Method and device for predicting future industrial evolution mechanism based on temporal hypergraph motif

Abstract

The present invention discloses a method and device for predicting the evolution mechanism of future industries based on temporal hypergraph motifs, belonging to the technical field of big data intelligent processing, including: collecting the original cooperation data set among a number of institutions in the target industrial field and generating a cooperation event data set containing timestamps; constructing a temporal cooperation hypergraph; determining the true number of P types of target temporal hypergraph motifs pre-selected in the temporal cooperation hypergraph, where P≥1; creating a set of null model random hypergraphs and counting the average number of each type of target temporal hypergraph motif in the set of null model random hypergraphs; calculating the abundance of each type of target temporal hypergraph motif according to the true number and the average number; constructing a feature profile of the temporal cooperation hypergraph based on the abundance and analyzing the change trend of the feature profile to predict the evolution mechanism of the target industry. This application can analyze the evolution mechanism of future industries more meticulously and intuitively characterize the interaction patterns among multiple innovation entities in the entire industry.

Description

Technical Field

[0001] The present invention relates to the technical field of big data intelligent processing, and in particular, to a method and device for predicting the future industrial evolution mechanism based on temporal hypergraph motifs. Background Technique

[0002] With the rapid development of technology, future industries represented by quantum information technology, brain-computer interfaces, humanoid robots, etc. are becoming the strategic high points of global technological competition. Such industries are usually spawned by the industrialization of major scientific and technological innovation achievements and have three characteristics: subversive technological breakthroughs, reconstruction of industrial ecosystems, and leading strategic value. However, the development of these industries also faces many challenges, such as high uncertainty in technology routes, strong heterogeneity of innovation entities, and large resource demand and complex allocation.

[0003] Specifically, the existing research and methods have the following deficiencies in analyzing the collaborative innovation model of future industries:

[0004] 1. Ignoring the temporal hypergraph structure. The development of future industries highly depends on the breakthroughs of future technologies, and the R & D of these technologies often requires the joint participation of multiple institutions, forming complex cluster-based collaborations. Classical network structures may not be able to capture the emergent behaviors caused by cluster-based collaborations, while the temporal hypergraph structure can more accurately model the non-linearity and temporality of institutional collaborations in future industries. However, the existing methods for constructing temporal hypergraphs have limitations. For example, the temporal academic hypergraph network inference method based on Motif proposed in Chinese patent document CN117350385A constructs the temporal hypergraph in the form of discrete snapshots, ignoring the connections between different snapshots.

[0005] 2. Ignoring the role of network motifs. A motif is a local topological primitive of a network and plays an important role in promoting the evolution of the network (Reference: Science (2002), 298(5594), 824 - 827p). There are two limitations in the prior art. One is emphasizing the global structure. For example, Chinese Patent Document CN118656713A predicts future industrial directions by analyzing the weighted degree centrality of the industrial cluster network. However, networks with similar global structures may have different local structures, and networks with different global structures may also exhibit similar local structures (Reference: Science (2004), 303(5663), 1538 - 1542). Network motifs provide a good entry point for studying local structures. Compared with the macroscopic global perspective, network motifs can study the interaction relationships between institutions in more detail at the mesoscale level. The other is ignoring the hypergraph structure. For example, Chinese Patent Document CN115766476A proposes a network similarity evaluation method based on network motifs, but does not introduce a temporal hypergraph structure to model the institutional cooperation network in future industries. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for predicting the future industrial evolution mechanism based on temporal hypergraph motifs, so as to solve the problem that the prior art cannot analyze the evolution mechanism of future industries and the interaction patterns between multiple innovation entities in detail due to ignoring the temporal hypergraph structure and the role of network motifs when analyzing the future industrial collaborative innovation model.

[0007] To achieve the above object, the present application adopts the following technical solutions:

[0008] In the first aspect of the present application, a method for predicting the future industrial evolution mechanism based on temporal hypergraph motifs is provided, including the following steps:

[0009] Collect the original cooperation data set among several institutions in the target industrial field, and preprocess the original cooperation data set to generate a cooperation event data set containing timestamps;

[0010] Construct a temporal cooperation hypergraph according to the cooperation event data set, where nodes represent institutions, hyperedges represent cooperation events, and are attached with corresponding timestamp marks;

[0011] Use the counting algorithm to determine the actual number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, where P ≥ 1;

[0012] Create a set of null model random hypergraphs, and count the average number of each type of target temporal hypergraph motif in the set of null model random hypergraphs;

[0013] Calculate the abundance of each type of target temporal hypergraph motif based on its true number in the temporal cooperation hypergraph and its average number in the null model random hypergraph set;

[0014] Construct a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs, and analyze the changing trend of the feature profile to predict the evolution mechanism of the target industry.

[0015] In the second aspect of this application, a device for predicting the future industry evolution mechanism based on temporal hypergraph motifs is provided, including:

[0016] A preprocessing module, configured to collect the original cooperation data set among several institutions in the target industry field, and preprocess the original cooperation data set to generate a cooperation event data set including timestamps;

[0017] A hypergraph construction module, configured to construct a temporal cooperation hypergraph according to the cooperation event data set, where nodes represent institutions, hyperedges represent cooperation events, and are attached with corresponding timestamp marks;

[0018] A motif detection module, configured to use a counting algorithm to determine the true number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, where P≥1;

[0019] A statistics module, configured to create a null model random hypergraph set, and count the average number of each type of target temporal hypergraph motif in the null model random hypergraph set;

[0020] An abundance calculation module, configured to calculate the abundance of each type of target temporal hypergraph motif based on its true number in the temporal cooperation hypergraph and its average number in the null model random hypergraph set;

[0021] A prediction module, configured to construct a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs, and analyze the changing trend of the feature profile to predict the evolution mechanism of the target industry.

[0022] In the third aspect of this application, an electronic device is provided, including a memory and a processor, where the memory is used to store one or more computer instructions, and where the one or more computer instructions are executed by the processor to implement a method for predicting the future industry evolution mechanism based on temporal hypergraph motifs as described in any one of the above.

[0023] In the fourth aspect of this application, a computer-readable storage medium storing a computer program is provided, where the computer program causes a computer to execute to implement a method for predicting the future industry evolution mechanism based on temporal hypergraph motifs as described in any one of the above.

[0024] The present invention has the following beneficial effects:

[0025] 1. By introducing the temporal hypergraph structure, the assumption of pairwise interactions between nodes in the classical network structure is extended to higher-order interactions among multiple nodes, and the temporality and dynamics of these higher-order interactions are considered, which can analyze the evolution mechanism of future industries in more detail.

[0026] 2. By introducing network motifs, the future industrial cooperation network is deconstructed at the mesoscopic level. Compared with the previous analysis of the macroscopic global structure, the motif-based analysis can consider the influence of specific types of local topological structures on the network formation mechanism and network dynamics; at the same time, as a special subgraph, motifs can also intuitively represent the interaction patterns among multiple innovation entities in the entire industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0028] Figure 1 is a flowchart of a method for predicting the evolution mechanism of future industries based on temporal hypergraph motifs provided by an embodiment of the present application;

[0029] Figure 2 is a schematic structural diagram of a device for predicting the evolution mechanism of future industries based on temporal hypergraph motifs provided by an embodiment of the present application;

[0030] Figure 3 is a schematic diagram of an electronic device for implementing a method for predicting the evolution mechanism of future industries based on temporal hypergraph motifs provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] To make the technical solutions of the present application clearer, the following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments. The terms "first", "second", etc. in the claims and the description of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such terms can be interchanged under appropriate circumstances. This is only a way of distinguishing objects with the same attributes when describing the embodiments of the present application. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, so that a process, method, system, product or device including a series of units does not have to be limited to those units, but may include other units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0032] This embodiment provides a prediction method for the future industrial evolution mechanism based on temporal hypergraph motifs, as Figure 1 shown, which includes the following steps:

[0033] S110. Collect the original cooperation data set among several institutions in the target industrial field, and preprocess the original cooperation data set to generate a cooperation event data set containing timestamps.

[0034] Among them, collecting the original cooperation data set among several institutions in the target industrial field includes:

[0035] Determine the target industry and extract the domain keywords corresponding to the target industry;

[0036] Using the domain keywords as the retrieval condition, obtain the list of institutions related to the target industry;

[0037] Based on the list of institutions, collect the cooperation events among various institutions to form the original cooperation data set. The first step of data collection is to determine the future industry to be investigated, that is, the target industry, and then obtain the data that can characterize the cooperation relationship among different institutions in the target industry to form the original cooperation data set. Obviously, the cooperation relationship among institutions will affect the data collection result. Therefore, it is also necessary to determine the representation form of the cluster cooperation relationship. In this embodiment, the representation form of the cooperation relationship among institutions includes, but is not limited to, co-authoring a paper, jointly applying for a patent, jointly writing a book, or jointly participating in an industrial alliance.

[0038] Taking the co-authorship relationship as an example of the representation form, after determining the target industry and the cooperation form, first rely on the domain expert knowledge to determine the domain keywords of the target industry, and then based on the domain keywords, retrieve relevant papers through public data platforms such as OpenAlex, Dimensions, Web of Science, Scopus, and PubMed, and only retain specific types of literature to form the original cooperation data set, including journal papers (article), conference papers (conference), books (book), and book chapters (book chapter), etc. Here, the data collection method and the retained literature types are all related to the cooperation form among institutions.

[0039] After completing the retrieval, key information of papers is screened, such as title, abstract, journal name (conference name), author, affiliated institution, publication year, and Digital Object Identifier (DOI), etc., to preprocess the original cooperation dataset. Among them, DOI is a standardized coding system used to uniquely and permanently identify digital resources, similar to a "digital ID card", aiming to provide a globally unique and resolvable identifier for various objects in the digital environment (such as academic papers, e-books, research reports, datasets, etc.).

[0040] In other cooperation forms such as supply chain networks, production cooperation networks, and merger and acquisition networks, after determining the target industry, semantic analysis technology can be used to deeply analyze the literature materials related to the target industry, accurately extract the domain keywords corresponding to the target industry, and then, with the help of a network search engine or a professional industry database, using the domain keywords as the retrieval conditions, obtain a list of institutions closely related to the target industry. Then, based on the list of institutions, through multi-channel information collection, including news reports, corporate announcements, and data mining technology, etc., comprehensively obtain the cooperation events among various institutions to form the original cooperation dataset.

[0041] Among them, preprocessing the original cooperation dataset generates a cooperation event dataset containing timestamps, including:

[0042] Disambiguate the institution names to ensure that the same institution has a unified name in all cooperation events;

[0043] Assign corresponding timestamps to each cooperation event according to the preset time unit;

[0044] When multiple cooperation events have the same timestamp, configure a unique timestamp for the multiple cooperation events through random sorting.

[0045] The first step of data preprocessing is to disambiguate the institution names. In the original cooperation dataset, the same institution may correspond to multiple institution names, such as inconsistent abbreviations, former names, and incorrect signatures. At the same time, the signing methods of different levels of sub-institutions within the same institution may also be different, which is likely to lead to data chaos and analysis difficulties. Therefore, in data preprocessing, it is necessary to disambiguate the institution names. Specifically, first identify all the variant institution names in the original cooperation dataset, and then select a standard name for each institution. If the institution has levels, merge its sub-institutions into the same level, so as to ensure that the same institution mentioned in all cooperation events uses the same name, which is convenient for subsequent data analysis and visualization.

[0046] It should be noted here that since there must theoretically be an affiliated institution in the paper, before disambiguating the institution names, it is also necessary to screen out the data with missing affiliated institutions, and then retrieve them again in other public databases. If the missing institution information is obtained, it is used to fill in the missing values. If it cannot be obtained, the authors with the same name are retrieved from the collected data, and the non-missing affiliated institutions of these authors in other data are used to fill in the blanks. If it still cannot be obtained, this piece of data is deleted. That is to say, the supplementation of missing affiliated institutions is not a necessary step in data preprocessing, but is selected according to the actual situation.

[0047] Finally, a corresponding timestamp is assigned to each paper according to the year. Specifically, different time units such as days, weeks, months, years and corresponding time windows can be determined according to the occurrence frequency and change degree of the cooperation behavior between institutions. However, in any case, it must be ensured that each cooperation relationship has its corresponding timestamp. When the time unit is relatively large, such as years, there may be multiple cooperation relationship timestamps that are the same. In this case, a random sorting method is adopted to sort each identical timestamp and configure a unique timestamp. For example, if the timestamps of cooperation events A, B, and C are all 2025, the randomly assigned timestamps can be: cooperation event A corresponds to April 5, 2025; cooperation event B corresponds to April 6, 2025; cooperation event C corresponds to April 7, 2025.

[0048] Preprocessing operations such as disambiguating institution names and timestamp assignment have significantly improved the quality and consistency of the data, and provided a solid foundation for subsequent data analysis.

[0049] S120. Construct a temporal cooperation hypergraph based on the cooperation event dataset, where the nodes represent institutions, the hyperedges represent cooperation events, and are attached with corresponding timestamp marks.

[0050] Define the hypergraph according to the cluster cooperation relationship between institutions , where, represents the node set, and each node represents an institution, represents the hyperedge set, and , each element in represents a hyperedge, which is mathematically represented as a non-empty subset containing any number of nodes. Define the temporal hypergraph on the node set , which represents an ordered sequence of temporal hyperedges. In this embodiment, the temporal hypergraph is also called a temporal cooperation hypergraph, representing the cooperation relationship between institutions. In this temporal cooperation hypergraph, each temporal hyperedge is represented as , where the node set it contains , is the timestamp corresponding to the temporal hyperedge, that is, the time point when the cooperation event occurs. At the same time, the timestamp of each temporal hyperedge is unique. Taking the co-authorship relationship as an example, a paper is regarded as a hyperedge composed of its publishing institutions. The number of nodes included in this hyperedge is the number of publishing institutions of the paper. After adding the year as the timestamp, each hyperedge becomes a temporal hyperedge. If and only if they have the same node set, that is two temporal hyperedges and are regarded as duplicates. Define the set of temporal hyperedges containing the node set as . When ignoring the timestamps and duplicate temporal hyperedges, the temporal hypergraph becomes a static hypergraph . If and only if , .

[0051] Suppose that Institution 1 and Institution 2 jointly published Paper 1 in 2021, Institution 2, Institution 3 and Institution 4 jointly published Paper 2 in 2022, Institution 1 and Institution 3 jointly published Paper 3 in 2023, Institution 4 and Institution 5 jointly invented Paper 4 in 2024, and Institution 1, Institution 2, Institution 3 and Institution 4 jointly invented Paper 5 in 2025. Then, in the constructed temporal cooperation hypergraph , the node set , where correspond to Institution 1, Institution 2, Institution 3, Institution 4 and Institution 5 respectively. The hyperedge corresponding to Paper 1 has a timestamp of 2021; the hyperedge corresponding to Paper 2 has a timestamp of 2022; the hyperedge corresponding to Paper 3 has a timestamp of 2023; the hyperedge corresponding to Paper 4 has a timestamp of 2024; the hyperedge corresponding to Paper 5 has a timestamp of 2025.

[0052] Classical network structures can only model pairwise interactions between nodes, while hypergraphs can incorporate a wider range of group interactions into the model. Temporal hypergraphs further consider the dynamics of such group interactions and generalize the assumption of pairwise interactions between nodes in classical network structures to higher-order interactions among multiple nodes, enabling a more detailed analysis of the evolution mechanism of future industries.

[0053] In this embodiment, through the temporal cooperation hypergraph, it is possible to analyze how the cooperation patterns among different innovation entities evolve over time, and it is also possible to identify core institutions or key cooperation patterns.

[0054] S130. Determine the true number of the pre-selected P types of target temporal hypergraph motifs in the temporal cooperation hypergraph using a counting algorithm, where P ≥ 1.

[0055] After obtaining the temporal cooperation hypergraph, the true number of each type of target temporal hypergraph motif (temporal hypergraph motif, abbreviated as TH-motif) in the temporal cooperation hypergraph can be determined by the exact counting algorithm THyMe+ or the approximate counting algorithm THyMe-A*.

[0056] It should be noted here that due to the complexity of the hypergraph structure, after considering the temporality, only the temporal hypergraph motifs formed by three temporal hyperedges are detected. This temporal hypergraph motif describes the structure and temporal relationship presented by the sequence formed by three connected temporal hyperedges within a given time range. The structure and temporal relationship presented by the sequence formed by three connected temporal hyperedges.

[0057] When considering the temporal hypergraph motifs formed by three temporal hyperedges, after excluding redundant cases, there are a total of 96 motifs containing three temporal hyperedges. Given three connected temporal hyperedges , where , and , according to the number of repeated temporal hyperedges included, these motifs can be divided into three categories: (1) Ternary TH-motif, containing three non-repeated temporal hyperedges, that is ; (2) Binary TH-motif, containing two repeated temporal hyperedges, such as ; (3) Unary TH-motif, containing three repeated temporal hyperedges, that is . In this embodiment, the ternary temporal hypergraph motif, the binary temporal hypergraph motif, and the unary temporal hypergraph motif are all target temporal hypergraph motifs, that is, P = 3.

[0058] Then, perform motif quantity detection.

[0059] Among them, when the counting algorithm uses the exact counting algorithm based on temporal hypergraph projection, that is, THyMe+, to determine the true number of the pre-selected P types of target temporal hypergraph motifs in the temporal cooperation hypergraph, where P ≥ 1, it includes:

[0060] Given a time window δ and initialize the projection graph , where the node v ∈ corresponds to the set of hyperedges with the same node set in the temporal cooperation hypergraph , and the set of timestamps carried by each node v = , the edge ∈ if and only if ∩ ≠ 。

[0061] Update based on the insertion module and deletion module :

[0062] In the insertion module, given any temporal hyperedge , if there is no hyperedge with the same node set in the current time window, then the corresponding nodes are added as new nodes to , and after the insertion , find its adjacent nodes , that is, find the temporal hyperedges with the same nodes as in the temporal cooperation hypergraph , then create an associated connecting edge , finally, add the corresponding timestamp to .

[0063] In the deletion module, if the selected temporal hyperedge is the only temporal hyperedge with the node set in the current time window , then the nodes represented by it in and the associated connecting edges are deleted from and respectively. Before that, the timestamp is deleted from .

[0064] When the projection graph is updated, start the counting step. Given any temporal hyperedge , the THyMe+ algorithm counts the motifs containing in the time window , where is the last hyperedge in the motif, and THyMe+ looks for any two temporal hyperedges connected to in the time window

[0065] Next, calculate the number of ternary temporal hypergraph motifs, binary temporal hypergraph motifs, and unary temporal hypergraph motifs respectively.

[0066] First, calculate the number of ternary temporal hypergraph motifs. Given a set of three connected temporal hyperedges , because is the last hyperedge to appear, the set can be obtained from two sequences or , where , , since , , and , at this time the number of motifs can be calculated from and the number of combinations of the timestamps contained therein. Denote the number of 3 - time - series hyper - graph motifs as , , where represents the indicator function, and denote two types of 3 - time - series hyper - graph motifs that may be formed when is the last - occurring time - series hyper - edge, considering the timings of and and in the time - series hyper - edge set

[0067] Next, calculate the number of 2 - time - series hyper - graph motifs. There are two repeated hyper - edges in such motifs, which is equivalent to enumerating all adjacent edges in . Similar to calculating the 3 - time - series hyper - graph motifs, consider three sequences . , and , where , , since , , the number of motifs can also be obtained by calculating the number of combinations of the timestamps it contains, that is , and .

[0068] Finally, calculate the number of 1 - time - series hyper - graph motifs. Given the sequence , where , . Because , , and , at this time .

[0069] Assume that in the time - series cooperation hyper - graph , the node set , A, B, and C represent institutions, and the time - series hyper - edge set contains , , , , and set the time window δ = 3.

[0070] Initially, the projection graph It is empty.

[0071] The insertion module inserts the temporal hyperedges one by one: Insert , and check the current time window to see if there is already the same set of nodes . Since there are no other hyperedges currently, is added to , and is updated. , , ; Insert , and check the current time window to see if there is already the same set of nodes . Since there are no other hyperedges currently, is added to , and is updated. , , , ; Insert , and check the current time window to see if there is already the same set of nodes . Since there are no other hyperedges currently, is added to , and is updated. , , , , ; Since δ = 3, is not inserted here.

[0072] The deletion module checks the hyperedges outside the time window: When time advances, if the timestamp of a certain hyperedge exceeds the current time window, it is deleted from . Taking the current time t = 5 as an example, the time window is . The timestamp of the hyperedge is not within the window, so it is deleted, and is updated. , , , , . Inside the time window , count the motifs containing .

[0073] Taking as an example, first calculate the ternary temporal hyperedge motif:

[0074] The hyperedges adjacent to are respectively and calculate the possible sequences whose timestamp combinations = 2, = 3, = 4, ; and the possible sequence whose timestamp combinations = 3, = 2, = 4, .

[0075] Calculate the binary temporal hyperedge motif again: Check the case of duplicate hyperedges such as and calculate the number of timestamp combinations .

[0076] Finally, calculate the unary temporal hyperedge motif: For a single hyperedge , the number of timestamp combinations = .

[0077] For a real hypergraph, when the hypergraph scale is large, i.e., the number of hyperedges is large, or the time window spanned is long, the exact counting algorithm often takes a long time and has higher requirements for computing resources. At this time, according to the idea of random sampling, the THyMe-A* algorithm can be used to obtain an unbiased estimate of the number of each type of temporal hypergraph motif. The core of the THyMe-A* algorithm is to repeatedly sample time windows of length , and then accurately count the number of temporal hypergraph motifs in each time window according to the THyMe+ algorithm, and finally obtain an unbiased estimate of the number of temporal hypergraph motifs over the entire time window.

[0078] Specifically, when the counting algorithm uses an approximate counting method based on random sampling, the counting algorithm is used to determine the true number of P preselected target temporal hypergraph motifs in the temporal cooperation hypergraph, where P ≥ 1, including:

[0079] Set the sampling time window length ∆ and the number of sampling times s.

[0080] For each sampling, perform the following steps:

[0081] a. Randomly and uniformly select a temporal hyperedge from the set of temporal hyperedges ;

[0082] b. According to the timestamp t of the selected temporal hyperedge , randomly select a timestamp from within the time window ;

[0083] c. According to the timestamp , use the THyMe+ algorithm to accurately count each target temporal hypergraph motif formed within the sub-window to obtain ;

[0084] d. Calculate the probability that the temporal hyperedge appears in the time window , where is the number of hyperedges within the sub-window;

[0085] e. Adjust according to to obtain .

[0086] Sum up the estimators of the number of each type of temporal hypergraph motif obtained in each sampling , and then multiply by to obtain the final unbiased estimator of the number of each type of temporal hypergraph motif, which is also regarded as the true number of each type of target temporal hypergraph motif.

[0087] Still taking the above temporal cooperation hypergraph as an example, assume the time window , and set the number of samplings s = 10. Randomly and uniformly select a temporal hyperedge . For , its timestamp , so the time window is = . Then randomly select the timestamp , and the sub-window is . At this time, it is necessary to use the above THyMe+ algorithm to calculate the number of all possible temporal hypergraph motifs within this sub-window.

[0088] Taking the calculation of the number of ternary temporal hypergraph motifs as an example, check and within this sub-window , and calculate the combination number (because ), that is . Because within the sub-window contains , , so , , and then calculate to obtain .

[0089] The results obtained in each of the 10 samplings are similar. For example, if each , then the total estimator is ​​, that is to say, in this embodiment, the true number of ternary temporal hypergraph motifs is 2.

[0090] Deconstructing the future industrial cooperation network from the mesoscopic level, compared with the previous analysis of the macroscopic global structure, the motif-based analysis can consider the influence of specific types of local topological structures on the network formation mechanism and network dynamics; at the same time, as a special subgraph, motifs can intuitively represent the interaction patterns among multiple innovation entities in the entire industry.

[0091] In this embodiment, by analyzing which nodes frequently appear in various motifs, institutions or organizations that play key roles in the industrial development process can be identified. For example, in the scientific and technological field, certain research institutions may play an important role in promoting technological progress due to their extensive interdisciplinary cooperation.

[0092] S140. Create a set of null-model random hypergraphs and count the average number of each type of target temporal hypergraph motif in the set of null-model random hypergraphs.

[0093] Furthermore, use the null-model generation method to construct a set of null-model random hypergraphs, where each random hypergraph retains the basic statistical characteristics of the temporal cooperation hypergraph.

[0094] Calculate the number of occurrences of each type of target temporal hypergraph motif in each random hypergraph according to the above counting method and record the average value.

[0095] Use the Chung-Lu model or other suitable methods to generate a set of null-model random hypergraphs. Each set of hypergraphs retains the basic statistical characteristics (such as degree distribution) of the original temporal cooperation hypergraph, but is randomized in other aspects.

[0096] In each null-model random hypergraph, calculate the number of occurrences of each type of target temporal hypergraph motif through the above-mentioned exact counting algorithm THyMe+ or approximate counting algorithm THyMe-A*, and then calculate the average number of each type of target temporal hypergraph motif in all null-model random hypergraphs.

[0097] S150. Calculate the abundance of each type of target temporal hypergraph motif according to the true number of each type of target temporal hypergraph motif in the temporal cooperation hypergraph and the average number in the set of null-model random hypergraphs.

[0098] Through the formula Calculate the abundance of each type of temporal hypergraph motif, where represents the true number of the temporal hypergraph motif in represents the mean value of the temporal hypergraph motif generated in all random networks is an adjustment factor (usually set to 1) to avoid a zero denominator.

[0099] In this embodiment, the statistical significance of each type of target temporal hypergraph motif can also be determined through a preset p-value, so as to identify which cooperation patterns are significantly different from random expectations in the actual network. This method not only helps to understand the structural characteristics of complex networks, but also provides data-driven support for policy making.

[0100] Suppose motif types 1 and 2 are significantly more than random expectations in the actual network, indicating that these two cooperation patterns may be important characteristics of the industry. For example, motif type 1 is a "strong core" structure, and the institutions participating in cooperation and their mutual cooperation relationships remain unchanged within a given time window, indicating that there are stable industrial clusters in the industry; motif type 2 is a "core-periphery" structure, and some institutions maintain stable cooperation relationships within a given time window, while there are some temporarily participating institutions in the remaining stages, indicating that there are stable core institutions in the industry, and at the same time there are peripheral institutions playing auxiliary roles; although motif type 3 also has statistical significance, its p-value is close to the threshold, indicating that the importance of this cooperation pattern is relatively low. Based on this analysis, for cooperation patterns significantly more than random expectations such as the "strong core" structure and the "core-periphery" structure, these patterns can be further strengthened through policy support such as funding cross-institutional projects, and resources can be preferentially tilted towards institutions showing significant cooperation patterns to maximize innovation output.

[0101] S160. Construct a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs, and analyze the change trend of the feature profile to predict the evolution mechanism of the target industry.

[0102] Furthermore, calculate the normalized abundances of each type of target temporal hypergraph motif, and use the normalized abundances as dimensions to construct the feature profile of the temporal cooperation hypergraph;

[0103] Analyze the differences in the feature profile at different times to predict the evolution mechanism of the target industry.

[0104] Given that there may also be comparisons between different hypergraphs based on temporal hypergraph motifs, such as cross-industry comparisons or international comparisons within the same industry, it is necessary to construct the feature profile of a single temporal cooperation hypergraph. In this embodiment, for each temporal cooperation hypergraph, since there are 3 different types of target temporal hypergraph motifs, the feature profile of each temporal hypergraph can be represented by a 3-dimensional vector, and each dimension represents the normalized abundance of a type of target temporal hypergraph motif This normalization process ensures fair comparison of networks of different scales.

[0105] Conduct in-depth analysis on the constructed feature profiles, including but not limited to time series analysis, pattern recognition, and correlation analysis with other variables, etc., to understand the changing trends of feature profiles in different periods, and by analyzing the changes of feature profiles over time, reveal some potential laws or trends, and then predict the future evolution direction or mechanism of the target industry. For example, certain types of cooperation models may indicate the arrival of the technology innovation cycle or significant changes in the market structure, thereby providing reference opinions for policy-making.

[0106] For example, through analysis, it is found that there are obvious differences in the feature profiles before and after 2018. Before 2018, the network was mainly dominated by a few large institutions, forming a highly centralized cooperation model; after 2018, there was extensive cooperation among more small and medium-sized institutions, showing a more decentralized but active cooperation ecosystem. This change may indicate a major change in the market structure of this industry, that is, from monopoly to diversified competition. Based on this, policymakers focus on the development needs of small and medium-sized enterprises and create a fair competition environment for them to promote the healthy and stable development of the market.

[0107] As Figure 2 shown, this embodiment also provides a prediction device for the future industrial evolution mechanism based on time-series hypergraph motifs, including:

[0108] A preprocessing module, configured to collect the original cooperation data set among several institutions in the target industrial field, and preprocess the original cooperation data set to generate a cooperation event data set containing timestamps;

[0109] A hypergraph construction module, configured to construct a time-series cooperation hypergraph according to the cooperation event data set, where the nodes represent institutions, the hyperedges represent cooperation events, and are attached with corresponding timestamp marks;

[0110] A motif detection module, configured to use a counting algorithm to determine the true number of P types of target time-series hypergraph motifs preselected in the time-series cooperation hypergraph, P≥1;

[0111] A statistics module, configured to create a set of null model random hypergraphs and count the average number of each type of target time-series hypergraph motif in the set of null model random hypergraphs;

[0112] An abundance calculation module, configured to calculate the abundance of each type of target time-series hypergraph motif according to the true number of each type of target time-series hypergraph motif in the time-series cooperation hypergraph and the average number in the set of null model random hypergraphs;

[0113] A prediction module, configured to construct a feature profile of the time-series cooperation hypergraph based on the abundances of all target time-series hypergraph motifs, and analyze the changing trend of the feature profile to predict the evolution mechanism of the target industry.

[0114] This embodiment is used to implement the method provided in the above embodiment and has the corresponding beneficial effects of the above method. For the technical details not described in detail in this embodiment, reference may be made to the methods provided in all the foregoing embodiments of the present invention.

[0115] As Figure 3 shown, this embodiment further provides an electronic device, including a memory 301 and a processor 302. The memory 301 is used to store one or more computer instructions. Among them, the one or more computer instructions are executed by the processor 302 to implement the above-mentioned method for predicting the future industrial evolution mechanism based on temporal hypergraph motifs.

[0116] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described electronic device can refer to the corresponding process in the foregoing method embodiment and will not be elaborated herein.

[0117] This embodiment further provides a computer-readable storage medium storing a computer program, and the computer program, when executed by a computer, implements the above-mentioned method for predicting the future industrial evolution mechanism based on temporal hypergraph motifs.

[0118] Exemplarily, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the memory 301 and executed by the processor 302, and the I / O interface transmission of data is completed by the input interface 305 and the output interface 306 to complete the present invention. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0119] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device may include, but is not limited to, the memory 301 and the processor 302. Those skilled in the art can understand that this embodiment is only an example of the computer device and does not constitute a limitation on the computer device. It may include more or fewer components, or combine some components, or different components. For example, the computer device may further include an input device 307, a network access device, a bus, etc.

[0120] The processor 302 may be a Central Processing Unit (CPU), or may also be other general-purpose processors 302, Digital Signal Processors (DSPs) 302, Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor 302 may be a microprocessor 302 or the processor 302 may also be any conventional processor 302, etc.

[0121] The memory 301 may be an internal storage unit of the computer device, such as the hard disk or memory of the computer device. The memory 301 may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 301 may also include both the internal storage unit and the external storage device of the computer device. The memory 301 is used to store computer programs and other programs and data required by the computer device. The memory 301 may also be used to temporarily store data in the output device 308, and the aforementioned storage media include various media that can store program codes, such as USB flash drives, mobile hard disks, Read-Only Memory ROM 303, Random Access Memory RAM 304, diskettes, or optical discs.

[0122] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A prediction method for the future industrial evolution mechanism based on temporal hypergraph motifs, characterized in that Including the following steps: Collect the original cooperation data set among several institutions in the target industrial field, and preprocess the original cooperation data set to generate a cooperation event data set including timestamps; Construct a temporal cooperation hypergraph according to the cooperation event data set, where nodes represent institutions, hyperedges represent cooperation events, and are attached with corresponding timestamp marks; Use a counting algorithm to determine the true number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, P≥1; Create a set of null model random hypergraphs, and count the average number of each type of target temporal hypergraph motif in the set of null model random hypergraphs; Calculate the abundance of each type of target temporal hypergraph motif according to the true number in the temporal cooperation hypergraph and the average number in the set of null model random hypergraphs; Construct a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs, and analyze the change trend of the feature profile to predict the evolution mechanism of the target industry; The constructing a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs, and analyzing the change trend of the feature profile to predict the evolution mechanism of the target industry includes: For each temporal cooperation hypergraph, calculate the normalized abundance of its target temporal hypergraph motifs, and map the normalized abundances to the corresponding dimensions of a three-dimensional feature vector to obtain the feature profile of the temporal cooperation hypergraph. Among them, each dimension of the three-dimensional feature vector corresponds to the normalized abundance of a type of target temporal hypergraph motif, and there are three different types of target temporal hypergraph motifs for each temporal cooperation hypergraph; Analyze the differences of the feature profile in different periods to predict the evolution mechanism of the target industry.

2. The prediction method for the future industrial evolution mechanism based on the temporal hypergraph motif according to claim 1, wherein The collecting the original cooperation data set among several institutions in the target industrial field includes: Determine the target industry, and extract the domain keywords corresponding to the target industry; Use the domain keywords as retrieval conditions to obtain a list of institutions related to the target industry; Based on the list of institutions, collect the cooperation events among various institutions to form an original cooperation data set.

3. A prediction method for the future industrial evolution mechanism based on temporal hypergraph motifs according to claim 2, characterized in that, The preprocessing the original cooperation data set to generate a cooperation event data set including timestamps includes: Perform disambiguation processing on the institution names to ensure that the same institution has a unified name in all cooperation events; Assign corresponding timestamps to each cooperation event according to a preset time unit; When multiple cooperation events have the same timestamp, assign unique timestamps to the multiple cooperation events respectively by means of random sorting.

4. A prediction method for the future industrial evolution mechanism based on temporal hypergraph motifs according to claim 1, characterized in that When the counting algorithm adopts an exact counting algorithm based on temporal hypergraph projection, the using a counting algorithm to determine the true number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, P≥1, includes: Given a time window δ and initialize the projection graph , where node v ∈ corresponds to the set of hyperedges with the same node set in the temporal cooperation hypergraph , the set of timestamps carried by each node v = , edge ∈ if and only if ∩ ≠ ; Update based on the insertion module and deletion module: In the insertion module, for a given temporal hyperedge , if there is no hyperedge with the same node set within the current time window, then add the -corresponding nodes to , and find the adjacent nodes of the newly added nodes to create associated connecting edges, and add the timestamp to ; In the deletion module, if the given temporal hyperedge is the only temporal hyperedge in the current time window that has the node set , then remove the node and all its associated incident edges from , and delete the -corresponding timestamp in . Within the specified time window respectively count the ternary, binary, and unary temporal hypergraph motifs that contain .

5. The prediction method for the future industrial evolution mechanism based on the temporal hypergraph motif according to claim 4, wherein When the counting algorithm adopts an approximate counting method based on random sampling, the using a counting algorithm to determine the true number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, P≥1, includes: Set the sampling time window length Δ and the number of sampling times s; For each sampling, perform the following steps: a. Select a temporal hyperedge randomly and uniformly from the temporal hyperedge set ; ; b. According to the timestamp t of the selected temporal hyperedge, randomly select a timestamp within the time window ; c. Use the counting algorithm described in claim 4 to accurately count the temporal hypergraph motifs within the sub-window to obtain ; d. Calculate the probability that a timing hyperedge appears in the selected time window , where is the number of hyperedges within the sub-window; e. Adjust the counting result to ; Sum the estimators of each sampling and multiply by to obtain an unbiased estimate of each type of temporal hypergraph motif, and use the unbiased estimate as its true quantity.

6. A method for predicting the future industrial evolution mechanism based on temporal hypergraph motifs according to any one of claims 4 or 5, characterized in that, Creating a set of zero-model random hypergraphs and counting the average number of each type of target temporal hypergraph motif in the set of zero-model random hypergraphs includes: Constructing a set of zero-model random hypergraphs using a zero-model generation method, where each random hypergraph retains the basic statistical characteristics of the temporal cooperation hypergraph; Calculating the number of occurrences of each type of target temporal hypergraph motif in each random hypergraph according to the counting method described in any one of claims 4 or 5, and recording the average value.

7. A prediction device for the future industrial evolution mechanism based on temporal hypergraph motifs, characterized in that, Including: A preprocessing module for collecting an original cooperation dataset among several institutions in a target industrial field and preprocessing the original cooperation dataset to generate a cooperation event dataset including timestamps; A hypergraph construction module for constructing a temporal cooperation hypergraph according to the cooperation event dataset, where nodes represent institutions, hyperedges represent cooperation events, and are attached with corresponding timestamp marks; A motif detection module for using a counting algorithm to determine the true number of P types of target temporal hypergraph motifs preselected in the temporal cooperation hypergraph, where P≥1; A statistics module for creating a set of zero-model random hypergraphs and counting the average number of each type of target temporal hypergraph motif in the set of zero-model random hypergraphs; An abundance calculation module for calculating the abundance of each type of target temporal hypergraph motif according to the true number of the target temporal hypergraph motif in the temporal cooperation hypergraph and the average number in the set of zero-model random hypergraphs; A prediction module for constructing a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs and analyzing the changing trend of the feature profile to predict the evolution mechanism of the target industry; The prediction module constructing a feature profile of the temporal cooperation hypergraph based on the abundances of all target temporal hypergraph motifs and analyzing the changing trend of the feature profile to predict the evolution mechanism of the target industry includes: For each temporal cooperation hypergraph, calculating the normalized abundance of its target temporal hypergraph motifs and mapping the normalized abundances to the corresponding dimensions of a three-dimensional feature vector to obtain the feature profile of the temporal cooperation hypergraph, where each dimension of the three-dimensional feature vector corresponds to the normalized abundance of a type of target temporal hypergraph motif, and there are three different types of target temporal hypergraph motifs for each temporal cooperation hypergraph; Analyzing the differences in the feature profile at different times to predict the evolution mechanism of the target industry.

8. An electronic device, characterized in that, Including a memory and a processor, where the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement a method for predicting the future industry evolution mechanism based on temporal hypergraph motifs as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, The computer program, when executed by a computer, implements a method for predicting the future industry evolution mechanism based on temporal hypergraph motifs as described in any one of claims 1 to 6.

Images