CN105069129A - Self-adaptive multi-label prediction method - Google Patents

Abstract

The invention discloses an adaptive multi-label prediction method, which is characterized by following steps: 1. Obtain an initialization example set; 2. Obtain leader examples, outsider examples and voter examples in the initialization example set; 3. Obtain voter examples 4. Use the support vector machine to roughly classify the prediction examples; 5. Perform multi-label prediction on the prediction examples. The invention can accurately add labels to network information, improve the accuracy, universality, interpretability and transferability of multi-label prediction, thereby realizing intelligent information classification and processing in a big data environment.

Description

Adaptive Multi-label Prediction Method

technical field

The invention belongs to the field of intelligent information classification and processing, and in particular relates to an adaptive multi-label prediction method applicable to fast clustering of multimedia information in a big data environment and finding density peak points.

Background technique

With the rapid development of the network, the amount of information is increasing geometrically. The current Weibo, forums, WeChat, online video, online shopping and social networks all need tags to facilitate users' search and classification. Accurate and detailed tags On the one hand, it allows users to quickly find what they need. On the other hand, businesses can also use tags to classify users and recommend products that cater to their tastes for different user groups, thereby preventing users from browsing a large amount of irrelevant information and making valuable content is drowned in the ocean of information. Conversely, if merchants cannot properly deal with the problem of information overload, it will eventually lead to the continuous loss of consumers.

At present, the methods of adding multi-labels to information mainly include decomposing multi-labels into independent single labels for labeling and transforming multi-labels into sorting between labels for labeling. Converting to a single label completely ignores the relationship between multiple labels, and the accuracy is low; the sorting between labels not only requires a lot of calculations, but also needs to further determine whether it is the previous label or the post label after the label is sorted The degree of similarity is higher, so there is also the defect of low accuracy.

Compared with the present invention, present processing method has following shortcoming:

1. The current network information uses computer learning methods to make predictions for a single label, that is, the recognition problem. However, due to the correlation between multiple labels in the information, the method of decomposing multiple labels into a single multiple label is used. The accuracy of the label is low and cannot achieve practical purposes.

2. The current multi-label prediction technology can only deal with a given static data set. If new information is considered, it often needs to re-learn and reset parameters. It cannot automatically adjust parameters as the data changes. Therefore, Weak generalization and poor universality.

3. Converting the multi-label prediction of information into the order relationship between labels not only requires a lot of calculations, but also has poor interpretability and low prediction accuracy.

4. Most of the existing multi-label prediction technologies are designed to improve a certain evaluation label, ignoring other standards, which results in poor portability and is only suitable for use in data sets that meet certain conditions.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the existing technology and provide an adaptive multi-label prediction method, in order to accurately add labels to network information and improve the accuracy, universality and interpretability of multi-label prediction And transferability, so as to realize intelligent information classification and processing in the big data environment.

The present invention adopts following technical scheme for solving technical problems:

The feature of a kind of self-adaptive multi-label prediction method of the present invention is to carry out as follows:

Step 1: Obtain the initialization example set D:

Step 1.1. Establish the original example set D′={inst′ ₁ , inst′ ₂ ,…,inst′ _a ,…,inst′ _num′ } from num′ known objects, where inst′ _a represents the ath known object The corresponding original example; 1≤a≤num'; and inst' _a = {attr'_a;lab' _a }; attr' _a represents the attribute set of the a-th known object feature; lab' _a represents The label set of the a-th known object semantics; and attr' _a ={attr' _a,1 ,attr' _a,2 ,...,attr' _a,n }; attr' _a,n represents the a-th The nth attribute of the known object; n is the number of attributes of the ath known object; lab′ _a ={lab′ _a,1 ,lab′ _a,2 ,…,lab′ _a,x ,…,lab′ _{a, m} }; lab′ _{a, x} represents the x-th label of the a-th known object; m is the number of labels of the a-th known object; 1≤x≤m; and: lab′ _{a, x} = 1 means that the semantics of the a-th known object conforms to the x-th label; lab′ _{a, x} = 0 means that the semantics of the a-th known object does not conform to the x-th label;

Step 1.2: Normalize the attribute sets {attr' ₁ , attr' ₂ , ..., attr' _a , ..., attr'_num' } of num' known object features in the original example set D' Processing, obtain normalized attribute sets of num′ known object features {attr″ ₁ , attr″ ₂ ,…, attr″ _a ,…, attr″ _num′ }; when the normalized When the m label values corresponding to the attribute set arrt″ _a of the a-th known object feature are all 0, delete the original example to which the a-th known object belongs after the normalization; thereby obtaining the Initialization example set D={inst ₁ , inst ₂ ,...,inst _i ,...,inst _num }; inst _i represents the example corresponding to the i-th known object after initialization; and inst _i ={attr _i ; lab _i }; attr _i represents the attribute set of the i-th example feature after initialization; lab _i represents the label set of the i-th example semantics after initialization; 1≤i≤num;

Step 2: Solve the clustering degree of each example in the initialization example set D, so as to determine the leader example, outsider example and voter example in the initialization example set D:

Step 2.1, using the m labels of each example in the num examples in the initialization example set D as m-dimensional coordinates, thereby obtaining the Euclidean distance d _ik of the i-th example inst _i and the k-th example inst _k label ;1≤k≤num and k≠i;

Step 2.2, define the number of iterations γ; and initialize γ=1; define the cluster to which the i-th example inst _i belongs to as clu _i ;

Step 2.3, using formula (1) to obtain the degree of internal aggregation of the i-th example inst _i of the γ-th iteration To obtain the degree of inner aggregation of the num examples of the γ-th iteration and record the maximum degree of internal polymerization as

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In formula (1), is the threshold of the γth iteration; when $d_{ik}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ when $d_{ik}>d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$

Step 2.4, use formula (2) or formula (3) to obtain the difference degree of the i-th example inst _i of the γ-th iteration To obtain the degree of difference of num examples of the γth iteration ${\mathrm{\&delta;}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&delta;}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{\&delta;}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_{\mathrm{no}um}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)}=\underset{k=1}{\overset{\mathrm{no}um}{\&Sigma;}}max\left(d_{ik}\right),$ when ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(2\right)$

when ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}\&NotEqual;{\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(3\right)$

Step 2.5, the degree of difference for the num examples of the γth iteration Perform normalization processing to obtain the normalized difference degree ${\mathrm{\&delta;}}^{\&prime;\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&delta;}}_1^{\&prime;\left(\mathrm{\&gamma;}\right)},{\mathrm{\&delta;}}_2^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_i^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_{\mathrm{no}um}^{\&prime;\left(\mathrm{\&gamma;}\right)}\};$

Step 2.6, using formula (4) to obtain the clustering degree of the i-th example inst _i of the γ-th iteration To obtain the clustering degree of num examples of the γth iteration ${\mathrm{sco}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{sco}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{sco}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}...,{\mathrm{sco}}_{\mathrm{no}um}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{<span\; class="notranslate">\; sco</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Step 2.7, sort the clustering degree sco ^(γ) of the num examples of the γth iteration in descending order, and obtain the clustering degree sequence And let the inner aggregation degree corresponding to the clustering degree sequence sco' ^(γ) be ${\mathrm{\&rho;}}^{\&prime;\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&rho;}}_1^{\&prime;\left(\mathrm{\&gamma;}\right)},{\mathrm{\&rho;}}_2^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_{\mathrm{no}um}^{\&prime;\left(\mathrm{\&gamma;}\right)}\};$ means when ${\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{sco}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)}$ The degree of inner aggregation of the i-th example inst _i of the γ-th iteration when ; 1≤t≤num;

Step 2.8, initialization t=1;

Step 2.9, Judgment and Whether ≥num×3% is true, if true, the threshold of the γth iteration is a valid value, and after recording t, go to step 2.10; otherwise, judge Whether it is true, if true, assign t+1 to t, and repeat step 2.9; otherwise, modify the threshold Assign γ+1 to γ, and return to step 2.3;

Step 2.10, if the degree of internal aggregation of the i-th example inst _i of the γ-th iteration Is it satisfied If it is satisfied, then the i-th example inst _i is an outlier example, and the cluster clu _i to which the i-th example inst _i belongs = -1; otherwise, determine Whether it is true, if true, then the i-th example inst _i is a leader example, and let clu _i =i, otherwise, the i-th example inst _i is a voter example;

Step 2.11, counting the number of leader examples and the number of voter examples, and recording them as N and M respectively;

Step 2.12, record N leader example sets as 1≤α≤N; then the degree of internal aggregation corresponding to the N leader example sets D ^(l) is Denotes the αth leader instance The degree of internal polymerization; the label set corresponding to the N leader example set D ^(l) is ${\mathrm{lab}}^{\left(l\right)}=\{{\mathrm{lab}}_1^{\left(l\right)},{\mathrm{lab}}_2^{\left(l\right)},...,{\mathrm{lab}}_{\mathrm{\&alpha;}}^{\left(l\right)},...,{\mathrm{lab}}_N^{\left(l\right)}\};$ Denotes the αth leader instance The label set of; The clustering corresponding to the N leaders example set D ^(l) is Denotes the αth leader instance belongs to the cluster;

Step 2.13, record M voter example sets as 1≤β≤M; then the degree of inner aggregation corresponding to the M voter example set D ^(v) is Denotes the βth voter instance The degree of internal aggregation; the label set corresponding to the M voter example set D ^(v) is ${\mathrm{lab}}^{\left(v\right)}=\{{\mathrm{lab}}_1^{\left(v\right)},{\mathrm{lab}}_2^{\left(v\right)},...,{\mathrm{lab}}_{\mathrm{\&beta;}}^{\left(v\right)},...,{\mathrm{lab}}_m^{\left(v\right)}\};$ Denotes the βth voter instance label set; the cluster corresponding to the M voter example set D ^(v) is Denotes the βth voter instance belongs to the cluster;

Step 3: Obtain the cluster clu ^(v) to which the M voter example sets D ^(v) belong:

Step 3.1, define the number of iterations χ; and initialize χ=1; and define the zth transit example inst _z ; z≥0; and initialize α=1, β=1, z=0;

Step 3.2, select any α-th leader example from the N leader example set D ^(l) Obtain the αth leader instance as With the βth voter example of the χth iteration Euclidean distance of labels

Step 3.3, if , then assign β+1 to β, and judge whether β≤M holds true, if true, repeat step 3.3; otherwise, go to step 3.5; if When , judge the β-th voter example of the χ-th iteration belongs to the cluster Is it empty, if it is empty, go to step 3.4; otherwise, it represents the βth voter example of the χth iteration belongs to the cluster The value of is the subscript of the existing leader instance in the χ iteration, denoted as Execute step 3.11;

Step 3.4, the αth leader example The subscript α ^(l) of is assigned to And assign z+1 to z, let Indicates that the β-th voter instance of the χ-th iteration will be The subscript β _χ in , the label set degree of internal polymerization and belong to the cluster Both are assigned to the z-th transition example of the χ-th iteration The subscript, label set, degree of inner aggregation and belonging cluster; and assign β+1 to β; judge whether β≤M holds true, if true, go to step 3.3; otherwise go to step 3.5;

Step 3.5, if z≤0, then execute step 3.14; otherwise, assign χ+1 to χ, and set in turn assigned to Let β = 1; and obtain the β-th voter instance of the χ-th iteration With the χth iteration zth transition example Euclidean distance of labels And assign z-1 to z;

Step 3.6, if , then assign β+1 to β, and judge whether β≤M holds true, if true, repeat step 3.6; otherwise, go to step 3.5; if When , judge the β-th voter example of the χ-th iteration belongs to the cluster Is it empty, if it is empty, go to step 3.7; otherwise, it represents the βth voter example of the χth iteration belongs to the cluster The value of is the subscript of the existing leader instance in the χ iteration, denoted as Execute step 3.8;

Step 3.7, transfer the z-th transfer example of the x-th iteration The subscript z ^(χ) of is assigned to And assign z+1 to z, let And assign β+1 to β; and judge whether β≤M is true, if it is true, repeat step 3.6; otherwise, execute step 3.5;

Step 3.8, using formula (5) to obtain the βth voter example of the χth iteration The influence of the existing leader example with the χth iteration

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Step 3.9, use formula (6) to obtain the βth voter example of the χth iteration The z-th transition example with the χ-th iteration influence

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

Step 3.10, if Then assign β+1 to β, and execute step 3.6; otherwise, let And assign z+1 to z, let And assign β+1 to β, and judge whether β≤M holds true, if true, go to step 3.6; otherwise go to step 3.5;

Step 3.11, use formula (7) to obtain the βth voter example of the χth iteration The influence of the existing leader example with the χth iteration

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

Step 3.12, use formula (8) to obtain the β-th voter example of the χ-th iteration Example with the αth leader influence

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

Step 3.13, if Then assign β+1 to β, and execute step 3.3; otherwise, assign the αth leader instance The subscript α ^(l) of is assigned to And assign z+1 to z, let And assign β+1 to β, and judge whether β≤M is true, if it is true, go to step 3.3; otherwise go to step 3.5;

Step 3.14, assigning α+1 to α; and judging whether α≤N is true, if true, set β=1, and execute step 3.2; otherwise, execute step 3.15;

Step 3.15, the corresponding clusters of the M voter example sets D ^(v) during the χ iteration Sequentially assign values to the corresponding clusters of the M voter example sets D ^(v)

Step 3.16. Determine whether there are still voter examples whose clusters are empty, and if so, set the cluster value of the voter examples whose clusters are empty to -1;

Step 4: Roughly classify the predicted examples using support vector machines:

4.1. Establish a prediction example set P={instp ₁ , instp ₂ ,…,instp _j ,…,instp _nump } consisting of nmp prediction examples; instp _j represents the jth prediction example; 1≤j≤nump; and have instp _j = {attrp _j ; labp _j }; arrtp _j represents the attribute set of the jth prediction example instp _j ; labp _j represents the label set of the jth prediction example instp _j ; record the jth prediction example instp _j The cluster to which it belongs is clup _j ;

4.2. Use the num clusters {clu ₁ ,clu ₂ ,...,clu _i ,...,clu _num } corresponding to the initialization example set D as training labels, and use the num clusters in the initialization example set D that have been The attribute sets {attr ₁ , attr ₂ ..., _{attrp 2} ..., _{attrp j ,} ..., attrp _nump } as a prediction sample, and train with a support vector machine method to obtain nmp prediction labels, and assign the nmp prediction labels to the nump belonging clusters of the prediction example set P; thereby completing the prediction Coarse classification of the example set P;

Step 5, perform multi-label prediction on nmp prediction examples;

Step 5.1. Integrate the num examples in the initialization example set D and the nmp examples in the prediction example set P into an updated example set for the ψth time ${\mathrm{D.}}_{\mathrm{no}ew}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1,{\mathrm{inst}}_2,...,{\mathrm{inst}}_i,...,{\mathrm{inst}}_{\mathrm{no}um};{\mathrm{instp}}_1,{\mathrm{instp}}_2,...,{\mathrm{instp}}_j,...,{\mathrm{instp}}_{\mathrm{no}ump}\},$ recorded as ${\mathrm{D.}}_{\mathrm{no}ew}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1^{\left(\mathrm{\&psi;}\right)},{\mathrm{inst}}_2^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{\mathrm{\&Omega;}}^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{\mathrm{no}um+\mathrm{no}ump}^{\left(\mathrm{\&psi;}\right)}\};$ Indicates the Ω-th ψ-th update example; 1≤Ω≤num+nump;

Step 5.2, the ψth update example set The n attributes of each example in the num+nump update examples are respectively used as n-dimensional coordinates, so as to obtain the Ω-th ψ-th update example Update example with ξ-th ψ-th Euclidean distance of attributes 1≤ξ≤num+nump and ξ≠Ω;

Step 5.3, using formula (9) to obtain the Ω-th ψ-th update example The attribute aggregation degree of So as to obtain the attribute aggregation degree of the num+nump update examples of the ψth update

${\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; \&Omega;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&psi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; p</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; \&Omega;</span>}\mathrm{<span\; class="notranslate">\; \&xi;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&psi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

when $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ when $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}>d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{\mathrm{\&Omega;\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$

Step 5.4, initialize j=1;

Step 5.5, if the cluster of the j-th prediction example instp _j in the prediction example set P is clup _j and the cluster of the i-th known example inst _i in the initialization example set D is clu _i ; Then use formula (10) to obtain the influence gra _ij of the i-th known example inst _i and the j-th predicted example instp _j :

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span>$

In formula (10), Γ _i represents the known example inst _i updating the example set at the ψth time The attribute aggregation degree of the corresponding update example, Γ _j indicates that the predicted example instp _j updates the example set at the ψth time The attribute aggregation degree of the corresponding update example, d _ij represents the Euclidean distance between the i-th known example inst _i and the j-th predicted example instp _j attribute;

Step 5.6, repeat step 5.5, so as to obtain the influence of the jth prediction example instp _j and other known examples of the initialization example set D, and record the maximum influence gra _max ;

Step 5.7. If gra _ij = gra _max , then set labp _j = lab _i , indicating each label in the label set labp _j of the prediction example set P and each label in the label set lab _i of the initialization example set D The same, so as to obtain the prediction example of the jth multi-label prediction;

Step 5.8, assign j+1 to j, and judge whether j≤nump is true, if true, return to step 5.5 for execution, otherwise, it means that the multi-label prediction of nump prediction examples is completed;

The characteristics of the self-adaptive multi-label prediction method of the present invention are:

In the step 5, it also includes step 5.9, assigning the label set of the prediction example set P that has completed the multi-label prediction to the corresponding ψth update example set , so as to obtain the ψ+1th update example set Update the example set by the ψ+1th time as a new set of initialization examples for adaptive multi-label prediction.

When a new prediction example with the same object features and the same object semantics appears, it only needs to start from step 4 to complete the multi-label prediction for the new prediction example.

In the step 2.9, modify the threshold The rule is: if then will Subtract τ ₂ and assign to Otherwise, will Add τ ₂ and assign to _{0.1≤τ2≤0.5} , 75% _≤τ1 <100%.

Compared with the prior art, the beneficial effects of the present invention are reflected in:

1. The present invention adopts the method of rough classification first and then precise prediction. With the help of the self-adaptability contained in the present invention, through multiple rounds of iterations, the prediction label is continuously evolved, and then more accurate prediction than the existing multi-label prediction technology can be obtained. The result, is a method that can be put into practical use.

2. Through the initialization example set, the present invention can determine different initialization example sets according to different known object characteristics and semantics, so that the present invention can be widely applied to most application environments of existing network platforms, from simple text data to Audio, even images, can make better label predictions, which is more universal than existing technologies.

3. The present invention expresses the degree of cohesion of the example by calculating the degree of cohesion, obtains the degree of coupling by calculating the degree of difference, and calculates the degree of clustering based on the degree of cohesion and degree of difference. Each parameter has a practical meaning , fully considers the data classification requirements of high cohesion and low coupling, and is easy to understand and explain, so that while ensuring the high prediction accuracy of the present invention, the present invention has strong portability and can be used in various Under the condition of multi-label prediction.

4. The present invention can accurately find examples of leaders in various product fields through the degree of inner aggregation; for microblogs, forums and social networks, this method can accurately find key users with the greatest influence in different topic fields, and through their behavior A detailed study of , can predict possible trends in the field and provide accurate recommendations to users in the field.

5. By calculating the influence between examples and examples, the present invention can not only be used for multi-label prediction, but also can make analogies to examples with known labels of the same semantics, and find examples that are very similar to the multi-label examples of this example. Recommended To users, improve user experience.

6. The present invention adopts the method of selecting the label set of the known example that is most similar to the predicted example as the label set of the predicted example when determining the multi-label of the predicted example, and can recommend the user group of the known example to the newly emerging Forecast example; it can find a more accurate market positioning for new products and discover potential users for them.

7. The present invention enriches the existing training set and improves the accuracy of the next round of prediction due to the method of adding the prediction examples that complete the multi-label prediction to the initialization example set, so that the present invention has adaptive learning ability , in the face of newly added examples, the existing data set can be further improved, and with the increase of known label examples, the prediction accuracy of this method will be further improved.

Detailed ways

In this embodiment, an adaptive multi-label prediction method is performed in the following steps:

Step 1: Obtain the initialization example set D:

Step 1.1. Establish the original example set D′={inst′ ₁ , inst′ ₂ ,…,inst′ _a ,…,inst′ _num′ } from num′ known objects, where inst′ _a represents the ath known object The corresponding original example; 1≤a≤num′; and inst′ _a = {attr′ _a ; lab′ _a }; attr′ _a represents the attribute set of the a-th known object feature; lab′ _a represents the a-th A label set of known object semantics; and attr′ _a ={attr′ _a,1 ,attr′ _a,2 ,…,attr′ _a,n }; attr′ _a,n represents the ath known object The nth attribute; n is the number of attributes of the ath known object, lab′ _a = {lab′ _a,1 ,lab′ _a,2 ,…,lab′ _a,x ,…,lab′ _a,m } ;lab′ _a,x represents the xth label of the ath known object; m is the label number of the ath known object; 1≤x≤m; and: lab′ _a,x = 1 represents the ath The semantics of a known object conforms to the x-th label; lab′ _{a, x} = 0 means that the semantics of the a-th known object does not conform to the x-th label; assuming that the known object is a picture, the color difference, size, etc. need to be described in detail Object features are used as an attribute set, and accurate and detailed numbers are used as the values of each attribute; object semantics such as landscape pictures and animal pictures are used as a label set, and 0 indicates that it does not meet the label, and 1 indicates that it meets the label. ;

Step 1.2, normalize the attribute sets {attr' ₁ , attr' ₂ , ..., attr' _a , ..., attr'_num' } of num' known object features in the original example set D'; In the normalization process, take the attribute set attr′ _a of the a-th known object feature as an example, that is, first record the attribute set {attr′ _a,1 ,attr′ _a,2 ,…,attr′ _a,n } attribute attr′ _a,max with the largest median value, and then use the largest attribute attr′ _a,max as the denominator, and perform division calculation with each attribute in the attribute set to obtain the a-th known object after normalization The attribute set attr″ _a of the feature; and so on to obtain the attribute set {attr″ ₁ , attr″ ₂ ,…,attr″ _a ,…,attr″ _num′ } of num′ known object features after normalization ; When the m label values corresponding to the attribute set arrt″ _a of the a-th known object feature after normalization are all 0, delete the original example to which the a-th known object belongs after normalization; thereby obtaining num The initialization example set D={inst ₁ , inst ₂ ,...,inst _i ,...,inst _num } composed of examples; inst _i represents the example corresponding to the i-th known object after initialization; and inst _i ={ attr _i ; lab _i }; attr _i represents the attribute set of the i-th example feature after initialization; lab _i represents the label set of the i-th example semantics after initialization; 1≤i≤num; as shown in Table 1:

Table 1: The data table of the i-th example inst _i of the initialization example set D

attri _i,1 … attri _i,n lab _i,1 … lab _i,m ρ _{i δi} sco _i clu _i inst _i

Step 2: Solve the clustering degree of each example in the initialization example set D, so as to determine the leader example, outsider example and voter example in the initialization example set D:

Step 2.1, the m labels of each example in the num examples in the initialization example set D are respectively used as m-dimensional coordinates, thereby obtaining the Euclidean distance d _ik between the i-th example inst _i and the k-th example inst _k label; 1 ≤k≤num and k≠i; for example, to solve the Euclidean distance d ₁₂ between the first example and the second example label, both the first example and the second example have m labels with the same name, but due to the value are not necessarily the same, they are represented as the label set lab ₁ = {lab _1,1 ,lab _1,2 ,...,lab _1,m } of the first example and the label set lab ₂ ={lab ₂ of the second example _,1 ,lab _2,2 ,…,lab _2,m }, then the Euclidean distance d ₁₂ of the label is $d_{12}=\sqrt{{({\mathrm{lab}}_{1,1}-{\mathrm{lab}}_{2,1})}^2+\mathrm{...}+{({\mathrm{lab}}_{1,m}-{\mathrm{lab}}_{2,m})}^2};$

Step 2.2, define the number of iterations γ; and initialize γ=1; define the cluster of the i-th example inst _i as clu _i ;

Step 2.3, using formula (1) to obtain the degree of internal aggregation of the i-th example inst _i of the γ-th iteration To obtain the degree of inner aggregation of the num examples of the γ-th iteration and record the maximum degree of internal polymerization as

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In formula (1), is the threshold of the γth iteration; when $d_{\mathrm{ik}}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ when $d_{ik}>d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$

Step 2.4, use formula (2) or formula (3) to obtain the difference degree of the i-th example inst _i of the γ-th iteration To obtain the degree of difference of num examples of the γth iteration ${\mathrm{\&delta;}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&delta;}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{\&delta;}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_{\mathrm{no}um}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)}=\underset{k=1}{\overset{\mathrm{no}um}{\&Sigma;}}max\left(d_{ik}\right),$ when ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(2\right)$

when ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}\&NotEqual;{\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(3\right)$

Step 2.5, the degree of difference for the num examples of the γ-th iteration Perform normalization processing to obtain the normalized difference degree With the help of steps 2.4 and 2.5, the normalized difference There is a large distinction, so that a few are close to 1, and most of the values are less than 0.5, which will help the selection of leader examples;

Step 2.6, use formula (4) to obtain the clustering degree of the i-th example inst _i of the γ-th iteration To obtain the clustering degree of num examples of the γth iteration ${\mathrm{sco}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{sco}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{sco}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}...,{\mathrm{sco}}_{\mathrm{no}um}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{<span\; class="notranslate">\; sco</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Step 2.7. Arrange the clustering degrees sco ^(γ) of the num examples of the γth iteration in descending order to obtain the clustering degree sequence And let the inner aggregation degree corresponding to the clustering degree sequence sco′ ^(γ) be ${\mathrm{\&rho;}}^{\&prime;\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&rho;}}_1^{\&prime;\left(\mathrm{\&gamma;}\right)},{\mathrm{\&rho;}}_2^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_{\mathrm{no}um}^{\&prime;\left(\mathrm{\&gamma;}\right)}\};$ means when ${\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{sco}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)}$ Inner aggregation degree of the i-th example inst _i of the γ-th iteration; 1≤t≤num;

Step 2.8, initialization t=1;

Step 2.9, Judgment and Whether ≥num×3% is true, if true, the threshold of the γth iteration is a valid value, and after recording t, go to step 2.10; otherwise, judge Whether it is true, if true, assign t+1 to t, and repeat step 2.9; otherwise, modify the threshold modify threshold The rule is: if then will Subtract τ ₂ and assign to Otherwise, will Add τ ₂ and assign to 0.1≤τ ₂ ≤0.5, 75%≤τ ₁ <100%; assign γ+1 to γ, and return to step 2.3; judge and In the condition of ≥num×3%, 1.25 and 3% are not fixed. The present invention is based on the fact that the number of examples is 10,000 and the number of labels is less than 20. There will be a better solution. When the number of examples and the number of labels change , can be modified as appropriate, and the principle is to ensure that only a small number of examples whose clustering degree is much greater than other examples are selected as leader examples in the following steps;

Step 2.10, if the degree of internal aggregation of the i-th example inst _i of the γ-th iteration Is it satisfied If it is satisfied, then the i-th example inst _i is an outlier example, and the cluster clu _i to which the i-th example inst _i belongs = -1; otherwise, judge Whether it is true, if true, then the i-th example inst _i is a leader example, and let clu _i =i, otherwise, the i-th example inst _i is a voter example;

Step 2.11, count the number of leader examples and the number of voter examples, and record them as N and M respectively;

Step 2.12, record N leader example sets as 1≤α≤N; then the degree of internal aggregation corresponding to the N leader example set D ^(l) is Denotes the αth leader instance The degree of internal aggregation; the label set corresponding to the N leader example set D ^(l) is ${\mathrm{lab}}^{\left(l\right)}=\{{\mathrm{lab}}_1^{\left(l\right)},{\mathrm{lab}}_2^{\left(l\right)},...,{\mathrm{lab}}_{\mathrm{\&alpha;}}^{\left(l\right)},...,{\mathrm{lab}}_N^{\left(l\right)}\};$ Denotes the αth leader instance label set; the cluster corresponding to the N leader example set D ^(l) is Denotes the αth leader instance belongs to the cluster;

Step 2.13, record M voter example sets as 1≤β≤M; then the degree of inner aggregation corresponding to the set of M voter examples D ^(v) is Denotes the βth voter instance The degree of inner aggregation; the label set corresponding to the set of M voter examples D ^(v) is ${\mathrm{lab}}^{\left(v\right)}=\{{\mathrm{lab}}_1^{\left(v\right)},{\mathrm{lab}}_2^{\left(v\right)},...,{\mathrm{lab}}_{\mathrm{\&beta;}}^{\left(v\right)},...,{\mathrm{lab}}_m^{\left(v\right)}\};$ Denotes the βth voter instance label set; the cluster corresponding to the M voter example set D ^(v) is Denotes the βth voter instance belongs to the cluster;

Step 3: Obtain the cluster clu ^(v) of M voter example set D ⁽ v):

Step 3.1, define the number of iterations χ; and initialize χ=1; and define the zth transfer example inst _z ; z≥0; and initialize α=1, β=1, z=0; the zth transfer example inst _z stores The structure is similar to the commonly used stack structure. In order to clarify the expression, the present invention introduces the number of iterations ^χ at the same time to distinguish the transfer examples when z is the same; The values are all empty;

Step 3.2, select any α-th leader example from the N leader example set D ^(l) Obtain the αth leader example as With the βth voter example of the χth iteration The Euclidean distance of the label

Step 3.3, if , then assign β+1 to β, and judge whether β≤M holds true, if true, repeat step 3.3; otherwise, go to step 3.5; if When , judge the β-th voter example of the χ-th iteration belongs to the cluster Is it empty, if it is empty, go to step 3.4; otherwise, it represents the βth voter example of the χth iteration belongs to the cluster The value of is the subscript of the existing leader instance in the χ iteration, denoted as Execute step 3.11; for example, the existing leader example of the χ iteration is inst ₉ , then

Step 3.4, the αth leader example The subscript α ^(l) of is assigned to And assign z+1 to z, let Indicates that the β-th voter instance of the χ-th iteration will be The subscript β _χ in , the label set degree of internal polymerization and belong to the cluster Both are assigned to the z-th transition example of the χ-th iteration The subscript, label set, degree of inner aggregation and belonging cluster; and assign β+1 to β; judge whether β≤M holds true, if true, go to step 3.3; otherwise go to step 3.5; Indicates that one example is equal to another example, it only means that the corresponding values of the two examples are the same, that is, assign the subscript, label set, degree of inner aggregation, and cluster of the example on the right side of the equal sign to the subscript of the example on the left side of the equal sign, label set, degree of inner aggregation and belonging cluster;

Step 3.5, if z≤0, then execute step 3.14; otherwise, assign χ+1 to χ, and set in turn assigned to For other parameters related to χ, it is also necessary to assign the value associated with χ-1 to the corresponding value associated with χ, so as to maintain the coherence and consistency of the data, for example Let β = 1; and obtain the β-th voter instance of the χ-th iteration With the χth iteration zth transition example The Euclidean distance of the label And assign z-1 to z;

Step 3.6, if , then assign β+1 to β, and judge whether β≤M holds true, if true, repeat step 3.6; otherwise, go to step 3.5; if When , judge the β-th voter example of the χ-th iteration belongs to the cluster Is it empty, if it is empty, go to step 3.7; otherwise, it represents the βth voter example of the χth iteration belongs to the cluster The value of is the subscript of the existing leader instance in the χ iteration, denoted as Execute step 3.8;

Step 3.7, transfer the z-th transfer example of the x-th iteration The subscript z ^(χ) of is assigned to And assign z+1 to z, let And assign β+1 to β; and judge whether β≤M is true, if it is true, repeat step 3.6; otherwise, execute step 3.5;

Step 3.8, using formula (5) to obtain the βth voter example of the χth iteration The influence of the existing leader example with the χth iteration

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Equation (5) can be extended to the calculation of the influence of any two examples with the same semantics, only need to know the degree of inner aggregation of the two examples and the Euclidean distance between the labels of the two examples, or the aggregation degree of attributes of the two examples and The Euclidean distance between the two attributes, the influence between the two examples can be obtained by applying the formula (5);

Step 3.9, use formula (6) to obtain the β-th voter example of the χ-th iteration The z-th transition example with the χ-th iteration influence

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

Step 3.10, if Then assign β+1 to β, and execute step 3.6; otherwise, let And assign z+1 to z, let And assign β+1 to β, and judge whether β≤M is true, if it is true, go to step 3.6; otherwise go to step 3.5;

Step 3.11, use formula (7) to obtain the βth voter example of the χth iteration The influence of the existing leader example with the χth iteration

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

Step 3.12, use formula (8) to obtain the β-th voter example of the χ-th iteration Example with the αth leader influence

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; \&chi;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

Step 3.13, if Then assign β+1 to β, and execute step 3.3; otherwise, assign the αth leader instance The subscript α ^(l) of is assigned to And assign z+1 to z, let And judge whether β≤M is true, if it is true, assign β+1 to β, and execute step 3.3; otherwise, execute step 3.5;

Step 3.14, assigning α+1 to α; and judging whether α≤N is true, if true, set β=1, and execute step 3.2; otherwise, execute step 3.15;

Step 3.15, the clusters corresponding to the M voter example sets D ^(v) in the χth iteration Sequentially assign values to the corresponding clusters of M voter example sets D ^(v)

Step 3.16. Determine whether there are still voter examples whose clusters are empty, and if so, set the cluster value of the voter examples whose clusters are empty to -1; therefore, the possible value of the cluster of voter examples The number of is N+1, corresponding to the value of the cluster to which the N leader examples belong and the case where the cluster to which they belong is -1;

Step 4: Roughly classify the predicted examples using support vector machine:

4.1. Establish a prediction example set P={instp ₁ , instp ₂ ,…,instp _j ,…,instp _nump } consisting of nmp prediction examples; instp _j represents the jth prediction example; 1≤j≤nump; and have instp _j = {attrp _j ; labp _j }; arrtp _j represents the attribute set of the jth prediction example instp _j ; labp _j represents the label set of the jth prediction example instp _j ; record the belonging cluster of the jth prediction example instp _j The class is clup _j ; in the present invention, the predicted example and the known example must be the same object, that is, the characteristics and semantics of the object are the same, for example, if the known example is a picture, the predicted example must also be a picture, and the color difference, size, etc. The detailed description of the object features is used as an attribute set, and the yes-or-no object semantics such as landscape pictures and animal pictures are used as a label set. The two example sets have the same attribute set and label set, but the values are different for clarity. , the present invention uses different symbols to distinguish when discussing;

4.2. Use the num clusters {clu ₁ ,clu ₂ ,…,clu _i ,…,clu _num } corresponding to the initialization example set D as training labels to initialize the attributes of num known objects in the example set D Set {attr ₁ , attr ₂ ..., attr _i , ..., attr _num } as a training sample; take the nump attribute set {attrp ₁ , attrp ₂ ..., attrp _j , ..., attrp _nump } of the prediction example set P as a prediction sample , and use the support vector machine method to train, obtain nmp prediction labels, and assign the nump prediction labels to the nump clusters of the prediction example set P; thus complete the rough classification of the prediction example set P; the support vector machine method usually There are three inputs, namely training label, training sample and prediction sample, so as to obtain an output, that is, the prediction label;

Step 5, perform multi-label prediction on nmp prediction examples;

Step 5.1. Integrate the num examples in the initialization example set D and the nmp examples in the prediction example set P into an updated example set for the ψth time ${\mathrm{D.}}_{\mathrm{no}ew}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1,{\mathrm{inst}}_2,...,{\mathrm{inst}}_i,...,{\mathrm{inst}}_{\mathrm{no}um};{\mathrm{instp}}_1,{\mathrm{instp}}_2,...,{\mathrm{instp}}_j,...,{\mathrm{instp}}_{\mathrm{no}ump}\},$ recorded as ${\mathrm{D.}}_{\mathrm{no}ew}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1^{\left(\mathrm{\&psi;}\right)},{\mathrm{inst}}_2^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{\mathrm{\&Omega;}}^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{\mathrm{no}um+\mathrm{no}ump}^{\left(\mathrm{\&psi;}\right)}\};$ Indicates the Ω-th ψ update example; 1≤Ω≤num+nump; ψ is the number of updates, the update mainly includes integrating the existing initialization examples and prediction examples into an example set, and the prediction example set that will complete the multi-label prediction The label set of P is assigned to the corresponding ψ-th update example set, ψ is initialized to 1, and ψ+1 is assigned to ψ after each update is completed;

Step 5.2, the ψth update example set The n attributes of each example in the num+nump update examples are respectively used as n-dimensional coordinates, so as to obtain the Ω-th ψ-th update example Update example with ξ-th ψ-th Euclidean distance of attributes 1≤ξ≤num+nump and ξ≠Ω;

Step 5.3, using formula (9) to obtain the Ω-th ψ-th update example The degree of attribute aggregation So as to obtain the attribute aggregation degree of the num+nump update examples of the ψth update

${\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; \&Omega;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&psi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; p</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; \&Omega;</span>}\mathrm{<span\; class="notranslate">\; \&xi;</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&psi;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

when $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ when $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}>d_c^{\left(\mathrm{\&gamma;}\right)}$ hour, $f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$ The formula for solving the attribute aggregation degree is similar to the inner aggregation degree formula, but the Euclidean distance of the label is changed to the Euclidean distance of the attribute;

Step 5.4, initialize j=1;

Step 5.5, if the cluster of the j-th prediction example instp _j in the prediction example set P is clup _j and the cluster of the i-th known example inst _i in the initialization example set D is clu _i ; Then use formula (10) to obtain the influence gra _ij of the i-th known example inst _i and the j-th predicted example instp _j :

${\mathrm{<span\; class="notranslate">\; gra</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; \&Gamma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span>$

In formula (10), Γ _i represents the known example inst _i updating the example set at the ψth time The attribute aggregation degree of the corresponding update example, Γ _j indicates that the predicted example instp _j updates the example set at the ψth time The attribute aggregation degree of the corresponding update example, d _ij represents the Euclidean distance between the i-th known example inst _i and the j-th predicted example instp _j attribute;

Step 5.6, repeat step 5.5, so as to obtain the influence of the jth prediction example instp _j and other known examples of the initialization example set D, and record the maximum influence gra _max ;

Step 5.7. If gra _ij = gra _max , then set labp _j = lab _i , indicating each label in the label set labp _j of the prediction example set P and each label in the label set lab _i of the initialization example set D The same, so as to obtain the prediction example of the jth multi-label prediction;

Step 5.8, assign j+1 to j, and judge whether j≤nump is true, if true, return to step 5.5 for execution, otherwise, it means that the multi-label prediction of nump prediction examples is completed;

Step 5.9: Assign the label set of the predicted example set P that has completed the multi-label prediction to the corresponding updated example set for the ψth time , so as to obtain the ψ+1th update example set Update the example set by the ψ+1th time Adaptive multi-label prediction is performed as a new initialization example set to enrich the existing training set and improve the accuracy of the next round of prediction. When there are new prediction examples with the same object features and the same object semantics, just Starting from step 4, the multi-label prediction of the new prediction example can be completed.

Claims (4)

1. self-adaptation many Tag Estimations method, is characterized in that carrying out as follows:

Step 1: obtain initialization example set D:

Step 1.1, to be set up by the individual known object of num ' original illustration collection D '=inst ' ₁, inst ' ₂..., inst ' _a..., inst ' _{num '}, inst ' _arepresent the original illustration corresponding to a known object; 1≤a≤num '; And have inst ' _a=attr ' _a; Lab ' _a; Attr ' _arepresent the property set of described a known object feature; Lab ' _arepresent the tally set of described a known object semanteme; And have attr ' _a=attr ' _{a, 1}, attr ' _{a, 2}..., attr ' _a,n; Attr ' _a,nrepresent the n-th attribute of a known object; N is the attribute number of a known object; Lab ' _a=lab ' _{a, 1}, lab ' _{a, 2}..., lab ' _a,x..., lab ' _a,m; Lab ' _a,xrepresent an xth label of a known object; M is the number of tags of a known object; 1≤x≤m; And have: lab ' _a,x=1 represents that a known object semanteme meets an xth label; Lab ' _a,x=0 represents that a known object semanteme does not meet an xth label;

Step 1.2, to the property set of the num ' individual known object feature in described original illustration collection D ' attr ' ₁, attr ' ₂..., attr ' _a..., attr ' _{num '}be normalized respectively, obtain the individual known object feature of num ' after normalized property set attr " ₁, attr " ₂..., attr " _a..., attr " _{num '}; As the property set arrt of a known object feature after described normalization _a" when m corresponding label value is 0, delete the original illustration belonging to a known object after described normalization; Thus obtain the initialization example set D={inst of num example formation ₁, inst ₂..., inst _i..., inst _num; Inst _irepresent the example corresponding to i-th known object after initialization; And have inst _i={ attr _i; lab _i; Attr _irepresent the property set of i-th exemplary characteristics after initialization; lab _irepresent the tally set of described i-th exemplary semantic after initialization; 1≤i≤num;

Step 2: the clustering degree solving each example in described initialization example set D, thus determine the leader's example in initialization example set D, example not in the know and voter's example:

Step 2.1, using m label of each example in num example in described initialization example set D as m dimension coordinate, thus obtain i-th example inst _iwith a kth example inst _kthe Euclidean distance d of label _ik; 1≤k≤num and k ≠ i;

Step 2.2, definition iterations γ; And initialization γ=1; Define described i-th example inst _iaffiliated cluster be clu _i;

Step 2.3, formula (1) is utilized to obtain i-th example inst of the γ time iteration _ithe interior degree of polymerization thus obtain the interior degree of polymerization of num example of the γ time iteration and the degree of polymerization in maximum is designated as

${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}=\underset{k=1}{\overset{num}{\&Sigma;}}f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})---\left(1\right)$

In formula (1), it is the threshold value of the γ time iteration; When $d_{ik}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ Time, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ When $d_{ik}>d_c^{\left(\mathrm{\&gamma;}\right)}$ Time, $f(d_{ik}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$

Step 2.4, formula (2) or formula (3) is utilized to obtain i-th example inst of the γ time iteration _idiversity factor thus obtain the diversity factor of num example of the γ time iteration ${\mathrm{\&delta;}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&delta;}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{\&delta;}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_{num}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{\&delta;}}_i^{\left(\mathrm{\&gamma;}\right)}=\underset{k=1}{\overset{num}{\&Sigma;}}max\left(d_{ik}\right),$ When ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(2\right)$

when ${\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}\&NotEqual;{\mathrm{\&rho;}}_{\max }^{\left(\mathrm{\&gamma;}\right)}---\left(3\right)$

Step 2.5, diversity factor δ to num example of described the γ time iteration ^(γ)be normalized, obtain the diversity factor after normalization ${\mathrm{\&delta;}}^{\&prime;\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&delta;}}_1^{\&prime;\left(\mathrm{\&gamma;}\right)},{\mathrm{\&delta;}}_2^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_i^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&delta;}}_{num}^{\&prime;\left(\mathrm{\&gamma;}\right)}\};$

Step 2.6, formula (4) is utilized to obtain i-th example inst of the γ time iteration _iclustering degree thus obtain the clustering degree of num example of the γ time iteration ${\mathrm{sco}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{sco}}_1^{\left(\mathrm{\&gamma;}\right)},{\mathrm{sco}}_2^{\left(\mathrm{\&gamma;}\right)},...,{\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}...,{\mathrm{sco}}_{num}^{\left(\mathrm{\&gamma;}\right)}\}:$

${\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{\&rho;}}_i^{\left(\mathrm{\&gamma;}\right)}\&times;{\mathrm{\&delta;}}_i^{\&prime;\left(\mathrm{\&gamma;}\right)}---\left(4\right)$

Step 2.7, clustering degree sco to num example of described the γ time iteration ^(γ)carry out descending sort, obtain clustering degree series and order and described clustering degree series sco ' ^(γ)the corresponding interior degree of polymerization is ${\mathrm{\&rho;}}^{\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&rho;}}_1^{\&prime;\left(\mathrm{\&gamma;}\right)},{\mathrm{\&rho;}}_2^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_{num}^{\&prime;\left(\mathrm{\&gamma;}\right)}\};$ represent and work as ${\mathrm{sco}}_i^{\left(\mathrm{\&gamma;}\right)}={\mathrm{sco}}_t^{\&prime;\left(\mathrm{\&gamma;}\right)}$ Time i-th example inst of the γ time iteration _ithe interior degree of polymerization; 1≤t≤num;

Step 2.8, initialization t=1;

Step 2.9, judgement and whether set up, if set up, then the threshold value of the γ time iteration for effective value, and after recording t, perform step 2.10; Otherwise, judge whether set up, if set up, then by t+1 assignment to t, and repeated execution of steps 2.9; Otherwise, amendment threshold value by γ+1 assignment to γ, and return execution step 2.3;

If i-th of step 2.10 the γ time iteration example inst _ithe interior degree of polymerization whether meet if meet, then described i-th example inst _ifor example not in the know, and make described i-th example inst _iaffiliated cluster clu _i=-1; Otherwise, judge whether set up, if set up, then i-th example inst _ifor leader's example, and make clu _i=i, otherwise, i-th example inst _ifor voter's example;

Step 2.11, add up the number of described leader's example and the number of described voter's example, and be designated as N and M respectively;

Step 2.12, remember that N number of leader's example set is $D^{\left(l\right)}=\{{\mathrm{inst}}_1^{\left(l\right)},{\mathrm{inst}}_2^{\left(l\right)},...,{\mathrm{inst}}_{\mathrm{\&alpha;}}^{\left(l\right)},...,{\mathrm{inst}}_N^{\left(l\right)}\},$ 1≤α≤N; Then with described N number of leader example set D ^(l)the corresponding interior degree of polymerization is represent α leader's example the interior degree of polymerization; With described N number of leader example set D ^(l)corresponding tally set is ${\mathrm{lab}}^{\left(l\right)}=\{{\mathrm{lab}}_1^{\left(l\right)},{\mathrm{lab}}_2^{\left(l\right)},...,{\mathrm{lab}}_{\mathrm{\&alpha;}}^{\left(l\right)},...,{\mathrm{lab}}_N^{\left(l\right)}\};$ represent α leader's example tally set; With described N number of leader example set D ^(l)corresponding affiliated cluster is ${\mathrm{clu}}^{\left(l\right)}=\{{\mathrm{clu}}_1^{\left(l\right)},{\mathrm{clu}}_2^{\left(l\right)},...,{\mathrm{clu}}_{\mathrm{\&alpha;}}^{\left(l\right)},...,{\mathrm{clu}}_N^{\left(l\right)}\};$ represent α leader's example affiliated cluster;

Step 2.13, note M voter's example set are $D^{\left(v\right)}=\{{\mathrm{inst}}_1^{\left(v\right)},{\mathrm{inst}}_2^{\left(v\right)},...,{\mathrm{inst}}_{\mathrm{\&beta;}}^{\left(v\right)},...,{\mathrm{inst}}_M^{\left(v\right)}\},$ 1≤β≤M; Then with described M voter example set D ^(v)the corresponding interior degree of polymerization is ${\mathrm{\&rho;}}^{\left(v\right)\left(\mathrm{\&gamma;}\right)}=\{{\mathrm{\&rho;}}_1^{\left(v\right)\left(\mathrm{\&gamma;}\right)},{\mathrm{\&rho;}}_2^{\left(v\right)\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_{\mathrm{\&beta;}}^{\left(v\right)\left(\mathrm{\&gamma;}\right)},...,{\mathrm{\&rho;}}_M^{\left(v\right)\left(\mathrm{\&gamma;}\right)}\};$ represent β voter's example the interior degree of polymerization; With described M voter example set D ^(v)corresponding tally set is ${\mathrm{lab}}^{\left(v\right)}=\{{\mathrm{lab}}_1^{\left(v\right)},{\mathrm{lab}}_2^{\left(v\right)},...,{\mathrm{lab}}_{\mathrm{\&beta;}}^{\left(v\right)},...,{\mathrm{lab}}_M^{\left(v\right)}\};$ represent β voter's example tally set; With described M voter example set D ^(v)corresponding affiliated cluster is ${\mathrm{clu}}^{\left(v\right)}=\{{\mathrm{clu}}_1^{\left(v\right)},{\mathrm{clu}}_2^{\left(v\right)},...,{\mathrm{clu}}_{\mathrm{\&beta;}}^{\left(v\right)},\mathrm{...},{\mathrm{clu}}_M^{\left(u\right)}\};$ represent β voter's example affiliated cluster;

Step 3: obtain described M voter example set D ^(v)affiliated cluster clu ^(v):

Step 3.1, definition iterations χ; And initialization χ=1; And define z transfer example inst _z; Z>=0; And initialization α=1, β=1, z=0;

Step 3.2, from described N number of leader example set D ^(l)in choose wantonly α leader's example obtaining described α leader's example is with β voter's example of the χ time iteration the Euclidean distance of label

If step 3.3 time, then by β+1 assignment to β, and judge whether β≤M sets up, if set up, repeated execution of steps 3.3; Otherwise perform step 3.5; If time, judge β voter's example of the χ time iteration affiliated cluster whether be empty, if it is empty, then perform step 3.4; Otherwise, represent β voter's example of the χ time iteration affiliated cluster value be the subscript of the χ time existing leader's example of iteration, be designated as perform step 3.11;

Step 3.4, by α leader's example subscript α ^(l)assignment is given and by z+1 assignment to z, order represent β voter's example of the χ time iteration in subscript β _χ, tally set the interior degree of polymerization with affiliated cluster equal assignment gives z transfer example of the χ time iteration subscript, tally set, the interior degree of polymerization and affiliated cluster; And by β+1 assignment to β; Judge whether β≤M sets up, if set up, then perform step 3.3; Otherwise perform step 3.5;

If step 3.5 z≤0, then perform step 3.14; Otherwise, by χ+1 assignment to χ, and will assignment is given successively make β=1; And obtain β voter's example of described the χ time iteration with the χ time iteration z transfer example the Euclidean distance of label and by z-1 assignment to z;

If step 3.6 time, then by β+1 assignment to β, and judge whether β≤M sets up, if set up, repeated execution of steps 3.6; Otherwise perform step 3.5; If time, judge β voter's example of the χ time iteration affiliated cluster whether be empty, if it is empty, then perform step 3.7; Otherwise, represent β voter's example of the χ time iteration affiliated cluster value be the subscript of the χ time existing leader's example of iteration, be designated as perform step 3.8;

Step 3.7, by z transfer example of the χ time iteration subscript z ^(χ)assignment is given and by z+1 assignment to z, order and by β+1 assignment to β; And judge whether β≤M sets up, if set up, then repeated execution of steps 3.6; Otherwise perform step 3.5;

Step 3.8, formula (5) is utilized to obtain β voter's example of the χ time iteration with the influence power of the existing leader's example of described the χ time iteration

${\mathrm{gra}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&epsiv;}}^{\left(v\right)\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}=\frac{{\mathrm{\&rho;}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}}^{\left(v\right)}\&times;{\mathrm{\&rho;}}_{\mathrm{\&epsiv;}}^{\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}}{d_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&epsiv;}}^{\left(v\right)\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}}---\left(5\right)$

Step 3.9, formula (6) is utilized to obtain β voter's example of the χ time iteration with z transfer example of the χ time iteration influence power

${\mathrm{gra}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}z}^{\left(v\right)\left(\mathrm{\&chi;}\right)}=\frac{{\mathrm{\&rho;}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}}^{\left(v\right)}\&times;{\mathrm{\&rho;}}_z^{\left(\mathrm{\&chi;}\right)}}{d_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}z}^{\left(v\right)\left(\mathrm{\&chi;}\right)}}---\left(6\right)$

If step 3.10 then by β+1 assignment to β, and perform step 3.6; Otherwise, order and by z+1 assignment to z, order and by β+1 assignment to β, and judge whether β≤M sets up, if set up, then perform step 3.6; Otherwise perform step 3.5;

Step 3.11, formula (7) is utilized to obtain β voter's example of the χ time iteration with the influence power of the existing leader's example of described the χ time iteration

${\mathrm{gra}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&epsiv;}}^{\left(v\right)\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}=\frac{{\mathrm{\&rho;}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}}^{\left(v\right)}\&times;{\mathrm{\&rho;}}_{\mathrm{\&epsiv;}}^{\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}}{d_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&epsiv;}}^{\left(v\right)\left({\mathrm{\&beta;}}_{\mathrm{\&chi;}}\right)}}---\left(7\right)$

Step 3.12, formula (8) is utilized to obtain β voter's example of the χ time iteration with α leader's example influence power

${\mathrm{gra}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&alpha;}}^{\left(v\right)\left(l\right)}=\frac{{\mathrm{\&rho;}}_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}}^{\left(v\right)}\&times;{\mathrm{\&rho;}}_{\mathrm{\&alpha;}}^{\left(l\right)}}{d_{{\mathrm{\&beta;}}_{\mathrm{\&chi;}}\mathrm{\&alpha;}}^{\left(v\right)\left(l\right)}}---\left(8\right)$

If step 3.13 then by β+1 assignment to β, and perform step 3.3; Otherwise, by α leader's example subscript α ^(l)assignment is given and by z+1 assignment to z, order and by β+1 assignment to β, and judge whether β≤M sets up, if set up, then perform step 3.3; Otherwise perform step 3.5;

Step 3.14, by α+1 assignment to α; And judge whether α≤N sets up, if set up, make β=1, and perform step 3.2; Otherwise perform step 3.15;

Step 3.15, by M voter example set D described during the χ time iteration ^(v)corresponding affiliated cluster assignment gives described M voter example set D successively ^(v)corresponding affiliated cluster $\{{\mathrm{clu}}_1^{\left(v\right)},{\mathrm{clu}}_2^{\left(v\right)},...,{\mathrm{clu}}_{\mathrm{\&beta;}}^{\left(v\right)},\mathrm{...},{\mathrm{clu}}_M^{\left(v\right)}\};$

Step 3.16, to judge whether also to exist affiliated cluster be empty voter's example, if exist, then to arrange affiliated cluster be the value of the affiliated cluster of empty voter's example is-1;

Step 4; Support vector machine is adopted to carry out rough sort to prediction example:

4.1, the prediction example set P={instp be made up of nump prediction example is set up ₁, instp ₂..., instp _j..., instp _nump; Instp _jrepresent a jth prediction example; 1≤j≤nump; And have instp _j={ attrp _j; Labp _j; Arrtp _jrepresent a jth prediction example instp _jproperty set; Labp _jrepresent a jth prediction example instp _jtally set; Remember a described jth prediction example instp _jaffiliated cluster be clup _j;

4.2, with num the affiliated cluster { clu that described initialization example set D is corresponding ₁, clu ₂..., clu _i..., clu _numas training label, with the property set { attr of num known object in described initialization example set D ₁, attr ₂, attr _i..., attr _numas training sample; With nump the property set { attrp of described prediction example set P ₁, attrp ₂, attrp _j..., attrp _numpas forecast sample, and train with support vector machine method, obtain nump and predict label, give nump of described prediction example set P affiliated cluster by described nump prediction label difference assignment; Thus the rough sort completed described prediction example set P;

Step 5, to nump prediction example carry out many Tag Estimations;

Step 5.1, be upgrade example set the ψ time by nump exemplary integrated in num example in described initialization example set D and described prediction example set P $D_{new}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1,{\mathrm{inst}}_2,...,{\mathrm{inst}}_i,...,{\mathrm{inst}}_{num};{\mathrm{instp}}_1,{\mathrm{instp}}_2,...,{\mathrm{instp}}_j,...,{\mathrm{instp}}_{nump}\},$ Be designated as $D_{new}^{\left(\mathrm{\&psi;}\right)}=\{{\mathrm{inst}}_1^{\left(\mathrm{\&psi;}\right)},{\mathrm{inst}}_2^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{\mathrm{\&Omega;}}^{\left(\mathrm{\&psi;}\right)},...,{\mathrm{inst}}_{num+nump}^{\left(\mathrm{\&psi;}\right)}\};$ represent that Ω upgrades example the ψ time; 1≤Ω≤num+nump;

Step 5.2, described the ψ time renewal example set middle num+nump upgrade in example n attribute of each example respectively as n dimension coordinate, thus obtain Ω the ψ time renewal example example is upgraded the ψ time with ξ the Euclidean distance of attribute 1≤ξ≤num+nump and ξ ≠ Ω;

Step 5.3, formula (9) is utilized to obtain Ω the ψ time renewal example the attribute degree of polymerization thus obtain the attribute degree of polymerization of num+nump the renewal example upgraded for the ψ time

${\mathrm{\&Gamma;}}_{\mathrm{\&Omega;}}^{\left(\mathrm{\&psi;}\right)}=\underset{\mathrm{\&xi;}=1}{\overset{num+nump}{\&Sigma;}}f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})---\left(9\right)$

When $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}\&le;d_c^{\left(\mathrm{\&gamma;}\right)}$ Time, $f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=1;$ When $d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}>d_c^{\left(\mathrm{\&gamma;}\right)}$ Time, $f(d_{\mathrm{\&Omega;}\mathrm{\&xi;}}^{\left(\mathrm{\&psi;}\right)}-d_c^{\left(\mathrm{\&gamma;}\right)})=0;$

Step 5.4, initialization j=1;

If a jth prediction example instp in the described prediction example set P of step 5.5 _jaffiliated cluster be clup _jwith i-th known example inst in described initialization example set D _iaffiliated cluster be clu _iidentical; Formula (10) is then utilized to obtain i-th known example inst _iexample instp is predicted with jth _jinfluence power gra _ij:

${\mathrm{gra}}_{ij}=\frac{{\mathrm{\&Gamma;}}_i\&times;{\mathrm{\&Gamma;}}_j}{d_{ij}}---\left(10\right)$

In formula (10), Γ _irepresent known example inst _iexample set is upgraded at the ψ time the corresponding attribute degree of polymerization upgrading example, Γ _jrepresent prediction example instp _jexample set is upgraded at the ψ time the corresponding attribute degree of polymerization upgrading example, d _ijrepresent described i-th known example inst _iexample instp is predicted with jth _jthe Euclidean distance of attribute;

Step 5.6, repetition step 5.5, thus obtain a jth prediction example instp _jwith the influence power of described other known example of initialization example set D, and record maximum effect power gra _max;

If step 5.7 gra _ij=gra _max, then labp is made _j=lab _i, represent the tally set labp of described prediction example set P _jin each label and the tally set lab of described initialization example set D _iin each label identical, thus obtain the prediction example of jth many Tag Estimations;

Step 5.8, by j+1 assignment to j, and judge whether j≤nump sets up, if set up, then return step 5.5 and perform, otherwise, has represented many Tag Estimations nump being predicted to example;

2. self-adaptation many Tag Estimations method according to claim 1, is characterized in that: in described step 5, also comprises step 5.9, by the described the ψ time renewal example set of tally set assignment to described correspondence completing the prediction example set P of many Tag Estimations in, thus obtain ψ+1 renewal example set example set is upgraded+1 time with described ψ the many Tag Estimations of self-adaptation are carried out as new initialization example set.

3. self-adaptation many Tag Estimations method according to claim 1 and 2, it is characterized in that: when occur new there is the prediction example of identical characteristics of objects and identical Object Semanteme time, only need can complete from step 4 and many Tag Estimations are carried out to new prediction example.

4. self-adaptation many Tag Estimations method according to claim 1, is characterized in that in described step 2.9, amendment threshold value rule be: if then will deduct τ ₂assignment is given otherwise, will add τ ₂assignment is given 0.1≤τ ₂≤ 0.5,75%≤τ ₁< 100%.