A method and system for identifying the integrated circuit industry chain based on knowledge graph

Abstract

The embodiment of the present invention provides an integrated circuit industry chain identification method and system based on a knowledge graph, the method comprising: establishing an integrated circuit industry chain ontology model; establishing an integrated circuit industry relationship graph; establishing an identification model for identifying whether a target enterprise belongs to the integrated circuit primary industry according to the integrated circuit industry chain ontology model and the integrated circuit industry relationship graph; determining whether the enterprise to be identified belongs to the integrated circuit primary industry according to the identification model. The embodiment of the present invention implements a method for constructing an integrated circuit industry chain graph and establishes a knowledge graph containing information on each industry link.

Description

Technical Field

[0001] The present invention relates to a knowledge graph, and in particular to a method and system for identifying an integrated circuit industry chain based on a knowledge graph. Background Art

[0002] The knowledge graph is essentially a semantic network, a graph-based data structure consisting of nodes and edges. In the knowledge graph, each node represents an "entity" that exists in the real world, and each edge is a "relationship" between entities.

[0003] An enterprise-level knowledge graph is actually the implementation of a graph in a certain industry field in an enterprise. The implementation of the industrial chain graph is mainly based on the relevant technologies of big data and artificial intelligence. It models the existing entities in the industry (industry classification or main products, etc.) and relationships (upstream and downstream relationships in the industry, the industry to which the enterprise belongs, the enterprise's main records, etc.) and the attributes of the entity relationships, and finally forms a complete set of professional industrial chain graphs.

[0004] Industrial chain is a concept in industrial economics. It is a chain-like relationship between various industrial sectors based on certain technical and economic connections, and objectively formed according to specific logical relationships and time-space layout relationships. Industrial chain is a concept that includes four dimensions: value chain, enterprise chain, supply and demand chain, and space chain. These four dimensions form an industrial chain in the equilibrium process of mutual docking. This "docking mechanism" is the internal model of the industrial chain. As an objective law, it regulates the formation of the industrial chain like an "invisible hand". The essence of the industrial chain is to describe a group of enterprises with some kind of internal connection. It is a relatively macro concept with two-dimensional attributes: structural attributes and value attributes. There are a large number of upstream and downstream relationships and mutual value exchanges in the industrial chain. The upstream link delivers products or services to the downstream link, and the downstream link feeds back information to the upstream link.

[0005] The industrial chain has many applications in commercial banks: including fundamental research to assist investment banking business and fundamental risk control; event transmission warning for credit business and identification of industrial credit concentration; upstream and downstream supply chain financial business in the marketing field and upstream and downstream business matchmaking.

[0006] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0007] The data on the industrial chain in commercial banks comes from the research on enterprises and industries by customer managers. It is manually determined and organized and has two main flaws: first, the coverage of enterprises is insufficient; second, the industry classification of the industrial chain is relatively simple. Summary of the invention

[0008] The present invention provides an integrated circuit industry chain identification method and system based on knowledge graph to solve the problem of the relationship between the integrated circuit industry chain and enterprises and the industry chain.

[0009] To achieve the above objectives, in a first aspect, an example of the present invention provides an integrated circuit industry chain identification method based on a knowledge graph, comprising:

[0010] Establish an integrated circuit industry chain ontology model;

[0011] Establish a relationship map for the integrated circuit industry;

[0012] According to the integrated circuit industry chain ontology model and the integrated circuit industry relationship map, establishing an identification model for identifying whether the target enterprise belongs to the integrated circuit primary industry;

[0013] According to the identification model, determine whether the enterprise to be identified belongs to the first-level integrated circuit industry.

[0014] In a second aspect, an embodiment of the present invention provides an integrated circuit industry chain identification system based on a knowledge graph, which includes:

[0015] The first building module is used to build an integrated circuit industry chain ontology model;

[0016] The second building module is used to build a relationship map of the integrated circuit industry;

[0017] A third establishment module is used to establish an identification model for identifying whether a target enterprise belongs to the integrated circuit primary industry according to the integrated circuit industry chain ontology model and the integrated circuit industry relationship map;

[0018] The first identification module is used to determine whether the enterprise to be identified belongs to the first-level integrated circuit industry according to the identification model.

[0019] Optionally, the identification system may further include: a fourth establishment module, used to establish an identification model for identifying whether the target enterprise belongs to the secondary industry of integrated circuits.

[0020] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the integrated circuit industry chain identification method based on the knowledge graph as described above.

[0021] In a fourth aspect, an embodiment of the present invention provides a computer device, comprising:

[0022] one or more processors;

[0023] A storage device for storing one or more programs;

[0024] When the one or more programs are executed by the one or more processors, the one or more processors implement the integrated circuit industry chain identification method based on the knowledge graph as described above.

[0025] The above technical solution has the following beneficial effects:

[0026] Through the method of the embodiment of the present invention, the scope of the primary and secondary industries of integrated circuits can be formed, enterprises in the integrated circuit industry can be accurately identified, and the coverage of enterprises in the primary and secondary industries of integrated circuits can be enriched, providing a strong basis for supply chain finance, industrial policy formulation and industrial risk prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 is a flow chart of an integrated circuit industry chain automatic identification method according to an embodiment of the present invention;

[0029] Figure 2 is a sample data acquisition flow chart of an embodiment of the present invention;

[0030] Figure 3 is a data preprocessing flow chart of an embodiment of the present invention;

[0031] Figure 4 This is a prediction effect evaluation diagram of a binary classification model for the integrated circuit primary industry according to an embodiment of the present invention;

[0032] Figure 5 is a processing flow chart of a multi-classification model as an example in an embodiment of the present invention;

[0033] Figure 6 This is an example diagram of using a text matching method in an embodiment of the present invention to quickly identify the secondary industry to which an enterprise belongs;

[0034] Figure 7 It is a functional block diagram of an integrated circuit industry chain identification system based on a knowledge graph according to an embodiment of the present invention;

[0035] Figure 8 It is a functional block diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] The integrated circuit industry chain identification method described in the present invention is based on the industry chain knowledge graph data constructed by commercial banks through big data and artificial intelligence technology, from which the relationship between enterprises and industries is mined to make up for the two deficiencies mentioned above.

[0038] In the era of big data, knowledge graphs have been used more and more widely, because graph databases such as knowledge graphs bring about the relationship between research objects, which is information that is difficult to extract from previous table databases. Enterprises do not exist independently, but always exist in a community network, such as a holding relationship network, an industrial supply chain relationship network, etc. Even if enterprises with similar fundamentals are in different positions in the network, they will have completely different properties. Therefore, if the enterprise under study can be put into its own network for analysis, the enterprise can be more accurately portrayed, and the interactive analysis of enterprise group events can also be better. For example, if a company in a holding relationship network has financial fraud and its stock price collapses, what impact will it have on other related companies? Or if a raw material or supplier changes in an industrial chain network, what impact will it have on the entire network? How is the event risk transmitted through what path, and how far is the transmission range? These predictions can enable knowledge graph users to have an overall grasp of risk events before the market, which is of great significance in the field of finance and capital markets. The basis for establishing the above predictions is to establish a knowledge graph that contains information from various industrial links.

[0039] At present, the main problem in establishing industrial knowledge graphs is that the upstream and downstream relationships of the integrated circuit and other industrial chains are numerous, and there are also many companies involved. How to establish the connection between enterprises and industrial chain nodes is one of the main technical problems solved by the embodiments of the present invention. The embodiments of the present invention study the construction method of the integrated circuit industry chain graph and establish a knowledge graph containing information on each industrial link.

[0040] The industrial chain is centered on products, so the knowledge graph expands outward from the integrated circuit as the center. At one point, it is the production process of the product, including IC design, wafer manufacturing, packaging and testing, etc. The subsequent expansion is the detailed products and process composition of each production link. Finally, the leaf nodes of the graph are the participating companies in the industrial chain. It is centered on products, but the ultimate focus is on the companies in the industrial chain. In this way, all node attributes, edge types, edge weights, edge directions and other information can be stored through the knowledge graph, and data can be retrieved as required.

[0041] The embodiment of the present invention is to automatically generate an integrated circuit knowledge graph based on the enterprise of industrial and commercial data. If the existing data is complete, then all the data can be directly put into the integrated circuit industry chain knowledge graph. The industrial and commercial data has a large number of enterprises whose industry chains and secondary industry affiliations are missing, which makes the industry chain knowledge graph incomplete. Therefore, the embodiment of the present invention determines whether the enterprise is an integrated circuit. If it is an integrated circuit, it will again determine which secondary industry of the integrated circuit it belongs to. Figure 1 1 is a flow chart of an automatic identification method of an integrated circuit industry chain based on a knowledge graph provided by an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0042] Step S1: Establish an integrated circuit industry chain ontology model.

[0043] In this step, the integrated circuit industry chain ontology model defines two entities and two relationships. The two entities include industries and enterprises, and the two relationships include parent-child industry relationships and enterprise-industry relationships.

[0044] Step S2: Establish an integrated circuit industry relationship map.

[0045] In this step, based on the Shenwan industry sub-categories and / or global industry classification standards, an integrated circuit industry entity can be established, which includes primary and secondary industries, and the parent-child industry relationship or upstream-downstream relationship between industries is established to form an integrated circuit industry relationship map. It reflects the common impact of one or more industries on one or more other industries, as well as the chain effect of multiple industries forming a chain.

[0046] S2.1. Establish industry entity nodes (as shown in Table 1):

[0047] Table 1

[0048]

[0049]

[0050] S2.2. Establish industry relationships, including parent-child industry relationships or upstream and downstream industry relationships (as shown in Table 2).

[0051] Table 2

[0052] Upstream Industry Downstream industries Information materials Discrete Device IDM Information materials Refrigeration devices Information materials Semiconductor Manufacturing Semiconductor Equipment Software Semiconductor Manufacturing Chip Design Integrated circuit packaging test Discrete Device IDM Integrated circuit packaging test Refrigeration devices Integrated circuit packaging test Semiconductor Manufacturing Integrated circuit packaging test

[0053] Step S3: Establish a model to automatically identify whether an enterprise belongs to the first-level integrated circuit industry.

[0054] Through the full amount of enterprise data disclosed by the Industrial and Commercial Bureau, a binary classification machine learning recognition model is established to automatically identify enterprises belonging to the first-level integrated circuit industry. The main recognition methods are as follows:

[0055] S3.1. Positive and negative sample definition process.

[0056] Figure 2 is a sample data acquisition flow chart of an embodiment of the present invention. Figure 2 As shown in the figure, the data sources come from manually labeled enterprise data and industrial and commercial enterprise data. Among the two types of data, those with clear integrated circuit labels (first-level industry labels) are defined as positive samples, and the enterprise belongs to the first-level integrated circuit industry; enterprises that are identified as non-first-level integrated circuit industries and have industry labels, and are clearly not in the integrated circuit industry, are defined as negative samples, and the enterprise belongs to other industries; enterprises with missing industry labels in industrial and commercial enterprise data, that is, enterprises without industry labels, are data to be predicted. Since the amount of negative sample data is much larger than that of positive samples, data is obtained comprehensively for positive samples, and negative samples are sampled according to a certain ratio of positive and negative samples.

[0057] S3.2, feature data preprocessing process. The feature processing step performs data feature preprocessing on the first sample data including the positive sample and the negative sample, and the data feature preprocessing refers to vectorization processing on the text feature data.

[0058] Figure 3 FIG. 1 is a flowchart of a method for vectorizing text feature data according to an embodiment of the present invention. Figure 3 As shown, the text vectorization process may include the following steps:

[0059] S3.2.1, word segmentation processing: use the word segmentation algorithm to perform word segmentation processing on the business description text. This step performs word segmentation on the business description text, and various words are obtained after word segmentation for subsequent text vectorization processing. The accuracy of the word segmentation is found to be about 97% through sampling test; a variety of word segmentation modules can be used here to perform word segmentation processing on the text, including but not limited to the use of the jieba module.

[0060] S3.2.2, stop word filtering: Use the stop word list to filter the words obtained by word segmentation. In this step, all Chinese and English punctuation marks, elements that are single characters after word segmentation, mainly prepositions, auxiliary words, etc., and words with a frequency of less than 10 times in the entire word library are deleted as stop words.

[0061] S3.2.3, word_to_ix text encoding processing: Use the word_to_ix encoding table to number all the words that appear in the business description text. This step maps words into digital numbers, and numbers all the words that appear in the business description text, which is convenient for subsequent text vectorization processing. In addition, "Eng", "Digit", "Un", and "0" are added and numbered in the table. "Eng" and "Digit" represent the English and numbers that appear in the business description text. The third one represents "Un". If some words appear in the data to be predicted but not in the training data, they are replaced with the "Un" symbol. The last one is used for length filling, a total of 36697 non-repeating words, numbered 1-36697 respectively; the word_to_ix encoding table is used to encode Chinese text data into an index table of 1 to N numbers, so as to facilitate the processing of Chinese into vector space processing. Use the numbers in the word_to_ix encoding table to replace the corresponding words of the business description text after word segmentation and stop word filtering.

[0062] S3.2.4, text length truncation and padding: The business description text after text encoding is truncated and padded so that each business description text after processing includes the same number of words. In this step, since the data entering the model is a matrix, the number of columns must be equal, and the length of each business description is different, so truncation and padding are performed. Because the average length of the business description text is 91 words, the columns of the input matrix are set to 100 words, that is, if a business description text exceeds 100 words, it will be truncated to 100, and if it is less than 100 words, it will be padded with 0 to the 100th word;

[0063] S3.2.5, text vectorization processing: perform text vectorization processing on the data obtained after text length truncation and padding to form a two-dimensional matrix. For example, a 100*36697 two-dimensional matrix feature is formed as sample data input. In this embodiment, the text feature vectorization representation can adopt any of the following methods: bag of words model, one-hot, word2vec, Glove, Emlo, Bert.

[0064] S3.2.6, after determining the positive sample label, negative sample label and the two-dimensional matrix features obtained through feature data preprocessing, the second sample data including the positive sample label, negative sample label and two-dimensional matrix is ​​divided into a training set and a test set according to a certain ratio. In this step, the positive sample and the negative sample bring in the label value of the training sample, and the two-dimensional matrix feature data is the feature variable, which together constitute the training data for the machine learning model.

[0065] S3.3. Training of the binary classification recognition model for the integrated circuit primary industry.

[0066] In this training of the integrated circuit primary industry binary classification model, the following hyperparameters were iterated through the grid search method to find their optimal configuration combination:

[0067] learning rate: learning rate, the step size control coefficient in gradient descent, set to 0.02;

[0068] max_depth: The maximum depth of the tree. Since Xgboost is a tree ensemble model, controlling the depth of the tree can control the degree of model fitting. Insufficient depth may lead to underfitting, while too large depth may lead to overfitting. It is set to 16 in this training;

[0069] n_estimator: the total number of iterations, i.e. the number of decision trees, is set to 120;

[0070] silent: specifies whether to output intermediate results. When silent = False, the intermediate process is not output (default); when silent = True, the intermediate process is output. Set to True;

[0071] objective[default=reg:linear], defines the learning task and the corresponding learning goal. The default is linear regression, but this model is a probability prediction for a binary classification problem, so the parameter is set to "binary:logistic", which means a binary classification logistic regression problem with output as probability.

[0072] S3.4. Model evaluation.

[0073] Figure 4 This is a prediction effect evaluation diagram of the integrated circuit primary industry binary classification model according to an embodiment of the present invention. Figure 4The classification result of the xgboost model shows that the AUC value is close to 0.82. The AUC value is the sum of the model accuracy and recall rate. The value is between 0 and 1. The higher the value, the better the model effect. However, if it is too high, it means that the model may be overfitted. In the field of social sciences, the AUC value of the project is generally between 0.72 and 0.85, which is an acceptable range. If it exceeds 0.9, the possibility of overfitting is very high. Of course, in the industrial field, the AUC value exceeds 0.97, which is also a normal range. Looking at the accuracy of the positive sample alone, it is 0.879 and the recall rate is 0.738, indicating that 73.8% of the positive samples are recognized by the model, and the rest are missed by the model. Among the samples recognized by the model, the accuracy rate is 87.9%. The specific number of samples can be seen in the confusion matrix below. In particular, 1583 positive samples were mistakenly judged as negative samples by the model, and 616 negative samples were mistakenly judged as positive samples by the model, and the rest were judged accurately. Support indicates the probability that items X and Y appear in the total data set at the same time.

[0074] S4. Model prediction: Based on the identification model, determine whether the enterprise to be identified belongs to the primary industry of integrated circuits.

[0075] In this step, the trained binary classification model can be used to perform model prediction on the predicted data, and enterprises with a probability greater than a certain prediction probability threshold can be set as enterprises in the first-level integrated circuit industry, so that all enterprises belonging to the first-level integrated circuit industry in the industrial and commercial enterprise data can be identified.

[0076] In some optional embodiments, the embodiments of the present invention further include the following processing steps:

[0077] Step S5: Establish a multi-classification model to identify the secondary industry of integrated circuits.

[0078] Figure 5 FIG. 1 is a processing flow chart of a multi-classification model as an example of an embodiment of the present invention. Figure 5 As shown, it includes the following processing steps:

[0079] S5.1. Use the text matching method to quickly identify the secondary industry label of integrated circuits to which the enterprise belongs, and use it as a training sample for the machine learning multi-classification model.

[0080] In the previous steps, all companies belonging to the primary industry of integrated circuits have been identified. Next, we will first use the text matching method to quickly identify the secondary industry of integrated circuits and establish sample labels for the machine learning multi-classification model.

[0081] The basic principle of text matching method: A subdivided industry is usually an entity such as semiconductor refrigeration devices, or an entity and a verb: semiconductor discrete device manufacturing. The text matching method only needs to match the entity and the verb.

[0082] Synonyms: chip, integrated circuit, semiconductor; design, research and development, development; production, manufacturing; materials, raw materials, chemicals.

[0083] Special note: [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]* represents Chinese or English and comma or bracket matches 0 or more times, [\uff1a\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]* represents Chinese or English and colon, comma or bracket matches 0 or more times. The reason for using it is that the business scope contains entities listed by comma followed by a verb or verb followed by a colon to list multiple entities. For example, Figure 6 As shown, Figure 6 This is an example diagram of using the text matching method to quickly identify the secondary industry to which an enterprise belongs in an embodiment of the present invention.

[0084] Information material:

[0085] Belong to the following industrial and commercial industries:

[0086] 26Chemical raw materials and chemical products manufacturing industry.

[0087] 261 Basic chemical raw materials manufacturing.

[0088] 2619 Manufacture of other basic chemical raw materials.

[0089] 2669 Manufacture of other special chemical products.

[0090] 3985 Manufacturing of special electronic materials.

[0091] Also matches:

[0092] Electronics [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Raw Materials|Electronics [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Materials|Electronics [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Chemicals|Semiconductors [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Raw Materials|Semiconductors [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Materials|Semiconductors [\u4e00-\u9fa5a-zA-Z\u3001\uff08\uff09]*Chemicals.

[0093] Electronic information chemical raw materials:

[0094] All information material companies except semiconductor silicon materials and semiconductor packaging materials.

[0095] Semiconductor silicon material, matching word: silicon.

[0096] Semiconductor packaging materials, matching word: packaging.

[0097] S5.2. Based on the machine learning multi-classification method and the above training samples, a multi-classification model for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs is established, so as to accurately identify the secondary industry classification of all integrated circuit enterprises. The positive samples of the secondary industry of integrated circuits marked in step S5.1 provide sample input for the machine learning algorithm of step S5.2.

[0098] S5.2.1. Steps for defining samples of multi-classification machine learning models and steps for defining enterprise data to be predicted.

[0099] Definition of multi-classification model training samples: Enterprises belonging to the first-level integrated circuit industry in the industrial and commercial data, and enterprises defined in the second-level industry are identified through the above-mentioned text matching method.

[0100] Definition of enterprise data to be predicted: Enterprises derived from industrial and commercial data, which are judged by the model to be in the first-level integrated circuit industry, but whose second-level industries are not defined.

[0101] S5.2.2. Establish a multi-classification machine learning model.

[0102] For the data after data feature preprocessing, hyperparameter setting and model training are performed according to the Xgboost multi-classification model.

[0103] booster: gbtree, optional parameters are gbtree, dart, gblinear. The first two are tree models, and the last one is a linear model.

[0104] Objective: multi:softmax, multi-classification problem.

[0105] num_class:20, the number of categories. This model training is for 20 integrated circuit secondary industries.

[0106] Gamma: 0.05, a parameter used to control whether to perform post-pruning. The larger the value, the more conservative it is and the less likely it is to overfit. However, if the value is too large, there is a risk of underfitting.

[0107] max_depth: 12, the depth of the tree to be built. The larger the depth, the easier it is to overfit.

[0108] subsample:0.4, randomly sample training samples, that is, row sampling when generating trees.

[0109] colsample_bytree: 0.7, column sampling when generating a tree.

[0110] min_child_weight: 12, controls the complexity of the model. If it is too small, it may be underfitting, and if it is too large, it may be overfitting. A trade-off must be made.

[0111] silent:1, if set to 1, no running information will be output.

[0112] eta:0.005, learning rate, learning rate.

[0113] seed:700, random seed, and then the model can be reproduced.

[0114] nthread: 8, number of cup threads.

[0115] S5.2.3. Model evaluation.

[0116] Since the confusion matrix of the model evaluation for multi-classification problems is too complex, the model evaluation method of the present invention directly performs sampling tests on the predicted data to check the accuracy of the model standard in the sampled objects.

[0117] The integrated circuit industry chain identification effect of the embodiment of the present invention is as follows:

[0118] The method used in the present invention can greatly expand the scale of enterprises in various industries of integrated circuits, providing a strong basis for supply chain finance, industrial policy formulation and industrial risk prevention.

[0119]

[0120] S5.3. Multi-classification model prediction.

[0121] Through the trained multi-classification model, the predicted data is predicted by the model, and the enterprises with a probability greater than a certain prediction threshold are set as enterprises in the secondary industry of integrated circuits, and all enterprises belonging to the secondary industry of integrated circuits in the industrial and commercial enterprise data are identified. The relationship between enterprises and industries is established to supplement the industrial chain map.

[0122] Figure 7 is a functional block diagram of an integrated circuit industry chain identification system based on a knowledge graph according to an embodiment of the present invention. Figure 7 As shown, the identification system includes:

[0123] The first building module is used to build an integrated circuit industry chain ontology model;

[0124] The second building module is used to build a relationship map of the integrated circuit industry;

[0125] A third establishment module is used to establish an identification model for identifying whether a target enterprise belongs to the integrated circuit primary industry according to the integrated circuit industry chain ontology model and the integrated circuit industry relationship map;

[0126] The first identification module is used to determine whether the target enterprise to be identified belongs to the first-level integrated circuit industry according to the identification model.

[0127] In some embodiments, the above-mentioned first establishment module can be specifically used to: establish an integrated circuit industry chain ontology model, and the integrated circuit industry chain ontology model includes: two entities and two relationships; the two entities include: industry and enterprise; the two relationships include: parent-child industry relationship, enterprise and industry relationship.

[0128] In some embodiments, the above-mentioned second establishment module can be specifically used to: establish an integrated circuit industry entity according to the industry classification standard, the integrated circuit industry entity includes a primary industry and a secondary industry, and establish a parent-child industry relationship between industries to form an integrated circuit industry relationship map.

[0129] In some embodiments, the third establishment module mentioned above can be specifically used to: establish a binary classification machine learning recognition model based on industrial and commercial data to identify whether the enterprise belongs to the first-level industry of integrated circuits.

[0130] In some embodiments, the third establishing module may be specifically used to:

[0131] Definition of positive and negative samples: Positive samples and negative samples are obtained from data sources including manually labeled enterprise data and industrial and commercial enterprise data; the industry label corresponding to the positive sample is the integrated circuit level industry label, and the industry label corresponding to the negative sample is not the integrated circuit level industry label;

[0132] Feature processing: performing data feature preprocessing on the first sample data including the positive sample and the negative sample, wherein the data feature preprocessing refers to performing vectorization processing on the text feature data; wherein the vectorization processing includes:

[0133] Use word segmentation algorithm to segment the business description text;

[0134] Use the stop word list to filter the words obtained by the word segmentation process;

[0135] Using the word_to_ix encoding table to number all words that appear in the business description text; the word_to_ix encoding table is used to encode Chinese text data into an index table of 1 to N numbers; using the numbers in the word_to_ix encoding table, replacing the corresponding words in the business description text after word segmentation and stop word filtering;

[0136] Performing text length truncation and padding on the business operation description texts after text encoding processing, so that each business operation description text after processing includes the same number of words;

[0137] Perform text vectorization on the data obtained after text length truncation and padding to form a two-dimensional matrix;

[0138] After obtaining the second sample data including the positive sample, the negative sample and the two-dimensional matrix, dividing the second sample data into a training set and a test set according to a preset ratio;

[0139] Model training: performing integrated circuit binary classification machine learning recognition model training according to the classification algorithm, the training set and the test set, iterating multiple hyperparameters through the grid search method to determine the optimal configuration combination of the multiple hyperparameters.

[0140] In some embodiments, the classification algorithm includes: an Xgboost classification algorithm, a logistic regression classification algorithm, or a random forest classification algorithm.

[0141] In some embodiments, the identification system also includes: a fourth establishment module, used to establish a multi-classification model for identifying the secondary industry of integrated circuits; and a second identification module, used to determine the secondary industry of the integrated circuit industry to which the enterprise to be identified belongs based on the multi-classification model.

[0142] In some embodiments, the fourth establishment module mentioned above can be specifically used to: identify the integrated circuit secondary industry label to which the enterprise belongs based on the text matching method as a training sample for the machine learning multi-classification model; and establish a multi-classification model for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs based on the machine learning multi-classification method and the training sample.

[0143] In some embodiments, the fourth establishment module can be specifically used to: obtain training samples, all training samples of the multi-classification model are derived from the industrial and commercial data that belong to the integrated circuit primary industry and are labeled in the secondary industry through text matching, and the data to be predicted are derived from the industrial and commercial data and are determined by the multi-classification model to belong to the integrated circuit primary industry, but the secondary industry is unclear; for the data after data feature preprocessing, multiple hyperparameters of the multi-classification model are set according to the Xgboost algorithm.

[0144] The embodiment of the present invention also provides a computer device, such as Figure 8 As shown, it includes one or more processors 301 , a communication interface 302 , a memory 303 and a communication bus 304 , wherein the processor 301 , the communication interface 302 , and the memory 303 communicate with each other via the communication bus 304 .

[0145] Memory 303, used for storing computer programs;

[0146] The processor 301 is used to implement the following steps when executing the computer program stored in the memory 303: establish an integrated circuit industry chain ontology model; establish an integrated circuit industry relationship map; establish an identification model for identifying whether an enterprise belongs to the integrated circuit primary industry based on the integrated circuit industry chain ontology model and the integrated circuit industry relationship map; determine whether the enterprise to be identified belongs to the integrated circuit primary industry based on the identification model.

[0147] In some embodiments, when the computer program is executed by the processor 301 to establish the integrated circuit industry chain ontology model, the processor 301 specifically performs the following steps: establish the integrated circuit industry chain ontology model, and the integrated circuit industry chain ontology model includes: two entities and two relationships; the two entities include: industry and enterprise; the two relationships include: parent-child industry relationship, enterprise and industry relationship.

[0148] In some embodiments, when the computer program is executed by the processor 301 to establish the step of establishing the integrated circuit industry relationship map, the processor 301 specifically performs the following steps: according to the industry classification standard, an integrated circuit industry entity is established, and the integrated circuit industry entity includes a primary industry and a secondary industry, and a parent-child industry relationship between industries is established to form an integrated circuit industry relationship map.

[0149] In some embodiments, when the computer program is executed by the processor 301 to establish a recognition model for identifying whether an enterprise belongs to the first-level industry of integrated circuits, the processor 301 specifically performs the following steps: based on industrial and commercial data, a binary classification machine learning recognition model is established to identify whether the enterprise belongs to the first-level industry of integrated circuits.

[0150] In some embodiments, when the computer program is executed by the processor 301 to establish a binary classification machine learning recognition model based on the industrial and commercial data to identify whether the enterprise belongs to the integrated circuit primary industry, the processor 301 specifically performs the following steps:

[0151] Definition of positive and negative samples: Positive samples and negative samples are obtained from data sources including manually labeled enterprise data and industrial and commercial enterprise data; the industry label corresponding to the positive sample is the integrated circuit level industry label, and the industry label corresponding to the negative sample is not the integrated circuit level industry label;

[0152] Feature processing: performing data feature preprocessing on the first sample data including the positive sample and the negative sample, wherein the data feature preprocessing refers to performing vectorization processing on the text feature data; wherein the vectorization processing includes:

[0153] Use word segmentation algorithm to segment the business description text;

[0154] Use the stop word list to filter the words obtained by the word segmentation process;

[0155] Using the word_to_ix encoding table to number all words that appear in the business description text; the word_to_ix encoding table is used to encode Chinese text data into an index table of 1 to N numbers; using the numbers in the word_to_ix encoding table, replacing the corresponding words in the business description text after word segmentation and stop word filtering;

[0156] Performing text length truncation and padding on the business operation description texts after text encoding processing, so that each business operation description text after processing includes the same number of words;

[0157] Perform text vectorization on the data obtained after text length truncation and padding to form a two-dimensional matrix;

[0158] After obtaining the second sample data including the positive sample, the negative sample and the two-dimensional matrix, dividing the second sample data into a training set and a test set according to a preset ratio;

[0159] Model training: performing integrated circuit binary classification machine learning recognition model training according to the classification algorithm, the training set and the test set, iterating multiple hyperparameters through the grid search method to determine the optimal configuration combination of the multiple hyperparameters.

[0160] The classification algorithm may include: an Xgboost classification algorithm, a logistic regression classification algorithm, or a random forest classification algorithm.

[0161] In some embodiments, when the computer program is executed by the processor 301, the processor 301 also performs the following steps: establishing a multi-classification model for identifying the secondary industry of integrated circuits; and determining the secondary industry of the integrated circuit industry to which the enterprise to be identified belongs based on the multi-classification model.

[0162] In some embodiments, when the computer program is executed by the processor 301 to establish a multi-classification model for identifying the secondary industry of integrated circuits, the processor specifically performs the following steps: based on the text matching method, identifying the secondary industry label of the integrated circuit to which the enterprise belongs as a training sample for the machine learning multi-classification model; based on the machine learning multi-classification method and the training samples, establishing a multi-classification model for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs.

[0163] In some embodiments, when the computer program is executed by the processor 301 to establish a multi-classification model for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs according to the machine learning multi-classification method and the training samples, the processor 301 specifically performs the following steps: obtaining training samples, all training samples of the multi-classification model are derived from the industrial and commercial data that belong to the first-level industry of integrated circuits and have labeled secondary industries through text matching, and the data to be predicted are derived from the industrial and commercial data and are determined by the multi-classification model to belong to the first-level industry of integrated circuits, but the secondary industry is unclear; for the data after data feature preprocessing, multiple hyperparameters of the multi-classification model are set according to the Xgboost algorithm.

[0164] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used for communication between the above electronic device and other devices.

[0165] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0166] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0167] An embodiment of the present invention also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above-mentioned integrated circuit industry chain identification methods based on the knowledge graph are implemented.

[0168] Those skilled in the art may also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention may be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units, and steps described above have generally described their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present invention.

[0169] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0170] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, and readable storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0171] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for identifying the integrated circuit industry chain based on knowledge graph, characterized in that: include: Establish an integrated circuit industry chain ontology model; Establish a relationship map for the integrated circuit industry; According to the integrated circuit industry chain ontology model and the integrated circuit industry relationship map, establishing an identification model for identifying whether an enterprise belongs to the integrated circuit primary industry; According to the identification model, determining whether the enterprise to be identified belongs to the first-level integrated circuit industry; Establish an identification model for identifying whether an enterprise belongs to the first-level integrated circuit industry, including: Based on the industrial and commercial data, a binary classification machine learning recognition model is established to identify whether the enterprise belongs to the first-level integrated circuit industry; Based on the industrial and commercial data, a binary classification machine learning recognition model is established to identify whether the enterprise belongs to the first-level integrated circuit industry, including: Definition of positive and negative samples: Positive samples and negative samples are obtained from data sources including manually labeled enterprise data and industrial and commercial enterprise data; the industry label corresponding to the positive sample is the integrated circuit level industry label, and the industry label corresponding to the negative sample is not the integrated circuit level industry label; Feature processing: performing data feature preprocessing on the first sample data including the positive sample and the negative sample, wherein the data feature preprocessing refers to performing vectorization processing on the text feature data; wherein the vectorization processing includes: Use word segmentation algorithm to segment the business description text; Use the stop word list to filter the words obtained by the word segmentation process; Using the word_to_ix encoding table to number all words that appear in the business description text; the word_to_ix encoding table is used to encode Chinese text data into an index table of 1 to N numbers; using the numbers in the word_to_ix encoding table, replacing the corresponding words in the business description text after word segmentation and stop word filtering; Performing text length truncation and padding on the business operation description texts after text encoding processing, so that each business operation description text after processing includes the same number of words; Perform text vectorization on the data obtained after text length truncation and padding to form a two-dimensional matrix; After obtaining the second sample data including the positive sample, the negative sample and the two-dimensional matrix, dividing the second sample data into a training set and a test set according to a preset ratio; Model training: performing integrated circuit binary classification machine learning recognition model training according to the classification algorithm, the training set and the test set, iterating multiple hyperparameters through a grid search method to determine the optimal configuration combination of the multiple hyperparameters; The step of establishing the stop word list includes: taking all Chinese and English punctuation marks, elements that are single characters after word segmentation, and words with a frequency of appearance less than 10 times in the entire vocabulary as stop words, and generating a stop word list based on the stop words; wherein the elements include: prepositions and auxiliary words.

2. The method according to claim 1, characterized in that The establishment of the integrated circuit industry chain ontology model specifically includes: An integrated circuit industry chain ontology model is established, wherein the integrated circuit industry chain ontology model includes: two entities and two relationships; the two entities include: industry and enterprise; the two relationships include: parent-child industry relationship, enterprise and industry relationship.

3. The method according to claim 1, characterized in that The establishment of the integrated circuit industry relationship map specifically includes: According to the industry classification standards, an integrated circuit industry entity is established, which includes primary industries and secondary industries, and parent-child industry relationships are established between industries to form an integrated circuit industry relationship map.

4. The method according to claim 1, characterized in that: The classification algorithm includes: Xgboost classification algorithm, logistic regression classification algorithm, or random forest classification algorithm.

5. The method according to claim 1, characterized in that Also includes: A multi-classification model for identifying the secondary industry of integrated circuits is established; based on the multi-classification model, the secondary industry of the integrated circuit industry to which the enterprise to be identified belongs is determined.

6. The method according to claim 5, characterized in that The establishment of a multi-classification model for identifying the secondary industry of integrated circuits specifically includes: Based on the text matching method, the integrated circuit secondary industry label to which the enterprise belongs is identified as a training sample for the machine learning multi-classification model; According to the machine learning multi-classification method and the training samples, a multi-classification model is established for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs.

7. The method according to claim 6, characterized in that The multi-classification model for identifying the secondary industry of the integrated circuit industry to which the enterprise belongs is established based on the machine learning multi-classification method and the training samples, specifically including: Obtain training samples. All training samples of the multi-classification model are derived from enterprises that belong to the integrated circuit primary industry in the industrial and commercial data and are labeled in the secondary industry through text matching. The data to be predicted are enterprises that are derived from the industrial and commercial data and are determined by the multi-classification model to belong to the integrated circuit primary industry, but the secondary industry is unclear; For the data after data feature preprocessing, multiple hyperparameters of the multi-classification model are set according to the Xgboost algorithm.

8. An integrated circuit industry chain identification system based on knowledge graph, characterized in that: include: The first building module is used to build an integrated circuit industry chain ontology model; The second building module is used to build a relationship map of the integrated circuit industry; A third establishment module is used to establish an identification model for identifying whether a target enterprise belongs to the integrated circuit primary industry according to the integrated circuit industry chain ontology model and the integrated circuit industry relationship map; A first identification module, used to determine whether the enterprise to be identified belongs to the integrated circuit primary industry according to the identification model; Establish an identification model for identifying whether an enterprise belongs to the first-level integrated circuit industry, including: Based on the industrial and commercial data, a binary classification machine learning recognition model is established to identify whether the enterprise belongs to the first-level integrated circuit industry; Based on the industrial and commercial data, a binary classification machine learning recognition model is established to identify whether the enterprise belongs to the first-level integrated circuit industry, including: Definition of positive and negative samples: Positive samples and negative samples are obtained from data sources including manually labeled enterprise data and industrial and commercial enterprise data; the industry label corresponding to the positive sample is the integrated circuit level industry label, and the industry label corresponding to the negative sample is not the integrated circuit level industry label; Feature processing: performing data feature preprocessing on the first sample data including the positive sample and the negative sample, wherein the data feature preprocessing refers to performing vectorization processing on the text feature data; wherein the vectorization processing includes: Use word segmentation algorithm to segment the business description text; Use the stop word list to filter the words obtained by the word segmentation process; Using the word_to_ix encoding table to number all words that appear in the business description text; the word_to_ix encoding table is used to encode Chinese text data into an index table of 1 to N numbers; using the numbers in the word_to_ix encoding table, replacing the corresponding words in the business description text after word segmentation and stop word filtering; Performing text length truncation and padding on the business operation description texts after text encoding processing, so that each business operation description text after processing includes the same number of words; Perform text vectorization on the data obtained after text length truncation and padding to form a two-dimensional matrix; After obtaining the second sample data including the positive sample, the negative sample and the two-dimensional matrix, dividing the second sample data into a training set and a test set according to a preset ratio; Model training: performing integrated circuit binary classification machine learning recognition model training according to the classification algorithm, the training set and the test set, iterating multiple hyperparameters through a grid search method to determine the optimal configuration combination of the multiple hyperparameters; The step of establishing the stop word list includes: taking all Chinese and English punctuation marks, elements that are single characters after word segmentation, and words with a frequency of appearance less than 10 times in the entire vocabulary as stop words, and generating a stop word list based on the stop words; wherein the elements include: prepositions and auxiliary words.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the integrated circuit industry chain identification method based on knowledge graph is implemented as described in any one of claims 1 to 7.

10. A computer device, characterized in that: It includes: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the integrated circuit industry chain identification method based on the knowledge graph as described in any one of claims 1-7.

Images