A method for identifying organizational structures based on the first-order difference and the method of measures of difference

Abstract

The present invention relates to an organizational structure recognition method based on the first-order difference and the method of measures of dispersion. For the undirected graph corresponding to the social network organization, the average shortest path of each node therein is traversed, and the first-order difference measure of dispersion method of the sequence to be analyzed corresponding to each average shortest path is used to identify the sources of each multi-core organization. Finally, starting from the sources of each multi-core organization, the top-down iterative direct node method is applied to sequentially divide the organizational structure, so as to obtain the organizational structure of the social network organization; the design scheme optimizes the traditional way that requires manual intervention for the source of the organizational structure, uses the first-order difference measure of dispersion method of the average shortest path to identify the source of the organization, and applies the idea of minimum cost to iteratively direct nodes in sequence, which can achieve the efficient sorting of any type of social network organizational structure and effectively solve the problems of insufficient division efficiency, accuracy and adaptability in this field at present.

Description

Technical Field

[0001] The present invention relates to an organizational structure recognition method based on the first-order difference and the method of measures of dispersion, and relates to the technical field of discovering the organizational structure of social networks. Background Art

[0002] In the field of discovering the organizational structure of complex social networks, especially in community attribute networks, it is found that the people in the community do not work together equally, but have different divisions of labor and unequal statuses. In order to accurately divide the organizational structure of the community, many ideas have been proposed, such as the hierarchical community discovery algorithm based on in-degree and out-degree leadership nodes, the influence propagation model based on pagerank, the adaptive information entropy model with penalty and balancing functions, etc. However, when fitting the selection of leadership nodes in different communities, these models all require manual intervention and cannot accurately divide the overlapping nodes that are connected at different levels. Therefore, in the face of complex group attribute networks, how to automatically determine multi-core leadership nodes and accurately divide the chaotic architecture of relationships is the core problem we need to solve.

[0003] In the discovery of community organizational structure, the traditional method is to select the organizational source under manual qualitative intervention and determine the lower layer by the tightness of in-out link association, which is difficult to achieve fully automatic division of community structures in different scenarios. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an organizational structure recognition method based on the first-order difference and the method of measures of dispersion, which adopts a new design logic to solve the deficiencies in division efficiency, accuracy and adaptability in the current field and efficiently realizes the recognition of the organizational structure.

[0005] The present invention adopts the following technical solutions to solve the above technical problems: The present invention designs an organizational structure recognition method based on the first-order difference and the method of measures of dispersion for realizing the recognition of the organizational structure for the target social network organization, including the following steps:

[0006] Step A. Based on the call connection data between the personnel in the target social network organization, taking the personnel as nodes and connecting the nodes with call connections between each other with undirected edges, construct an undirected graph corresponding to the target social network organization, and then enter Step B;

[0007] Step B. For each node in the undirected graph, obtain the shortest path hop count from the node to each node with which it has a direct or indirect edge connection, and then obtain the average shortest path hop count of the shortest path hop counts corresponding to the node, that is, obtain the average shortest path hop count corresponding to each node, and sort the average shortest path hop counts from small to large to form a sequence to be analyzed, and then enter Step C;

[0008] Step C. Obtain a sequence of differences between the adjacent average shortest path hop counts of each group in the sequence to be analyzed according to the first-order difference of the sequence to be analyzed, that is, the difference sequence, and proceed to Step D;

[0009] Step D. Based on the method of measuring the amount of difference with a preset number of sub-differences, obtain the central value QD and the standard deviation PD of the difference sequence, and then proceed to Step E;

[0010] Step E. According to the central value QD and the standard deviation PD of the difference sequence, obtain the target threshold threshold_value corresponding to the sequence to be analyzed, and further obtain each multi-core tissue source in the sequence to be analyzed, and then proceed to Step F;

[0011] Step F. Based on the undirected graph, starting from each multi-core tissue source respectively, obtain the organizational structure of the target social network organization through an iterative outward manner.

[0012] As a preferred technical solution of the present invention: In the said Step D, based on the preset stable data segmentation range under the preset number N sub-difference method, select each difference in the preset stable data segmentation range of the N equal parts of the difference sequence, obtain the average value of each difference to form the central value QD of the difference sequence, and obtain the standard deviation of each difference to form the standard deviation PD of the difference sequence, and then proceed to Step E.

[0013] As a preferred technical solution of the present invention: In the said Step D, based on the second stable data segmentation to the seventh stable data segmentation under the octal difference method, select each difference in the second stable data segmentation to the seventh stable data segmentation of the eight equal parts of the difference sequence, obtain the average value of each difference to form the central value QD of the difference sequence, and obtain the standard deviation of each difference to form the standard deviation PD of the difference sequence, and then proceed to Step E.

[0014] As a preferred technical solution of the present invention: In the said Step E, according to the central value QD and the standard deviation PD of the difference sequence, according to the following formula:

[0015] threshold_value = QD + 2.5·PD

[0016] Obtain the target threshold threshold_value corresponding to the sequence to be analyzed, and sequentially select the node corresponding to the average shortest path hop count that is greater than the target threshold threshold_value in the sequence to be analyzed as the source breakpoint, and define the source breakpoint and each node sequentially located before the source breakpoint as each multi-core tissue source.

[0017] As a preferred technical solution of the present invention: in step F, based on the undirected graph, when iterating outward from each multi-core organizational source in turn, if the directly obtained node in the iteration is discovered for the first time, this directly obtained node is used as the next-layer node; if the directly obtained node in the iteration is not discovered for the first time, this directly obtained node is ignored; iterate in this way to obtain the organizational structure of the target social network organization.

[0018] For the organizational structure recognition method based on the first-order difference and the method of measures of dispersion of the present invention, compared with the prior art by adopting the above technical solution, it has the following technical effects:

[0019] The organizational structure recognition method designed by the present invention based on the first-order difference and the method of measures of dispersion traverses the average shortest path of each node in the undirected graph corresponding to the social network organization, and through the first-order difference method of measures of dispersion of the sequence to be analyzed corresponding to each average shortest path, identifies each multi-core organizational source. Finally, starting from each multi-core organizational source, the organizational structure is divided in turn by applying the method of iterating directly obtained nodes from top to bottom, and the organizational structure of the social network organization is obtained; the design solution optimizes the traditional way that requires manual intervention for the organizational structure source, adopts the first-order difference method of measures of dispersion of the average shortest path to identify the organizational source, and applies the idea of minimum cost to iterate directly obtained nodes in turn, which can realize the efficient arrangement of any type of social network organizational structure and effectively solve the problems of insufficient division efficiency, accuracy and adaptability in this field at present. Brief Description of the Drawings

[0020] Figure 1 It is a schematic diagram of the average shortest path and its first-order difference of 50 nodes in the application embodiment of the present invention;

[0021] Figure 2 It is a schematic diagram of the organizational structure of the target social network organization in the application embodiment of the present invention. Detailed Embodiments

[0022] The following further describes in detail the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0023] The present invention designs an organizational structure recognition method based on the first-order difference and the method of measures of dispersion for realizing the recognition of the organizational structure of the target social network organization. In actual application, the following steps A to F are specifically executed.

[0024] Step A. Based on the call connection data between each person in the target social network organization, taking the person as a node and connecting the nodes with call connections between each other with undirected edges, construct the undirected graph corresponding to the target social network organization, and then enter step B.

[0025] Step B. For each node in the undirected graph, obtain the shortest path hop counts from the node to each node directly or indirectly connected by an edge, and then obtain the average shortest path hop count of each corresponding shortest path hop count of the node, that is, obtain the average shortest path hop counts corresponding to each node respectively. Then, sort the average shortest path hop counts from small to large to form a sequence to be analyzed, and then proceed to Step C.

[0026] Step C. According to the first-order difference of the sequence to be analyzed, obtain the sequence of differences between adjacent average shortest path hop counts in each group in the sequence to be analyzed, that is, the difference sequence, and then proceed to Step D.

[0027] In the actual implementation example, as Figure 1 shown, it is the schematic diagram of the average shortest path and its first-order difference of 50 nodes. Among them, the abscissa represents each node, the left main coordinate axis is the average shortest path hop count, and the right secondary coordinate axis is the difference hop count of the first-order difference.

[0028] Step D. Based on the method of difference measures with a preset number of sub-differences, obtain the central value QD and the standard deviation PD of the difference sequence, and then proceed to Step E.

[0029] In actual applications, in the above Step D, specifically design the preset stable data segmentation range based on the method of sub-differences with a preset number N. Select each difference in the preset stable data segmentation range under the N-equal division of the difference sequence, obtain the average value of each difference, form the central value QD of the difference sequence, and obtain the standard deviation of each difference, form the standard deviation PD of the difference sequence.

[0030] Regarding the preset stable data segmentation range under the method of sub-differences with a preset number N here, in specific selection, such as using

[0031] the method of difference measures based on octal differences to select the source of multi-core tissue. Among them, the method of difference measures is a statistic representing the degree of data dispersion, reflecting the degree to which each variable value is far from its central value. Based on comprehensive business experience, for octal differences, the data fluctuation range in the front is relatively large, and the source of the organization will also be generated here. Therefore, in the application of octal differences, use the data at and to evaluate the difference measure of the overall data; that is, specifically based on the second stable data segmentation to the seventh stable data segmentation under the octal difference method, select each difference in the second stable data segmentation to the seventh stable data segmentation under the eight-equal division of the difference sequence, obtain the average value of each difference, form the central value QD of the difference sequence. While considering the central value of the difference sequence, the design also needs to add a fluctuation situation to obtain a reliable confidence interval. Therefore, further obtain the standard deviation of each difference, form the standard deviation PD of the difference sequence, and then proceed to Step E.

[0032] Step E. Based on the central value QD and the standard deviation PD of the difference sequence, obtain the target threshold value threshold_value corresponding to the sequence to be analyzed, and further obtain each multi-core tissue source in the sequence to be analyzed, and then proceed to Step F.

[0033] In specific implementation, in Step E here, specifically based on the central value QD and the standard deviation PD of the difference sequence, according to the following formula:

[0034] threshold_value = QD + 2.5·PD

[0035] Obtain the target threshold value threshold_value corresponding to the sequence to be analyzed, and sequentially select the node corresponding to the average shortest path hop count that is greater than the target threshold value threshold_value in the sequence to be analyzed as the source breakpoint, and define the source breakpoint and each node sequentially before the source breakpoint as each multi-core tissue source.

[0036] In actual implementation, based on Figure 1 the specific data shown, after executing Step E here, the first three nodes are obtained as multi-core tissue sources.

[0037] Step F. Based on the undirected graph, starting from each multi-core tissue source respectively, obtain the organizational structure of the target social network organization through an iterative manner outward. In actual implementation, here in the process of iterating outward respectively starting from each multi-core tissue source based on the undirected graph, if the directly obtained node in the iteration is discovered for the first time, then use this directly obtained node as the next-layer node; if the directly obtained node in the iteration is not discovered for the first time, then ignore this directly obtained node; iterate in this way to obtain the organizational structure of the target social network organization.

[0038] Then continue Figure 1 the three multi-core tissue sources obtained through Step E based on the data shown, and further perform the iterative operation of the above Step F, then the organizational structure of the target social network organization in this embodiment can be obtained. Finally, through i2 software mapping, the three-dimensional organizational structure of the entire planar community network can be obtained, as Figure 2 shown.

[0039] The above technical solution designs an organizational structure recognition method based on the first-order difference and the method of difference quantity. For the undirected graph corresponding to the social network organization, traverse the average shortest path of each node therein, and through the first-order difference quantity method of the sequence to be analyzed corresponding to each average shortest path, identify the sources of each multi-core organization. Finally, starting from the sources of each multi-core organization respectively, use the method of iteratively directly connecting nodes from top to bottom to divide the organizational structure in sequence, and obtain the organizational structure of the social network organization; the design solution optimizes the way that the traditional organizational structure source requires manual intervention, uses the first-order difference quantity method of the average shortest path to identify the organizational source, and applies the idea of minimum cost to iteratively directly connect nodes in sequence, which can achieve the efficient sorting of any type of social network organizational structure and effectively solve the problems of insufficient division efficiency, accuracy and adaptability in this field at present.

[0040] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for identifying an organizational structure based on the first-order difference and the method of measures of dispersion, which is used to identify the organizational structure for a target social network organization. Characterized in that: It includes the following steps: Step A. Based on the call connection data between each person in the target social network organization, taking the persons as nodes and connecting the nodes with undirected edges between those with call connections, construct an undirected graph corresponding to the target social network organization, and then proceed to Step B; Step B. For each node in the undirected graph respectively, obtain the shortest path hop counts from the node to each node directly or indirectly connected to it, and then obtain the average shortest path hop count of each of these shortest path hop counts corresponding to the node, that is, obtain the average shortest path hop counts corresponding to each node respectively, and arrange the average shortest path hop counts in ascending order to form a sequence to be analyzed, and then proceed to Step C; Step C. According to the first-order difference of the sequence to be analyzed, obtain a sequence of differences between the adjacent average shortest path hop counts in each group in order in the sequence to be analyzed, that is, the difference sequence, and proceed to Step D; Step D. Obtain the central value of the difference sequence based on the difference measure method with a preset number of differences and the standard deviation , and then proceed to Step E; Step E. According to the central value of the difference sequence , and the standard deviation , obtain the target threshold corresponding to the sequence to be analyzed , and further obtain each multi-nuclear tissue source in the sequence to be analyzed, and then enter Step F; Step F. Based on the undirected graph, starting from each multi-core organizational source respectively, through an iterative manner outward, obtain the organizational structure of the target social network organization; In the above step E, according to the central value of the difference sequence , and the standard deviation , according to the following formula: ； Obtain the target threshold corresponding to the sequence to be analyzed , and sequentially select the node corresponding to the average shortest path hop count that is greater than the target threshold in the sequence to be analyzed as the source breakpoint. Define the source breakpoint and each node that is sequentially before the source breakpoint as each multi-core organization source 2. The method for identifying an organizational structure based on the first-order difference and the method of measures of dispersion according to claim 1, Characterized in that: In the said step D, based on a preset number select a difference sequence within the preset stable data segmentation range under the difference method equally divide each difference within the preset stable data segmentation range, obtain the average value of each difference, and form the central value of the difference sequence and obtain the standard deviation of each difference to form the standard deviation of the difference sequence , and then proceed to step E.

3. The method for identifying an organizational structure based on the first-order difference and the method of measures of dispersion according to claim 2, Characterized in that: In the said step D, based on the second to seventh stationary data segments under the octant difference method, select each difference value in the second to seventh stationary data segments under the octant division of the difference sequence, obtain the average value of these difference values, and form the central value of the difference sequence , and obtain the standard deviation of these difference values to form the standard deviation of the difference sequence , and then proceed to step E 4. The method for identifying an organizational structure based on the first-order difference and the method of measures of dispersion according to claim 1, Characterized in that: In the said Step F, based on the undirected graph, starting from each multi-core organizational source respectively, during the iterative process outward, if the directly obtained node is discovered for the first time, then take this directly obtained node as the next-layer node; If the directly obtained node is not discovered for the first time, then ignore this directly obtained node; iterate in this way to obtain the organizational structure of the target social network organization.

Images