CN106897705B - Ocean observation big data distribution method based on incremental learning - Google Patents

Ocean observation big data distribution method based on incremental learning Download PDF

Info

Publication number
CN106897705B
CN106897705B CN201710117922.4A CN201710117922A CN106897705B CN 106897705 B CN106897705 B CN 106897705B CN 201710117922 A CN201710117922 A CN 201710117922A CN 106897705 B CN106897705 B CN 106897705B
Authority
CN
China
Prior art keywords
data
observation
incremental
classification
calculating
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
Application number
CN201710117922.4A
Other languages
Chinese (zh)
Other versions
CN106897705A (en
Inventor
黄冬梅
贺琪
随宏运
何盛琪
石少华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Ocean University
Original Assignee
Shanghai Ocean University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Ocean University filed Critical Shanghai Ocean University
Priority to CN201710117922.4A priority Critical patent/CN106897705B/en
Publication of CN106897705A publication Critical patent/CN106897705A/en
Application granted granted Critical
Publication of CN106897705B publication Critical patent/CN106897705B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)
  • Multimedia (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a marine observation big data distribution method based on incremental learning, which comprises the following steps: s1: inputting an incremental marine observation data set to be laid out; s2: initializing a storage capacity; s3: calculating the data value of the data in the incremental data set; s4: dividing all data in the incremental dataset; s5: training the incremental data set by using an incremental learning method; s6: laying out the trained data; s7: outputting the incremental marine observation data set after layout; the incremental learning method in step S5 is a support vector machine incremental learning method. The method has the advantages that the classification accuracy is ensured, and meanwhile, the expenditure of training time and the response time of a user for accessing data are reduced; solving the problem of excessive matching by using an incremental learning algorithm of a support vector machine; effectively compressing the size of the sample set and discarding useless samples.

Description

Ocean observation big data distribution method based on incremental learning
Technical Field
The invention relates to the technical field of marine data distribution, in particular to a marine observation big data distribution method based on incremental learning.
Background
With the gradual promotion of the strong national strategy of oceans in China, the rapid development of scientific big data technology injects scientific power into the oceanic economic industry. In addition, the orbit of the special satellites such as the ocean one-number A star and the ocean one-number B star successfully optimizes the ocean three-dimensional observation road network in China, so that the real-time multi-modal ocean data with high precision, high frequency and large coverage is increased in a geometric series explosion manner. The characteristics of the science and discipline of ocean science and the multi-source of the ocean data acquisition means lead the ocean data to have the characteristics of massiveness, multi-dimension, real time, strong correlation and the like, so that the ocean data becomes a model of big data. The method for effectively storing and managing the marine observation big data and constructing the marine big data service is a key way for mining the value of marine data.
Data distribution is a key problem in data storage, and is to divide data into a series of disjoint data segments or regions, and to place the data segments or regions on each data node in a distributed manner according to a certain data allocation strategy. In the data distribution process, a good slicing implementation strategy is the key of data distribution. The existing data slicing strategy (such as round robin division) is suitable for a general relational database with a fixed mode, and has a remarkable effect on general data. However, the multi-modal real-time ocean observation big data has special properties, so that the traditional slicing strategy ignores the characteristics of the traditional slicing strategy when slicing the ocean observation big data, and certain practicability is lacked. Therefore, the data value of the ocean data itself needs to be further considered and analyzed, and the data can be effectively distributed and stored.
In addition, with the rapid development of large-scale marine stereoscopic observation technology, in the actual marine observation process, the information of marine observation big data is not acquired at one time, and new data is continuously added. In the face of large data observed at sea, which is greatly increasing, it is obviously unrealistic if a great deal of time is spent on modeling and storing all the data again or mining the data each time. Incremental learning can effectively solve the problems, so that the storage and management of ocean big data can be better towards service and practicability.
The main purpose of data distribution is to reduce data access across logical partitions or physical nodes by allowing as much data as possible to be stored locally through reasonable distribution of the data.
Under the strategic demands of the ocean strong nation and the rapid development of novel information technology, the mining and management of ocean big data can provide important information resources for the research of the observation of ocean environment, the detection of ocean resources, the early warning and forecast of ocean disasters and the like. However, with the diversification and deep layout of marine observation means and equipment, such as buoy, satellite, remote sensing, observation station and other real-time data source acquisition, the security level of data volume increases, so that the traditional data distribution strategy has certain limitations on the storage and management of marine data.
In the face of mass ocean observation data which grows rapidly, how to effectively utilize the learning result of the historical data to carry out efficient analysis on the newly added data is achieved, so that repeated training and learning on historical samples are avoided, the accurate data classification result is the key for distributing the ocean observation data, and the problem can be well solved by incremental learning. At present, the incremental learning algorithm is well applied in some fields. In the process of distributing the ocean data, the good dynamic self-adaptability can bring better influence on the data distribution effect and the response time of a user for accessing the data in the face of real-time updated observation data. Therefore, in the face of continuously real-time updated marine observation big data, it is important to introduce the idea of incremental learning into the data distribution of the marine big data.
The chinese invention patent CN201610561677.1, published as 2016.12.14, discloses an SAR image classification method based on SPM and depth increment SVM. However, the method cannot be adapted to ocean data and cannot achieve the technical effect of the invention.
Therefore, an incremental learning-based marine observation big data distribution method which reduces the overhead of training time and the response time of user access data and solves excessive matching is needed, and no report on the method is found at present.
Disclosure of Invention
The invention aims to provide a marine observation big data distribution method based on incremental learning, aiming at the defects in the prior art.
In order to achieve the purpose, the invention adopts the technical scheme that:
a marine observation big data distribution method based on incremental learning comprises the following steps:
s1: inputting an incremental marine observation data set to be laid out;
s2: initializing a storage capacity;
s3: calculating the data value of the data in the incremental data set;
s4: dividing all data in the incremental dataset;
s5: training the incremental data set by using an incremental learning method;
s6: laying out the trained data;
s7: outputting the incremental marine observation data set after layout;
the incremental learning method in step S5 is a support vector machine incremental learning method.
The data value calculation in step S3 includes calculating timeliness, calculating relevance, and calculating regionality.
The division in step S4 is to initially divide all data in the data set by using a k-means method, and divide the data set into an active area and an inactive area.
The layout in step S6 is to lay out the trained data according to the active area and the inactive area.
The calculation method of step S3 includes the following steps:
s31: calculating the timeliness
Calculating the timeliness of the marine observation big data by using a TF-IDF weighting technology, wherein the calculation formula is as follows:
Figure BDA0001236143720000031
wherein N is the total data volume of the ocean observation large data set, and N isiRepresenting the number of datasets containing an attribute item d of observed data, tfi(d) Representing the frequency of occurrence of the observation data attribute item d in the dataset, Wi(d) Representing the weight of the attribute item d.
S32: calculating relevance
Is provided with
Figure BDA0001236143720000034
Respectively representing the number of application observationsAccording to dkAnd dmThe observation task of (2), then the data d is observedkAnd dmDegree of correlation S betweenijThe calculation formula of (a) is as follows:
Figure BDA0001236143720000032
s33: calculating regional characteristics
Calculating the distance L between each observation position in each observation area by using a Euclidean distance calculation methodmnThe calculation formula is as follows:
Lmn＝√(xm-xn)2+(ym-yn)2(3)
wherein L ismnRepresenting the distance, x, between observation point m and observation point nmAnd xnLongitude values, y, representing observation point m and observation point n, respectivelymAnd ynAnd respectively represent latitude values of observation point m and observation point n.
Introducing a normalization variable, and taking the ratio of the relative position to the maximum distance value in the whole interval as the distance related value RL of the observation pointmnThe calculation formula is as follows:
Figure BDA0001236143720000033
wherein RLmnMax { L12, L13, L } which represents a distance correlation value between observation point m and observation point n14,……,LmnMeans take the maximum value between each distance value.
S34: calculating data value
For ocean observation big data, according to the frequency of the used big data, the downloading frequency of the big data, the importance degree of a data user, the production cost of a data product and other factors, the weighting silver of each factor is properly selected to calculate the data value, and the calculation formula is as follows:
Vi(d)＝Wi(d)×k1+Si(d)×k2+RLi(d)×k3+C (5)
wherein, Vi(d) Data value, W, representing observed datai(d) Representing the timeliness of the observed data, Si(d) Representing the relevance of observed data, RLi(d) Regionality of belt pack observation data, k1 is Wi(d) K2 is Si(d) K3 is RLi(d) The weighting factor C represents a penalty factor of data value, and is obtained by integrating the attention of users observing data, the time spent on completing data acquisition, the manpower involved and the links spent on producing data.
The working flow of step S5 is as follows:
s51: inputting newly-added ocean big data sample set Bi(i＝1,2,3,……,n)；
S52: judging whether the newly added sample meets the KKT condition:
s521: if the KKT condition is met, classifying a Support Vector Machine (SVM) according to the KKT condition, and then entering a step S56;
s522: if the KKT condition is not met, the process proceeds to step S53;
s53: judgment BiWhether all are on the classification plane:
s531: if B isiAll on the classification surface, classifying the classification surface as a sample on the classification interval, and then proceeding to step S56;
s532: if B isiIf not, the process proceeds to step S54;
s54: judgment BiWhether all are at the edge of the classification plane or the original classification is wrong:
s541: if B isiIf the samples are at the edge of the classification surface or the original classification is wrong, classifying the samples as samples in the classification interval, and then entering the step S56;
s542: if B isiIf the classification is not at the edge of the classification surface or the original classification is correct, the step S55 is executed;
s55: training a sample set according to the data value, namely dividing the sample set by using a k-means method;
s56: and outputting the incremental sample set.
The invention has the advantages that:
1. the classification accuracy is ensured, and meanwhile, the overhead of training time and the response time of user access data are reduced;
2. solving the problem of excessive matching by using an incremental learning algorithm of a support vector machine;
3. effectively compressing the size of the sample set and discarding useless samples.
Drawings
FIG. 1 is a flow chart of a marine observation big data distribution method based on incremental learning.
FIG. 2 is a schematic diagram of an incremental learning method of the marine observation big data distribution method based on incremental learning.
Detailed Description
The following detailed description of the present invention will be made with reference to the accompanying drawings.
The reference numerals and components referred to in the drawings are as follows:
example 1
Referring to fig. 1, the flow of the ocean observation big data distribution method based on incremental learning of the present invention is as follows:
s1: inputting an incremental marine observation data set to be laid out;
s2: initializing a storage capacity;
s3: calculating the data value of the data in the incremental data set;
s4: dividing all data in the incremental dataset;
s5: training the incremental data set by using an incremental learning method;
s6: laying out the trained data;
s7: and outputting the incremental marine observation data set after layout.
The data value calculation in step S3 includes calculating timeliness, calculating relevance, and calculating regionality.
The division in step S4 is to initially divide all data in the data set by using a k-means method, and divide the data set into an active area and an inactive area.
The incremental learning method in step S5 is a support vector machine incremental learning method.
The layout in step S6 is to lay out the trained data according to the active area and the inactive area.
The ocean observation big data distribution method based on incremental learning has the advantages that the accuracy of classification is guaranteed, and meanwhile, the expenditure of training time and the response time of user access data are reduced; solving the problem of excessive matching by using an incremental learning algorithm of a support vector machine; effectively compressing the size of the sample set and discarding useless samples.
Example 2
The invention discloses a data value calculation method of a marine observation big data distribution method based on incremental learning, which comprises the following steps:
s31: calculating the timeliness
The timeliness of the data is one of the important factors affecting the data distribution. In the field of marine research, the timeliness problem of marine big data is very important, especially for incremental marine observation big data. In designing the distribution strategy, the time length of data storage and the data access frequency are important factors for judging whether the data is valuable or not. Data has different meanings at different stages, and is more frequently called by the user when the data has just been stored in the data storage system. Over time, the data becomes historical data relative to the data just stored, and the number of times the historical data is called by the user is reduced sharply. Therefore, ensuring the timeliness of data is a very important issue.
Calculating the timeliness of the marine observation big data by using a TF-IDF weighting technology, wherein the calculation formula is as follows:
Figure BDA0001236143720000061
wherein N is the total data volume of the ocean observation large data set, and N isiRepresenting the number of datasets containing an attribute item d of observed data, tfi(d) Representing the frequency of occurrence of the observation data attribute item d in the dataset, Wi(d) Weight value representing attribute item d。
S32: calculating relevance
The marine data of each observation point has various attributes including longitude, latitude, temperature, humidity, salinity, atmospheric pressure, fluorescence intensity, etc., and there is a certain relation between the attributes of these observation data. Therefore, when laying out the ocean observation big data, it is necessary to consider the correlation between the observation data.
Is provided with
Figure BDA0001236143720000063
Respectively representing application observations dkAnd dmThe observation task of (2), then the data d is observedkAnd dmDegree of correlation S betweenijThe calculation formula of (a) is as follows:
Figure BDA0001236143720000062
s33: calculating regional characteristics
The spatiality of ocean data leads to the regionality of ocean data. When each ocean data is observed, the longitude value and the latitude value of the data must be determined, which is also characterized by the spatial correlation of the ocean data. By analyzing the distance correlation degree of the observation points, the data with the close observation points can be effectively placed in the same data center as much as possible for later trimming some isolated data with the longer distance, so that the storage capacity of the data center is reduced.
Calculating the distance L between each observation position in each observation area by using a Euclidean distance calculation methodmnThe calculation formula is as follows:
Lmn＝√(xm-xn)2+(ym-yn)2(3)
wherein L ismnRepresenting the distance, x, between observation point m and observation point nmAnd xnLongitude values, y, representing observation point m and observation point n, respectivelymAnd ynAnd respectively represent latitude values of observation point m and observation point n.
Introduction of normalization variables, using phasesThe ratio of the position to the maximum distance value in the whole interval is used as the distance related value RL of the observation pointmnThe calculation formula is as follows:
Figure BDA0001236143720000071
wherein RLmnRepresents a distance correlation value between observation point m and observation point n, max { L }12,L13,L14,……,LmnMeans take the maximum value between each distance value.
S34: calculating data value
For ocean observation big data, according to the frequency of the used big data, the downloading frequency of the big data, the importance degree of a data user, the production cost of a data product and other factors, the weighting silver of each factor is properly selected to calculate the data value, and the calculation formula is as follows:
Vi(d)＝Wi(d)×k1+Si(d)×k2+RLi(d)×k3+C (5)
wherein, Vi(d) Data value, W, representing observed datai(d) Representing the timeliness of the observed data, Si(d) Representing the relevance of observed data, RLi(d) Regionality of belt pack observation data, k1 is Wi(d) K2 is Si(d) K3 is RLi(d) The weighting factor C represents a penalty factor of data value, and is obtained by integrating the attention of users observing data, the time spent on completing data acquisition, the manpower involved and the links spent on producing data.
Example 3
Referring to fig. 2, the incremental learning method of the marine observation big data distribution method based on incremental learning of the present invention is as follows:
s51: inputting newly-added ocean big data sample set Bi(i＝1,2,3,……,n)；
S52: judging whether the newly added sample meets the KKT condition:
s521: if the KKT condition is met, classifying a Support Vector Machine (SVM) according to the KKT condition, and then entering a step S56;
s522: if the KKT condition is not met, the process proceeds to step S53;
s53: judgment BiWhether all are on the classification plane:
s531: if B isiAll on the classification surface, classifying the classification surface as a sample on the classification interval, and then proceeding to step S56;
s532: if B isiIf not, the process proceeds to step S54;
s54: judgment BiWhether all are at the edge of the classification plane or the original classification is wrong:
s541: if B isiIf the samples are at the edge of the classification surface or the original classification is wrong, classifying the samples as samples in the classification interval, and then entering the step S56;
s542: if B isiIf the classification is not at the edge of the classification surface or the original classification is correct, the step S55 is executed;
s55: training a sample set according to the data value, namely dividing the sample set by using a k-means method;
s56: and outputting the incremental sample set.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and additions can be made without departing from the method of the present invention, and these modifications and additions should also be regarded as the protection scope of the present invention.

Claims (1)

1. A marine observation big data distribution method based on incremental learning is characterized by comprising the following steps:
s1: inputting an incremental marine observation data set to be laid out;
s2: initializing a storage capacity;
s3: calculating the data value of the data in the incremental data set;
s4: dividing all data in the incremental dataset;
s5: training the incremental data set by using an incremental learning method;
s6: laying out the trained data;
s7: outputting the incremental marine observation data set after layout;
wherein, the incremental learning method in the step S5 is a support vector machine incremental learning method;
the data value calculation in the step S3 includes calculating timeliness, calculating relevance, and calculating regionality;
the division in the step S4 is to initially divide all data in the data set by using a k-means method, and divide the data set into an active area and an inactive area;
the layout in step S6 is to lay out the trained data according to the active area and the inactive area;
the calculation method of step S3 includes the following steps:
s31: calculating the timeliness
Calculating the timeliness of the marine observation big data by using a TF-IDF weighting technology, wherein the calculation formula is as follows:
Figure FDA0002233565050000011
wherein N is the total data volume of the ocean observation large data set, and N isiRepresenting the number of datasets containing an attribute item d of observed data, tfi(d) Representing the frequency of occurrence of the observation data attribute item d in the dataset, Wi(d) Representing the weight of the attribute item d;
s32: calculating relevance
Let Tdk，TdmRespectively representing application observations dkAnd dmThe observation task of (2), then the data d is observedkAnd dmDegree of correlation S betweenijThe calculation formula of (a) is as follows:
Figure FDA0002233565050000012
s33: calculating regional characteristics
Calculating the distance between each observation position in each observation area by using a Euclidean distance calculation methodLmnThe calculation formula is as follows:
Lmn＝√[(xm-xn)2+(ym-yn)2](3)
wherein L ismnRepresenting the distance, x, between observation point m and observation point nmAnd xnLongitude values, y, representing observation point m and observation point n, respectivelymAnd ynRespectively representing latitude values of an observation point m and an observation point n;
introducing a normalization variable, and taking the ratio of the relative position to the maximum distance value in the whole interval as the distance related value RL of the observation pointmnThe calculation formula is as follows:
Figure FDA0002233565050000021
wherein RLmnRepresents a distance correlation value between observation point m and observation point n, max { L }12,L13,L14,……,LmnMeans take the maximum value between each distance value;
s34: calculating data value
For ocean observation big data, according to the frequency of the used big data, the downloading frequency of the big data, the importance degree of a data user and the production cost factors of a data product, a weighting factor of each factor is properly selected to calculate the data value, and the calculation formula is as follows:
Vi(d)＝Wi(d)×k1+Si(d)×k2+RLi(d)×k3+C (5)
wherein, Vi(d) Data value, W, representing observed datai(d) Representing the timeliness of the observed data, Si(d) Representing the relevance of observed data, RLi(d) Regionality of belt pack observation data, k1 is Wi(d) K2 is Si(d) K3 is RLi(d) C represents a penalty factor for the value of the data, the user's attention from the observed data, the time elapsed for completion of data acquisition, the manpower involved and the data productionComprehensively obtaining links;
the working flow of step S5 is as follows:
s51: inputting newly-added ocean big data sample set Bi(i＝1,2,3,……,n)；
S52: judging whether the newly added sample meets the KKT condition:
s521: if the KKT condition is met, classifying a Support Vector Machine (SVM) according to the KKT condition, and then entering a step S56;
s522: if the KKT condition is not met, the process proceeds to step S53;
s53: judgment BiWhether all are on the classification plane:
s531: if B isiAll on the classification surface, classifying the classification surface as a sample on the classification interval, and then proceeding to step S56;
s532: if B isiIf not, the process proceeds to step S54;
s54: judgment BiWhether all are at the edge of the classification plane or the original classification is wrong:
s541: if B isiIf the samples are at the edge of the classification surface or the original classification is wrong, classifying the samples as samples in the classification interval, and then entering the step S56;
s542: if B isiIf the classification is not at the edge of the classification surface or the original classification is correct, the step S55 is executed;
s55: training a sample set according to the data value, namely dividing the sample set by using a k-means method;
s56: and outputting the incremental sample set.
CN201710117922.4A 2017-03-01 2017-03-01 Ocean observation big data distribution method based on incremental learning Active CN106897705B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710117922.4A CN106897705B (en) 2017-03-01 2017-03-01 Ocean observation big data distribution method based on incremental learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710117922.4A CN106897705B (en) 2017-03-01 2017-03-01 Ocean observation big data distribution method based on incremental learning

Publications (2)

Publication Number Publication Date
CN106897705A CN106897705A (en) 2017-06-27
CN106897705B true CN106897705B (en) 2020-04-10

Family

ID=59185404

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710117922.4A Active CN106897705B (en) 2017-03-01 2017-03-01 Ocean observation big data distribution method based on incremental learning

Country Status (1)

Country Link
CN (1) CN106897705B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110598860B (en) * 2019-08-06 2023-02-24 山东省科学院海洋仪器仪表研究所 Multi-station online wave cycle data prediction diagnosis method
CN111783869A (en) * 2020-06-29 2020-10-16 杭州海康威视数字技术股份有限公司 Training data screening method and device, electronic equipment and storage medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005022449A1 (en) * 2003-08-25 2005-03-10 Siemens Medical Solutions Usa, Inc. Greedy support vector machine classification for feature selection applied to the nodule detection problem
KR20080078292A (en) * 2007-02-23 2008-08-27 재단법인서울대학교산학협력재단 Domain density description based incremental pattern classification method
CN101944122A (en) * 2010-09-17 2011-01-12 浙江工商大学 Incremental learning-fused support vector machine multi-class classification method
CN103605631A (en) * 2013-11-20 2014-02-26 温州大学 Increment learning method on the basis of supporting vector geometrical significance
CN103780501A (en) * 2014-01-03 2014-05-07 濮阳职业技术学院 Peer-to-peer network traffic identification method of inseparable-wavelet support vector machine
CN104866869A (en) * 2015-05-29 2015-08-26 武汉大学 Time sequence SAR (Synthetic Aperture Radar) image classification method on the basis of distribution difference and incremental learning
CN105913073A (en) * 2016-04-05 2016-08-31 西安电子科技大学 SAR image target identification method based on depth increment support vector machine

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005022449A1 (en) * 2003-08-25 2005-03-10 Siemens Medical Solutions Usa, Inc. Greedy support vector machine classification for feature selection applied to the nodule detection problem
KR20080078292A (en) * 2007-02-23 2008-08-27 재단법인서울대학교산학협력재단 Domain density description based incremental pattern classification method
CN101944122A (en) * 2010-09-17 2011-01-12 浙江工商大学 Incremental learning-fused support vector machine multi-class classification method
CN103605631A (en) * 2013-11-20 2014-02-26 温州大学 Increment learning method on the basis of supporting vector geometrical significance
CN103780501A (en) * 2014-01-03 2014-05-07 濮阳职业技术学院 Peer-to-peer network traffic identification method of inseparable-wavelet support vector machine
CN104866869A (en) * 2015-05-29 2015-08-26 武汉大学 Time sequence SAR (Synthetic Aperture Radar) image classification method on the basis of distribution difference and incremental learning
CN105913073A (en) * 2016-04-05 2016-08-31 西安电子科技大学 SAR image target identification method based on depth increment support vector machine

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Modeling and Analysis in Marine Big Data：Advances and Challenges;Dongmei Huang 等;《Mathematical Problems in Engineering》;20150101;第1-13页 *
一种基于聚类分析的增量支持向量机入侵监测方法;王亚兵;《电脑知识与技术》;20140731;第10卷(第19期);摘要、第1.2节、第2.1节、第2.4节 *

Also Published As

Publication number Publication date
CN106897705A (en) 2017-06-27

Similar Documents

Publication Publication Date Title
Martınez et al. Scalable learning of Bayesian network classifiers
CN102119389B (en) Automatic image annotation using semantic distance learning
Kumar et al. A novel clustering technique for efficient clustering of big data in Hadoop Ecosystem
CN106202335B (en) A kind of traffic big data cleaning method based on cloud computing framework
CN106951498A (en) Text clustering method
CN105184307A (en) Medical field image semantic similarity matrix generation method
CN106779087A (en) A kind of general-purpose machinery learning data analysis platform
CN103064903B (en) Picture retrieval method and device
CN107038505B (en) Ore finding model prediction method based on machine learning
CN104820905A (en) Space trajectory big data analysis-based person management and control method and system
CN109492796A (en) A kind of Urban Spatial Morphology automatic Mesh Partition Method and system
CN113434736A (en) Multi-dimensional mixed indexing method and system for remote sensing big data
CN106250925B (en) A kind of zero Sample video classification method based on improved canonical correlation analysis
CN106897705B (en) Ocean observation big data distribution method based on incremental learning
CN103761286B (en) A kind of Service Source search method based on user interest
CN106022359A (en) Fuzzy entropy space clustering analysis method based on orderly information entropy
CN112241676A (en) Method for automatically identifying terrain sundries
CN111626311B (en) Heterogeneous graph data processing method and device
CN106611016A (en) Image retrieval method based on decomposable word pack model
CN104778480A (en) Hierarchical spectral clustering method based on local density and geodesic distance
Liu et al. A novel DBSCAN clustering algorithm via edge computing-based deep neural network model for targeted poverty alleviation big data
CN109409407A (en) A kind of industry monitoring data clustering method based on LE algorithm
CN105513051B (en) A kind of Processing Method of Point-clouds and equipment
CN106775694A (en) A kind of hierarchy classification method of software merit rating code product
CN116304213A (en) RDF graph database sub-graph matching query optimization method based on graph neural network

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