A single-pass based multi-dimensional reference pattern statistical analysis method and system
Patent Information
- Application Number
- CN202610847058.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-12
- Publication Date
- 2026-08-18
AI Technical Summary
[0005]本发明要解决的技术问题是：在多维引用模式统计中，现有逐维度多轮遍历方案存在重复计算开销大导致接口响应延迟高的缺陷，提供一种基于单次遍历的多维引用模式统计分析方法及系统
第一，O(n)时间复杂度消除重复遍历开销。本发明对引用网络所有节点执行单次遍历，在遍历的每一次迭代中同时更新四个独立维度的统计容器——分别更新引用方向计数器、年份计数器、期刊计数器和引用次数累加器。时间复杂度为O(n)，与统计维度的数量k无关，而多轮独立遍历方案的时间复杂度为O(k×n)。当引用网络包含200个节点时，四维度统计的单次遍历方案相比四轮逐维度方案减少约75%的遍历操作。在实际Web服务接口的响应时间测试中，200节点网络的多维统计从逐维度方案的约12毫秒降低至约3毫秒。
Smart Images
Figure CN122594548A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of academic literature bibliometric analysis technology, specifically to a method and system for statistical analysis of multidimensional citation patterns based on a single traversal. Background Technology
[0002] Citation pattern statistics in academic literature are a fundamental analytical tool in bibliometrics and scientometrics. By analyzing multiple dimensions of data, such as publication year, journal distribution, citation direction, and citation intensity, researchers can understand the development process, core publication venues, knowledge evolution direction, and overall influence level of a research field. The collection of these statistical results constitutes a comprehensive knowledge portrait of that research field.
[0003] In existing technologies, multidimensional citation pattern statistics are typically implemented using a sequential, independent traversal approach. One approach involves multiple rounds of traversal: a separate full-network traversal is performed for each dimension requiring statistical analysis. For example, the first round might analyze citation direction distribution, the second year distribution, the third journal distribution, and the fourth calculating the average number of citations. Each round requires traversing the node list from the beginning, resulting in a time complexity of O(k×n), where k is the number of statistical dimensions and n is the total number of nodes. Another approach uses database aggregation queries: the citation network node data is imported into a relational database or data analysis framework, and statistical data for each dimension is obtained through multiple independent aggregation queries.
[0004] The drawback of Method 1 lies in the repetitive computational overhead of multiple rounds of traversal. When the citation network is large (more than 200 nodes), four independent traversals mean that the direction attribute, year, journal, and citation count fields of each node are read four times, resulting in low computational efficiency. In the web service scenario of a literature analysis system, this method increases interface response latency, affecting user experience. The drawback of Method 2 is the introduction of database dependency and query serialization overhead. For temporary citation network data in memory (a network built in a single user session, which does not need to be persisted), the path of importing into the database and then querying introduces unnecessary serialization, deserialization, and network round-trip overhead, which contradicts the goal of lightweight system architecture. Summary of the Invention
[0005] The technical problem this invention aims to solve is that existing multi-dimensional reference pattern statistics methods suffer from high overhead due to repeated calculations, leading to high interface response latency. This invention provides a multidimensional reference pattern statistical analysis method and system based on a single traversal.
[0006] The technical solution of this invention is as follows: All nodes in the citation network are acquired and a four-dimensional statistical container is initialized. The four-dimensional statistical container includes a first-dimensional counter for counting by citation direction, a second-dimensional counter for counting by publication year, a third-dimensional counter for counting by publication journal name, and a fourth-dimensional accumulator for accumulating the citation counts of all nodes. A single traversal is performed on all nodes in the citation network. When traversing each node, the first-dimensional counter is updated with the current node's direction attribute as the key, the second-dimensional counter is updated with the current node's publication year as the key, the third-dimensional counter is updated with the current node's publication journal name as the key, and the fourth-dimensional accumulator is updated with the current node's citation count. After the single traversal is completed, a multi-dimensional citation pattern analysis report is generated based on the cumulative data of the four-dimensional statistical containers. The analysis report includes the number of forward-cited papers and the number of backward-cited papers, the distribution range and frequency distribution of publication years, the names and corresponding frequencies of the top few high-frequency journals arranged in descending order of frequency, and the average citation count of each node.
[0007] The beneficial effects of this invention are as follows: First, the O(n) time complexity eliminates the overhead of repeated traversals. This invention performs a single traversal of all nodes in the citation network, updating four independent statistical containers in each iteration—the citation direction counter, year counter, journal counter, and citation count accumulator, respectively. The time complexity is O(n), independent of the number of statistical dimensions k, while the time complexity of the multi-round independent traversal scheme is O(k×n). When the citation network contains 200 nodes, the single traversal scheme of four-dimensional statistics reduces traversal operations by approximately 75% compared to the four-round dimension-by-dimensional scheme. In actual Web service interface response time tests, the multi-dimensional statistics of a 200-node network decreased from approximately 12 milliseconds in the dimension-by-dimensional scheme to approximately 3 milliseconds.
[0008] Second, the four-dimensional synchronous statistical output provides a panoramic view of the domain's knowledge. The four-dimensional statistical results generated in a single traversal naturally constitute the four components of this panoramic view: the citation direction distribution reveals the relative weight of the seed papers' roles as knowledge contributors and knowledge aggregators within the citation network; the publication year distribution reveals the time span and development activity of the citation network (an increasing trend in year frequency corresponds to an active period in the field, while a decreasing trend corresponds to a mature period); the publication journal distribution reveals the main publication venues of researchers in the citation network; and the average number of citations per node reveals the average influence level of papers in the citation network. This four-dimensional data aggregation provides a systematic overview of the research field.
[0009] Third, a hash key-value counter enables automatic grouping of dimension fields. This invention employs a hash mapping data structure to implement year and journal counters, using the year and journal name as hash keys and the count frequency as the hash value. During traversal, a corresponding key-value pair is automatically created and initialized to 1 for each newly appearing year or journal name, and an increment operation is performed on existing keys. This hash mapping mechanism completes dimension grouping and counting in a single O(1) operation, without needing to pre-enumerate all possible years or journal names before traversal.
[0010] Fourth, lightweight in-memory statistics require no external dependencies. The statistical operations of this invention are performed entirely in memory. The data structure of the statistical container is a native hash map and scalar accumulator from the programming language, eliminating the need to import relational databases or data analysis frameworks. This design allows multidimensional citation pattern statistics to be seamlessly embedded into the web service processing flow of a literature analysis system, avoiding the serialization overhead and external dependency coupling caused by data import and export. Attached Figure Description
[0011] Figure 1 This is a flowchart of a multidimensional reference pattern statistical analysis method based on a single traversal in an embodiment of the present invention.
[0012] Figure 2 This is a schematic diagram of the knowledge panoramic portrait formed by the four-dimensional statistical results in an embodiment of the present invention.
[0013] The meanings of the markings in the diagram are as follows: Figure 1 Steps are as follows: 101 - Obtain all nodes in the citation network; 102 - Initialize the four-dimensional statistical containers; 103 - Perform a single traversal loop on all nodes in the citation network; 104 - Update the first-dimensional counter using the current node's direction attribute as the key; 105 - Update the second-dimensional counter using the current node's publication year as the key; 106 - Update the third-dimensional counter using the current node's journal name as the key; 107 - Update the fourth-dimensional accumulator using the current node's citation count; 108 - Generate a multi-dimensional citation pattern analysis report based on the cumulative data in the four-dimensional statistical containers after the single traversal; 109 - Export the number of forward-cited and backward-cited papers from the first-dimensional counter; 110 - Export the publication year distribution range and frequency distribution from the second-dimensional counter; 111 - Export the names of the top few high-frequency journals sorted by frequency in descending order from the third-dimensional counter; 112 - Divide the total value of the fourth-dimensional accumulator by the total number of nodes to obtain the average citation count per node.
[0014] Figure 2201 - Dimension 1 (Citation direction / Knowledge role diagnosis), 202 - Dimension 2 (Publication year / Development process stage diagnosis), 203 - Dimension 3 (Published journal / Publication venue diagnosis), 204 - Dimension 4 (Average citations / Overall influence diagnosis), Center - Knowledge panoramic portrait. Detailed Implementation
[0015] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0016] This embodiment provides a method for statistical analysis of multidimensional citation patterns based on a single traversal, executed by the processor of a terminal device. The method receives previously constructed directed graph data of a citation network as input. This citation network contains a list of nodes, each node having a direction attribute, the year of publication, the name of the journal, and the number of global citations.
[0017] The specific structure of the input citation network is as follows: one seed paper (direction attribute="seed", depth=0), a set of forward citation nodes (direction attribute="forward", depth≥1), and a set of backward citation nodes (direction attribute="backward", depth≥1). The total number of network nodes depends on the quota settings during construction, typically between 30 and 200.
[0018] like Figure 1 As shown, step 102: Initialize four-dimensional statistical containers. The first-dimensional container is a reference direction counter (corresponding to...). Figure 1 Section 104, "Update the first-dimensional counter using the current node's direction attribute as the key," uses a hash map as its data structure, initialized as an empty map, with expected keys including "forward" and "backward." The second-dimensional container is a year counter (corresponding to...). Figure 1 Example 105, "Update the second-dimensional counter using the publication year of the current node as the key," uses a hash map as its data structure, initialized as an empty map. The key is a four-digit integer year, and the value is the number of times the node appears in that year. The third-dimensional container is a journal counter (corresponding to...). Figure 1 Section 106, "Update the third-dimensional counter using the journal name published by the current node as the key," uses a hash map as its data structure, initialized as an empty map. The key is the journal name string, and the value is the number of times that journal appears in the node. The fourth-dimensional container is a citation count accumulator (corresponding to...). Figure 1 In section 107, "Update the fourth dimension accumulator with the current node's reference count", the data structure is a numeric variable with an initial value of 0.
[0019] like Figure 1 As shown, step 103: Perform a single node traversal. Traverse the list of nodes in the referenced network, and perform the following four operations sequentially on each node encountered (corresponding to...). Figure 1The four parallel operation boxes 104, 105, 106, and 107 inside the middle loop 103 are connected by horizontal arrows to indicate synchronous execution within a single traversal.
[0020] Operation 1: Read the direction attribute field value of the current node. If the value is "forward", increment the value corresponding to the key "forward" in the first dimension counter by 1; if the value is "backward", increment the value corresponding to the key "backward" by 1. After traversal, the two count values in the first dimension counter are the number of forward-referenced papers and the number of backward-referenced papers.
[0021] Operation 2: Read the publication year field value of the current node. Use this year value as the hash key to query the second-dimensional counter. If the key already exists, increment its corresponding count by 1; if the key does not exist, insert a new key-value pair with the current year as the key and 1 as the value. After traversal, the second-dimensional counter stores the distribution of the number of papers corresponding to each publication year in this citation network.
[0022] Operation 3: Read the journal name field value of the current node. Use the journal name string as the hash key to query the third-dimensional counter. If the key already exists, increment its corresponding count by 1; if the key does not exist, insert a new key-value pair with the current journal name as the key and 1 as the value. After traversal, the third-dimensional counter stores the distribution of the number of papers corresponding to each journal name in the citation network.
[0023] Operation 4: Read the reference count field value of the current node and add it to the fourth-dimensional accumulator. After traversal, the fourth-dimensional accumulator stores the sum of the reference counts of all nodes.
[0024] like Figure 1 As shown by the dashed return arrow and the "Loop Traversal" label, the above four operations are performed simultaneously in each loop iteration, and the loop exits after all nodes are completed.
[0025] like Figure 1 and Figure 2 As shown, step 108: Generate a multi-dimensional reference pattern analysis report. After a single traversal is completed, an analysis report is constructed based on the cumulative data in the four-dimensional statistical containers. The statistical results of the four dimensions are aggregated into a single report. Figure 2 The image shown is a panoramic view of the domain knowledge.
[0026] like Figure 2As shown in section 201, "Dimension 1: Citation Direction," step 109 is executed from the first dimension counter to extract the number of forward-cited papers and the number of backward-cited papers. The ratio of the number of forward-cited papers to the number of backward-cited papers is calculated. If the ratio is greater than 1, it indicates that the subsequent research citing the seed paper is more than the previous research cited by the seed paper, and the seed paper mainly acts as a knowledge output in this network; if the ratio is less than 1, it indicates that the seed paper mainly acts as a knowledge aggregator in this network. This dimension realizes knowledge role diagnosis.
[0027] like Figure 2 As shown in section 202, "Dimension 2: Publication Year," step 110 is executed from the second dimension counter to extract the year distribution. The minimum and maximum values among all year keys are calculated to form a year range. The year keys are arranged in chronological order, and the year-frequency distribution sequence is output. The changing trend of the frequency distribution is observed: if the frequency in recent years shows an increasing trend, it indicates that the field is in a period of active development; if the frequency in recent years shows a decreasing or stable trend, it indicates that the field has entered a mature or stable period. This dimension enables the diagnosis of development process stages.
[0028] like Figure 2 As shown in section 203, "Dimension 3: Published Journals," step 111 is performed from the third-dimensional counter to extract the journal distribution. All key-value pairs in the third-dimensional counter are sorted in descending order of frequency. The top 5 journal names and their corresponding frequencies are used to construct a list of top-tier publication locations. In the biomedical field, the presence of top-tier comprehensive journals such as Nature, Cell, Science, or NEJM among the top 5 journals further indicates a high overall quality level of research in this citation network. This dimension enables publication location diagnosis.
[0029] like Figure 2 As shown in section 204, "Dimension 4: Average Citations," step 112 is performed from the fourth-dimensional accumulator to extract the average number of citations per node. The total value of the fourth-dimensional accumulator is divided by the total number of nodes in the citation network (excluding seed papers), and the quotient is the average number of citations per node. This metric reflects the average global impact level of papers in this citation network. This dimension enables overall impact diagnosis.
[0030] Taking a seed paper in the field of tumor immunology (DOI "10.1038 / s41568-020-00320-0") as an example, four-dimensional statistical analysis yielded the following comprehensive knowledge profile: 38 papers were cited forward and 14 backward, with a forward / backward ratio of 2.71, indicating that the seed paper received high attention in the field and primarily acted as a knowledge contributor; the years ranged from 2012 to 2024, with papers published between 2021 and 2023 accounting for 61% of the total, indicating that the field is in a period of active development; the top five publishing journals were Cancer Discovery, Nature Medicine, Cancer Cell, Journal of Clinical Oncology, and Science Translational Medicine, all top journals in the field of oncology; the average number of citations per node was 327, higher than the average citation level of biomedical papers (approximately 50), indicating a high overall quality of papers in this citation network. This comprehensive knowledge profile provides researchers with a systematic reference for understanding the knowledge architecture of this subfield of tumor immunology.
[0031] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A single-pass based multi-dimensional reference pattern statistical analysis method, characterized in that, include: Get all nodes in the referenced network; Initialize four-dimensional statistical containers, which include a first-dimensional counter for counting by citation direction, a second-dimensional counter for counting by publication year, a third-dimensional counter for counting by publication journal name, and a fourth-dimensional accumulator for accumulating the citation counts of all nodes. A single traversal is performed on all nodes in the reference network. When traversing each node, the first dimension counter is updated with the direction attribute of the current node as the key, the second dimension counter is updated with the publication year of the current node as the key, the third dimension counter is updated with the publication journal name of the current node as the key, and the fourth dimension accumulator is updated with the citation count of the current node. After the single traversal is completed, a multidimensional reference pattern analysis report is generated based on the accumulated data in the first dimension counter, the second dimension counter, the third dimension counter, and the fourth dimension accumulator. The multidimensional citation pattern analysis report includes all of the following: the number of forward-cited papers and the number of backward-cited papers derived from the first dimension counter; the distribution range and frequency distribution of publication years derived from the second dimension counter; the names and corresponding frequencies of the top few high-frequency journals derived from the third dimension counter in descending order of frequency; and the average number of citations per node obtained by dividing the total value of the fourth dimension accumulator by the total number of nodes.
2. The method according to claim 1, characterized in that, After the single traversal is completed, the second dimension counter also generates a range of publication years for the referenced network, which is determined by the minimum and maximum values of all year keys in the second dimension counter.
3. The method according to claim 1, characterized in that, After the third-dimensional counter is sorted in descending order of frequency, the top 5 high-frequency journal names and their corresponding frequencies are selected as the core publication information of the citation network and written into the multidimensional citation pattern analysis report.
4. The method according to claim 1, characterized in that, The first dimension counter uses the citation direction attribute value as the key and the number of times the citation direction attribute value appears in all nodes as the value. The citation direction attribute value includes forward direction and backward direction. The ratio of the number of forward-cited papers to the number of forward-cited papers is used to characterize the relative weight of the seed paper's role as a knowledge outputter and knowledge aggregator in the citation network.
5. The method according to claim 1, characterized in that, The multidimensional citation pattern analysis report also correlates the frequency distribution of each year in the publication year distribution range with the development stage of the field, where the increasing trend of year frequency corresponds to the active period of the field, and the decreasing trend of year frequency corresponds to the mature period of the field.
6. A multidimensional reference pattern statistical analysis system based on a single traversal, characterized in that, include: The node acquisition module is used to acquire all nodes in the referenced network; The container initialization module is used to initialize four-dimensional statistical containers, which include a first-dimensional counter for counting by citation direction, a second-dimensional counter for counting by publication year, a third-dimensional counter for counting by publication journal name, and a fourth-dimensional accumulator for accumulating the citation count of all nodes. The single-pass statistics module is used to perform a single pass through all nodes in the reference network. When passing through each node, the first dimension counter is updated with the direction attribute of the current node as the key, the second dimension counter is updated with the publication year of the current node as the key, the third dimension counter is updated with the publication journal name of the current node as the key, and the fourth dimension accumulator is updated with the citation count of the current node. The report generation module is used to generate a multidimensional reference pattern analysis report based on the accumulated data in the first dimension counter, the second dimension counter, the third dimension counter, and the fourth dimension accumulator after the single traversal is completed. The multidimensional citation pattern analysis report includes: the number of forward-cited papers and the number of backward-cited papers, the distribution range and frequency distribution of publication years, the names and corresponding frequencies of the top few high-frequency journals arranged in descending order of frequency, and the average number of citations per node.
7. The system according to claim 6, characterized in that, After the single traversal is completed, the second dimension counter also generates a range of publication years for the referenced network, which is determined by the minimum and maximum values of all year keys in the second dimension counter.
8. The system according to claim 6, characterized in that, The first dimension counter uses the citation direction attribute value as the key, and the ratio of the number of forward-cited papers to the number of backward-cited papers is used to characterize the relative weight of the seed paper's role as a knowledge outputter and a knowledge aggregator in the citation network.
9. A terminal device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1 to 5.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 5.