CN107426040B - A kind of prediction technique of network behavior - Google Patents
A kind of prediction technique of network behavior Download PDFInfo
- Publication number
- CN107426040B CN107426040B CN201710854916.7A CN201710854916A CN107426040B CN 107426040 B CN107426040 B CN 107426040B CN 201710854916 A CN201710854916 A CN 201710854916A CN 107426040 B CN107426040 B CN 107426040B
- Authority
- CN
- China
- Prior art keywords
- data
- time
- kick
- network behavior
- behavior
- 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.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Algebra (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of prediction techniques of network behavior, belong to user's behaviors analysis technical field, comprising: pre-process to the network behavior data of extraction, obtain taking the behavior frequency in the unit time as the target network behavioral data for fluctuating data；Standardize to target network behavioral data, and folding is carried out to the data after standardization and obtains folding matrix, the kick folded in matrix is detected；According to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data, the time-varying random differential equation models of network behavior are established, and to the time-varying parameter in model and determine parameter and estimate；Parameter is estimated using the dynamic prediction mode continuous iteration data training pattern in preceding several weeks, then predicts the track in next week, is predicted with the network behavior to user.The network behavior stochastic model that the present invention establishes can predict well the Future Trajectory of network behavior, and can reflect the multifrequency nature that network behavior has.
Description
Technical field
The invention belongs to user's behaviors analysis technical fields, more particularly, to a kind of prediction technique of network behavior.
Background technique
With the development of internet technology with it is universal, more and more people are handled by internet progress Traditional affair,
Especially " internet+" continues deeply and promotes, and merging for internet and social life is accelerated, so that the behavior mould of the mankind
Formula produces huge variation.However, the following network security situation is but as the depth that people participate in internet increases
It is more severe.The universal law of network behavior is furtherd investigate, the kinetic mechanism of network behavior behind is grasped, can be Network Abnormal
Behavioral value provides theoretical foundation, and then generates important function to Support network security.
The existing modeling to network behavior is the time interval level from network behavior.A large number of studies show that network
The time interval of behavior has the heterogeneous character for deviateing Poisson, and the time interval sequence for showing as behavior obeys power-law distribution.
For this characteristic of network behavior, the queuing model of task based access control priority is suggested, this class model regards behavior as one
A task, it is intended to behavior is described in the form being lined up by the priority of task, reappear the heterogeneity of behavior.However very
Multi net voting behavior such as web page browsing, online game etc. belongs to interest drive type, and the queuing model of task based access control priority is obvious
It can not effectively portray.
As the memory characteristic in network behavior is constantly verified, different from the memory effect mould of task queue's model classification
Type is suggested one after another, and the memory characteristic of this class model subordinate act sets out, it is believed that people can according to the height of historical behavior frequency come
Change the probability that current behavior occurs.In addition, the similar model based on interest is also suggested therewith, this class model thinks interest
Variation will affect the generation of behavior.The interest of model hypothesis people constantly changes at any time, and the frequency of behavior is dependent on interest
Variation, while also adversely affecting interest.However even most interested thing also will receive circadian rhythm and the external world of people
The influence of the factors such as environment.Only consider that the model of interest can not effectively portray the network behavior for showing cyclic swing.It is a kind of
Periodical cascade process model is suggested, which considers that Email behavior is influenced by day alternates with night to show week
Phase wave characteristic portrays behavior with cascade Poisson process, it is intended to explain the origin of behavior heterogeneous character.But also have many
Positive research show, after weeding out the influence from circadian rhythm from network behavior data, behavior it is heterogeneous still
In the presence of.This is that is to say, bright, it is intended to which it is defective that the heterogeneity of behavior, which is only attributed to people's own physiological rhythm and pace of moving things,.
Importantly, whether the time interval of network behavior, which obeys power-law distribution, has many disputes.More and more
Positive research discovery, logarithm normal distribution, Weibull distribution, Gamma distribution, or even segmentation distribution can preferably describe network
Behavior.Therefore, it needs to carry out the potential kinetic mechanism of network behavior and modeling method from the angle other than time interval
Research.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of prediction technique of network behavior,
Thus it solves existing for the analysis mode of the network behavior of the existing time interval level from network behavior to network behavior
The technical problem of inaccuracy is described.
To achieve the above object, the present invention provides a kind of prediction techniques of network behavior, comprising:
(1) the network behavior data of acquisition are pre-processed, is obtained using the frequency of behavior in the unit time as fluctuation number
According to target network behavioral data；
(2) standardized by the mean value of target network behavioral data and standard deviation to target network behavioral data, and right
Data after standardization obtain the folding matrix of M*N by Zhou Jinhang folding, detect to the kick folded in matrix,
In, M indicates the quantity of the unit time in one week, and N indicates the data acquired always all numbers；
(3) it according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data, builds
The time-varying random differential equation models of vertical network behavior, and to the time-varying parameter in model and determine parameter and estimated to obtain target
Model；
(4) it is predicted by network behavior track of the object module to next week of user.
Preferably, step (2) specifically includes:
(2.1) byTo jth weekly data Yij, i ∈ [1, M], j ∈ [1, N] standardize, wherein μjTable
Show the mean value of jth weekly data sequence, σjIndicate the standard deviation of j weekly data sequence；
(2.2) if annual acquire N weekly data sequence altogether, length N*M then advises every weekly data of acquisition
Data sequence X after being standardized after generalizedt, t ∈ [1, M × N] is configured to the folding square of M*N then by all folded datas
Battle array Xij, i ∈ [1, M], j ∈ [1, N], to l, l ∈ LiIfIt is then considered as kick, wherein Li={ lj=
Xij, j ∈ [1, N] } and indicate to moment i each in week, i ∈ [1, M] there is the time series that a length is N,It is the time
Sequence LiMean value,It is time series LiStandard deviation, ξ is preset value.
Preferably, step (3) specifically includes:
(3.1) according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data,
Establish the time-varying random differential equation models of network behavior:
Wherein, k and d is constant, for adjusting two power
Weight, a(t)With c(t)It is time-varying parameter, a(t)It is the mean value that t moment returns, c(t)It is diffusion amplitude, N(t)Be parameter be β(t)Time-varying
Compound poisson process, J(t)It is the kick amplitude of t moment, obedience mean value is μt, standard deviation σtNormal distribution, WtIndicate one
Wiener-Hopf equation；
(3.2) byTo time-varying parameter a(t)Estimated, byTime-varying is joined
Number c(t)Estimated, wherein XtjThe folding matrix of (t ∈ [1, M], j ∈ [1, N]) expression M*N；
(3.3) according to the kick position detected in kick detection-phase and amplitude, byTo time-varying parameter
β(t)Estimated, byTo time-varying parameter μ(t)And σ(t)Estimated,
In, ncount(t)Detect that all numbers of kick, N are total all numbers, μ for t moment(t)With σ(t)What respectively t moment detected is all prominent
The mean value and variance of jump, XJump(tj)It is the kick point being detected in t moment in N number of week；
(3.4) it is combined and is estimated by Markov Chain Monte Carlo (Markov Chain Monte Carlo, MCMC) method
Time-varying parameter estimate parameter k, d is determined.
Preferably, step (4) specifically includes:
(4.1) the dynamic prediction process of model: using first N weeks data training pattern of the continuous iteration of dynamic prediction method
Estimate parameter, then the track for predicting the N+1 weeks；
(4.2) it measures the index of model prediction performance: computing repeatedly model P times, mean value is sought to the P track predicted
Mean value track is obtained, and using mean value track as final prediction locus.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
(1) present invention considers influence of many factors to network behavior, establishes network row from the track level of network behavior
For stochastic model.The network behavior model can be suitably used for all types of network behaviors；
(2) network behavior model proposed by the present invention can predict well the Future Trajectory of network behavior, and can reflect
The multifrequency nature that network behavior has.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of the prediction technique of network behavior disclosed by the embodiments of the present invention；
Fig. 2 is that the microblog data in a kind of pair of network behavior data disclosed by the embodiments of the present invention carries out pretreated method
Flow diagram；
Fig. 3 is a kind of pair of microblogging group behavior data normalization disclosed by the embodiments of the present invention and kick detection schematic diagram,
Wherein, (a) indicates granularity initial data in 2014；(b) indicate that data after trend are gone in standardization；(c) it indicates to detect before going trend
The kick signal arrived；(d) kick for going to detect after trend is indicated；
Fig. 4 is disclosed by the embodiments of the present invention a kind of according to the time-varying parameter that 50 weeks microblog datas are calculated before 2015
a(t)、c(t)；
Fig. 5 is one kind disclosed by the embodiments of the present invention by taking 2015 annual data of microblogging as an example, to the 21st week prediction locus and
Real trace comparison diagram, MAPE 15.14%, RMSE 979.78；
Fig. 6 is a kind of prediction result figure by taking 2015 annual data of microblogging as an example disclosed by the embodiments of the present invention, wherein (a)
Indicate the psd of truthful data；(b) psd of prediction data is indicated；(c) pdf of truthful data is indicated；(d) prediction data is indicated
pdf。
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
The present invention provides a kind of prediction technique of network behavior from the fluctuation track level of network behavior.According to existing reality
Card research, network behavior generally have the characteristics such as heterogeneous, cyclic swing and Memorability.The present invention occurs with the unit time
Behavior frequency sequence portray network behavior (referred to as action trail), and according to above-mentioned behavioral trait, by action trail
Variation is summarized as the cyclic swing influenced by circadian rhythm, the random fluctuation being affected by the external environment and the shadow by emergency event
Loud kick, finally establishes the random differential equation models of behavior accordingly, and according to the random differential equation models of foundation to
It is predicted the network behavior track in next week at family.
It is as shown in Figure 1 a kind of flow diagram of the prediction technique of network behavior disclosed by the embodiments of the present invention, in Fig. 1
Shown in method, comprising the following steps:
(1) the network behavior data of extraction are pre-processed, is obtained using the frequency of behavior in the unit time as fluctuation number
According to target network behavioral data；
In an optional embodiment, as shown in Fig. 2, illustrating process of data preprocessing, comprising the following steps:
Remove the invalid information in primitive network behavioral data, such as the record and network of system message, timing error
Content of the act is empty record；
Extract the timestamp information that network behavior content ID, User ID and network behavior occur；
It is sorted according to network behavior content ID, removes the duplicate record of network behavior content ID；
The timestamp information that network behavior occurs is converted into Beijing time；
Temporally information resequences to network behavior data, counts the quantity of network behavior in per unit time, finally
Being formed with the behavior frequency in the unit time is the target network behavioral data for fluctuating data.
(2) standardized by the mean value of target network behavioral data and standard deviation to target network behavioral data, and right
Data after standardization obtain the folding matrix of M*N by Zhou Jinhang folding, detect to the kick folded in matrix,
In, M indicates the unit time quantity in one week, and N indicates the data acquired always all numbers；
Wherein, the trend that data have will greatly affect the detection of kick, and standardize to data to go well
Except the long-term trend in data.
In an optional embodiment, step (2) is specifically included:
(2.1) byTo jth weekly data Yij, i ∈ [1, M], j ∈ [1, N] standardize, wherein μjTable
Show the mean value of jth weekly data sequence, σjIndicate the standard deviation of j weekly data sequence；
(2.2) if annual acquire N weekly data sequence altogether, length N*M then advises every weekly data of acquisition
Data sequence X after being standardized after generalizedt, t ∈ [1, M × N] is configured to the folding square of M*N then by all folded datas
Battle array Xij, i ∈ [1, M], j ∈ [1, N], to l, l ∈ LiIfIt is then considered as kick, wherein Li={ lj=
Xij, j ∈ [1, N] } and indicate to moment i each in week, i ∈ [1, M] there is time series (the shared M item that a length is N
Time series),It is time series LiMean value,It is time series LiStandard deviation, ξ is preset value, it is preferable that ξ is taken as
3。
As shown in figure 3, being the standardization of a kind of pair of micro blog network behavioral data and kick detection disclosed by the embodiments of the present invention
Schematic diagram, wherein the unit time is hour, and M value is 168 at this time, and (a) indicates granularity initial data in 2014；(b) rule are indicated
Generalized goes data after trend；(c) it indicates that the kick detected before trend is gone to illustrate；(d) kick for going to detect after trend is indicated.
As seen from Figure 3, after data go trend, there is preferable kick detection effect.
(3) it according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data, builds
The time-varying random differential equation models of vertical network behavior, and to the time-varying parameter in model and determine parameter and estimated to obtain target
Model；
In an optional embodiment, step (3) is specifically included:
(3.1) according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data,
Establish the time-varying random differential equation models of network behavior:
Wherein, k and d is constant, for adjusting two power
Weight, a(t)With c(t)It is time-varying parameter, a(t)It is the mean value that t moment returns, c(t)It is diffusion amplitude, N(t)Be parameter be β(t)Time-varying
Compound poisson process, J(t)It is the kick amplitude of t moment, obedience mean value is μt, standard deviation σtNormal distribution, WtIndicate one
Wiener-Hopf equation；
Wherein, this model portrays network behavior in terms of three:
A: drift term k (a(t)-Xt) dt is time-varying equation of linear regression, portray the daily fluctuation rule of network behavior.Due to
The periodic law of human behavior constantly changes as time goes by, therefore, will return mean value a herein(t)It is set as time-varying parameter；
B: diffusion termPortray network behavior is influenced by external factor in mood, weather etc., generation it is small
Amplitude random fluctuation；The fluctuation at per moment by the moment in historical track regular behavioral implications, therefore, random fluctuation amplitude
c(t)It is set as time-varying parameter, determines amplitude of the parameter d for population equilibrium random fluctuation；
C. kick item J(t)dN(t)Portray network behavior is influenced by emergency event or hot topic, and the violent wave generated
It is dynamic, due to the Poisson characteristics that emergency event typically exhibits, portrayed herein with compound poisson process.
(3.2) byTo time-varying parameter a(t)Estimated, byTime-varying is joined
Number c(t)Estimated, wherein XtjThe folding matrix of (t ∈ [1, M], j ∈ [1, N]) expression M*N；
Be as shown in Figure 4 it is disclosed by the embodiments of the present invention it is a kind of according to 50 weeks microblog datas before 2015 be calculated when
Variable element a(t)、c(t)。
(3.3) according to the kick position detected in kick detection-phase and amplitude, byTo time-varying parameter
β(t)Estimated, byTo time-varying parameter μ(t)And σ(t)Estimated,
In, ncount(t)Detect that all numbers of kick, N are total all numbers, μ for t moment(t)With σ(t)What respectively t moment detected is all prominent
The mean value and variance of jump, XJump(tj)It is the kick point being detected in t moment in N number of week；
(3.4) time-varying parameter of estimation is combined to estimate to parameter k, d is determined by Markov Chain Monte Carlo MCMC methodology
Meter.
(4) it is predicted by network behavior track of the object module to next week of user.
In an optional embodiment, step (4) is specifically included:
(4.1) the dynamic prediction process of model: continuous using dynamic prediction method for the estimated performance for investigating model comprehensively
Iteration estimates parameter, then the track for predicting the N+1 weeks with first N weeks data training pattern；Preferably, N is no less than 10.
(4.2) it measures the index of model prediction performance: computing repeatedly model P times, mean value is sought to the P track predicted
Mean value track is obtained, and using mean value track as final prediction locus.Preferably, P value is not less than 1000 times.
Using prediction locus and actual path relative error index (Mean Absolute Percentage Error,
MAPE) and absolute error index (Root Mean Squared Error, RMSE) carries out qualitative portray to estimated performance.Index
It calculates as follows:
Model prediction performance is portrayed from data probability distributions level and frequency domain level using PDF and PSD curve.Such as
It is one kind disclosed by the embodiments of the present invention shown in Fig. 5 by taking 2015 annual data of microblogging as an example, to the 21st week prediction locus and really
Track comparison diagram, wherein the unit time is hour, and M value is 168, MAPE 15.14%, RMSE 979.78 at this time；Such as figure
It is a kind of prediction result figure by taking 2015 annual data of microblogging as an example disclosed by the embodiments of the present invention shown in 6, wherein (a) indicates true
The psd of real data；(b) psd of prediction data is indicated；(c) pdf of truthful data is indicated；(d) pdf of prediction data is indicated.From
As can be seen that can be very good the Future Trajectory of prediction network behavior using the method for the present invention in figure.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (3)
1. a kind of prediction technique of network behavior characterized by comprising
(1) the network behavior data of acquisition are pre-processed, obtains taking the frequency of behavior in the unit time as fluctuation data
Target network behavioral data；
(2) standardized by the mean value of target network behavioral data and standard deviation to target network behavioral data, and to specification
Data after change obtain the folding matrix of M*N by Zhou Jinhang folding, detect to the kick folded in matrix, wherein M
Indicate the quantity of the unit time in one week, N indicates the data acquired always all numbers, and the kick indicates that data generation is more than pre-
If the variation of range；
(3) according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data, net is established
The time-varying random differential equation models of network behavior, and to the time-varying parameter in model and determine parameter and estimated to obtain target mould
Type；
(4) it is predicted by network behavior track of the object module to next week of user；
Wherein, the step (3) specifically includes:
(3.1) it according to the periodic law, stochastic volatility and kick characteristic shown in target network behavioral data, establishes
The time-varying random differential equation models of network behavior:
Wherein, k and d is constant, for adjusting two weights,
a(t)With c(t)It is time-varying parameter, a(t)It is the mean value that t moment returns, c(t)It is diffusion amplitude, N(t)Be parameter be β(t)Time-varying it is multiple
Close Poisson process, J(t)It is the kick amplitude of t moment, obedience mean value is μt, standard deviation σtNormal distribution, WtIndicate a dimension
Receive process, Xt, t ∈ [1, M × N] indicate acquisition every weekly data standardized after standardized after data sequence；
(3.2) byTo time-varying parameter a(t)Estimated, byTo time-varying parameter c(t)
Estimated, wherein XtjThe folding matrix of (t ∈ [1, M], j ∈ [1, N]) expression M*N；
(3.3) according to the kick position detected in kick detection-phase and amplitude, byTo time-varying parameter β(t)Into
Row estimation, byTo time-varying parameter μ(t)And σ(t)Estimated, wherein
ncount(t)Detect that all numbers of kick, N are total all numbers, μ for t moment(t)With σ(t)Respectively t moment all kicks for detecting
Mean value and variance, XJump(tj)It is the kick point being detected in t moment in N number of week；
(3.4) time-varying parameter of estimation is combined to estimate parameter k, d is determined by Markov Chain Monte Carlo MCMC methodology.
2. the method according to claim 1, wherein step (2) specifically includes:
(2.1) byTo jth weekly data Yij, i ∈ [1, M], j ∈ [1, N] standardize, wherein μjIndicate jth
The mean value of weekly data sequence, σjIndicate the standard deviation of j weekly data sequence；
(2.2) if annual acquire N weekly data sequence altogether, length N*M then standardizes to every weekly data of acquisition
Data sequence X after being standardized afterwardst, t ∈ [1, M × N] is configured to the folding matrix of M*N then by all folded datas
Xij, i ∈ [1, M], j ∈ [1, N], to l, l ∈ LiIfIt is then considered as kick, wherein Li={ lj=Xij,
J ∈ [1, N] } indicate to moment i each in week, i ∈ [1, M] there is the time series that a length is N,It is time sequence
Arrange LiMean value,It is time series LiStandard deviation, ξ is preset value.
3. according to the method described in claim 2, it is characterized in that, step (4) specifically includes:
(4.1) the dynamic prediction process of model: estimated using first N weeks data training pattern of the continuous iteration of dynamic prediction method
Parameter, then predict the N+1 weeks track；
(4.2) it measures the index of model prediction performance: computing repeatedly model P times, mean value is sought to the P track predicted and is obtained
Mean value track, and using mean value track as final prediction locus.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710854916.7A CN107426040B (en) | 2017-09-20 | 2017-09-20 | A kind of prediction technique of network behavior |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710854916.7A CN107426040B (en) | 2017-09-20 | 2017-09-20 | A kind of prediction technique of network behavior |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107426040A CN107426040A (en) | 2017-12-01 |
| CN107426040B true CN107426040B (en) | 2019-08-30 |
Family
ID=60433680
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710854916.7A Expired - Fee Related CN107426040B (en) | 2017-09-20 | 2017-09-20 | A kind of prediction technique of network behavior |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107426040B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108228887B (en) * | 2018-01-31 | 2019-12-03 | 百度在线网络技术（北京）有限公司 | Method and apparatus for generating information |
| CN109597833A (en) * | 2018-10-15 | 2019-04-09 | 平安科技（深圳）有限公司 | Event prediction method, apparatus, computer equipment and storage medium based on big data |
| CN110149343B (en) * | 2019-05-31 | 2021-07-16 | 国家计算机网络与信息安全管理中心 | Abnormal communication behavior detection method and system based on flow |
| CN116975300B (en) * | 2023-09-22 | 2024-01-26 | 中国铁塔股份有限公司吉林省分公司 | Information mining method and system based on big data set |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102075352A (en) * | 2010-12-17 | 2011-05-25 | 北京邮电大学 | Method and device for predicting network user behavior |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3283517A4 (en) * | 2015-04-16 | 2019-03-06 | Alder Biopharmaceuticals, Inc. | Anti-pacap antibodies and uses thereof |
-
2017
- 2017-09-20 CN CN201710854916.7A patent/CN107426040B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102075352A (en) * | 2010-12-17 | 2011-05-25 | 北京邮电大学 | Method and device for predicting network user behavior |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107426040A (en) | 2017-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107426040B (en) | A kind of prediction technique of network behavior | |
| CN104216954B (en) | The prediction meanss and Forecasting Methodology of accident topic state | |
| CN111144542B (en) | Oil well productivity prediction method, device and equipment | |
| CA2985028C (en) | Gating decision system and methods for determining whether to allow material implications to result from online activities | |
| Zhang et al. | Resource requests prediction in the cloud computing environment with a deep belief network | |
| D'Amico et al. | Wind speed modeled as an indexed semi‐Markov process | |
| Wang et al. | Forecasting crude oil price and stock price by jump stochastic time effective neural network model | |
| CN105760649B (en) | A kind of credible measure towards big data | |
| CN111460294A (en) | Message pushing method and device, computer equipment and storage medium | |
| McCaffery et al. | Improved analysis of lek count data using N‐mixture models | |
| CN110083518B (en) | AdaBoost-Elman-based virtual machine software aging prediction method | |
| CN110460458A (en) | Based on multistage markovian Traffic anomaly detection method | |
| Yang et al. | Prediction method of PM2. 5 concentration based on decomposition and integration | |
| Liu et al. | Sparse-gev: Sparse latent space model for multivariate extreme value time serie modeling | |
| Sauer et al. | Using multi‐species occupancy models in structured decision making on managed lands | |
| Sæther et al. | Rarity, life history and scaling of the dynamics in time and space of British birds | |
| Ren et al. | Robustness verification for machine-learning-based power system dynamic security assessment models under adversarial examples | |
| CN115221396A (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
| Chen et al. | A deep learning ensemble approach for predicting tropical cyclone rapid intensification | |
| Połap et al. | Energy consumption prediction model for smart homes via decentralized federated learning with LSTM | |
| JP2023535851A (en) | METHOD, DEVICE, TERMINAL DEVICE, AND STORAGE MEDIUM FOR DATA PROCESSING MODEL BY PRIVACY PROTECTION | |
| Song et al. | A Novel Outlier Detection Method of Long‐Term Dam Monitoring Data Based on SSA‐NAR | |
| Zang et al. | The credit risk assessment of P2P lending based on BP neural network | |
| Lindén et al. | Estimating environmental effects on population dynamics: consequences of observation error | |
| Du et al. | Microblog bursty feature detection based on dynamics model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190830 Termination date: 20200920 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |