Artificial intelligence-based scientific and technological achievement transformation data intelligent monitoring method and system

Abstract

The application relates to the technical field of artificial intelligence, in particular to a scientific and technological achievement transformation data intelligent monitoring method and system based on artificial intelligence. The method comprises the following steps: collecting historical transformation data, randomly masking a certain word item in the historical transformation data, calculating the semantic difference degree between the masked transformation data and the historical transformation data, and selecting a first masking word; expanding the first masking word to both sides until the semantic difference degree is greater than a preset threshold to obtain a first masking area; adding the word item of the first masking word until the smallest semantic difference degree is greater than the preset threshold to obtain a second masking area; constructing a first structure vector according to the intersection and union ratio of the first masking area and the second masking area; grouping all the first structure vectors; obtaining the keywords of the historical transformation data; obtaining a second structure vector of current transformation data, selecting the most similar group, and judging whether the current transformation data is abnormal. The embodiment of the application can realize content anomaly monitoring of the transformation data.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method and system for intelligent monitoring of data related to the transformation of scientific and technological achievements based on artificial intelligence. Background Technology

[0002] Technology transfer refers to the subsequent testing, development, application, and promotion of scientific and technological achievements with practical value generated from scientific research and technological development in order to improve productivity, ultimately leading to the formation of new products, new processes, new materials, and the development of new industries. Promoting technology transfer and accelerating the industrialization of scientific and technological achievements has become a new trend in the science and technology policies of countries around the world.

[0003] With the continued efforts in reforming the science and technology system, especially the introduction of major reforms in resource allocation, planning management, and the transformation of scientific and technological achievements, as well as the rise of mass entrepreneurship and innovation, the speed at which scientific and technological achievements are transformed into real productive forces is accelerating. At the same time, monitoring data on the transformation of scientific and technological achievements is gradually being put on the agenda.

[0004] Existing technologies for intelligent monitoring of technology transfer data typically rely on manual methods. This approach requires significant human resources, and the inherent subjectivity of human monitoring can lead to biased monitoring results. Therefore, some existing technologies have proposed using artificial intelligence networks for process monitoring. However, current monitoring methods only target the transfer process and do not consider data anomalies in the technology transfer data itself. This poses a potential risk of data tampering, which could hinder the technology transfer process. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention aims to provide an intelligent monitoring method and system for the transformation of scientific and technological achievements based on artificial intelligence. The specific technical solution adopted is as follows:

[0006] In a first aspect, one embodiment of the present invention provides an intelligent monitoring method for data on the transformation of scientific and technological achievements based on artificial intelligence, the method comprising the following steps:

[0007] Collect historical conversion data of historical achievement conversion files, randomly occlude a word in the historical conversion data, calculate the semantic difference between the occluded conversion data and the historical conversion data, and take the word corresponding to the smallest semantic difference as the first occluded word.

[0008] Expand the first occluded word to both sides, and then occlude the historical conversion data until the semantic difference is greater than a preset threshold. Use the occluded word at this time as the first occluded area.

[0009] Add a term to the first occluded word and occlude the historical conversion data. When adding a term, traverse the unoccluded terms and select the second occluded word. Continue to expand the term of the second occluded word until the minimum semantic difference is greater than the preset threshold. Use the occluded word at this time as the second occluded region.

[0010] A first structural vector is constructed based on the intersection-union ratio of the first occlusion region and the second occlusion region; the first structural vectors of all the historical conversion data are clustered into multiple groups; the terms contained in the first structural vectors are removed from each historical conversion data to obtain the keywords of the historical conversion data;

[0011] Obtain the second structure vector of the current conversion data, and select the group most similar to the second structure vector; compare the keywords of the current conversion data with the keywords corresponding to each of the first structure vectors in the group. When the maximum similarity obtained by the comparison is below the similarity threshold, the current conversion data is abnormal data.

[0012] Preferably, the historical conversion data collection steps include:

[0013] Establish a historical document database for the transformation of scientific and technological achievements, using the subject text of each document as its identifier, and the word vector corresponding to the identifier as the historical transformation data.

[0014] Preferably, the calculation process for the semantic dissimilarity includes:

[0015] The historical conversion data after the words are masked is reconstructed by an autoencoder to obtain the masked conversion data. The Euclidean distance between the corresponding word vectors of the historical conversion data and the masked conversion data is calculated as the semantic difference.

[0016] Preferably, the step of obtaining the first occluded area includes:

[0017] Centered on the first occluded word, the word length of the first occluded word is used as the expansion size to expand to both sides to obtain the first expanded word. The first expanded word is used to occlude the historical conversion data to obtain the first conversion data. The semantic difference between the word vectors of the historical conversion data and the first conversion data is calculated and compared with a preset threshold. When the semantic difference is not greater than the preset threshold, the expansion continues to expand to both sides according to the expansion size, centered on the first expanded word, until the semantic difference is greater than the preset threshold, thus obtaining the first occluded region.

[0018] Preferably, the step of obtaining the second occluded area includes:

[0019] A second expanded term is expanded by randomly adding a term to the first occluded word. The second expanded term is used to occlude the historical conversion data to obtain the second conversion data. The semantic difference between the word vectors of the historical conversion data and the second conversion data is calculated. Unoccluded words are traversed, and the occluded word corresponding to the minimum semantic difference is taken as the second occluded word. When it is not greater than a preset threshold, the second occluded word is expanded by randomly adding a term until the semantic difference is greater than the preset threshold, thus obtaining the second occluded region.

[0020] Preferably, the construction steps of the first structure vector include:

[0021] The intersection-union ratio is calculated based on the number of terms corresponding to the first occluded region and the second occluded region;

[0022] Obtain the difference positions of the difference terms in the intersection and union of the first occluded region and the second occluded region, and construct the first structure vector based on the difference positions and the intersection-union ratio.

[0023] Preferably, the first structure vector of all the historical transformation data is clustered into multiple groups, including:

[0024] The initial number of clusters is preset, all the first structure vectors are clustered to obtain multiple initial categories, the difference of the first structure vectors in each initial category is calculated, and then the category difference between all the initial category differences is obtained;

[0025] Gradually increase the initial number of categories, calculate the category differences sequentially, select the number of categories corresponding to the smallest category difference as the ideal number, and use the clustering result corresponding to the ideal number as the grouping result.

[0026] Preferably, selecting the group most similar to the second structural vector includes:

[0027] Obtain the central structure vector at the geometric center of each group, calculate the similarity between each central structure vector and the second structure vector, and take the group with the maximum similarity as the group most similar to the second structure vector.

[0028] Secondly, another embodiment of the present invention provides an intelligent monitoring system for technology transfer data based on artificial intelligence, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-mentioned intelligent monitoring method for technology transfer data based on artificial intelligence.

[0029] The embodiments of the present invention have at least the following beneficial effects:

[0030] By analyzing historical conversion data, a first structural vector is obtained, and the first structural vector is grouped. Then, a second structural vector of the current conversion data is obtained, and the group most similar to the second structural vector is selected. The group is then compared with the second structural vector by keywords to determine whether the current conversion data is abnormal. This embodiment of the invention can quickly match the unknown conversion files to the corresponding group and perform keyword comparison, thereby realizing abnormal monitoring of conversion data and improving the efficiency of intelligent data monitoring. Attached Figure Description

[0031] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 A flowchart illustrating the steps of an artificial intelligence-based intelligent monitoring method for the transformation of scientific and technological achievements, provided as an embodiment of the present invention. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the intelligent monitoring method and system for the transformation of scientific and technological achievements based on artificial intelligence proposed in this invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] This invention applies intelligent monitoring of conversion data to the preprocessing stage of the conversion process, that is, to perform preliminary intelligent monitoring of the conversion data in order to facilitate the smooth progress of the subsequent conversion process.

[0036] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent monitoring method and system for the transformation of scientific and technological achievements based on artificial intelligence provided by this invention.

[0037] Please see Figure 1 The diagram illustrates a flowchart of a method for intelligent monitoring of technology transfer data based on artificial intelligence, according to an embodiment of the present invention. The method includes the following steps:

[0038] Step S001: Collect historical conversion data, randomly occlude a word in the historical conversion data, calculate the semantic difference between the occluded conversion data and the historical conversion data, and take the word with the smallest semantic difference as the first occluded word.

[0039] The specific steps include:

[0040] 1. Collect historical conversion data.

[0041] Establish a historical document database for the transformation of scientific and technological achievements, using the subject text of each document as its identifier, and the corresponding word vector as historical transformation data.

[0042] In this embodiment of the invention, the title of the file is used as the identifier of the file. In other embodiments, the title summary or other text indicating the subject of the file may also be used as the identifier.

[0043] As an example, the title of this invention, "Intelligent Monitoring Method and System for Technology Transfer Data Based on Artificial Intelligence," obtains its word vector through Natural Language Processing (NLP) as [based on, artificial intelligence, of, technology transfer data, intelligent, monitoring, method, and system]. This word vector is the transfer data of this invention.

[0044] 2. Obtain the first masked word.

[0045] The historical conversion data of the occluded terms is reconstructed by an autoencoder to obtain the occluded conversion data. The Euclidean distance between the corresponding word vectors of the historical conversion data and the occluded conversion data is calculated as the semantic difference.

[0046] An autoencoder is constructed, with a single historical transformation data as input and reconstructed transformation data as output. The specific training process uses the historical transformation data as the training dataset and the Euclidean distance between the input and output as the reconstruction loss of the autoencoder.

[0047] The trained autoencoder is used to reconstruct a single historical conversion data point, obtaining the reconstructed conversion data x′ before occlusion. The historical conversion data is then randomly occluded by covering words with blanks; single words in the historical conversion data are randomly occluded and input into the trained autoencoder to obtain the reconstructed conversion data x′ after occlusion. m , x′ and x′ m Compare them and calculate their Euclidean distance dis(x′,x′). m The semantic dissimilarity is defined as follows: ) represents the m-th term being occluded. Occlusion is applied to each term in the transformed data, resulting in multiple dis(x′,x′) values. m ), select min[dis(x′,x′)m The corresponding masking term is used as the first masking term.

[0048] Step S002: Expand the first occluded word to both sides, and then occlude the historical conversion data until the semantic difference is greater than the preset threshold. Use the occluded word at this time as the first occluded area.

[0049] The specific steps include:

[0050] Centered on the first occluded word, the word length of the first occluded word is used as the expansion size to expand to both sides to obtain the first expanded word. The first expanded word is used to occlude historical conversion data to obtain the first conversion data. The semantic difference between the word vectors of the historical conversion data and the first conversion data is calculated and compared with a preset threshold. When the semantic difference is not greater than the preset threshold, the expansion continues to expand to both sides with the first expanded word as the center, until the semantic difference is greater than the preset threshold, thus obtaining the first occluded region.

[0051] The occlusion is expanded based on the first occluded word, i.e., the size of the occluded part changes. The size of the occluded part of the first occluded word is 1, and the expansion size is also 1. The size change of each expansion is L = L + 2, that is, expanding to the left and right by one word position respectively. After the first expansion, the first expanded word is obtained. The historical conversion data is occluded using the first expanded word and fed into the autoencoder to obtain the reconstructed first conversion data. The Euclidean distance dis1(x′,x″) between the first expanded word and the reconstructed historical conversion data is calculated. m ), where x″ m This represents the first transformation data reconstructed after the first expanded term centered on the m-th term is occluded in the first expansion case. A preset threshold m for Euclidean distance is set. dis Compare dis1(x′,x″) m ) and preset threshold m dis The size of dis1(x′,x″) m )>m dis The first expanded term is the first occluded region; otherwise, continue expanding outwards from the first expanded term as the center, resulting in dis2(x′,x″). m ), build about c-dis c (x′,x″ m The scatter plot of ) is given, where c represents the transformation data at the c-th expansion, and the first time the condition dis is satisfied is obtained. c (x′,x″ m )>m dis The c value is selected, and c-1 is chosen as the final expansion number. The occluded term at this time is then taken as the first occluded region.

[0052] Step S003: Add the first occlusion term and occlude the historical conversion data. When adding the term, traverse the unoccluded terms and select the second occlusion term. Continue to expand the term of the second occlusion term until the minimum semantic difference is greater than the preset threshold. Use the occlusion term at this time as the second occlusion area.

[0053] The specific steps include:

[0054] For the first occluded word, a term is randomly added to expand it into a second expanded term. The second expanded term is used to occlude historical conversion data to obtain the second conversion data. The semantic difference between the word vectors of the historical conversion data and the second conversion data is calculated. Unoccluded words are traversed, and the occluded word corresponding to the minimum semantic difference is taken as the second occluded word. When it is not greater than a preset threshold, the second occluded word is expanded by randomly adding a term until the minimum semantic difference is greater than the preset threshold, thus obtaining the second occluded region.

[0055] Based on the first occluded word, a word is randomly added to obtain the second expanded word. The historical transformation data is then occluded using the second expanded word and input into the autoencoder to obtain the second transformation data. The semantic difference between the word vectors of the historical transformation data and the second transformation data is then calculated.

[0056] Iterate through all unmasked words to obtain multiple second expanded terms and corresponding second conversion data. Calculate the semantic difference between the word vectors of historical conversion data and each second conversion data. Select the masked word corresponding to the minimum semantic difference as the second masked word. If the semantic difference of the second masked word is not greater than a preset threshold m... dis Continue adding a term to obtain the third occluded word, and so on, until the minimum semantic difference is greater than the preset threshold.

[0057] When an additional occluded term is added, the unoccluded terms are iterated over each time, and the occluded term corresponding to the minimum semantic difference is selected to update the conversion data each time, until the minimum semantic difference is greater than the preset threshold. At this point, the occluded term is taken as the second occlusion area.

[0058] Step S004: Construct a first structural vector based on the intersection-union ratio of the first and second occlusion regions; divide the first structural vectors of all historical conversion data into multiple groups through clustering; remove the terms contained in the first structural vector from each historical conversion data to obtain the keywords of the historical conversion data.

[0059] The specific steps include:

[0060] 1. Construct the first structure vector.

[0061] The intersection-union ratio (IUR) is calculated based on the number of terms corresponding to the first and second occluded regions. The difference positions of the differing terms in the intersection and union of the first and second occluded regions are obtained, and a first structural vector is constructed based on the difference positions and the IUR.

[0062] Obtain the intersection-union ratio (IUGR) of the first and second occluded regions. The IUGR is calculated based on the number of covered terms. Obtain the location information of the differing terms within the region after subtracting the intersection from the union. The location information is represented by labels.

[0063] Obtain the number of differing terms in the first and second occlusion regions corresponding to all historical conversion data, find the maximum number Z, construct a difference location description for each historical conversion data as a vector of Z elements. If the number of differing terms in a certain historical conversion data is less than Z, zero-padding is performed. The intersection-union ratio and the difference location description are concatted to obtain a first structure vector of Z+1 elements.

[0064] 2. Group all first structure vectors.

[0065] The initial number of clusters is preset. All first structure vectors are clustered to obtain multiple initial clusters. The difference of the first structure vectors in each initial cluster is calculated, and then the category difference between all initial category differences is obtained. The number of initial clusters is gradually increased, and the category difference is calculated in turn. The number of clusters corresponding to the smallest category difference is selected as the ideal number, and the clustering result corresponding to the ideal number is used as the grouping result.

[0066] The K-means clustering algorithm is used to cluster all first structure vectors with a preset number of initial categories to obtain multiple initial categories.

[0067] As an example, in this embodiment of the invention, the initial number of categories is 3, that is, the initial K is set to 3.

[0068] It should be noted that the distance metric does not use Euclidean distance, which is the standard K-means clustering algorithm, but rather cosine similarity.

[0069] The intra-class variance of each initial cluster set is obtained as the difference of the first structure vector in each initial category. The variance between intra-class variances is calculated as the category difference.

[0070] Adjust K = K + 1, and perform the same processing steps as described above to obtain multiple class differences, and fit a value related to K - σ. 2 The description curve is used to select the K′ corresponding to the minimum point as the optimized K, and the cluster set corresponding to K′ clustering is used as the final grouping result.

[0071] 3. Remove the terms contained in the first structural vector from each historical conversion data to obtain the keywords of the historical conversion data.

[0072] The first structural vector is obtained and keywords are selected, which facilitates subsequent clustering and grouping based on structural representation and comparison with the keyword information of the current conversion data.

[0073] Step S005: Obtain the second structure vector of the current conversion data, select the group most similar to the second structure vector; compare the keywords of the current conversion data with the keywords corresponding to each first structure vector in the group. When the maximum similarity obtained by comparison is below the similarity threshold, the current conversion data is abnormal data.

[0074] The specific steps include:

[0075] 1. Obtain the second structure vector of the current transformed data, and select the group that is most similar to the second structure vector.

[0076] Obtain the current transformation data, and use the same method to obtain the second structure vector of the current transformation data as the structure representation.

[0077] Simultaneously, the central structure vector at the geometric center of each group is obtained, and the similarity between each central structure vector and the second structure vector is calculated. The group with the maximum similarity is taken as the group most similar to the second structure vector.

[0078] As an example, in this embodiment of the invention, the maximum similarity is the minimum value of the Euclidean distance. In other embodiments, other methods that can calculate vector similarity, such as cosine similarity, can also be used.

[0079] 2. Determine if the current conversion data is abnormal.

[0080] The keyword information of the current conversion data is compared with the keyword information of each historical data in the selected cluster set using cosine similarity s. n The calculation is performed where n represents the keyword information of the nth historical data in the cluster set, with a value range of [0,1]. The closer to 1, the more similar the data is, and the closer to 0, the less similar the data is.

[0081] Get the maximum value among multiple cosine similarities, max(s) n And set a similarity threshold m. s If max(s) n )≤m s If the current data is abnormal, it will be saved and an alert will be issued for subsequent manual verification; otherwise, it will be judged as normal data and the subsequent process will continue.

[0082] It should be noted that m sAs an example, the empirical threshold is set to m in this embodiment of the invention. s =0.6.

[0083] In summary, this embodiment of the invention collects historical conversion data of historical achievement conversion files, randomly occludes a term in the historical conversion data, calculates the semantic difference between the occluded conversion data and the historical conversion data, and selects the term corresponding to the smallest semantic difference as the first occluded term; expands the first occluded term to both sides, and then occludes the historical conversion data again until the semantic difference is greater than a preset threshold, and uses the occluded term at this point as the first occluded region; adds terms of the first occluded term, and occludes the historical conversion data again. When adding terms, iterates through unoccluded terms, and then selects the second occluded term. Continues to expand the second occluded term through terms until the smallest semantic difference is reached. If the semantic difference is greater than a preset threshold, the occluded word at that time is used as the second occluded region. A first structural vector is constructed based on the intersection-union ratio of the first and second occluded regions. The first structural vectors of all historical conversion data are clustered into multiple groups. The terms contained in the first structural vectors are removed from each historical conversion data to obtain the keywords of the historical conversion data. The second structural vector of the current conversion data is obtained, and the group most similar to the second structural vector is selected. The keywords of the current conversion data are compared with the keywords corresponding to each first structural vector in the group. When the maximum similarity obtained by the comparison is below the similarity threshold, the current conversion data is considered abnormal data. This embodiment of the invention can quickly match the unknown conversion files to the corresponding group and perform keyword comparison, thereby realizing the abnormal monitoring of conversion data.

[0084] This invention also proposes an intelligent monitoring system for technology transfer data based on artificial intelligence, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described method. Since the intelligent monitoring method for technology transfer data based on artificial intelligence has been described in detail above, it will not be repeated here.

[0085] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent monitoring of data on the transformation of scientific and technological achievements based on artificial intelligence, characterized in that, The method includes the following steps: Collect historical conversion data of historical achievement conversion files, randomly occlude a word in the historical conversion data, calculate the semantic difference between the occluded conversion data and the historical conversion data, and take the word corresponding to the smallest semantic difference as the first occluded word. Expand the first occluded word to both sides, and then occlude the historical conversion data until the semantic difference is greater than a preset threshold. Use the occluded word at this time as the first occluded area. Add a term to the first occluded word and occlude the historical conversion data. When adding a term, traverse the unoccluded terms and select the second occluded word. Continue to expand the term of the second occluded word until the minimum semantic difference is greater than the preset threshold. Use the occluded word at this time as the second occluded region. A first structural vector is constructed based on the intersection-union ratio of the first occlusion region and the second occlusion region; the first structural vectors of all the historical conversion data are clustered into multiple groups; the terms contained in the first structural vectors are removed from each historical conversion data to obtain the keywords of the historical conversion data; Obtain the second structure vector of the current conversion data, and select the group most similar to the second structure vector; compare the keywords of the current conversion data with the keywords corresponding to each of the first structure vectors in the group. When the maximum similarity obtained by the comparison is below the similarity threshold, the current conversion data is abnormal data. The steps for obtaining the first occluded area include: Centered on the first occluded word, the word length of the first occluded word is used as the expansion size to expand to both sides to obtain the first expanded word. The first expanded word is used to occlude the historical conversion data to obtain the first conversion data. The semantic difference between the word vectors of the historical conversion data and the first conversion data is calculated and compared with a preset threshold. When the semantic difference is not greater than the preset threshold, the expansion continues to expand to both sides according to the expansion size, centered on the first expanded word, until the semantic difference is greater than the preset threshold, to obtain the first occluded region. The steps for obtaining the second occlusion area include: A second expanded term is expanded by randomly adding a term to the first occluded word. The second expanded term is used to occlude the historical conversion data to obtain the second conversion data. The semantic difference between the word vectors of the historical conversion data and the second conversion data is calculated. Unoccluded words are traversed, and the occluded word corresponding to the minimum semantic difference is taken as the second occluded word. When it is not greater than a preset threshold, the second occluded word is expanded by randomly adding a term until the semantic difference is greater than the preset threshold, thus obtaining the second occluded region.

2. The method according to claim 1, characterized in that, The steps for collecting the historical conversion data include: Establish a historical document database for the transformation of scientific and technological achievements, using the subject text of each document as its identifier, and the word vector corresponding to the identifier as the historical transformation data.

3. The method according to claim 1, characterized in that, The calculation process for the semantic dissimilarity includes: The historical conversion data after the words are masked is reconstructed by an autoencoder to obtain the masked conversion data. The Euclidean distance between the corresponding word vectors of the historical conversion data and the masked conversion data is calculated as the semantic difference.

4. The method according to claim 1, characterized in that, The steps for constructing the first structure vector include: The intersection-union ratio is calculated based on the number of terms corresponding to the first occluded region and the second occluded region; Obtain the difference positions of the difference terms in the intersection and union of the first occluded region and the second occluded region, and construct the first structure vector based on the difference positions and the intersection-union ratio.

5. The method according to claim 1, characterized in that, The first structure vector, which represents all the historical transformation data, is clustered into multiple groups, including: The initial number of clusters is preset, all the first structure vectors are clustered to obtain multiple initial categories, the difference of the first structure vectors in each initial category is calculated, and then the category difference between all the initial category differences is obtained; Gradually increase the initial number of categories, calculate the category differences sequentially, select the number of categories corresponding to the smallest category difference as the ideal number, and use the clustering result corresponding to the ideal number as the grouping result.

6. The method according to claim 1, characterized in that, The selection of the group most similar to the second structure vector includes: Obtain the central structure vector at the geometric center of each group, calculate the similarity between each central structure vector and the second structure vector, and take the group with the maximum similarity as the group most similar to the second structure vector.

7. An intelligent monitoring system for technology transfer data based on artificial intelligence, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

Images