the holiday item time sequence and the lower limit threshold value in the period to be measured as a lower limit interval of income fluctuation.
Specifically, the period to be measured may be a specified period in a history period, or may be a specified period in a future period. If the period to be measured is a designated period in the historical period, the income fluctuation interval can be as shown in formula (14):
wherein the revenue fluctuation threshold [ Delta ]lower,Δupper]，
Expressed as the first mean value timing, [ y ]history_lower(t),yhistory_upper(t)]I.e. the revenue fluctuation interval for the historical period.
If the time interval to be measured is a designated time interval in the future time interval, the income fluctuation interval can be as shown in formula (15):
wherein, gfuture(t) is expressed as a trend term time sequence, h, for a future time periodfuture(t) is expressed as the holiday term timing for the future period, and the revenue fluctuation threshold is expressed as [ Delta ]lower,Δupper]，
Represented as a first mean value time sequence,[yfuture_lower(t),yfuture_upper(t)]i.e. the revenue fluctuation interval in the future time period.
Step 207: and detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and acquiring a detection result.
In one example, inputting the time period to be measured into a time sequence decomposition model, and acquiring a income time sequence corresponding to the time period to be measured; judging whether the income time sequence corresponding to the time period to be detected exceeds an income fluctuation interval or not; if so, the detection result indicates that the income time sequence corresponding to the time period to be detected is abnormal, otherwise, the detection result indicates that the income time sequence corresponding to the time period to be detected is normal.
Specifically, the historical income time sequence y (t) is compared with the income fluctuation interval in the historical time period, and if y (t) is greater than yhistory_upper(t) or less than yhistory_lowerAnd (t), outputting an alarm when the income time sequence corresponding to the time t is abnormal. Timing of future revenues yfuture(t) comparing with the income fluctuation interval of the future period if yfuture(t) is greater than yfuture_upper(t) or less than yfuture_lowerAnd (t), if the income time sequence at the future time t is abnormal, outputting early warning.
The anomaly detection is performed at the historical revenue sequence of the communication operator consummated in the present example below.
The total monthly revenue timing and holiday factors for the carrier from 2016 month 1 to 2019 month 5 are collected and partitioned into a training set and a test set. The historical income time sequence from 2016 to 2018 is used as a training set, and based on the method in the example, the income fluctuation interval of the historical time period and the income fluctuation interval of the future time period are acquired, and meanwhile, the income time sequence in the future time period can be acquired after the future time period is input into the time sequence decomposition model. And predicting the income of 2019, and completing the detection of abnormal fluctuation of the income time sequence of the historical time period and the income time sequence of the future time period.
A schematic diagram of results from 2018 and 1 to 2019 and 12 is shown in fig. 3, where a bold solid line represents an actual collected revenue time sequence, a thin solid line represents a revenue time sequence obtained through a time sequence decomposition model, that is, a fitting/predicted value, and a thin dotted line represents a revenue fluctuation interval.
In this example, the actual income of the month from 2019 to 2019, month 1 to 5, is used as a test set, and the actual income of the corresponding month is compared with the actual income of the corresponding month, and the accuracy is calculated. As shown in table 1, the predicted value is the income time sequence obtained based on the income anomaly detection method in this example, and it can be seen that the accuracy of the predicted value in this example is higher than 95%, and the accuracy is 1- | predicted value/true value-1 |;
| actual value | Prediction value | Rate of |
|
| 201901 | 2.31E+09 | 2.34E+09 | 98.55％ |
| 201902 | 2.08E+09 | 2.11E+09 | 98.55％ |
| 201903 | 2.36E+09 | 2.4E+09 | 98.05％ |
| 201904 | 2.24E+09 | 2.33E+09 | 95.65％ |
| 201905 | 2.29E+09 | 2.35E+09 | 97.46％ |
TABLE 1
The upper and lower limits of revenue variation in revenue abnormality detection for the period to be measured are shown in table 2.
TABLE 2
The model result value in table 2 refers to the revenue of the period to be measured obtained with the time-series decomposition model in this example. In historical income, the actual income of 11 months in 2018 is higher than the upper limit value of the income fluctuation interval, and the income is regarded as abnormal alarm; in the income of the future time period, the expected income of the months such as 2 months and 6 months in 2019 exceeds the upper and lower limits of the income fluctuation interval, and the early warning is regarded as abnormal early warning. Abnormal early warning is generated on the prediction income of 2 months in 2019, the actual income value also meets the condition of abnormal fluctuation, and prediction and reminding of potential risks can be realized on the basis of accurate prediction.
Therefore, the method and the device realize accurate prediction of future income, accurately realize abnormal detection of history and future income, and improve early warning of future risks.
The steps of the above methods are divided for clarity, and the implementation may be combined into one step or split some steps, and the steps are divided into multiple steps, so long as the same logical relationship is included, which are all within the protection scope of the present patent; it is within the scope of the patent to add insignificant modifications to the algorithms or processes or to introduce insignificant design changes to the core design without changing the algorithms or processes.
The third embodiment of the present invention relates to a method for detecting revenue anomaly, and is a detailed process for acquiring a first mean value timing sequence when the historical period item in step 202 includes a plurality of periods, and the flow thereof can be as shown in fig. 4.
Step 301: and acquiring an average period mean value sequence of the m period sequences, wherein m is an integer larger than 0.
In one example, the periodic sequence is divided into n subsequences according to the period length of the periodic sequence, wherein n is an integer greater than 0; acquiring income average values corresponding to the same sampling time in each period length as an average value subsequence of each subsequence; and taking the permutation and combination of the n mean subsequences as an average periodic mean sequence of the periodic sequence.
Specifically, the cycle duration of different cycle sequences is different, for example, the cycle duration of a cycle sequence is 7 days, the cycle length of a month cycle sequence is 30 days, and the cycle duration of an year cycle is 365 days.
In this example, three periodic sequences are included, i.e., s (t) syear(t)+smonth(t)+sweek(t), the first mean time sequence s (t) can be expressed as formula (16):
the
The obtaining method and solving the
In much the same way.
Assuming that the time length of the historical income time sequence is L, the period duration of each period sequence is piAccording to the period duration p of the period termiThen the cycle timing can be divided into n-L/piSubsequences, where n is an integer. The same sample time can be calculated in each cycleAverage value of moments as average periodic subsequence
The average periodic subsequence may be represented by formula (17):
wherein i can be given as year, month and week.
Step 302: and taking the sum of the average period mean sequences of the m period sequences as a first mean sequence.
Namely obtain
And
and then, overlapping the average periodic mean sequences to obtain a first mean sequence.
s(t)＝syear(t)+smonth(t)+sweek(t) formula (18).
A fourth embodiment of the present invention relates to an apparatus for detecting an abnormal income, wherein a block diagram of a configuration of the apparatus for detecting an abnormal income 40 is shown in fig. 5, and the apparatus includes: a first acquisition module 401, a second acquisition module 402, a third acquisition module 403, and a detection module 404.
The first obtaining module 401 is configured to obtain, according to a preset time sequence decomposition model and a historical time period before the current day, a decomposition result of a historical income time sequence in the historical time period, where the decomposition result includes: the time sequence decomposition model is obtained by training a Prophet model based on a historical income time sequence; the second obtaining module 402 is configured to obtain a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result; the third acquisition module is used for acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model; the detection module is used for detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and obtaining a detection result.
It should be understood that this embodiment is an example of the apparatus corresponding to the first embodiment, and may be implemented in cooperation with the first embodiment. The related technical details mentioned in the first embodiment are still valid in this embodiment, and are not described herein again in order to reduce repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the first embodiment.
It should be noted that each module referred to in this embodiment is a logical module, and in practical applications, one logical unit may be one physical unit, may be a part of one physical unit, and may be implemented by a combination of multiple physical units. In addition, in order to highlight the innovative part of the present invention, elements that are not so closely related to solving the technical problems proposed by the present invention are not introduced in the present embodiment, but this does not indicate that other elements are not present in the present embodiment.
A fifth embodiment of the present invention relates to an electronic apparatus, as shown in fig. 6, including: at least one processor 501; and a memory 502 communicatively coupled to the at least one processor 501; the memory 502 stores instructions executable by the at least one processor 501, and the instructions are executable by the at least one processor 501 to enable the at least one processor 501 to perform the above-described revenue anomaly detection method.
The memory 502 and the processor 501 are connected by a bus, which may include any number of interconnected buses and bridges that link one or more processors and various circuits of the memory together. The bus may also link various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. The data processed by the processor is transmitted over a wireless medium via an antenna, which further receives the data and transmits the data to the processor.
The processor is responsible for managing the bus and general processing and may also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. And the memory may be used to store data used by the processor in performing operations.
A sixth embodiment of the present invention relates to a computer-readable storage medium storing a computer program which, when executed by a processor, implements the method of revenue anomaly detection described above.
Those skilled in the art can understand that all or part of the steps in the method of the foregoing embodiments may be implemented by a program to instruct related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, etc.) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.
Claims (10)
1. A method of revenue anomaly detection, comprising:
according to a preset time sequence decomposition model and a historical time period before the current day, obtaining a decomposition result of a historical income time sequence in the historical time period, wherein the decomposition result comprises: the time sequence decomposition model is obtained by training a Prophet model based on the historical income time sequence;
acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result;
acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model;
and detecting whether the income in the time period to be detected is abnormal according to the income fluctuation interval in the time period to be detected, and acquiring a detection result.
2. The method of revenue anomaly detection according to claim 1, wherein said deriving a revenue fluctuation threshold based on said historical revenue timing and said decomposition results comprises:
acquiring a first average value time sequence of the historical periodic item, wherein the first average value time sequence is used for representing a time sequence of income average values corresponding to the periodic regularity in the historical period;
taking the sum of the first average time sequence, the historical trend item time sequence and the historical holiday item time sequence as the historical income average time sequence in the historical time period;
acquiring the historical income mean value time sequence and a difference sequence between the historical income time sequences;
and acquiring the income fluctuation threshold according to the difference sequence.
3. The method of revenue anomaly detection according to claim 2, wherein the revenue fluctuation threshold comprises: an upper threshold and a lower threshold;
the obtaining the revenue fluctuation threshold value according to the difference sequence comprises:
obtaining the statistical distribution of the difference sequences;
acquiring a first difference value between a first numerical value corresponding to a first preset digit acquired in the statistical distribution and a numerical value corresponding to a second preset digit in the statistical distribution, wherein the second preset digit is smaller than the first preset digit;
taking the sum of the first value and the first difference as an upper threshold;
and taking a second difference value between the second value and the first difference value as a lower threshold value.
4. The method of revenue anomaly detection according to claim 2 or 3, wherein the historical periodic term timing comprises: m periodic sequences, m being an integer greater than 0;
the obtaining of the first mean time sequence of the history cycle items includes:
obtaining respective average period mean value sequences of the m period sequences;
and taking the sum of the average periodic mean sequences of the m periodic sequences as the first mean sequence.
5. The method of revenue anomaly detection according to claim 4, wherein said obtaining an average cycle mean sequence for each of the m cycle sequences comprises:
dividing the periodic sequence into n subsequences according to the period length of the periodic sequence, wherein n is an integer greater than 0;
acquiring a revenue average value corresponding to the same sampling time in each period length as an average value subsequence of each subsequence;
and combining the n permutation subsequences of the mean value as an average periodic mean value sequence of the periodic sequence.
6. The method of revenue anomaly detection according to claim 3, wherein the obtaining a revenue fluctuation interval of the time period to be measured according to the revenue fluctuation threshold, the time period to be measured and the time sequence decomposition model comprises:
acquiring a trend item time sequence and a holiday item time sequence in the time period to be detected according to the time sequence decomposition model and the time period to be detected;
obtaining the sum of the trend item time sequence, the first mean value time sequence, the holiday item time sequence and the upper limit threshold value in the period to be measured to obtain a fluctuating upper limit interval;
and taking the sum of the trend item time sequence, the first mean value time sequence, the holiday item time sequence and the lower limit threshold value in the period to be measured as a lower limit interval of income fluctuation.
7. The method for detecting income anomaly according to claim 1, wherein the step of detecting whether the income in the time period to be detected is abnormal according to the income fluctuation interval in the time period to be detected and obtaining the detection result comprises the following steps:
inputting the time period to be measured into the time sequence decomposition model, and acquiring a income time sequence corresponding to the time period to be measured;
judging whether the income time sequence corresponding to the time period to be detected exceeds the income fluctuation interval or not; if so, the detection result indicates that the income time sequence corresponding to the time period to be detected is abnormal, otherwise, the detection result indicates that the income time sequence corresponding to the time period to be detected is normal.
8. An apparatus for revenue anomaly detection, comprising: the device comprises a first acquisition module, a second acquisition module, a third acquisition module and a detection module;
the first obtaining module is used for obtaining a decomposition result of a historical income time sequence in a historical time period according to a preset time sequence decomposition model and the historical time period before the current day, and the decomposition result comprises: the time sequence decomposition model is obtained by training a Prophet model based on the historical income time sequence;
the second acquisition module is used for acquiring a revenue fluctuation threshold according to the historical revenue time sequence and the decomposition result;
the third acquisition module is used for acquiring a revenue fluctuation interval of the time period to be detected according to the revenue fluctuation threshold, the time period to be detected and the time sequence decomposition model;
the detection module is used for detecting whether the income in the time period to be detected is abnormal or not according to the income fluctuation interval in the time period to be detected, and obtaining a detection result.
9. An electronic device, comprising:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method of revenue anomaly detection according to any of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method of revenue anomaly detection of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011192328.XA CN112308414A (en) | 2020-10-30 | 2020-10-30 | Income abnormity detection method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011192328.XA CN112308414A (en) | 2020-10-30 | 2020-10-30 | Income abnormity detection method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112308414A true CN112308414A (en) | 2021-02-02 |
Family
ID=74332856
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011192328.XA Pending CN112308414A (en) | 2020-10-30 | 2020-10-30 | Income abnormity detection method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112308414A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112965876A (en) * | 2021-03-10 | 2021-06-15 | 中国民航信息网络股份有限公司 | Monitoring alarm method and device |
| CN113537577A (en) * | 2021-06-25 | 2021-10-22 | 中国铁路广州局集团有限公司 | Revenue prediction method, system, electronic device, and computer-readable storage medium |
| CN114896124A (en) * | 2022-05-16 | 2022-08-12 | 华能澜沧江水电股份有限公司 | System for monitoring of reporting an emergency and asking for help or increased vigilance based on chronogenesis decomposes |
| CN115158399A (en) * | 2022-06-14 | 2022-10-11 | 通号城市轨道交通技术有限公司 | Time sequence signal abnormity detection method and system |
| CN115994248A (en) * | 2023-03-24 | 2023-04-21 | 青岛精锐机械制造有限公司 | Data detection method and system for valve faults |
| CN116343953A (en) * | 2023-05-30 | 2023-06-27 | 苏州绿华科技有限公司 | Intelligent community management system based on artificial intelligence |
| CN116342151A (en) * | 2023-04-12 | 2023-06-27 | 余珊 | Monetary payment device amount tampering detection system |
| CN118190276A (en) * | 2024-05-17 | 2024-06-14 | 菏泽市产品检验检测研究院 | Pressure vessel fault diagnosis method based on machine learning |
-
2020
- 2020-10-30 CN CN202011192328.XA patent/CN112308414A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112965876A (en) * | 2021-03-10 | 2021-06-15 | 中国民航信息网络股份有限公司 | Monitoring alarm method and device |
| CN113537577A (en) * | 2021-06-25 | 2021-10-22 | 中国铁路广州局集团有限公司 | Revenue prediction method, system, electronic device, and computer-readable storage medium |
| CN114896124A (en) * | 2022-05-16 | 2022-08-12 | 华能澜沧江水电股份有限公司 | System for monitoring of reporting an emergency and asking for help or increased vigilance based on chronogenesis decomposes |
| CN114896124B (en) * | 2022-05-16 | 2024-08-20 | 华能澜沧江水电股份有限公司 | System for alarm monitoring based on time sequence decomposition |
| CN115158399A (en) * | 2022-06-14 | 2022-10-11 | 通号城市轨道交通技术有限公司 | Time sequence signal abnormity detection method and system |
| CN115158399B (en) * | 2022-06-14 | 2023-10-17 | 通号城市轨道交通技术有限公司 | Time sequence signal abnormality detection method and system |
| CN115994248A (en) * | 2023-03-24 | 2023-04-21 | 青岛精锐机械制造有限公司 | Data detection method and system for valve faults |
| CN116342151A (en) * | 2023-04-12 | 2023-06-27 | 余珊 | Monetary payment device amount tampering detection system |
| CN116342151B (en) * | 2023-04-12 | 2024-02-02 | 西安你左我右网络科技有限公司 | Monetary payment device amount tampering detection system |
| CN116343953A (en) * | 2023-05-30 | 2023-06-27 | 苏州绿华科技有限公司 | Intelligent community management system based on artificial intelligence |
| CN116343953B (en) * | 2023-05-30 | 2023-08-01 | 苏州绿华科技有限公司 | Intelligent community management system based on artificial intelligence |
| CN118190276A (en) * | 2024-05-17 | 2024-06-14 | 菏泽市产品检验检测研究院 | Pressure vessel fault diagnosis method based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112308414A (en) | Income abnormity detection method and device, electronic equipment and storage medium | |
| CN106991145B (en) | Data monitoring method and device | |
| US9448787B2 (en) | Methods and systems for analyzing software development risks | |
| CN106295995A (en) | A kind of data analysing method for determining the production schedule and equipment | |
| CN113962468A (en) | Energy consumption monitoring and statistics-based energy consumption carbon emission management method and system | |
| CN112070284A (en) | Screening method, device, equipment and storage medium for component prediction | |
| CN117934247B (en) | Carbon emission factor prediction method and device based on time sequence decomposition, electronic equipment and storage medium | |
| CN112785057A (en) | Component prediction method, device, equipment and storage medium based on exponential smoothing | |
| US20200050982A1 (en) | Method and System for Predictive Modeling for Dynamically Scheduling Resource Allocation | |
| CN112183868A (en) | Construction method of traffic flow prediction model and electronic equipment | |
| CN113642886B (en) | Method, system, medium and computing equipment for analyzing utilization rate of planned power grid line | |
| CN117952377A (en) | Digital twin discrete manufacturing workshop synchronous evolution method, device and medium | |
| Seshan et al. | LSTM-based autoencoder models for real-time quality control of wastewater treatment sensor data | |
| CN112053181A (en) | Agricultural product price prediction method and system | |
| CN111737233A (en) | Data monitoring method and device | |
| CN115913745A (en) | Vehicle safety event prediction method and device, computer equipment and storage medium | |
| CN116564102A (en) | Traffic prediction method of Prophet-deep AR model | |
| CN114118637A (en) | Method and device for building prediction model of accessory demand and computer equipment | |
| CN111898786A (en) | Electronic lock demand prediction method, system, equipment and storage medium | |
| CN118071179B (en) | Multi-factor pig collection quantity prediction method, device and storage medium | |
| CN116737784A (en) | Method, apparatus, device, medium and program product for detecting periodic fluctuation of data | |
| CN111339156B (en) | Method, apparatus and computer readable storage medium for long-term determination of business data | |
| Emmanuel et al. | Singular spectrum analysis: an application to Kenya’s industrial inputs price index | |
| EP4006791B1 (en) | Method for detecting anomalies in data communications | |
| CN115456681A (en) | Method and device for establishing base station electricity charge data estimation model and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210202 |
|
| RJ01 | Rejection of invention patent application after publication |