Method for identifying and optimizing organizational tacit collaboration relationship based on multi-modal social network mining

Abstract

The application discloses an organization implicit cooperation relationship recognition and optimization method based on multi-modal social network mining, comprising the following steps: multi-source cooperation data acquisition and preprocessing: collecting organization internal mail correspondence records, meeting participation logs, instant messaging interaction data and project management system data, anonymizing the data, and retaining department and role labels; constructing a double-level dynamic weighted directed graph: the node set includes department-level nodes and employee-level nodes, and the two layers of nodes are connected through membership; the edge set represents the cooperation interaction behavior between nodes, including communication frequency, task association and information transmission path, the application, by collecting multi-source heterogeneous data (mail correspondence, meeting logs, instant messaging records, project management data, etc.), and through anonymization, cleaning and processing, breaks down the department data dispersion barriers, provides a complete data basis for cooperation analysis, and avoids the analysis one-sidedness caused by relying on single structured data in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of cross-departmental collaboration technology, and in particular relates to a method for identifying and optimizing implicit collaborative relationships within organizations based on multimodal social network mining. Background Technology

[0002] In the field of cross-departmental collaboration, existing technologies primarily provide basic collaboration support through process management tools and organizational structure visualization systems. Process-based collaboration systems (such as OA systems) standardize collaboration steps through electronic approval workflows and task assignment, relying on pre-defined organizational structures to allocate permissions and tasks, ensuring standardized collaboration processes. Organizational chart-based optimization tools visualize departmental hierarchies and define collaboration scopes according to job responsibilities, providing a framework for cross-departmental task allocation. Furthermore, some project management software integrates data such as task progress and resource allocation to help track the execution status of cross-departmental projects. These technologies, to a certain extent, achieve structured management of the collaboration process, providing basic tools to support cross-departmental communication within the organization.

[0003] However, existing technologies have several limitations that make it difficult to meet the needs of efficient cross-departmental collaboration. First, data coverage is incomplete, relying heavily on single structured data (such as task forms) and failing to integrate unstructured collaborative information such as emails and instant messaging, leading to data fragmentation and information silos between departments. Second, the portrayal of collaborative relationships is static, defining collaboration paths solely based on formal organizational structures, failing to reflect the implicit networks in actual communication, and making it difficult to identify key coordinators and other important nodes. Third, there is a lack of dynamic early warning mechanisms, allowing only passive responses to collaboration bottlenecks caused by over-reliance on a few nodes, rather than proactive prediction. Fourth, team configuration relies on experience, failing to quantify member compatibility using historical collaboration data, increasing the cost of adjustment. Fifth, it is difficult to adapt to organizational changes; static models cannot dynamically adjust with collaborative behavior, causing optimization solutions to lag behind actual collaboration needs. These shortcomings significantly limit the effectiveness of existing technologies in improving the efficiency of cross-departmental collaboration. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying and optimizing implicit collaborative relationships in organizations based on multimodal social network mining, which solves the problems of one-sided data, static relationships, delayed early warning, blind configuration, and poor adaptability in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for identifying and optimizing implicit collaborative relationships within organizations based on multimodal social network mining includes the following steps:

[0007] S1. Multi-source collaborative data collection and preprocessing: Collect email records, meeting participation logs, instant messaging interaction data and project management system data within the organization, anonymize the data, and retain department and role labels;

[0008] S2. Construct a two-level dynamically weighted directed graph:

[0009] S2.1. The node set includes department-level nodes and employee-level nodes, and the two levels of nodes are connected through a hierarchical relationship;

[0010] S2.2. Edge sets represent the collaborative interaction behaviors between nodes, including communication frequency, task association, and information transmission paths;

[0011] S2.3. Edge weights are calculated using the formula: W ij =α·freq ij +β·urgency ij +γ·depth ij , where freq ij For interaction frequency, urgency ij For mission urgency, depth ij The information depth coefficient is α, β, and γ, which are weighting coefficients.

[0012] S3. Identification of Implicit Collaborative Relationships: Based on the aforementioned dynamic weighted directed graph, hub nodes and implicit coordinators are identified through node centrality analysis, and high-cohesion subgroups are detected through the Louvain algorithm;

[0013] S4. Collaboration efficiency optimization decision generation:

[0014] S4.1. Bottleneck warning: When the node betweenness centrality is greater than the preset threshold and the load growth rate is greater than 15% / month, a resource allocation warning is triggered;

[0015] S4.2. Team configuration optimization: Generate employee collaboration compatibility vectors based on historical collaboration data, calculate the cosine similarity between candidate member vectors, and select combinations with similarity greater than a threshold to generate recommended solutions;

[0016] S4.3. Path Optimization: For inefficient collaboration paths, add cross-departmental coordination nodes to restructure the collaboration process.

[0017] Preferably, the weighting coefficients α, β, and γ in step S2.3 are determined as follows: collect the organization's collaboration history data for the past 3 months, use the on-time delivery rate of tasks as the objective function, and use the least squares method to fit the optimal coefficients, wherein the value of β in cross-departmental collaboration is 1.2 times that of intra-departmental collaboration.

[0018] Preferably, the information depth coefficient in step S2.3 ijDetermining the density of technical terms based on the communication content: When technical terms account for more than 30%, the depth... ij The value is 0.8; when technical terms account for less than 10%, depth ij The value is 0.3; when the proportion is between 10% and 30%, linear interpolation is used for calculation.

[0019] Preferably, the node centrality analysis in step S3 includes the calculation of direct centrality and betweenness centrality. The hub node is the node with the highest direct centrality in the top 10%, and the implicit coordinator is the node with the highest betweenness centrality in the top 10%.

[0020] Preferably, the criteria for determining a high-cohesion subgroup in step S3 are: the total internal collaboration weight of the subgroup accounts for more than 60% of the total external collaboration weight of the subgroup, and the average information transmission delay within the subgroup is less than 24 hours.

[0021] Preferably, the load growth rate in step S4.1 is calculated as follows: Load growth rate = (current month's collaborative task processing volume - previous month's collaborative task processing volume) / previous month's collaborative task processing volume × 100%; wherein, the collaborative task processing volume is calculated by the frequency of cross-departmental communication involving nodes and the time spent on task approval with a weight of 3:7.

[0022] Preferably, the employee collaboration compatibility vector in step S4.2 is generated in the following way: based on the organization's employee collaboration records for the past 6 months, a 128-dimensional feature vector is generated by training with the Node2Vec algorithm, and the similarity threshold is set to 0.75.

[0023] Preferably, the criteria for determining an inefficient collaborative path in step S4.3 are: the average information transmission delay of the path is greater than 48 hours, and the total weight of the path is less than 60% of the average weight of other paths with the same starting point and ending point.

[0024] Preferably, the preprocessing in step S1 also includes data cleaning: removing duplicate records and invalid interaction data. Invalid interaction data includes instant messaging records with a single communication duration of less than 30 seconds and emails without substantial content.

[0025] Preferably, the optimization method employs a system including:

[0026] Data acquisition module: used to collect and anonymize multi-source collaborative data;

[0027] Network modeling module: used to construct and dynamically update a two-level dynamically weighted directed graph;

[0028] Relationship identification module: used to calculate node centrality and detect highly cohesive subgroups;

[0029] Optimization Decision Module: Used to generate bottleneck warnings, team configuration recommendations, and path optimization solutions;

[0030] Results output module: used to visualize the collaborative network topology and optimization decision results.

[0031] The technical effects and advantages of this invention's method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining are as follows:

[0032] 1. This invention solves the problem of information silos and achieves comprehensive coverage of collaborative data: by collecting multi-source heterogeneous data (email correspondence, meeting logs, instant messaging records, project management data, etc.), and after anonymization and cleaning, it breaks down the barriers of scattered departmental data, provides a complete data foundation for collaborative analysis, and avoids the one-sidedness of analysis caused by existing technologies that rely only on single structured data.

[0033] 2. This invention accurately depicts collaborative relationships and breaks through the limitations of organizational structure: by constructing a two-level dynamic weighted directed graph that covers both department-level (macro) and employee-level (micro) collaborative nodes, and combining dynamic weight dimensions such as interaction frequency, task urgency, and information depth, it quantifies the actual intensity of collaboration, solving the problem that traditional methods that rely solely on formal organizational structures cannot reflect implicit communication paths, and making the depiction of collaborative relationships more realistic.

[0034] 3. This invention dynamically identifies implicit collaborative networks and uncovers key collaborative nodes: By using node centrality analysis (direct centrality, betweenness centrality) and Louvain algorithm subgroup detection, it accurately identifies hub nodes, implicit coordinators, and highly cohesive subgroups, solving the problems of key coordinators not being identified and collaborative networks being opaque in existing technologies, and providing clear direction for resource allocation.

[0035] 4. This invention provides early warning of collaboration bottlenecks and improves the timeliness of decision-making: By monitoring betweenness centrality and load growth rate in conjunction, resource allocation warnings can be triggered before nodes are overloaded, solving the problem of decision-making delays caused by over-reliance on a few nodes in traditional methods, and realizing proactive intervention in collaboration bottlenecks rather than passive response.

[0036] 5. This invention optimizes team configuration and collaboration paths, reducing integration costs and communication delays: Based on compatibility vectors generated from historical collaboration data, it recommends highly matched teams, reducing integration costs caused by personnel mismatch; by reconstructing inefficient paths and adding coordination nodes, it shortens information transmission delays and solves the problems of resource mismatch and inefficient collaboration paths.

[0037] 6. This invention adapts to dynamic organizational changes and maintains the continuity of collaborative optimization: the dynamic weighted directed graph can be adjusted in real time as collaborative data is updated, and the two-level network model can respond to changes in departmental and personnel structures, solving the limitations of traditional static optimization tools that cannot adapt to organizational changes, and ensuring the long-term stability of collaborative optimization effects. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the method flow for identifying and optimizing implicit collaborative relationships in organizations based on multimodal social network mining, as proposed in this invention. Detailed Implementation

[0039] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0040] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0041] Example 1

[0042] refer to Figure 1 As shown, this embodiment provides a method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining, which is used for cross-departmental collaboration optimization in large manufacturing enterprises. Specific implementation details include:

[0043] Implementation scenario: A car manufacturing company with 1,200 employees, including four core departments: R&D, production, sales and purchasing, is experiencing inefficiencies in cross-departmental projects (such as new car development) due to information transmission delays and overloaded collaboration nodes (historical project delay rate is as high as 28%).

[0044] Implementation steps:

[0045] S1. Multi-source collaborative data acquisition and preprocessing:

[0046] Data collection:

[0047] Email system: Extracted cross-departmental email records from the past 3 months, totaling 15,600 entries (including sender, recipient, sending time, body content, and attachment subject);

[0048] Meeting system: Export cross-departmental meeting logs, totaling 210 sessions (including participating departments, participants, meeting duration, agenda items, and resolutions);

[0049] Instant messaging tools: Retrieved cross-department group chat and one-on-one chat logs via API interface, totaling 9,800 messages (including chat participants, time, content, and attachments);

[0050] Project Management System: Export cross-departmental task data, totaling 850 items (including task initiating department, executing department, person in charge, urgency level, completion deadline, and approval process).

[0051] Preprocessing:

[0052] Anonymization: Replace real names with "department code + employee number" (e.g., R&D department employees are marked as "R-001" and "R-002", and production department employees are marked as "P-001", etc.) to hide private information such as ID card number and mobile phone number;

[0053] Data cleaning: Removing invalid data, including:

[0054] Instant messaging records with a single chat duration of less than 30 seconds (a total of 1,200 records, mostly consisting of non-substantive interactions such as "received" and "okay");

[0055] Duplicate emails (800 emails in total, with the same content forwarded multiple times);

[0056] Records of tasks that were not actually performed (30 items in total, cancelled due to plan adjustments).

[0057] S2. Construct a two-level dynamically weighted directed graph:

[0058] S2.1. Node Definition:

[0059] Department-level nodes: R&D (R), Production (P), Sales (S), and Purchasing (C), totaling 4 macro nodes;

[0060] Employee-level nodes: Select core employees from each department who participate in cross-departmental collaboration (participating in cross-departmental tasks ≥ 5 times in the past 3 months), totaling 220 people (60 from R&D, 50 from Production, 50 from Sales, and 60 from Purchasing). The micro-nodes are marked as "department code-number" (e.g., R-001).

[0061] Membership relationship: Employee-level nodes point to their respective department-level nodes through directed edges (e.g., R-001→R).

[0062] S2.2. Definition of edge set:

[0063] Department-level edges: Represent collaboration and interaction between departments, such as the edge between R and P corresponding to task coordination and meeting communication between the R&D department and the production department;

[0064] Employee-level edges: Represent direct collaboration between employees, such as the edge between R-001 and P-005 corresponding to their email exchanges, meeting collaborations, etc.

[0065] Cross-level edges: Only subordinate relationships exist (employee → department), with no collaborative interaction edges.

[0066] S2.3. Edge weight calculation:

[0067] Weighting formula: W ij =α·freq ij +β·urgency ij +γ·depth ij , where α = 0.3, β = 0.4, γ = 0.3 (determined by fitting historical data from the past 3 months, with the on-time delivery rate of the task as the objective function, and using the least squares method);

[0068] Parameter calculation:

[0069] freq ij Interaction frequency (unit: times / month), such as R-001 and P-005 having 12 email interactions, 3 collaborative meetings, and 20 instant messaging interactions per month, with a total frequency of 12 + 3 + 20 = 35.

[0070] urgency ij Task urgency (0.5-1.0): Urgent tasks (completion time < 7 days) are assigned 1.0, regular tasks (7-30 days) are assigned 0.7, and low-priority tasks (> 30 days) are assigned 0.5. The average urgency of the tasks involved in the interaction is taken (e.g., 60% of the collaborative tasks between R-001 and P-005 are urgent tasks, urfency = 0.6 × 1.0 + 0.4 × 0.7 = 0.88).

[0071] depth ij Information depth coefficient (0.3-0.8), calculated based on the density of technical terms in the communication content (the technical terminology database contains 3,000 automotive R&D professional terms). For example, in emails R-001 and P-005, technical terms account for 35%, and depth = 0.8 (because the percentage is >30%).

[0072] Calculation result: W R-001,p-005 =0.3×35+0.4÷0.88+0.3×0.8≈10.5+0.35+0.24=11.09.

[0073] S3. Identification of Implicit Collaboration Relationships:

[0074] Node centrality analysis:

[0075] Direct centrality: Calculate the number of associated edges (total weight) of employee-level nodes. R-003 (R&D Department Project Manager) has a direct centrality of 85 (top 10%), and is designated as a hub node.

[0076] Betweenness centrality: The frequency with which a node is the intermediate point of the shortest path to other nodes. P-012 (process engineer in the production department) has a betweenness centrality of 0.72 (ranked in the top 10%), and is defined as an implicit coordinator (non-management but frequently involved in R&D-production technology docking).

[0077] High cohesive subgroup detection:

[0078] The Louvain algorithm was used to cluster employee-level nodes, resulting in a subgroup G1 = {R-001, R-003, P-012, C-005}.

[0079] Judgment: The total weight of internal collaboration within the subgroup is 280, and the total weight of external collaboration is 110, with a proportion of 280 / (280+110)≈72%>60%; the average information transmission delay within the subgroup (the average time from information transmission to feedback) is 18 hours<24 hours, which meets the criteria for a high-cohesion subgroup.

[0080] S4. Collaboration efficiency optimization decision generation:

[0081] S4.1. Bottleneck Warning:

[0082] The betweenness centrality of S-001 (Sales Director) is calculated to be 0.95 (preset threshold 0.7). The load growth rate = (June task processing volume - May task processing volume) / May task processing volume × 100% = (120 - 95) / 95 × 100% ≈ 26.3% > 15%, triggering a resource allocation warning.

[0083] Decision: It is recommended to add a Deputy Director of Sales to share 50% of the cross-departmental approval tasks.

[0084] S4.2. Team Configuration Optimization:

[0085] Generate collaboration compatibility vectors: Based on employee collaboration records over the past 6 months, train 128-dimensional feature vectors using the Node2Vec algorithm, with R-001 vector as V1, P-012 vector as V2, and C-005 vector as V3;

[0086] Similarity calculation: cos(V1,V2)=0.82, cos(V1,V3)=0.78, cos(V2,V3)=0.85, average similarity=0.82>0.75 (threshold);

[0087] Decision: Recommend G1 subgroup members to participate in the new model development project.

[0088] S4.3. Path Optimization:

[0089] Identifying inefficient paths: The average information transmission delay of "S→R" (from the sales department to the R&D department) is 52 hours > 48 hours, and the total path weight is 8.5 (the average weight of other paths with the same start and end point is 15, 8.5 < 15 × 60% = 9), so it is judged as an inefficient path;

[0090] Decision: Add a "Technology Market Liaison Position" (node ​​M-001) as a backup edge, reconstruct the path to "S→M-001→R", and shorten the information transmission delay to 22 hours.

[0091] Implementation results: Cross-departmental communication delays in new vehicle R&D projects decreased by 45%, and the average project delay rate dropped from 28% to 12%.

[0092] Example 2

[0093] This embodiment provides a method for identifying and optimizing implicit organizational cooperative relationships based on multimodal social network mining, used for dynamic calibration of weight coefficients. Specific implementation details include:

[0094] Implementation scenario: An e-commerce company (with distinct peak and off-peak seasons) needs to adjust the weighting coefficients according to quarterly business fluctuations (emphasizing task urgency during peak seasons and communication frequency during off-peak seasons).

[0095] Implementation steps:

[0096] S1. Data collection and preprocessing: Same as in Example 1, collect data for 3 months each during the peak and off-peak seasons (peak season is Q4, off-peak season is Q2).

[0097] S2. Weighting coefficient calibration:

[0098] Off-season (Q2): Based on "communication frequency", the fitted values ​​are α = 0.5, β = 0.2, and γ = 0.3 (β = 0.2 × 1.2 = 0.24 in cross-departmental collaboration);

[0099] Peak season (Q4): Based on "task urgency", the fitted values ​​are α = 0.2, β = 0.6, and γ = 0.2 (β = 0.6 × 1.2 = 0.72 in cross-departmental collaboration).

[0100] S3-S4: Same as Example 1, calculate weights according to the calibrated coefficients to generate optimization decisions.

[0101] Implementation results: During peak season, the response speed for urgent cross-departmental orders increased by 30%, and during off-season, the communication cost for routine tasks decreased by 20%.

[0102] Example 3

[0103] This embodiment provides a method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining, used for information depth coefficient quantification. Specific implementation details include:

[0104] Implementation scenario: A research institution (collaboration information includes technical reports and administrative notices), which needs to distinguish the impact of information complexity on collaboration.

[0105] Implementation steps:

[0106] S1. Data Collection: including technical meeting minutes (300 copies), administrative notices (200 copies), and project progress reports (500 copies).

[0107] S2. Calculation of information depth coefficient: Technical meeting minutes: technical terms account for 42% → depth = 0.8; Administrative notices: technical terms account for 8% → depth = 0.3; Project progress reports: technical terms account for 25% → linear interpolation: depth = 0.3 + (0.8 - 0.3) × (25% - 10%) / (30% - 10%) = 0.675.

[0108] S3-S4: Calculate weights based on the depth mentioned above, and prioritize the collaborative path resources for high-depth information.

[0109] Implementation results: Communication efficiency of the technical collaboration subgroup (relying on high-depth information) improved by 25%.

[0110] Example 4

[0111] This embodiment provides a method for identifying and optimizing implicit collaborative relationships in organizations based on multimodal social network mining, which is used for node centrality and subgroup detection. The specific implementation content includes: Implementation scenario: A group company (8 branches) needs to identify implicit collaborative networks across branches.

[0112] Implementation steps:

[0113] S1-S2: Construct a network containing branch office level (8 nodes) and employee level (300 nodes), and calculate the weights.

[0114] S3. Relationship Identification:

[0115] Hub nodes: Top 10% of direct centrality (30 people), such as the group's technical director (connecting 8 branch offices);

[0116] Implicit coordinators: Top 10% (30 people) in terms of betweenness centrality, such as branch office liaison specialists (high-frequency transfer of cross-branch information);

[0117] High cohesive subgroup: The weight of the subgroup is 68%, and the average delay is 16 hours → marked as "North China Regional Collaboration Group".

[0118] S4: Optimize resource allocation based on identification results and strengthen the authority of implicit coordinators.

[0119] Implementation results: Cross-branch project collaboration efficiency increased by 35%.

[0120] Example 5

[0121] This embodiment provides a method for identifying and optimizing implicit collaborative relationships within organizations based on multimodal social network mining, used for team configuration optimization. Specific implementation details include:

[0122] Implementation scenario: A software company is forming a new project team and needs to match highly compatible members based on historical data.

[0123] Implementation steps:

[0124] S1. Collection: Employee collaboration records for the past 6 months (including code submissions, test feedback, and meeting collaborations).

[0125] S2. Generate vectors: Train 128-dimensional vectors using Node2Vec, namely employee A (backend) vector V_A, employee B (test) vector V_B, and employee C (frontend) vector V_C.

[0126] S3. Calculate similarity: cos(V_A,V_B)=0.81>0.75, cos(V_A,V_C)=0.65<0.75.

[0127] S4. Team configuration: The "A+B" combination is recommended, as the historical collaboration bug rate is 40% lower than that of "A+C".

[0128] Implementation results: The integration period for the new project team was shortened from 4 weeks to 2 weeks.

[0129] Comparative Example 1

[0130] This comparison provides collaboration capabilities for traditional OA systems.

[0131] Implementation scenario: Using a certain OA system, tasks are assigned according to a fixed organizational structure, and information is transmitted by approval workflow.

[0132] Implementation steps:

[0133] S1: Only structured task data is collected (no unstructured data such as emails and instant messages);

[0134] S2: No dynamic weighting network, weights are fixed (only graded according to task type);

[0135] S3: Unable to identify implicit coordinators and highly cohesive subgroups (dependent on explicit organizational structure);

[0136] S4: Team configuration depends on the manager's experience, and bottleneck warnings are delayed (only triggered when a task is overdue).

[0137] Implementation results: The average delay rate for cross-departmental projects was 40%, and communication costs were 65% higher than in Example 1.

[0138] The comparison between Examples 1-5 and Comparative Example 1 is summarized as follows:

[0139] From the perspective of data collection, Examples 1-5 all adopted multi-source data fusion (email, meeting logs, instant messaging, project management system data), and covered structured and unstructured data by anonymizing and cleaning up invalid information (such as interactions without substantial content within 30 seconds). Comparative Example 1 only relied on structured task data from the OA system, lacking unstructured collaborative behavior information, resulting in insufficient data integrity.

[0140] In terms of network modeling, Examples 1-5 construct a two-level dynamic weighted directed graph containing department and employee nodes. The edge weights are calculated by weighting the interaction frequency, task urgency, and information depth coefficient (as in Example 2, the weight ratio is dynamically adjusted according to peak and off-peak seasons) to quantify the intensity of collaboration. Comparative Example 1 does not have a dynamic network model and only allocates tasks according to a fixed organizational structure. The weights are singular and static, which cannot reflect the actual intensity of collaboration.

[0141] In terms of relationship identification capabilities, Examples 1-5 identify hub nodes and implicit coordinators (such as the cross-branch liaison specialists who are not in management) through node centrality (direct centrality, betweenness centrality) and use the Louvain algorithm to detect highly cohesive subgroups (such as the R&D-production-procurement collaboration group in Example 1); Comparative Example 1 relies only on the explicit organizational structure and cannot identify implicit collaboration networks, resulting in key coordinators being overlooked.

[0142] In terms of optimizing the decision-making mechanism, Examples 1-5 achieve precise intervention: bottlenecks are warned by linking load growth rate with betweenness centrality (e.g., Example 1 triggers resource allocation for overloaded sales directors), highly compatible teams are recommended based on Node2Vec vector similarity (e.g., Example 5 shortens the break-in period by 50%), and coordination nodes are added to inefficient paths (e.g., Example 1 reduces information delay from 52 hours to 22 hours); In contrast, Example 1 relies on human experience for decision-making, with delayed warnings and blind team configuration, only responding passively after the task is overdue.

[0143] The implementation results show that cross-departmental collaboration efficiency in Examples 1-5 was significantly improved: the project delay rate decreased from 40% in Comparative Example 1 to 12%-28% (a 45% reduction in Example 1), cross-branch collaboration efficiency increased by 35% (Example 4), and communication efficiency among technical subgroups increased by 25% (Example 3). Comparative Example 1, due to incomplete data and rigid models, had communication costs 65% higher than in Example 1, and the optimization effect was limited.

[0144] In summary, Examples 1-5, through multi-source data fusion, dynamic weighted network modeling, implicit relationship identification, and precise optimization decision-making, solved the problems of single data, static model, and lagging optimization in Comparative Example 1, and verified the advancement and effectiveness of the technical solution of "integrating social network analysis to optimize cross-departmental collaboration" in the cited documents.

[0145] The above embodiments can be implemented in whole or in part by software, hardware, firmware or other arbitrary combinations. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0146] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0147] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0149] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining, characterized in that, Includes the following steps: S1. Multi-source collaborative data collection and preprocessing: Collect email records, meeting participation logs, instant messaging interaction data and project management system data within the organization, anonymize the data, and retain department and role labels; S2. Construct a two-level dynamically weighted directed graph: S2.

1. The node set includes department-level nodes and employee-level nodes, and the two levels of nodes are connected through a hierarchical relationship; S2.

2. Edge sets represent the collaborative interaction behaviors between nodes, including communication frequency, task association, and information transmission paths; S2.

3. The edge weight is calculated by weighting the interaction frequency, task urgency, and information depth coefficient according to the preset weight coefficient, where the interaction frequency, task urgency, and information depth coefficient each correspond to different weight proportions; S3. Identification of Implicit Collaborative Relationships: Based on the aforementioned dynamic weighted directed graph, hub nodes and implicit coordinators are identified through node centrality analysis, and high-cohesion subgroups are detected through the Louvain algorithm; S4. Collaboration efficiency optimization decision generation: S4.

1. Bottleneck warning: When the node betweenness centrality is greater than the preset threshold and the load growth rate is greater than 15% / month, a resource allocation warning is triggered; S4.

2. Team configuration optimization: Generate employee collaboration compatibility vectors based on historical collaboration data, calculate the cosine similarity between candidate member vectors, and select combinations with similarity greater than a threshold to generate recommended solutions; S4.

3. Path Optimization: For inefficient collaboration paths, add cross-departmental coordination nodes to restructure the collaboration process; In step S2.3, the information depth coefficient is determined based on the density of technical terms in the communication content: when the proportion of technical terms is greater than 30%, the information depth coefficient is 0.8; when the proportion of technical terms is less than 10%, the information depth coefficient is 0.3; when the proportion is between 10% and 30%, it is calculated by linear interpolation. In step S3, the node centrality analysis includes the calculation of direct centrality and betweenness centrality. The hub node is the node with the top 10% direct centrality, and the implicit coordinator is the node with the top 10% betweenness centrality. In step S3, the node centrality analysis includes the calculation of direct centrality and betweenness centrality. The hub node is the node with the top 10% direct centrality, and the implicit coordinator is the node with the top 10% betweenness centrality.

2. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in claim 1, characterized in that, The preset weighting coefficients in step S2.3 are determined in the following way: collect the organization's collaboration history data for the past 3 months, use the on-time delivery rate of tasks as the objective function, and use the least squares method to fit the optimal coefficients. The weight of task urgency in cross-departmental collaboration is 1.2 times that of intra-departmental collaboration.

3. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in claim 1, characterized in that, The load growth rate in step S4.1 is calculated as follows: Load growth rate = (current month's collaborative task processing volume - previous month's collaborative task processing volume) / previous month's collaborative task processing volume × 100%; where the collaborative task processing volume is calculated by weighting the frequency of cross-departmental communication involving nodes and the time spent on task approval at a ratio of 3:

7.

4. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in claim 1, characterized in that, In step S4.2, the employee collaboration compatibility vector is generated as follows: based on the organization's employee collaboration records for the past 6 months, a 128-dimensional feature vector is generated by training the Node2Vec algorithm, and the similarity threshold is set to 0.

75.

5. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in claim 1, characterized in that, The criteria for determining an inefficient collaboration path in step S4.3 are: the average information transmission delay of the path is greater than 48 hours, and the total weight of the path is less than 60% of the average weight of other paths with the same starting point and ending point.

6. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in claim 1, characterized in that, The preprocessing in step S1 also includes data cleaning: removing duplicate records and invalid interaction data. Invalid interaction data includes instant messaging records with a single communication duration of less than 30 seconds and emails without substantial content.

7. The method for identifying and optimizing implicit organizational collaborative relationships based on multimodal social network mining as described in any one of claims 1-6, characterized in that, The optimization method employs a system including: Data acquisition module: used to collect and anonymize multi-source collaborative data; Network modeling module: used to construct and dynamically update a two-level dynamically weighted directed graph; Relationship identification module: used to calculate node centrality and detect highly cohesive subgroups; Optimization Decision Module: Used to generate bottleneck warnings, team configuration recommendations, and path optimization solutions; Results output module: used to visualize the collaborative network topology and optimization decision results.

Images