CN110610192B - Spectrum space channel clustering method - Google Patents
Spectrum space channel clustering method Download PDFInfo
- Publication number
- CN110610192B CN110610192B CN201910727149.2A CN201910727149A CN110610192B CN 110610192 B CN110610192 B CN 110610192B CN 201910727149 A CN201910727149 A CN 201910727149A CN 110610192 B CN110610192 B CN 110610192B
- Authority
- CN
- China
- Prior art keywords
- channel
- clustering
- point
- channels
- similarity
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a spectrum space channel clustering method, which solves the technical problems that the clustering effect is unreasonable and reliable, the clustering time efficiency is high, and non-clustering channel points cannot be clustered.
Description
Technical Field
The invention relates to the field of clustering of channels in spectrum data, in particular to a spectrum space channel clustering method.
Background
In recent years, with the rapid development of wireless technologies and services, wireless spectrum becomes more and more crowded, spectrum resources are increasingly in short supply, and the hidden dangers of radio order and spectrum safety become more and more serious; it can be seen that there is a serious contradiction between the shortage of spectrum resources and the low utilization rate of spectrum resources, so that both domestic and foreign scholars are actively engaged in exploring a new spectrum resource allocation mechanism, and some scholars propose a dynamic spectrum allocation mechanism, which is different from the conventional fixed spectrum allocation mechanism in that, regarding a frequency band allocated to an authorized user, when the frequency band is in an idle state, an unauthorized user can utilize the frequency band, and such a change will improve the usage efficiency of the spectrum.
In the future, the spectrum sharing and the radio order management of the mobile communication system will be developed and transformed from the traditional fixed allocation and management mode to the flexible and dynamic direction, so how to make use of the electromagnetic spectrum situation to perform more intelligent and effective spectrum management and decision is an important subject of research in this field.
The existing channel clustering has the technical problems of unreasonable and reliable clustering effect, clustering time efficiency and incapability of clustering non-clustered channel points. The invention provides a spectrum space channel clustering method which solves the problems.
Disclosure of Invention
The invention aims to solve the technical problems that the clustering effect is unreasonable and reliable, the clustering time efficiency is high, and non-clustering channel points cannot be clustered in the prior art. A new spectrum space channel clustering method is provided, and the spectrum space channel clustering method has the characteristics of reasonability and reliability.
In order to solve the technical problems, the technical scheme is as follows:
a method of spectral-spatial channel clustering, the method comprising:
acquiring spectrum data through a spectrum receiver, preprocessing the data, and processing the spectrum data of all channels into a binary sequence, wherein 0 represents that the channel is idle and 1 represents that the channel is occupied;
step two, calculating the similarity between the information by using an improved channel similarity calculation formula introducing a time factor;
step three, obtaining similarity values of the channels through the step two, and calculating the distance between the channels by using an information entropy formula;
step four, clustering the channels by using the distance between the channels obtained in the step three and an improved DBSCAN-based channel clustering algorithm;
step five, using the clustering channel clustered in the step four as a clustering result, and calculating a representative channel of each class by using a PEG method and an MPEG method;
and step six, defining non-clustering points after the improved DBSCAN-based channel clustering algorithm is clustered as free points, and performing secondary clustering on the free points through a frequent pattern mining algorithm based on a representative channel occupation state sequence to obtain a final clustering result.
In the invention, a time factor is introduced into a traditional channel similarity calculation method, and the time complexity of a channel clustering algorithm based on minimum information entropy increment and a channel clustering algorithm based on density is considered to be higher, so that an improved channel clustering algorithm based on density is provided in the invention, and repeated inquiry of channel points in the traditional algorithm is reduced to improve the algorithm clustering efficiency. Then, aiming at the problem that the channel clustering algorithm based on the minimum information entropy increment and the channel clustering algorithm based on the density are not good for non-clustering channel points, the invention introduces a frequent pattern mining algorithm based on a representative channel occupation state sequence on the basis of the improved channel clustering algorithm based on the density to perform secondary clustering on non-clustering.
In the foregoing solution, for optimization, further, the second step includes defining similarity between information:
where ρ isijDenoted is the similarity coefficient, s m, for channel i and channel j]Indicating that at the mth timedrop slot, if ci[m]＝cj[m]Then s [ m ]]Is 1; otherwise, s [ m ]]Is 0;
weight [ m ] represents the channel similarity weight at the mth timestore time slot;
weight ═ {0.9,0.9+ x,0.9+2 x., 0.9+ (n-1) x } represents the weight of channel i with time slot size n, the distribution of the defined weights obeys the arithmetic series, and the sum of the weights is n;
m is a positive integer and n is a positive integer.
Further, the distance between the channels is calculated by using the information entropy formula, and the value calculated by subtracting the information entropy formula from 1 is the distance between the channels:
wherein the information entropy h (x) satisfies symmetry and is symmetric with respect to ρ ═ 0.5; when rho is larger than or equal to 0.5, if the correlation coefficient is higher, the space distance calculated by the information entropy formula is smaller; when rho is less than 0.5, if the channel correlation coefficient is higher, the space distance calculated by the information entropy formula is larger, and m and n are positive integers.
Further, the clustering the channels through the improved DBSCAN-based channel clustering algorithm in the fourth step includes:
step (1), setting a parameter eps in a channel clustering algorithm based on density as a preset value, and setting a parameter minPts as 2;
step (2), calculating the channel space distance between all channels, numbering each channel according to the channel sequence, and constructing a channel index sequence;
step (3), neighborhood query is started from the channel point m1 with the minimum index number, and if the channel point m1 is a core channel point, all nodes in the neighborhood range of the channel point m1 are marked by cluster 1;
step (4), traversing the next channel point n1 which does not enter a cluster and is not marked according to the channel index sequence, and performing neighborhood query on the channel point n 1;
if the channel point n1 is a core channel point and has no core channel point in the overlapped or overlapped area with the marked channel point m1, marking the channel point n 8926 by cluster 2;
if the node has a repeat point with the neighborhood of the channel point m1 and a core channel point s exists in the repeat point, marking the nodes in all neighborhoods of the channel point n1 with cluster 1;
if the channel point n1 has an overlapped node with the unable channel point and a core channel point exists in the overlapped node, the channel point n1 is marked and merged by using different category marks;
and (5) repeating the steps (1) to (4).
The invention has the beneficial effects that: the time factor is considered in the channel similarity calculation method, so that the calculation method is more reasonable and reliable; in addition, in the traditional channel clustering algorithm based on density, the clustering time efficiency is reduced by performing neighborhood query on the overlapped region again, the improved channel clustering algorithm based on density is provided, and then for the free points which do not enter the clustering cluster, the final clustering effect is better than that of the traditional channel clustering algorithm by utilizing a frequent pattern mining algorithm based on the representative channel occupation state sequence.
Drawings
The invention is further illustrated with reference to the following figures and examples.
Fig. 1 is a flowchart of a spectrum space channel clustering method in embodiment 1.
Fig. 2 is a flow chart of the steps of an improved spectrum space channel clustering system based on density and frequency pattern mining.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Example 1
The embodiment provides an improved spectrum space channel clustering method based on density and frequent pattern mining, which comprises the following specific implementation steps:
s1, preprocessing data: the spectrum data is collected by a spectrum receiver, and then the spectrum data of all channels are processed into a 0,1 sequence, wherein 0 represents the idle channel and 1 represents the occupied channel.
S2, calculating the similarity between channels: the similarity between the information is calculated using an improved channel similarity calculation formula that introduces a time factor.
S3, calculating the distance between channels: the similarity value between the channels is obtained through S2, and the distance between the channels is calculated using the information entropy formula.
S4, clustering channels: using the result obtained in S3, the channels are clustered by the improved DBSCAN-based channel clustering algorithm.
S5, calculating a representative channel of the clustering result: calculating The representative channel Of each class by using PEG (Prediction entry Of The group) and MPEG (minimum Prediction entry Of The group) methods.
S6, carrying out secondary clustering on the free points: aiming at non-clustering points after clustering by the DBSCAN, performing secondary clustering on free channels by a proposed frequent pattern mining algorithm based on the representative channel occupation state sequence, and finally obtaining a clustering result.
Specifically, the similarity calculation formula between the channels in S2 is specifically as follows:
the above formula rhoijDenoted is the similarity coefficient for channel i and channel j, where s m]Indicating that in the mth timedrop slot, if ci[m]＝cj[m]Then s [ m ]]Is 1; otherwise, s [ m ]]Is 0; and weight [ m ]]It represents its channel similarity weight at the mth time slot.
weigh＝{0.9,0.9+x,0.9+2x,...,0.9+(n-1)x}
weight represents a channel i with a time slot size of n, and the corresponding weight is what; assuming that the distribution of the weights is subject to the arithmetic progression, the sum of the weights is calculated to be n.
Specifically, the specific method for calculating the distance between the channels by using the information entropy formula is as follows:
then according to the nature of the information entropy h (x), which is symmetrical and symmetrical about ρ ═ 0.5; when rho is more than or equal to 0.5, if the correlation coefficient is higher, the space distance calculated by the information entropy formula is smaller, which is in line with the purpose and logic of the user, but when rho is less than 0.5, if the channel correlation coefficient is higher, the space distance calculated by the information entropy formula is larger, and then the value calculated by the information entropy is subtracted by 1 by utilizing the symmetry of the information entropy H (x).
Specifically, an improved density-based channel clustering algorithm is used for clustering the number of channels in a certain service, and the algorithm clustering process specifically comprises the following steps:
firstly, setting a parameter eps in a channel clustering algorithm based on density as a reasonable value, and setting a parameter minPts as 2;
secondly, calculating the channel space distance among all channels, and numbering each channel according to the channel sequence;
thirdly, starting neighborhood query from a channel point m1 with the minimum index number, and if m is a core channel point, marking all nodes in the neighborhood range by cluster 1;
fourthly, traversing the next channel point n1 which does not enter a cluster and is not marked according to the channel index sequence, then carrying out neighborhood query, and if the channel point n1 is a core point and has no core channel point in an overlapping or overlapping area with the marked channel m, marking the channel point by cluster 2; if there are duplicate points in the neighborhood of the channel point m1 and there is a core channel point s in these points, the nodes in all neighborhoods in the channel point n1 are marked with cluster1, if there are overlapping nodes with the channel points that cannot and there are core channel points in the overlapping nodes, they are marked and merged with different classes of marks.
And fifthly, repeating the process.
Specifically, The channels are clustered secondarily by a Frequent Pattern Mining Algorithm (RC-FPM) Based on a sequence Representing The occupation State of The channels, wherein The clustering process of The Algorithm comprises The following specific steps:
firstly, determining a hyperparameter min _ rep in an RC-FPM algorithm, wherein the parameter is used for judging whether a mode is a frequent mode;
secondly, sequentially traversing the representative channels in each cluster by using the clustering result of the improved density-based channel clustering algorithm;
thirdly, sequentially traversing i from 1 to the total column number N aiming at a two-dimensional matrix M consisting of the representative channel and each free channel, storing the occurrence times corresponding to M sub-matrixes of 2 rows and i columns in a hash table T (2, y), and if T (2, y-1) is empty, easily obtaining the occurrence times, wherein T (2, y) is also empty;
fourthly, representing M sub-matrixes in 2 rows and i columns as B (2, y), sequentially traversing M sub-matrix blocks B (2, y) in the hash table T (2, y), further sequentially judging whether the corresponding occurrence frequency is greater than a super parameter min _ rep, and if so, putting into an output frequent mode set; if not, the operation is not required;
fifthly, finding out a corresponding strong association rule according to the output frequency spectrum mode;
the embodiment is based on an improved spectrum space channel clustering system based on density and frequent pattern mining, and the system mainly comprises a data uploading module, a channel similarity and channel distance calculating module, a channel clustering module and a clustering result visualization module.
And the data uploading module inputs the occupation state sequence of each channel.
The channel similarity and channel distance calculation module calculates the similarity between two channels and further calculates the distance between two channels.
The channel clustering module clusters the channels according to an improved spectrum space channel clustering algorithm based on density and frequent pattern mining.
And the clustering result visualization module is used for visually displaying the clustering result by utilizing a visualization technology according to the clustering result.
An improved spectrum space channel clustering system based on density and frequency pattern mining mainly comprises: a processor for executing a program;
a memory for storing a program to be processed by the processor, wherein the program when executed has the main flow steps of:
s1 input spectrum channel occupation state sequence
S2, calculating the similarity between the channels, and further calculating the distance between the channels: the similarity between the information is calculated by using an improved channel similarity calculation formula introducing a time factor, and then the channel distance between the information is calculated by an information entropy formula.
S3, clustering by using an improved spectrum space channel clustering method based on density and frequent pattern mining: the improved DBSCAN-based channel clustering is firstly carried out, and then the rest non-clustered channels are processed through an RC-FPM algorithm for secondary clustering.
The computer readable storage medium stores the computer program, as shown in fig. 2, and the main steps of the improved spectrum space channel clustering system based on density and frequency pattern mining comprise the following steps:
s1 input spectrum channel occupation state sequence
S2, calculating the similarity between the channels, and further calculating the distance between the channels: the similarity between the information is calculated by using an improved channel similarity calculation formula introducing a time factor, and then the channel distance between the information is calculated by an information entropy formula.
S3, displaying the clustering result: the channels are clustered by the proposed improved channel clustering algorithm based on density and frequent pattern mining.
The embodiment realizes the mining and clustering of the correlation of the radio service channels, so that the states of other channels in the cluster can be predicted according to the states of the representative channels, the spectrum situation analysis and the spectrum management can be guided, and the method has strong practical significance in the future wireless cognitive communication system.
Although the illustrative embodiments of the present invention have been described above to enable those skilled in the art to understand the present invention, the present invention is not limited to the scope of the embodiments, and it is apparent to those skilled in the art that all the inventive concepts using the present invention are protected as long as they can be changed within the spirit and scope of the present invention as defined and defined by the appended claims.
Claims (4)
1. A spectrum space channel clustering method is characterized in that: the spectrum space channel clustering method comprises the following steps:
acquiring spectrum data through a spectrum receiver, preprocessing the data, and processing the spectrum data of all channels into a binary sequence, wherein 0 represents that the channel is idle and 1 represents that the channel is occupied;
step two, calculating the similarity between the information by using an improved channel similarity calculation formula introducing a time factor;
step three, obtaining similarity values of the channels through the step two, and calculating the distance between the channels by using an information entropy formula;
step four, clustering the channels by using the distance between the channels obtained in the step three and an improved DBSCAN-based channel clustering algorithm;
step five, using the clustering channel clustered in the step four as a clustering result, and calculating a representative channel of each class by using a PEG method and an MPEG method;
and step six, defining non-clustering points after the improved DBSCAN-based channel clustering algorithm is clustered as free points, and performing secondary clustering on the free points through a frequent pattern mining algorithm based on a representative channel occupation state sequence to obtain a final clustering result.
2. The method for clustering spectral spatial channels according to claim 1, wherein: step two includes defining the similarity between the information:
where ρ isijDenoted is the similarity coefficient, s m, for channel i and channel j]Indicating that at the mth timedrop slot, if ci[m]＝cj[m]Then s [ m ]]Is 1; otherwise, s [ m ]]Is 0;
weight [ m ] represents the channel similarity weight at the mth timestore time slot;
weight ═ {0.9,0.9+ x,0.9+2 x., 0.9+ (n-1) x } represents the weight of channel i with time slot size n, the distribution of the defined weights obeys the arithmetic series, and the sum of the weights is n;
m is a positive integer and n is a positive integer.
3. The method of spectral spatial channel clustering according to claim 2, characterized in that: the third step comprises: the distance between the channels is calculated by using an information entropy formula, and the value obtained by subtracting the information entropy formula from 1 is the distance between the channels:
wherein the information entropy h (x) satisfies symmetry and is symmetric with respect to ρ ═ 0.5; when rho is larger than or equal to 0.5, if the correlation coefficient is higher, the space distance calculated by the information entropy formula is smaller; when rho is less than 0.5, if the channel correlation coefficient is higher, the space distance calculated by the information entropy formula is larger, and m and n are positive integers.
4. The method of spectral spatial channel clustering according to claim 3, characterized by: the fourth step of clustering the channels through the improved DBSCAN-based channel clustering algorithm comprises the following steps:
step (1), setting a parameter eps in a channel clustering algorithm based on density as a preset value, and setting a parameter minPts as 2;
step (2), calculating the channel space distance between all channels, numbering each channel according to the channel sequence, and constructing a channel index sequence;
step (3), neighborhood query is started from the channel point m1 with the minimum index number, and if the channel point m1 is a core channel point, all nodes in the neighborhood range of the channel point m1 are marked by cluster 1;
step (4), traversing the next channel point n1 which does not enter a cluster and is not marked according to the channel index sequence, and performing neighborhood query on the channel point n 1;
if the channel point n1 is a core channel point and has no overlapping with the marked channel point m1 or has no core channel point in the overlapping area, the channel point n1 is marked by cluster 2;
if the neighborhood of the channel point n1 and the channel point m1 has a repeat point and a core channel point s exists in the repeat point, marking the channel points in all the neighborhoods of the channel point n1 by cluster 1;
if the channel point n1 has an overlapped node with the unable channel point and a core channel point exists in the overlapped node, the channel point n1 is marked and merged by using different category marks;
and (5) repeating the steps (1) to (4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910727149.2A CN110610192B (en) | 2019-08-07 | 2019-08-07 | Spectrum space channel clustering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910727149.2A CN110610192B (en) | 2019-08-07 | 2019-08-07 | Spectrum space channel clustering method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110610192A CN110610192A (en) | 2019-12-24 |
| CN110610192B true CN110610192B (en) | 2022-03-15 |
Family
ID=68891054
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910727149.2A Active CN110610192B (en) | 2019-08-07 | 2019-08-07 | Spectrum space channel clustering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110610192B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353530A (en) * | 2020-02-24 | 2020-06-30 | 电子科技大学 | Frequency spectrum channel clustering method based on ultralimit learning machine |
| CN112073135B (en) * | 2020-08-24 | 2022-08-19 | 广东工业大学 | Centralized spectrum sensing method, system, storage medium and communication equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103428704A (en) * | 2013-07-30 | 2013-12-04 | 北京邮电大学 | Method and device for sensing frequency spectra |
| WO2015059272A1 (en) * | 2013-10-24 | 2015-04-30 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| CN107657266A (en) * | 2017-08-03 | 2018-02-02 | 华北电力大学（保定） | A kind of load curve clustering method based on improvement spectrum multiple manifold cluster |
| CN109299698A (en) * | 2018-09-30 | 2019-02-01 | 北京交通大学 | A kind of wireless channel scene recognition method based on support vector machines |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10319113B2 (en) * | 2017-06-21 | 2019-06-11 | Foveon, Inc. | Method for recovering highlights and saturated regions and extending dynamic range in a digital image |
-
2019
- 2019-08-07 CN CN201910727149.2A patent/CN110610192B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103428704A (en) * | 2013-07-30 | 2013-12-04 | 北京邮电大学 | Method and device for sensing frequency spectra |
| WO2015059272A1 (en) * | 2013-10-24 | 2015-04-30 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| CN107657266A (en) * | 2017-08-03 | 2018-02-02 | 华北电力大学（保定） | A kind of load curve clustering method based on improvement spectrum multiple manifold cluster |
| CN109299698A (en) * | 2018-09-30 | 2019-02-01 | 北京交通大学 | A kind of wireless channel scene recognition method based on support vector machines |
Non-Patent Citations (2)
| Title |
|---|
| Online Calibration of Phasor Measurement Unit Using Density-Based Spatial Clustering;Xinan Wang 等;《IEEE Transactions on Power Delivery》;20171231;全文 * |
| 一种基于聚类分析的信道估计算法;樊同亮 等;《科技通报》;20140531;第30卷(第5期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110610192A (en) | 2019-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110610192B (en) | Spectrum space channel clustering method | |
| US8996436B1 (en) | Decision tree classification for big data | |
| US9147168B1 (en) | Decision tree representation for big data | |
| CN110569263B (en) | Real-time data deduplication counting method and device | |
| JP7486659B2 (en) | Energy saving processing method and processing device for base station | |
| Papadakis et al. | Progressive, holistic geospatial interlinking | |
| CN105117488A (en) | RDF data balance partitioning algorithm based on mixed hierarchical clustering | |
| Nikitopoulos et al. | Hot spot analysis over big trajectory data | |
| CN108549696B (en) | Time series data similarity query method based on memory calculation | |
| Chen | A diversified AHP-tree approach for multiple-criteria supplier selection | |
| CN103838680A (en) | Data caching method and device | |
| CN114429195A (en) | Performance optimization method and device for hybrid expert model training | |
| US11625451B2 (en) | Local search with global view for large scale combinatorial optimization | |
| CN113076319A (en) | Dynamic database filling method based on outlier detection technology and bitmap index | |
| CN102855278A (en) | Simulation method and system | |
| Malon et al. | Q-learning based radio channels utility evaluation algorithm for the local dynamic spectrum management in mobile ad-hoc networks | |
| Hu et al. | Rebalancing Strategy for Bike‐Sharing Systems Based on the Model of Level of Detail | |
| Zhang et al. | Monochromatic and bichromatic reverse top-k group nearest neighbor queries | |
| CN112601245A (en) | Method and device for evaluating utilization condition of capacity resources in long-term evolution area | |
| Migov et al. | Genetic algorithms for drain placement in wireless sensor networks optimal by the reliability criterion | |
| Zhang et al. | On convergence rate for multi-agent consensus: a community detection algorithm | |
| CN105868435A (en) | Efficient control method for realizing optical network construction on basis of linear correlation analysis | |
| CN112015922B (en) | Method, device, equipment and storage medium for retrieving multimedia file | |
| CN103338460A (en) | Method for calculating centrality of nodes of dynamic network environment | |
| CN115208818B (en) | QoS routing method based on genetic algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |