CN103209005B - The pre-examining system of frequency hop sequences of a kind of graphic based model - Google Patents
The pre-examining system of frequency hop sequences of a kind of graphic based model Download PDFInfo
- Publication number
- CN103209005B CN103209005B CN201310137018.1A CN201310137018A CN103209005B CN 103209005 B CN103209005 B CN 103209005B CN 201310137018 A CN201310137018 A CN 201310137018A CN 103209005 B CN103209005 B CN 103209005B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- prediction
- phase space
- hop sequences
- 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
Abstract
The present invention discloses the pre-examining system of frequency hop sequences of a kind of graphic based model, comprising: pre-processing module, for the original frequency hopping sequences intercepted and captured being carried out denoising, goes the process such as bandwidth; Prediction module, is connected with described pre-processing module, for phase space reconstruction and structure predictive model; Feedback adjustment module, is connected with described pre-processing module and prediction module, for accuracy detection, and feedback and model adjustment. The present invention adopts Cao method and auto-correlation method to solve Embedded dimensions m and time lag �� and then phase space reconstruction, and the Markov border based on the MMPC algorithm study query node improved builds predictive model. The Embedded dimensions m of the present invention and time lag �� is two key parameters of phase space reconfiguration, utilizes auto-correlation method and Cao method to obtain parameter more reliable and more stable, simplifies Bayesian network model by Markov border so that forecasting efficiency is higher.
Description
Technical field
The invention belongs to frequency hop sequences electric powder prediction, particularly relate to the pre-examining system of frequency hop sequences of a kind of graphic based model.
Background technology
Existing Chaotic time series forecasting method is, original chaos time sequence is embedded into m and ties up phase space, its essence is and utilize multiple difference delay points that in original series, certain sequence of points generates with it under delay time T, common reconstruct m ties up a phase point in phase space, because the choosing of time lag �� can be guaranteed between these time delay points interrelated, therefore Bayesian network can be utilized to portray this kind of cognation;
Phase space after reconstruct represents to be the matrix that a m �� n ties up, Existence dependency associate feature between each row vector, prior art:
(1) using the variable of each row vector as Bayesian network, K2 algorithm is adopted to carry out bayesian network structure learning, and then structure contains the directed acyclic graph of m node, directed acyclic graph is the relation of interdependence between row vector each in phase space intuitively graphically;
(2) learn Bayesian network parameter with maximum likelihood estimation, namely determine the conditional probability distribution table at each node place, wherein the probable value in probability distribution table, embody the dependence intensity between each row vector, and then complete the structure of bayes predictive model;
(3) phase space after reconstruct being carried out �� and walk continuation, the phase space after continuation is as follows:
To following phase point
X(n+l)=[xn+l,xn+l+��,��,xn+l+(m-2)��,xn+l+(m-1)��]TThe prediction of (1��l�ܦ�, n=N-(m-1) ��) is exactly target of prediction is the m component x of new phase pointn+l+(m-1)��Value, owing to the value of front m-1 the component of the new phase point of l is known, then can as evidence variable, namely
E={X1=xn+l,X2=xn+l+��,��,Xm-1=xn+l+(m-2)��}��
Obtain maximum a posteriori probability P (Xm=fiE) corresponding value fi, this value fiIt isPredictor, the maximum predicted step of the Bayesian network predictive model set up by the known prior art of l�ܦ� again is ��, and wherein �� is the time of lag of frequency hop sequences;
(4) newer real �� sequential value is extended for phase point, and adds in phase space, upgrade network parameter, then carry out predicting, the circulation of undated parameter go on, until prediction terminates;
The research of prior art is to liking chaos time sequence, and the value number of chaos time sequence is less than 64. When studying object and change into frequency hop sequences, the time lag �� and embedding window width �� of the C-C method assessment frequency hop sequences in prior artwFor a long time consuming time, unstable result; Cao method and auto-correlation method is utilized to solve Embedded dimensions m and time lag ��, after phase space reconfiguration, Bayesian network node number to be learned is had to equal Embedded dimensions m, and m >=5, node value number equals the frequency hopping code value number of frequency hop sequences, and generally it being greater than 64, the Bayesian network sample D that the K2 algorithm in prior art obtains after frequency hop sequences cannot be utilized to reconstruct learns bayesian network structure. .
Summary of the invention
The object of the embodiment of the present invention is to provide the pre-examining system of frequency hop sequences of a kind of graphic based model, it is intended to solve when the value condition of frequency hop sequences is many, and in prior art, the C-C method evaluation time postpones �� and embeds window width ��wFor a long time consuming time, unstable result, and the problem such as K2 algorithm inefficacy of study network structure.
The embodiment of the present invention realizes like this, the pre-examining system of frequency hop sequences of a kind of graphic based model, it is characterised in that, the pre-examining system of the frequency hop sequences of graphic based model comprises:
Pre-processing module, for being carried out denoising, go bandwidth etc. by the original frequency hopping sequences intercepted and captured, chooses appropriate one section frequency hop sequences { xi, i=1,2 ..., N as the training set data of model construction, using M frequency hopping code of training set data rear adjacent as model testing data;
Prediction module, is connected with described pre-processing module, for phase space reconstruction and structure predictive model;
Feedback adjustment module, is connected with described pre-processing module and prediction module, for accuracy detection, and feedback and model adjustment.
Further, phase space reconstruction adopts Cao method and auto-correlation method to solve Embedded dimensions m and time lag ��.
Further, build the Markov border that predictive model adopts the innovatory algorithm study query node based on MMPC, and using Markov border as Bayes's localized network structure.
Further, described pre-processing module also comprises:
The frequency hop sequences data collection module of original frequency hopping sequences is intercepted and captured with melodeon; For removing the data processing unit of original frequency hopping sequences noise, bandwidth etc., it is connected with described frequency hop sequences data collection module; Through normalization method obtain for the training set data unit of phase space reconstruction and model, be connected with described data processing unit; For predicting the detection model check data unit of accuracy detection, feedback and model adjustment, it is connected with described data processing unit; Through the proof data unit for predicting that phase space �� step continuation obtains, it is connected with described training set data unit.
Further, described prediction module also comprises:
The phase space data cell utilizing training set data study to obtain, is connected with described training set data unit; The Bayesian network unit obtained through local structure study and parameter learning on phase space reconstruction basis, is connected with described phase space data cell; Utilize the prediction ordered series of numbers unit of Bayesian network model prediction frequency hop sequences, it is connected with described Bayesian network unit and proof data unit.
Another object of the present invention is to provide the frequency hop sequences Forecasting Methodology of a kind of graphic based model, described frequency hop sequences Forecasting Methodology comprises the following steps:
Step 1, carries out denoising to the original frequency hopping sequences intercepted and captured, goes bandwidth etc., choose appropriate one section frequency hop sequences { xi, i=1,2 ..., N as the training set data of model construction, using M frequency hopping code of training set data rear adjacent as model testing data;
Step 2, utilizes auto-correlation method and Cao method to ask the time lag �� and Embedded dimensions m of phase space reconstruction, then according to these two parameters, using training set data reconstruct m �� n dimension matrix as phase space X, wherein
As the data sample D learnt for Bayesian network, the scale of data sample D is n;
Step 3, based on the innovatory algorithm study query node X of MMPCmMarkov border, recycling maximum likelihood estimation learns the parameter of each node, final obtains the local Bayesian network being used for the prediction of multi-frequency point frequency hop sequences;
Step 4, the prediction step �� of model=��, �� is time lag, according to Bayes's posteriority reasoning algorithm, calculatesWherein the span of l is 1��l�ܦ�, and namely l is the l time prediction of prediction in �� step prediction, and E (n+l) is query node XmMarkov boundary set in the value of each node. When P is maximum, fiIt is exactly xN+lPredictor;
Step 5, after every �� step prediction terminates, preserve the frequency hopping code of prediction, and true frequency hopping code corresponding in model inspection data is extended for phase point X (n+l) l=1 ..., ��, and add in phase space, upgrade network parameter, forward step 4 to, until M frequency hopping code of training set data rear adjacent predicted after terminate;
Step 6, utilizes model inspection Data Detection model prediction accuracy, if not reaching accuracy requirement, feedback and model adjustment, forward step 3 to, if reaching accuracy requirement, namely obtaining stable Bayesian network predictive model, predicting for frequency hop sequences.
Further, the delay values ri tried to achieve in auto-correlation methoddOn basis, time lag �� can regulate downwards, and namely the span of time lag �� is 2�ܦӡܦ�d��
Further, the node number that Markov border comprises is ��, and the span of parameter alpha is 2�ܦ���5.
The pre-examining system of frequency hop sequences of the graphic based model of the present invention, by adopting Cao method and auto-correlation method to solve Embedded dimensions m and time lag ��, based on the Markov border of the innovatory algorithm study query node of MMPC, and then phase space reconstruction and structure predictive model. The Embedded dimensions m of the present invention and time lag �� is two key parameters of phase space reconfiguration, and it is more reliable and more stable that auto-correlation method and Cao method make to try to achieve parameter, utilizes Markov border to simplify Bayesian network model so that forecasting efficiency is higher.
Accompanying drawing explanation
Fig. 1 is the structural representation of the pre-examining system of frequency hop sequences of the graphic based model that the embodiment of the present invention provides.
Fig. 2 is three integrant schematic diagram of the Bayesian network that the embodiment of the present invention provides;
Fig. 3 is that the chaos time sequence that the embodiment of the present invention provides maps higher-dimension phase space schematic diagram;
Fig. 4 be the embodiment of the present invention provide phase space as Bayesian network sample D schematic diagram;
Fig. 5 is the figure that predicts the outcome of the Bayesian network predictive model that the embodiment of the present invention provides;
Fig. 6 is the prediction accuracy analogous diagram of the Bayesian network predictive model that the embodiment of the present invention provides.
Embodiment
In order to make the object of the present invention, technical scheme and advantage clearly understand, below in conjunction with embodiment, the present invention is further elaborated. It is to be understood that specific embodiment described herein is only in order to explain the present invention, it is not intended to limit the present invention.
Fig. 1 shows the frequency hop sequences prediction system architecture of graphic based model provided by the invention. For convenience of explanation, illustrate only part related to the present invention.
The pre-examining system of frequency hop sequences of the graphic based model of the present invention, comprising:
Pre-processing module, for carrying out denoising to the frequency hop sequences intercepted and captured, go the process such as bandwidth;
Prediction module, is connected with described pre-processing module, for phase space reconstruction and structure predictive model;
Feedback adjustment module, is connected with described pre-processing module and prediction module, for accuracy detection, and feedback and model adjustment.
Present invention also offers the frequency hop sequences Forecasting Methodology of a kind of graphic based model, described frequency hop sequences Forecasting Methodology comprises the following steps:
Step 1, carries out denoising to the original frequency hopping sequences intercepted and captured, goes bandwidth etc., choose appropriate one section frequency hop sequences { xi, i=1,2 ..., N as the training set data of model construction, using M frequency hopping code of training set data rear adjacent as model testing data;
Step 2, utilizes auto-correlation method and Cao method to ask the time lag �� and Embedded dimensions m of phase space reconstruction, then according to these two parameters, using training set data reconstruct m �� n dimension matrix as phase space X, wherein
As the data sample D learnt for Bayesian network, the scale of data sample D is n;
Step 3, based on the innovatory algorithm study query node X of MMPCmMarkov border, recycling maximum likelihood estimation learns the parameter of each node, final obtains the local Bayesian network being used for the prediction of multi-frequency point frequency hop sequences;
Step 4, the prediction step �� of model=��, �� is time lag, according to Bayes's posteriority reasoning algorithm, calculatesWherein the span of l is 1��l�ܦ�, and namely l is the l time prediction of prediction in �� step prediction, and E (n+l) is query node XmMarkov boundary set in the value of each node. When P is maximum, fiIt is exactly xN+lPredictor;
Step 5, after every �� step prediction terminates, preserve the frequency hopping code of prediction, and true frequency hopping code corresponding in model inspection data is extended for phase point X (n+l) l=1 ..., ��, and add in phase space, upgrade network parameter, forward step 4 to, until M frequency hopping code of training set data rear adjacent predicted after terminate;
Step 6, utilizes model inspection Data Detection model prediction accuracy, if not reaching accuracy requirement, feedback and model adjustment, forward step 3 to, if reaching accuracy requirement, namely obtaining stable Bayesian network predictive model, predicting for frequency hop sequences.
As a prioritization scheme of the embodiment of the present invention, phase space reconstruction adopts Cao method and auto-correlation method to solve Embedded dimensions m and time lag ��. Wherein, time lag �� only can affect the Euclidean geometry shape of attraction of reconstruct when not getting optimum delay, and then affects the calculating of Embedded dimensions m, does not affect the kinetic property of the sub unambiguously reactive system of attraction of reconstruct. Therefore, the delay values ri tried to achieve in auto-correlation methoddOn basis, time lag �� can regulate downwards, and the span of �� is 2�ܦӡܦ�d��
As a prioritization scheme of the embodiment of the present invention, build the Markov border that predictive model adopts the innovatory algorithm study query node based on MMPC, and using Markov border as Bayes's localized network structure. Wherein, the node number that the Markov border of query node comprises is ��, parameter alpha and model training data volume close relation, and parameter alpha value is more big, builds the model training data volume needed for Bayesian network model more big. And in the actual environment of communication antagonism, it is necessary to model accepts a small amount of frequency hop sequences and can set up predictive model and implement interference prediction. Therefore, the optionally value of regulating parameter ��, and the span of parameter alpha is 2�ܦ���5.
As a prioritization scheme of the embodiment of the present invention, pre-processing module also comprises:
The frequency hop sequences data collection module of original frequency hopping sequences is intercepted and captured with melodeon; For removing the data processing unit of original frequency hopping sequences noise, bandwidth etc., it is connected with described frequency hop sequences data collection module; Through normalization method obtain for the training set data unit of phase space reconstruction and model, be connected with described data processing unit; For predicting the detection model check data unit of accuracy detection, feedback and model adjustment, it is connected with described data processing unit; Through the proof data unit for predicting that phase space �� step continuation obtains, it is connected with described training set data unit.
As a prioritization scheme of the embodiment of the present invention, prediction module also comprises:
The phase space data cell utilizing training set data study to obtain, is connected with described training set data unit; The Bayesian network unit obtained through local structure study and parameter learning on phase space reconstruction basis, is connected with described phase space data cell; Utilize the prediction ordered series of numbers unit of Bayesian network model prediction frequency hop sequences, it is connected with described Bayesian network unit and proof data unit.
Below in conjunction with accompanying drawing and concrete enforcement, the application principle of the present invention is further described.
As shown in Figure 1, the pre-examining system of frequency hop sequences of the graphic based model of the embodiment of the present invention comprises: pre-processing module 1, the frequency hop sequences intercepted and captured is carried out denoising, goes the process such as bandwidth, obtain the frequency hop sequences for model construction and prediction; Pre-processing module 1 comprises: frequency hop sequences data collection module 11, data processing unit 12, training set data unit 13, proof data unit 14, model testing data cell 15.
Prediction module 2, adopts Cao method and auto-correlation method to solve Embedded dimensions m and time lag ��, and then phase space reconstruction; Adopt the Markov border of the innovatory algorithm study query node based on MMPC, and using Markov border as Bayes's localized network structure, and then build predictive model. Prediction module 2 comprises: phase space data cell 21, Bayesian network unit 22, forecasting sequence unit 23.
Feedback adjustment module 3, accuracy detection, feedback and model adjustment.
The principle of the present invention is:
1, phase space reconstruction technique:
The research of the present invention is to liking during frequency hopping communicates the original frequency hopping sequences intercepted and captured, and existing result of study shows: frequency hop sequences has chaotic characteristic, chaotic characteristic shows as the trend irregularities etc. of the extreme sensitivity dependency to initial value, bigger linear complexity, sequence jump, and this is the key issue place that frequency hop sequences is difficult to prediction;
Phase space reconfiguration is the first step analyzing Time Chaotic Dynamical Systems, object is in higher-dimension phase space and recovers chaos attractor, because chaos attractor is as one of the feature of chaos system, embody the regularity of chaos system, mean that chaos system finally can fall into a certain specific track, phase space and original system (producing the Time Chaotic Dynamical Systems of frequency hop sequences) after reconstruct are differential homeomorphisms, higher-dimension phase space carries out frequency hop sequences prediction, is equivalent to the prediction in lower dimensional space; Concrete steps show complicated pseudo-randomness, the chaos characteristic such as non-linear according to frequency hop sequences in lower dimensional space, the Cao method in chaology and auto-correlation method is utilized to choose smallest embedding dimension number m and time lag �� respectively, then phase space reconstruction, and then launch the original power characteristic of frequency hop sequences.
2, graphical model coupling technology:
Bayesian network has solid data reasoning prediction theory as the graphical model that a class is special, the conditional probability table that it be correspond to by the directed acyclic graph and each node that represent dependence between variable forms, arc on figure represents the dependence between node qualitatively, conditional probability table then give dependence gone out quota portray.
Study Bayesian network is exactly determine the network structure of Bayesian network and corresponding parameter, and parameter learning has two kinds of basic skills, i.e. maximum likelihood estimation and Bayesian Estimation; Common structure learning algorithm has: K2 algorithm, climbing method, simulated annealing etc., and the Inference Forecast based on Bayesian network is by building on the basis of graphical model at data with existing, according to the data that the probabilistic relation Inference Forecast of parameter is new; " graphical model coupling technology " refers in phase space reconfiguration theoretical basis, a kind of Hopping time sequence multistep forecasting method based on Bayesian network model prediction is proposed, first using the phase space after reconstructing as priori data information, then learn bayesian network structure and parameter, finally by employing Bayesian network model, chaos time sequence is predicted; Bayesian network in graphical model is applied to frequency hop sequences prediction, and Embedded dimensions m and time lag �� is two key parameters of phase space reconfiguration, and it is more reliable and more stable that auto-correlation method and Cao method make to try to achieve parameter; Due to nodes XmWhen given Markov border, conditional sampling and other nodes, so Bayes's localized network and complete Bayesian network have identical Inference Forecast result, utilize Markov border to simplify Bayesian network model, and forecasting efficiency is higher.
As shown in Figure 2, Bayesian network is made up of a structure iron G and conditional probability distribution P, is designated as B=(G, P). Wherein, structure iron G is made up of node collection V and oriented arc collection E, i.e. G=(V, E) is a class directed acyclic graph, and namely all oriented arcs do not form a closed loop. It is the qualitative part of Bayesian network, and conditional probability P is then its quantitative part, is the prerequisite that network quantitative reasoning calculates. Node in set V represents stochastic variable or event, it is possible to be discrete or continuous print. In discrete Bayesian network there is different values in each node, is called the state of node, and conventional is two-value state, also has the state of more than three. The Two Variables that oriented arc in set E exists probabilistic relation and cause-effect relationship for connecting or event. Conditional probability P then indicates the degree that this kind connects.
As shown in Figure 3, time lag ��=7 of a dimension chaos time sequence, Embedded dimensions m=3, phase space reconstruction technique maps this group one-dimensional data to be tieed up in phase spaces to 3, and then recovers the chaos attractor of original system. Wherein, chaos attractor, as one of the feature of chaos system, embodies the regularity of chaos system, it is meant that chaos system finally can fall into a certain specific track. Phase space after reconstruct can by a matrix representation, and the line number of matrix is Embedded dimensions m. As the Embedded dimensions m > 3 of phase space, solid figure cannot be utilized to show its mimic diagram picture; When Embedded dimensions m��3 of phase space, it is possible to show the system trajectory recovered by phase space in solid space, such as chaos attractor in Fig. 3.
As shown in Figure 4, according to chaology, can be similar to by phase space reconfiguration and recover frequency hop sequences { xi, i=1,2 ..., the Nonlinear Dynamical Characteristics of N, the manifestation of phase space can be matrix X,
The each of matrix is classified as a phase point, and any phase point has m time delay point, choosing appropriate time lag �� can guarantee between time delay point interrelated, and this kind of cognation can utilize Bayesian network to portray, so whole phase point X (t)=[x in phase spacet,xt+��,��,xt+(m-2)��,xt+(m-1)��]T, t=1,2, n can carry out Bayesian network study as priori data information, and by the partial data that each phase point of phase space learns as Bayesian network, i-th component of each phase point is as a conception of history measured value of i-th node of Bayesian network. Therefore, the phase space of reconstruct as Bayesian network sample D, on this basis, can adopt the Markov border of the innovatory algorithm study query node based on MMPC, and using Markov border as Bayes's localized network structure, and then build Bayesian network predictive model.
As shown in Figure 5 and Figure 6, for the frequency hop sequences with 256 frequency numbers that a group is intercepted and captured, through processes such as data prediction, phase space reconfiguration, Bayesian network structure, feedback and model adjustment, a stable Bayesian network model can be obtained, and then for frequency hop sequences prediction, and we to predict accuracy and predict the judgement criteria of mean rate as model. Wherein, predict that accuracy refers to that predictor is completely identical with true value, there is not the frequency hopping code number fluctuated up and down and accounts for total per-cent; Prediction mean rate is the mean time predicting that single frequency hopping code is used.
Getting 1200 frequency hopping codes of frequency hop sequences as model training data, front 1000 frequency hopping codes of this section of frequency hop sequences do training set data, and remaining 200 frequency hopping codes are as model inspection data. 1000 frequency hopping codes of model training data rear adjacent, as prediction collection, utilize the phase space data after reconstruct to carry out Bayesian network study, and are predicted by frequency hop sequences according to Bayesian network posteriority reasoning algorithm. In Figure 5, red " o " represents that Bayesian network model predicts correct frequency hopping code, and blue " * " represents the true frequency hopping code of frequency hop sequences and accurately do not predicted by model. In figure 6, blue " " represents the prediction accuracy of model, and its consensus forecast accuracy is 50.8%. Prediction mean rate is 4.2 �� 10-4s��
The situation that the embodiment of the present invention is greater than 64 for Bayesian network interior joint value number, node number is greater than 4, the time lag �� of C-C method assessment frequency hop sequences and embed window width �� in prior artwFor a long time consuming time, unstable result; The Bayesian network sample D that the K2 algorithm of study network structure obtains after frequency hop sequences cannot be utilized to reconstruct learns to obtain Bayesian network; Complicated computing is needed, and result is unreliable when utilizing the learning method based on dependency test to learn complete bayesian network structure; The node number that the Markov border of query node comprises is ��, parameter alpha value is more big, build the model training data volume needed for Bayesian network model more big, and in the actual environment of communication antagonism, it is necessary to model accepts a small amount of frequency hop sequences and can set up predictive model and implement interference prediction. Therefore, utilize Markov border special property (in Bayesian network, given query node XiThe Markov border mb (X of (1��i��m)i), then XiConditional sampling is other nodes all in network) learn the Markov border of query node, using the Markov border of query node as model structure, simplify Bayesian network model, forecasting efficiency is higher.
The foregoing is only the better embodiment of the present invention, not in order to limit the present invention, all any amendment, equivalent replacement and improvement etc. done within the spirit and principles in the present invention, all should be included within protection scope of the present invention.
Claims (2)
1. the pre-examining system of the frequency hop sequences of a graphic based model, it is characterised in that, the pre-examining system of frequency hop sequences of the graphic based model set up by phase space reconfiguration is comprised:
Pre-processing module, carries out denoising to the original frequency hopping sequences intercepted and captured, goes bandwidth, choose appropriate one section frequency hop sequences { xi, i=1,2 ..., N as the training set data of model construction, using M frequency hopping code of training set data rear adjacent as model testing data;
Prediction module, is connected with described pre-processing module, for phase space reconstruction and structure predictive model;
Feedback adjustment module, is connected with described pre-processing module and prediction module, for accuracy detection, and feedback and model adjustment;
Embedded dimensions m and time lag �� needed for phase space reconstruction adopt Cao method and auto-correlation method to solve respectively;
Build the Markov border of the local of the Bayesian network needed for predictive model network structure by the innovatory algorithm study query node based on MMPC, and using Markov border as Bayes's localized network structure;
Described pre-processing module also comprises:
The frequency hop sequences data collection module of original frequency hopping sequences is intercepted and captured with melodeon; For removing the data processing unit of original frequency hopping sequences noise, bandwidth, it is connected with described frequency hop sequences data collection module; Through normalization method obtain for phase space reconstruction and build model training set data unit, be connected with described data processing unit; For predicting the detection model check data unit of accuracy detection, feedback and model adjustment, it is connected with described data processing unit; Through the proof data unit for predicting that phase space �� step continuation obtains, it is connected with described training set data unit;
Described prediction module also comprises:
The phase space data cell utilizing training set data study to obtain, is connected with described training set data unit; On phase space reconfiguration basis, through the Bayesian network unit that local structure study and parameter learning obtain, it is connected with described phase space data cell; Utilize the prediction ordered series of numbers unit of Bayesian network model prediction frequency hop sequences, it is connected with described Bayesian network unit and proof data unit.
2. the frequency hop sequences Forecasting Methodology of a graphic based model, it is characterised in that, described frequency hop sequences Forecasting Methodology comprises the following steps: frequency hop sequences Forecasting Methodology comprises the following steps:
Step 1, carries out denoising to the original frequency hopping sequences intercepted and captured, goes bandwidth, choose appropriate one section frequency hop sequences { xi, i=1,2 ..., N as the training set data of model construction, using M frequency hopping code of training set data rear adjacent as model testing data;
Step 2, utilizes auto-correlation method and Cao method to ask the time lag �� and Embedded dimensions m of phase space reconstruction, then according to these two parameters, using training set data reconstruct m �� n dimension matrix as phase space X, wherein
As the data sample D learnt for Bayesian network, the scale of data sample D is n;
Step 3, based on the innovatory algorithm study query node X of MMPCmMarkov border, recycling maximum likelihood estimation learns the parameter of each node, final obtains the local Bayesian network being used for the prediction of multi-frequency point frequency hop sequences;
Step 4, the prediction step �� of model=��, �� is time lag, according to Bayes's posteriority reasoning algorithm, calculatesWherein the span of l is 1��l�ܦ�, and namely l is the l time prediction of prediction in �� step prediction, and E (n+l) is query node XmMarkov boundary set in the value of each node; When P is maximum, fiIt is exactly xN+lPredictor;
Step 5, after every �� step prediction terminates, preserve the frequency hopping code of prediction, and true frequency hopping code corresponding in model inspection data is extended for phase point X (n+l), l=1, ��, and add in phase space, upgrade network parameter, forward step 4 to, until M frequency hopping code of training set data rear adjacent predicted after terminate;
Step 6, utilizes model inspection Data Detection model prediction accuracy, if not reaching accuracy requirement, feedback and model adjustment, forward step 3 to, if reaching accuracy requirement, namely obtaining stable Bayesian network predictive model, predicting for frequency hop sequences;
In the delay values ri that auto-correlation method is tried to achievedOn basis, time lag �� can regulate downwards, and the span of �� is 2�ܦӡܦ�d;
The node number that Markov border comprises is ��, and the span of parameter alpha is 2�ܦ���5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310137018.1A CN103209005B (en) | 2013-04-18 | 2013-04-18 | The pre-examining system of frequency hop sequences of a kind of graphic based model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310137018.1A CN103209005B (en) | 2013-04-18 | 2013-04-18 | The pre-examining system of frequency hop sequences of a kind of graphic based model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103209005A CN103209005A (en) | 2013-07-17 |
| CN103209005B true CN103209005B (en) | 2016-06-01 |
Family
ID=48756107
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310137018.1A Expired - Fee Related CN103209005B (en) | 2013-04-18 | 2013-04-18 | The pre-examining system of frequency hop sequences of a kind of graphic based model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103209005B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104779973B (en) * | 2014-01-13 | 2017-02-22 | 中国科学院沈阳自动化研究所 | Adaptive frequency hopping method based on Markovian decision |
| CN103973335B (en) * | 2014-05-07 | 2016-08-24 | 电子科技大学 | Synchronising frequency hopping sequence prediction method based on chaology |
| CN104917515A (en) * | 2015-05-19 | 2015-09-16 | 清华大学 | Noise tolerable universal digital logic circuit and construction method thereof |
| CN106650189A (en) * | 2015-10-30 | 2017-05-10 | 日本电气株式会社 | Causal relationship mining method and device |
| CN105894029B (en) * | 2016-03-31 | 2019-01-25 | 浙江大学 | A kind of adaptive motion track data denoising method solved based on Fermat point |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877077A (en) * | 2009-11-25 | 2010-11-03 | 天津工业大学 | Time series predicting model |
| CN102495937A (en) * | 2011-10-18 | 2012-06-13 | 南京信息工程大学 | Prediction method based on time sequence |
| CN102685766A (en) * | 2012-05-13 | 2012-09-19 | 西华大学 | Wireless network flow prediction method based on local minimax probability machine |
-
2013
- 2013-04-18 CN CN201310137018.1A patent/CN103209005B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877077A (en) * | 2009-11-25 | 2010-11-03 | 天津工业大学 | Time series predicting model |
| CN102495937A (en) * | 2011-10-18 | 2012-06-13 | 南京信息工程大学 | Prediction method based on time sequence |
| CN102685766A (en) * | 2012-05-13 | 2012-09-19 | 西华大学 | Wireless network flow prediction method based on local minimax probability machine |
Non-Patent Citations (1)
| Title |
|---|
| "基于贝叶斯网络的跳频序列多步预测";张恒伟等;《计算机应用研究》;20120131;第29卷(第1期);第237-240页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103209005A (en) | 2013-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103209005B (en) | The pre-examining system of frequency hop sequences of a kind of graphic based model | |
| Dawson et al. | Evaluation of artificial neural network techniques for flow forecasting in the River Yangtze, China | |
| Wang et al. | A new wind power prediction method based on chaotic theory and Bernstein Neural Network | |
| CN111079931A (en) | State space probabilistic multi-time-series prediction method based on graph neural network | |
| CN107274030B (en) | Runoff Forecast method and system based on hydrology variable year border and monthly variation characteristic | |
| CN113313947A (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| CN103117817B (en) | A kind of frequency spectrum detecting method under time-varying fading channels | |
| CN113762595B (en) | Traffic time prediction model training method, traffic time prediction method and equipment | |
| CN114636932A (en) | Method and system for predicting remaining service life of battery | |
| CN112437451B (en) | Wireless network flow prediction method and device based on generation countermeasure network | |
| EP3502978A1 (en) | Meta-learning system | |
| CN112272074B (en) | Information transmission rate control method and system based on neural network | |
| CN113852432A (en) | RCS-GRU model-based spectrum prediction sensing method | |
| Adnan et al. | New Artificial Neural Network and Extended Kalman Filter hybrid model of flood prediction system | |
| Elvers et al. | Short-term probabilistic load forecasting at low aggregation levels using convolutional neural networks | |
| Ouyang et al. | Model of selecting prediction window in ramps forecasting | |
| Tang et al. | Adaptive probabilistic vehicle trajectory prediction through physically feasible bayesian recurrent neural network | |
| CN106953801B (en) | Random shortest path realization method based on hierarchical learning automaton | |
| CN113784380B (en) | Topology prediction method adopting graph attention network and fusion neighborhood | |
| CN104092503A (en) | Artificial neural network spectrum sensing method based on wolf pack optimization | |
| CN107480786B (en) | Output state limitation-based recurrent neural network track likelihood probability calculation method | |
| CN109800517A (en) | Improved reverse modeling method for magnetorheological damper | |
| EP2386987B1 (en) | A method of reinforcement learning, corresponding computer program product, and data storage device therefor | |
| CN115310355A (en) | Multi-energy coupling-considered multi-load prediction method and system for comprehensive energy system | |
| CN115459982A (en) | Power network false data injection attack detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160601 Termination date: 20170418 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |