File classification management method based on RFID electronic tag

Abstract

The application relates to the technical field of text classification, and proposes an archive classification management method based on an RFID electronic tag, which comprises the following steps: extracting the content and archive information of an archive to be classified and managed, obtaining an RFID tag and a keyword feature map of the archive; determining an embedding vector of the archive and a feature vector of the keyword of the archive, determining the first similarity and the second similarity of two archives; determining the similarity evaluation parameters of each iteration, weighting and summing the first similarity and the second similarity of different archives according to the similarity evaluation parameters, and determining the adjustment distance of different archives; and respectively assigning an archive classification code to all archives according to the adjustment distance of different archives, so as to realize the archive classification management based on the RFID electronic tag. The application aims to improve the efficiency and effectiveness of solving the archive classification management.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of text classification, in particular to an archive classification management method based on RFID electronic tags. BACKGROUND

[0002] Matching the corresponding RFID electronic tags to the archives can help realize archive classification management. The archive name, number, category and other detailed information in the memory of multiple RFID electronic tags can be non-contact read by a handheld RFID reader, so as to improve the archive search speed and accuracy. Generally, the K-means algorithm is used to match the corresponding RFID electronic tags to the archives due to the high efficiency of the K-means algorithm in processing large-scale data sets and the prominent advantages of simple iteration.

[0003] However, the archive management needs to face a large number of archives and a large number of types, and the network topology structure corresponding to the keywords of different archives is complex, which often affects the effectiveness of the classification matching of the archives and the corresponding RFID electronic tags, and further affects the efficiency of the archive classification management. SUMMARY

[0004] The application provides an archive classification management method based on RFID electronic tags to solve the problems of complex relationship between keywords of archives and insufficient efficiency and effectiveness of archive classification management. The technical solution adopted is as follows:

[0005] One embodiment of the application provides an archive classification management method based on RFID electronic tags, which comprises the following steps:

[0006] Extracting the content and archive information of the archives to be classified and managed, and obtaining the RFID tags and keyword feature maps of the archives to be classified and managed;

[0007] According to the keyword feature map of the archive, the embedding vector of the archive and the feature vector of each keyword of the archive are determined, the first similarity between two archives is determined according to the similarity between the embedding vectors of the two archives, and the second similarity between the two archives is determined according to the difference between the feature vectors of the keywords of the two archives;

[0008] According to the first similarity and the second similarity of different archives, the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares in each iteration in the K-means algorithm clustering process are determined, the similarity evaluation parameters of the next iteration after the adjacent two iterations are determined according to the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares of the adjacent two iterations, the first similarity and the second similarity of different archives are weighted and summed according to the similarity evaluation parameters, and the adjusted distance of different archives is determined.

[0009] According to the adjustment distance of different archives, the archive classification code is respectively allocated to all archives, and the archive classification management based on the RFID electronic tag is realized.

[0010] Further, the RFID tag and the keyword feature map of the archives to be classified and managed are obtained by the following specific method:

[0011] The RFID reader is used to read the information of each archive to be classified and managed in sequence, and a unique identification code is allocated to each archive. The RFID reader is used to write the unique identification code and the information of the archive into the electronic tag chip of the archive, and the RFID tag of the archive is obtained.

[0012] The text data of the archive is extracted, and the keyword feature map of the archive is obtained according to the text data of the archive.

[0013] Further, the first similarity of the two archives is determined by the following specific method:

[0014] The cosine similarity of the embedding vectors of the two archives is recorded as the first similarity of the two archives.

[0015] Further, the second similarity of the two archives is determined according to the difference between the feature vectors of the keywords of the two archives, and the specific method includes:

[0016] The matching matrix of the two archives is constructed, and the element value of the first row and the first column in the matching matrix of the first archive to be classified and managed and the first archive to be classified and managed is:

[0017]

[0018] Among them, represents the feature vector of the first keyword of the first archive to be classified and managed, and the number of keywords contained in the first archive to be classified and managed is recorded as , ; represents the feature vector of the first keyword of the first archive to be classified and managed, and the number of keywords contained in the first archive to be classified and managed is recorded as , ; represents norm;

[0019] The matching matrix is supplemented to a square matrix by using the number 0; ​​​

[0020] Based on the matching matrix of the two files, determine the second similarity between the two files.

[0021] Furthermore, the specific method for determining the second similarity between the two files based on their matching matrices includes:

[0022] Retrieve the matching nodes of two files to be classified and managed, corresponding to elements in the matching matrix;

[0023] Interpolate the embedding vectors of all matching nodes from the two files. The mean of the norms is denoted as the second similarity between the two files.

[0024] Furthermore, the method for determining the sum of squares within the cosine cluster, the sum of squares within the keyword cluster, the sum of squares between cosine clusters, and the sum of squares between keyword clusters is as follows:

[0025] The first similarity of different files in the intra-cluster sum of squares of the clusters obtained in the same iteration is used as the distance between different files, and the cosine intra-cluster sum of squares of the clusters in the same iteration is obtained.

[0026] The second similarity of different files in the intra-cluster sum of squares of the clusters obtained in the same iteration is used as the distance between different files to obtain the intra-cluster sum of squares of the keywords of the clusters in the same iteration;

[0027] The first similarity of different files in the inter-cluster sum of squares of different clusters obtained in the same iteration is used as the distance between different files, and the cosine inter-cluster sum of squares of different clusters in the same iteration is obtained.

[0028] The second similarity of different files in the inter-cluster sum of squares of different clusters obtained in the same iteration is used as the distance between different files to obtain the keyword inter-cluster sum of squares of the clusters in the same iteration.

[0029] Furthermore, the method for determining the similarity evaluation parameters for the next iteration based on the sum of squares within the cosine cluster, the sum of squares within the keyword cluster, the sum of squares between cosine clusters, and the sum of squares between keyword clusters in two adjacent iterations includes the following specific methods:

[0030]

[0031] in, Indicates the first Changes in clustering quality of clusters in each iteration; Indicates the first Changes in the overall clustering quality of each iteration; Indicates the first The similarity evaluation parameters for the next iteration, where... Similarity evaluation parameters in the first iteration Similarity evaluation parameters in the second iteration The value is 0.5; This represents the preset smoothing factor.

[0032] Furthermore, the formulas for calculating the clustering quality change of the clusters and the overall clustering quality change are as follows:

[0033]

[0034]

[0035] in, Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares between keyword clusters in each iteration; Indicates the first The sum of squares between keyword clusters in each iteration.

[0036] Furthermore, the method for determining the adjustment distance of the different files is as follows:

[0037] For the same iteration process when using the K-means algorithm to cluster all archives, the similarity evaluation parameter is used as the weight of the first similarity of different archives, and the difference between the number 1 and the similarity evaluation parameter is used as the weight of the second similarity of different archives. The weighted sum is then calculated, and the result of the weighted sum is recorded as the adjusted distance of different archives.

[0038] Furthermore, the specific method for assigning file classification codes to all files based on the adjustment distance of different files includes:

[0039] The adjustment distance between different files is used as the distance between different files. The K-means algorithm is used to cluster all files to obtain clusters. All files in the same cluster are assigned the same file classification code.

[0040] The beneficial effects of this application are:

[0041] The application considers that the content in the archives to be classified and managed contains a large amount of information, which is manifested as that the keywords in the archives have a complex keyword topology structure. If K-means is directly used for clustering on the archives to be classified and managed, the keyword topology structure between the keywords in the archives will be directly converted into the Euclidean distance of the vector space, that is, the topology relationship is flattened, and the classification result of the archives obtained is often inaccurate. When the archives to be classified and managed are classified and managed, firstly, the first similarity of two archives is determined according to the similarity between the embedding vectors of the two archives, then the second similarity of the two archives is determined according to the difference between the feature vectors of the keywords of the two archives in combination with the keyword topology structure, the accuracy of the similarity evaluation between the archives is improved by avoiding the problems of too many keywords and too complex association between the keywords, and the accuracy of the classification management of the archives is further improved; further, the intra-cluster sum of squares of the same clustering cluster and the inter-cluster sum of squares of different clustering clusters are improved according to the first similarity and the second similarity of different archives respectively, the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares in each iteration in the clustering process of the K-means algorithm are determined respectively, and the similarity evaluation parameters in the former iteration and the latter iteration in the adjacent two iterations are determined according to the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares of the adjacent two iterations, the similarity evaluation parameters provide a basis for measuring the similarity between the archives, and the first similarity and the second similarity of different archives are weighted and summed according to the similarity evaluation parameters, so that the adjustment distance of different archives can be determined, that is, the total evaluation of the difference between different archives; finally, according to the adjustment distance of different archives, the archive classification codes are respectively allocated to all archives, the archive classification management based on the RFID electronic tag is realized, and the problems of complex relationship between the keywords of the archives and insufficient efficiency and effectiveness of the archive classification management are solved. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] Figure 1 The flow chart of the archive classification management method based on the RFID electronic tag provided by an embodiment of the present application;

[0044] Figure 2 The flow chart of the first similarity acquisition provided by an embodiment of the present application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0046] Please refer to Figure 1 which shows a flow chart of an archive classification management method based on an RFID electronic tag according to an embodiment of the present application. The method comprises the following steps:

[0047] In step S001, the content and archive information of an archive to be classified and managed are extracted, and the RFID tag and keyword feature map of the archive to be classified and managed are obtained.

[0048] The ISBN, title, author, publisher, publication time, call number and other archive information of the archive to be classified and managed are read in sequence using an RFID reader. An ISBN international standard book number and library internal index number mapping table is generated for each archive, and a mapping relationship table is established in a system database. The system database is a complete book information management system, which can realize management operations such as book information input, retrieval, change, deletion, and functions such as administrator and user account registration, borrowing and returning. A UID unique identification code is assigned to each archive, and the UID and archive information of the archive to be classified and managed are written into the electronic tag chip of the archive to be classified and managed by the RFID reader, so as to obtain the RFID tag of the archive to be classified and managed.

[0049] Since the archive to be classified and managed is divided into two types of text archives and electronic archives, the RFID tag of the text archive to be classified and managed is pasted on the surface of the archive, and the text archive to be classified and managed is placed in a text classification area; the RFID tag of the electronic archive to be classified and managed is saved in the attribute information of the archive, and the electronic archive to be classified and managed is placed in an electronic classification area. For each text archive in the classification area, the text data of the archive is extracted by text analysis technology. Specifically, a picture of the text archive is taken, and the text data in the picture is recognized by using OCR technology. The OCR technology for recognizing the text data in the picture is a known technology, and will not be described in detail. For each electronic archive in the classification area, the text data of the archive is directly read.

[0050] The text data of each archive to be classified and managed is processed by using a TextRank algorithm respectively, and the keyword feature map of each archive to be classified and managed is obtained.

[0051] It can be understood that the keyword feature map of each archive to be classified and managed is an undirected weighted graph, each keyword in the archive to be classified and managed corresponds to a node in the undirected weighted graph, the nodes are connected by edges, and the weight of the edge represents the co-occurrence frequency of the two nodes connected by the edge. Wherein, the keyword feature map obtained by using the TextRank algorithm is a known technology and will not be described again.

[0052] Up to now, the keyword feature map of each archive to be classified and managed is obtained.

[0053] Step S002, according to the keyword feature map of the archive, determine the embedding vector of the archive and the feature vector of each keyword of the archive, determine the first similarity of two archives according to the similarity between the embedding vectors of the two archives, and determine the second similarity of the two archives according to the difference between the feature vectors of the keywords of the two archives.

[0054] The content of the archive to be classified and managed contains a large amount of information, which is manifested as the complex keyword topology graph structure of the keywords in the archive. If K-means is directly used for clustering on the archive to be classified and managed, the keyword topology structure between the keywords in the archive will be directly converted into the Euclidean distance in the vector space, that is, the topology relationship is flattened, and the result of the archive classification obtained is often inaccurate. Therefore, when classifying and managing the archive to be classified and managed, the keyword topology graph structure can be combined to improve the accuracy of the archive classification and management, avoid the problem that the number of keywords is too large and the association between the keywords is too complex, and make the category to which the archive belongs difficult to identify.

[0055] The keyword feature map of all archives to be classified and managed is processed using the Node2vec graph embedding algorithm to obtain the embedding vector of each archive to be classified and managed and the feature vector of each keyword of the archive to be classified and managed.

[0056] Preferably, as an embodiment of the present application, the back-off probability of the Node2vec graph embedding algorithm is 1, the forward probability is 2, the walk strategy is controlled, and the embedding vector of each archive is obtained; using the Node2vec graph embedding algorithm to obtain the embedding vector and the feature vector of the embedding vector is a known technology and will not be described again.

[0057] The cosine similarity of the embedding vectors of two archives to be classified and managed is denoted as the first similarity of the two archives to be classified and managed.

[0058] Wherein, calculating the cosine similarity of two vectors is a known technology and will not be described again.

[0059] The first similarity of two archives to be classified and managed can evaluate the spatial similarity of the embedding vectors of the two archives. The first similarity acquisition flow chart is as shown in Figure 2 ​

[0060] When the keywords of two archives exist co-occurrence relationship, but the co-occurrence frequency is different, for example, both archives contain artificial intelligence, neural network, optimization algorithm and smart medical, the four keywords, the content of one archive is artificial intelligence and neural network theoretical system, and introduces the application of part of artificial intelligence in smart medical, the content of the other archive is to take artificial intelligence in smart medical as the main body, and introduces various application scenarios, the keywords of the two archives are the same, but the trend of keyword expression information is different, that is, the importance of keyword expression information to the meaning of archive content expression is different, at this time, using only the first similarity as the basis for archive classification is not accurate, and further evaluation of the importance of keyword expression information of different archives to the meaning of archive content expression is needed.

[0061] According to the difference between the feature vectors of the keywords of the two archives to be classified and managed, a matching matrix of the two archives to be classified and managed is constructed, the element value in the i-th row and the j-th column of the matching matrix of the first and the second archive to be classified and managed is:

[0062]

[0063] Wherein, represents the feature vector of the i-th keyword of the first archive to be classified and managed, and the number of keywords contained in the first archive to be classified and managed is denoted as ; ; represents the feature vector of the i-th keyword of the second archive to be classified and managed, and the number of keywords contained in the second archive to be classified and managed is denoted as ; ; represents norm.

[0064] The matching matrix is supplemented to a square matrix by filling with the number 0 on the right or the bottom of the matching matrix of the two archives to be classified and managed, wherein the number of elements contained in the row and the column of the matching matrix of the two archives to be classified and managed is the maximum value of the number of keywords contained in the two archives to be classified and managed.

[0065] Hungarian algorithm is used on the matching matrix of the two archives to be classified and managed to obtain the matched nodes of the two archives to be classified and managed corresponding to the elements in the matching matrix.​​​​​

[0066] the average of the norms between the embedding vectors of all matched nodes of the two archives to be classified and managed, denoted as the second similarity of the two archives to be classified and managed. The second similarity is determined by the matched nodes in the keyword feature map of the archive, and the similarity of the two archives to be classified and managed can be evaluated according to the keywords corresponding to the nodes and the co-occurrence frequency of the keywords, so as to avoid errors caused by different importance of keyword expression information to the meaning of archive content expression.

[0067] The first similarity and the second similarity of different archives to be classified and managed are obtained.

[0068] The first similarity and the second similarity of different archives to be classified and managed are obtained.

[0069] Step S003, according to the first similarity and the second similarity of different archives, respectively determine the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares in each iteration in the K-means algorithm clustering process, determine the similarity evaluation parameters of the latter iteration in the adjacent two iterations according to the cosine intra-cluster sum of squares, the keyword intra-cluster sum of squares, the cosine inter-cluster sum of squares and the keyword inter-cluster sum of squares of the adjacent two iterations, and determine the adjusted distance of different archives by weighted sum of the first similarity and the second similarity of different archives according to the similarity evaluation parameters.

[0070] When using the K-means algorithm to cluster all archives to be classified and managed, each iteration needs to divide the archives into different clustering clusters, and calculate the intra-cluster sum of squares of the same clustering cluster and the inter-cluster sum of squares of different clustering clusters. The process of using K-means algorithm to calculate the intra-cluster sum of squares of the same clustering cluster and the inter-cluster sum of squares of different clustering clusters in each iteration is a known technology and will not be described here.

[0071] The first similarity of different archives to be classified and managed in the intra-cluster sum of squares of the clustering cluster obtained in the same iteration is taken as the distance of different archives to be classified and managed, and the cosine intra-cluster sum of squares of the clustering cluster in the same iteration is obtained; the second similarity of different archives to be classified and managed in the intra-cluster sum of squares of the clustering cluster obtained in the same iteration is taken as the distance of different archives to be classified and managed, and the keyword intra-cluster sum of squares of the clustering cluster in the same iteration is obtained. The first similarity of different archives to be classified and managed in the inter-cluster sum of squares of different clustering clusters obtained in the same iteration is taken as the distance of different archives to be classified and managed, and the cosine inter-cluster sum of squares of different clustering clusters in the same iteration is obtained; the second similarity of different archives to be classified and managed in the inter-cluster sum of squares of different clustering clusters obtained in the same iteration is taken as the distance of different archives to be classified and managed, and the keyword inter-cluster sum of squares of the clustering cluster in the same iteration is obtained.

[0072] Based on the sum of squares within the cosine cluster, the sum of squares within the keyword cluster, the sum of squares between cosine clusters, and the sum of squares between keyword clusters in two adjacent iterations, the similarity evaluation parameters for the later iteration in two adjacent iterations are determined.

[0073]

[0074]

[0075]

[0076] in, Indicates the first Changes in clustering quality of clusters in each iteration; Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration; Indicates the first Changes in the overall clustering quality of each iteration Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares between keyword clusters in each iteration; Indicates the first The sum of squares between keyword clusters in each iteration; Indicates the first The similarity evaluation parameters for the next iteration, where... Similarity evaluation parameters in the first iteration Similarity evaluation parameters in the second iteration The value is directly assigned to 0.5; This represents the preset smoothing factor. In this embodiment, the smoothing factor is set to 0.1.

[0077] When using the K-means algorithm to cluster all files under classification management, the first... The next iteration At that time, the first In the next iteration, the first similarity of different archives to be classified and managed plays a more important role in measuring the similarity between archives; The next iteration At that time, the first The second similarity of the different archives to be classified and managed in the secondary iteration process plays a more important role in measuring the similarity between archives; the first similarity of the different archives to be classified and managed in the secondary iteration process plays a less important role in measuring the similarity between archives; and the third similarity of the different archives to be classified and managed in the secondary iteration process plays the least important role in measuring the similarity between archives. The first similarity of the different archives to be classified and managed in the secondary iteration process plays a more important role in measuring the similarity between archives; the second similarity of the different archives to be classified and managed in the secondary iteration process plays a less important role in measuring the similarity between archives; and the third similarity of the different archives to be classified and managed in the secondary iteration process plays the least important role in measuring the similarity between archives. The first similarity of the different archives to be classified and managed in the secondary iteration process plays a more important role in measuring the similarity between archives; the second similarity of the different archives to be classified and managed in the secondary iteration process plays a less important role in measuring the similarity between archives; and the third similarity of the different archives to be classified and managed in the secondary iteration process plays the least important role in measuring the similarity between archives. The first similarity of the different archives to be classified and managed in the secondary iteration process plays a more important role in measuring the similarity between archives; the second similarity of the different archives to be classified and managed in the secondary iteration process plays a less important role in measuring the similarity between archives; and the third similarity of the different archives to be classified and managed in the secondary iteration process plays the least important role in measuring the similarity between archives. The first similarity of the different archives to be classified and managed in the secondary iteration process plays a more important role in measuring the similarity between archives; the second similarity of the different archives to be classified and managed in the secondary iteration process plays a less important role in measuring the similarity between archives; and the third similarity of the different archives to be classified and managed in the secondary iteration process plays the least important role in measuring the similarity between archives.

[0078] For the same iteration process of using the K-means algorithm to cluster all the archives to be classified and managed, the similarity evaluation parameter is taken as the weight of the first similarity of the different archives to be classified and managed, the difference between the number 1 and the similarity evaluation parameter is taken as the weight of the second similarity of the different archives to be classified and managed, the first similarity and the second similarity of the different archives to be classified and managed are weighted and summed, and the result of the weighted sum is recorded as the adjusted distance of the different archives to be classified and managed.

[0079] Up to now, the adjusted distance of the different archives to be classified and managed is obtained.

[0080] Step S004: According to the adjusted distance of the different archives, all the archives are respectively assigned with an archive classification code, so as to realize the archive classification management based on the RFID electronic tag.

[0081] The adjusted distance of the different archives to be classified and managed is taken as the distance between the different archives to be classified and managed, the K-means algorithm is used to cluster all the archives to be classified and managed, and the clustering cluster is obtained.

[0082] Preferably, as an embodiment of the present application, the number of the clustering cluster is the number of the types of the archives to be classified and managed, and the value of the number of the clustering cluster in the embodiment is 300. The K-means algorithm is used to cluster the archives, and the clustering cluster is obtained, which is a well-known technology and will not be described here.

[0083] All the archives to be classified and managed in the same clustering cluster are assigned with the same archive classification code, the archive classification code is recorded in the RFID electronic tag, and the classification management is performed.

[0084] It can be understood that the archive classification codes of the archives in different clustering clusters are different.

[0085] It can be understood that the RFID tag of the file can be scanned by the handheld RFID device, the reliable reading in the multi-RFID tag environment is verified, the label integrity of the RFID tag is verified, and the label UID, classification code and storage location are verified, if the label reading fails or the information is inconsistent, an abnormal report is immediately pushed to the management terminal. Set read-write nodes at the links of borrowing, transferring, returning, etc. Each operation executes the process of reading the file UID, verifying the identity, automatically writing the operation record, and recording the file flow track and generating the flow log, synchronizing the flow log to the database to display the file state, realizing the classified management of the file. The process of operation record includes the borrower, operation time and operation type. The management system background summarizes the file distribution, borrowing statistics, abnormal warning and other data of each storage area, generates a visual table, supports multi-dimensional query and export according to file categories, time, users and the like, and realizes efficient and fast classified management of the whole life cycle of the file.

[0086] Thus, the classified management of the file based on the RFID electronic tag is realized.

[0087] The above only describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for classifying and managing archives based on RFID electronic tags, characterized in that, The method includes the following steps: Extract the content and information of the archives to be classified and managed, and obtain the RFID tags and keyword feature maps of the archives to be classified and managed; Based on the keyword feature map of the archives, the embedding vector of the archives and the feature vector of each keyword of the archives are determined. The first similarity between the two archives is determined based on the similarity between their embedding vectors. The second similarity between the two archives is determined based on the difference between their feature vectors of each keyword. Based on the first and second similarities of different files, the sum of squares within cosine clusters, the sum of squares within keyword clusters, the sum of squares between cosine clusters, and the sum of squares between keyword clusters are determined for each iteration in the K-means clustering process. Based on the sum of squares within cosine clusters, the sum of squares within keyword clusters, the sum of squares between cosine clusters, and the sum of squares between keyword clusters in two adjacent iterations, the similarity evaluation parameters for the next iteration in two adjacent iterations are determined. Based on the similarity evaluation parameters, the first and second similarities of different files are weighted and summed to determine the adjustment distance of different files. Based on the adjustment distance of different archives, each archive is assigned an archive classification code to achieve archive classification management based on RFID electronic tags; The second similarity between the two documents is determined based on the differences in the feature vectors of each keyword. The specific method includes constructing a matching matrix between the two documents, and... The and the first The matching matrix of the files to be classified and managed Line 1 Column element values for: in, Indicates the first The first batch of files awaiting classification management The feature vector of the i-th keyword, the i-th keyword The number of keywords contained in each file awaiting classification is denoted as: , ; Indicates the first The first batch of files awaiting classification management The feature vector of the i-th keyword, the i-th keyword The number of keywords contained in each file awaiting classification is denoted as: , ; express Norm; The matching matrix is ​​padded to a square matrix using the number 0; based on the matching matrices of the two files, the second similarity between the two files is determined, including the following specific methods: obtaining the matching nodes of the two files to be classified and managed corresponding to the elements in the matching matrix; and merging the embedding vectors of all matching nodes of the two files. The mean of the norm is denoted as the second similarity between the two files; The methods for determining the cosine cluster intra-cluster sum of squares, keyword cluster intra-cluster sum of squares, cosine cluster inter-cluster sum of squares, and keyword cluster inter-cluster sum of squares are as follows: Using the first similarity between different files in the intra-cluster sum of squares of clusters obtained in the same iteration as the distance between different files, the cosine cluster intra-cluster sum of squares of clusters obtained in the same iteration is obtained; using the second similarity between different files in the intra-cluster sum of squares of clusters obtained in the same iteration as the distance between different files, the keyword cluster intra-cluster sum of squares of clusters obtained in the same iteration is obtained; using the first similarity between different files in the inter-cluster sum of squares of different clusters obtained in the same iteration as the distance between different files, the cosine cluster inter-cluster sum of squares of clusters obtained in the same iteration is obtained; using the second similarity between different files in the inter-cluster sum of squares of different clusters obtained in the same iteration as the distance between different files, the keyword cluster inter-cluster sum of squares of clusters obtained in the same iteration is obtained. Based on the sum of squares within cosine clusters, the sum of squares within keyword clusters, the sum of squares between cosine clusters, and the sum of squares between keyword clusters from two adjacent iterations, the similarity evaluation parameters for the later iteration in two adjacent iterations are determined. The specific methods include: in, Indicates the first Changes in clustering quality of clusters in each iteration; Indicates the first Changes in the overall clustering quality of each iteration; Indicates the first The similarity evaluation parameters for the next iteration, where... Similarity evaluation parameters in the first iteration Similarity evaluation parameters in the second iteration The value is 0.5; Indicates the preset smoothing factor; The formulas for calculating the change in cluster quality of individual clusters and the overall change in cluster quality are as follows: in, Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The cosine sum of squares within the cluster of the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration; Indicates the first The sum of squares between cosine clusters in the next iteration Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares within the keyword cluster of the next iteration; Indicates the first The sum of squares between keyword clusters in each iteration; Indicates the first The sum of squares between keyword clusters in each iteration; The specific method for obtaining the RFID tags and keyword feature maps of the archives to be classified and managed is as follows: use an RFID reader to read the information of each archive to be classified and managed in sequence, assign a unique identification code to each archive, and write the unique identification code and archive information into the electronic tag chip of the archive through the RFID reader to obtain the RFID tag of the archive; extract the text data of the archive, and obtain the keyword feature map of the archive based on the text data of the archive.

2. The method for classifying and managing archives based on RFID electronic tags according to claim 1, characterized in that, The method for determining the first similarity between two files is as follows: The cosine similarity of the embedding vectors of two files is denoted as the first similarity between the two files.

3. The document classification and management method based on RFID electronic tags according to claim 1, characterized in that, The method for determining the adjustment distance for different files is as follows: For the same iteration process when using the K-means algorithm to cluster all archives, the similarity evaluation parameter is used as the weight of the first similarity of different archives, and the difference between the number 1 and the similarity evaluation parameter is used as the weight of the second similarity of different archives. The weighted sum is then calculated, and the result of the weighted sum is recorded as the adjusted distance of different archives.

4. The method for classifying and managing archives based on RFID electronic tags according to claim 1, characterized in that, Based on the adjustment distance of different archives, each archive is assigned an archive classification code, including the following specific methods: The adjustment distance between different files is used as the distance between different files. The K-means algorithm is used to cluster all files to obtain clusters. All files in the same cluster are assigned the same file classification code.

Images