Project cost allocation and calculation method and system based on research and development factor data

Abstract

The present application relates to the technical field of research and development project management, and particularly relates to a project cost apportionment and calculation method and system based on research and development element data. The method identifies shared elements by constructing a cross-project achievement delivery network, calculates direct and indirect contribution values to form an initial contribution matrix, further identifies bidirectional reference relationships of elements within a project, calculates a collaborative contribution increment, and performs nonlinear correction on the matrix. Finally, the total cost is apportioned according to the corrected matrix, and the apportioned result is locally updated when the data changes. The present application realizes fine and dynamic cost apportionment for complex achievement reuse and collaborative effect among research and development projects.

Description

Technical Field

[0001] This invention relates to the field of R&D project management technology, and in particular to a method and system for project cost allocation and calculation based on R&D element data. Background Technology

[0002] Traditional project cost allocation methods typically rely on direct aggregation or simple proportional allocation principles. A common practice is to allocate the total cost proportionally to each R&D project based on the explicit resource inputs directly consumed, such as labor hours, material costs, and equipment depreciation. Another common approach is to use activity-based costing (ABC), which identifies resource drivers and activity drivers to more accurately trace indirect costs to specific R&D projects.

[0003] Existing methods struggle to accurately identify and quantify the value contribution of cross-project technology reuse. Traditional direct aggregation or proportional allocation methods essentially assume that projects are isolated, completely ignoring the transmission and reuse effects of technological achievements within the project network. This leads to biases in the cost allocation of shared R&D elements, failing to reflect their true cross-project value and resulting in unfair allocation outcomes that may affect the accuracy of project evaluations and subsequent investment decisions.

[0004] Traditional linear and static accounting models are inadequate for handling the nonlinear synergistic effects arising from complex interactions between elements within a research and development project. In actual R&D, different technical modules or elements within a project may have bidirectional and cyclical promoting relationships. When R&D element data or inter-project reference relationships change, traditional methods often require recalculating globally, lacking an efficient local update mechanism. Therefore, a cost allocation method is needed that can more precisely characterize the R&D network structure, dynamically quantify synergistic effects, and support efficient updates. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for project cost allocation and calculation based on R&D element data, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a method for project cost allocation and calculation based on R&D element data, comprising:

[0007] Acquire research and development element data for multiple research and development projects, including element identifiers, associated projects, and resulting technological achievements;

[0008] Based on the citation relationships of technological achievements across different R&D projects, a cross-project achievement transfer network is constructed to identify shared elements reused across multiple R&D projects.

[0009] In the cross-project results delivery network, trace the delivery path of the technological achievements generated by shared elements, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the delivery path, and generate an initial contribution matrix.

[0010] Identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project deliverable network, calculate the synergistic contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix.

[0011] When R&D element data changes, the affected transmission path is located in the cross-project result transmission network, the contribution matrix is ​​locally updated and corrected along the affected transmission path, and the updated cost allocation for each R&D project is output.

[0012] Based on the citation relationships of technological achievements across different R&D projects, a cross-project deliverable network is constructed, identifying shared elements reused across multiple R&D projects, including:

[0013] Extract the citation relationships of technological achievements in R&D element data among different R&D projects. The citation relationships include the source project identifier, the target project identifier, and the citation time, and generate a set of citation relationships.

[0014] A directed graph is constructed based on the source project identifier and the target project identifier in the reference relationship set. The nodes of the directed graph represent technical achievements, and the edges represent reference relationships. The reference time in the reference relationship set is appended to the edges of the directed graph as a time attribute to form a cross-project achievement transfer network.

[0015] Traverse the nodes in the cross-project technology achievement transfer network, count the number of edges pointing to each node that contain different source project identifiers, and generate the cross-project citation count of the technology achievement corresponding to each node.

[0016] Nodes that are referenced across projects more than a preset threshold are filtered out. The R&D elements corresponding to the filtered nodes are extracted from the R&D element data, and the extracted R&D elements are marked as shared elements that can be reused across multiple R&D projects.

[0017] In a cross-project deliverables network, the delivery path of technological achievements generated by shared elements is traced. Based on the delivery path, the direct and indirect contributions of shared elements to each R&D project are calculated, generating an initial contribution matrix including:

[0018] In the cross-project results delivery network, locate the source node corresponding to the technical results generated by the shared elements, traverse from the source node to each target node, record the traversal path, trace the delivery path of the technical results generated by the shared elements, and generate a set of delivery paths.

[0019] Extract the technical feature description text corresponding to each node on each transmission path in the transmission path set, construct the semantic evolution graph between the technical feature description texts of adjacent nodes on each transmission path, and identify the technical fault nodes on the transmission path through the semantic evolution graph.

[0020] Remove transmission paths containing technology gap nodes from the transmission path set, and use the number of remaining single-hop transmission paths as the direct contribution value of the shared elements to the R&D project to which the target node belongs, and use the number of remaining multi-hop transmission paths as the indirect contribution value of the shared elements to the R&D project to which the target node belongs.

[0021] The matrix with the row index of CCB as the shared element identifier and the column index as the R&D project identifier is filled with the direct contribution value and the indirect contribution value to generate the initial contribution matrix.

[0022] Construct a semantic evolution graph of the technical feature description texts of adjacent nodes on each transmission path, and identify technical discontinuity nodes on the transmission path through the semantic evolution graph, including:

[0023] The text describing technical features is semantically vectorized and encoded to generate a set of node semantic vectors.

[0024] Based on the set of semantic vectors of nodes, construct the semantic vector sequence of nodes on each transmission path, calculate the vector angle between adjacent semantic vectors in the semantic vector sequence, use the vector angle as the edge weight between adjacent nodes, use the nodes in the set of transmission paths as graph nodes, connect adjacent graph nodes with edge weights, and construct a semantic evolution graph.

[0025] Nodes with edge weights exceeding a preset angle threshold in the semantic evolution graph are identified as candidate fault nodes. The research and development time interval and overlap of participants in the technological achievements corresponding to the preceding and following adjacent nodes of the candidate fault nodes are extracted.

[0026] When the research and development time interval exceeds a preset time threshold and the overlap of participants is lower than a preset overlap threshold, the candidate fault node is marked as a technology fault node.

[0027] This involves identifying multiple R&D elements within the same R&D project that have bidirectional referencing relationships from a cross-project deliverable network, calculating the synergistic contribution increment of these elements, performing a nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocating the total cost to each R&D project based on the corrected contribution matrix.

[0028] Traverse the nodes in the cross-project deliverable network, extract the project identifier of the corresponding technical deliverable of each node, and filter the nodes with the same project identifier to form a set of nodes within the project.

[0029] Detect bidirectional edge connections between nodes in the node set within the project, identify R&D element pairs with bidirectional reference relationships, mark them as collaborative element combinations, and generate a combination list;

[0030] Extract the bidirectional reference time window of each collaborative element combination in the combination list, calculate the overlapping area of ​​the bidirectional reference time window as the collaborative coefficient, and multiply the collaborative coefficient with the contribution value of the collaborative element combination at the corresponding position in the initial contribution matrix to generate the collaborative contribution increment.

[0031] The incremental contribution of collaboration is nonlinearly and recursively amplified based on the dependency depth between R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution of collaboration to generate a corrected contribution matrix.

[0032] The sum of the weights corresponding to each R&D project column in the modified contribution matrix is ​​used as the project weight. The proportion of the project weight to the total weight of all projects is used as the allocation ratio. The total cost is then allocated to each R&D project according to the allocation ratio.

[0033] The incremental contribution of collaboration is non-linearly recursively amplified based on the dependency depth among R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution, resulting in a corrected contribution matrix including:

[0034] Extract the nodes corresponding to the technological achievements generated by each R&D element within the collaborative element combination in the cross-project achievement delivery network, traverse the delivery path of each node in the cross-project achievement delivery network, and count the number of intermediate nodes in the delivery path from the starting node to the ending node as the dependency depth value.

[0035] A non-linear amplification function is constructed by using the dependency depth value as an exponential parameter. The collaborative contribution increment is then recursively amplified by exponential powers through the non-linear amplification function to generate the amplified collaborative contribution increment.

[0036] Extract the row and column coordinates of the collaborative element combination in the initial contribution matrix, and superimpose the amplified collaborative contribution increment onto the original contribution value at the corresponding position in the initial contribution matrix according to the row and column coordinates. Update the contribution allocation value of the R&D project to which the collaborative element combination belongs in the initial contribution matrix to generate the corrected contribution matrix.

[0037] When R&D element data changes, the affected delivery path is located in the cross-project outcome delivery network. The contribution matrix is ​​locally updated and corrected along the affected delivery path, and the updated allocated costs for each R&D project are output, including:

[0038] The R&D elements that have changed in the R&D element data are identified as changed elements. All transmission paths of the changed elements in the cross-project result transmission network are traced, and a set of affected transmission paths is generated.

[0039] The impact attenuation coefficient is calculated based on the length of each transmission path in the set of affected transmission paths, and the impact attenuation coefficient is applied to the original contribution value corresponding to the affected transmission path for recalculation.

[0040] The recalculated contribution values ​​are updated to the corresponding positions in the corrected contribution matrix to generate a partially updated corrected contribution matrix. The sum of the weight values ​​corresponding to each R&D project column in the partially updated corrected contribution matrix is ​​calculated as the weight of the updated project. The total cost is redistributed to each R&D project according to the proportion of the weight of the updated project to the total weight of all updated projects.

[0041] A second aspect of this invention provides a project cost allocation and calculation system based on R&D element data, comprising:

[0042] The data acquisition unit is used to acquire R&D element data of multiple R&D projects, wherein the R&D element data includes element identifier, project to which it belongs, and the resulting technological achievements.

[0043] The shared identification unit is used to build a cross-project results transfer network based on the citation relationship of technological achievements in different R&D projects, and to identify shared elements reused in multiple R&D projects.

[0044] The contribution calculation unit is used to trace the transmission path of technological achievements generated by shared elements in a cross-project results transmission network, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the transmission path, and generate an initial contribution matrix.

[0045] The collaborative correction unit is used to identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project result delivery network, calculate the collaborative contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the collaborative contribution increment, generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix.

[0046] The dynamic update unit is used to locate the affected transmission path in the cross-project result transmission network when the R&D element data changes, locally update and correct the contribution matrix along the affected transmission path, and output the updated cost allocation for each R&D project.

[0047] A third aspect of the present invention provides an electronic device, comprising:

[0048] processor;

[0049] Memory used to store processor-executable instructions;

[0050] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0051] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0052] This method achieves refined cost allocation for R&D projects by constructing a cross-project deliverable network and identifying shared elements. It fully considers the reuse of R&D elements across multiple projects, avoiding biases caused by neglecting indirect contributions in traditional cost allocation methods, thus improving the rationality and accuracy of cost attribution. By identifying R&D elements with bidirectional referencing relationships within the same project and calculating the incremental collaborative contribution, the initial contribution matrix is ​​nonlinearly corrected. This mechanism captures the synergistic effects between R&D elements, reflecting the additional value generated by element combinations. The corrected contribution matrix more closely reflects the nonlinear cost impact of element interactions in actual R&D, making the cost allocation results more scientific and credible. It effectively reduces system computational overhead, adapts to frequent changes in the relationship between elements and deliverables during R&D, and ensures the timeliness of cost allocation results. This advances cost allocation from simple direct aggregation to a level based on networked contribution tracing and nonlinear collaborative correction, providing reliable data support for cost management, resource optimization, and benefit evaluation of R&D projects. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the project cost allocation and calculation method based on R&D element data, as described in an embodiment of the present invention.

[0054] Figure 2 This is a flowchart illustrating the cost allocation process for R&D projects based on collaborative contribution correction, as described in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the project cost allocation and calculation method based on R&D element data according to an embodiment of the present invention. Figure 1 As shown, the methods for project cost allocation and calculation based on R&D element data include:

[0058] Acquire research and development element data for multiple research and development projects, including element identifiers, associated projects, and resulting technological achievements;

[0059] Based on the citation relationships of technological achievements across different R&D projects, a cross-project achievement transfer network is constructed to identify shared elements reused across multiple R&D projects.

[0060] In the cross-project results delivery network, trace the delivery path of the technological achievements generated by shared elements, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the delivery path, and generate an initial contribution matrix.

[0061] Identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project deliverable network, calculate the synergistic contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix.

[0062] When R&D element data changes, the affected transmission path is located in the cross-project result transmission network, the contribution matrix is ​​locally updated and corrected along the affected transmission path, and the updated cost allocation for each R&D project is output.

[0063] Based on the citation relationships of technological achievements across different R&D projects, a cross-project deliverable network is constructed, identifying shared elements reused across multiple R&D projects, including:

[0064] Extract the citation relationships of technological achievements in R&D element data among different R&D projects. The citation relationships include the source project identifier, the target project identifier, and the citation time, and generate a set of citation relationships.

[0065] A directed graph is constructed based on the source project identifier and the target project identifier in the reference relationship set. The nodes of the directed graph represent technical achievements, and the edges represent reference relationships. The reference time in the reference relationship set is appended to the edges of the directed graph as a time attribute to form a cross-project achievement transfer network.

[0066] Traverse the nodes in the cross-project technology achievement transfer network, count the number of edges pointing to each node that contain different source project identifiers, and generate the cross-project citation count of the technology achievement corresponding to each node.

[0067] Nodes that are referenced across projects more than a preset threshold are filtered out. The R&D elements corresponding to the filtered nodes are extracted from the R&D element data, and the extracted R&D elements are marked as shared elements that can be reused across multiple R&D projects.

[0068] After acquiring R&D element data from multiple R&D projects, it is necessary to extract the citation relationships between technological achievements from the underlying data. Specifically, for each R&D project, the technical documents, code comments, design documents, and technical reports in its R&D element data are scanned, and text parsing technology is used to identify statements that explicitly cite technological achievements from other projects. Extracting citation relationships requires recording three key pieces of information: the source project identifier (identifying the R&D project number that initiated the citation), the target project identifier (identifying the R&D project number to which the cited technological achievement belongs), and the citation time (recording the specific timestamp of the citation). In practice, the citation time can be obtained from the document's modification time, the commit time in the version control system, or the time information in the technical review records. All extracted citation relationships are then aggregated into a citation relationship set, where each record contains three fields: source project identifier, target project identifier, and citation time.

[0069] When constructing a cross-project deliverable network based on a set of reference relationships, a directed graph data structure is used for representation. Each record in the reference relationship set is traversed, and the technical achievement corresponding to the referenced target project is used as a node in the directed graph. If the node does not already exist in the graph, it is created. For the relationship between the source project identifier and the referenced target project identifier, a directed edge is created from the node corresponding to the referenced target project to the node corresponding to the source project. The direction of this directed edge represents the direction of the technical achievement delivery. The reference time of the record in the reference relationship set is appended as a time attribute to the newly created directed edge, forming a time-labeled directed edge. This process is repeated until all records in the reference relationship set have been processed, ultimately forming a complete cross-project deliverable network. The node set of this network covers all technical achievements that are referenced or initiate references between projects, while the edge set fully depicts the delivery relationships and timing of technical achievements between different R&D projects.

[0070] After the cross-project deliverables network is constructed, shared elements reused across multiple R&D projects are identified by analyzing the network topology. Specifically, an empty statistics table is initialized to record the cross-project citation count for each node in the network. Starting from any node in the network, all incoming edges ending at that node are retrieved. For each incoming edge, the project identifier of the starting node is extracted and added to the current node's source project set. To avoid duplicate counting of multiple citations of the same technical achievement by the same project, the source project set uses a set data structure to ensure the uniqueness of the project identifier. The number of different project identifiers in the source project set is counted, and this number is recorded as the cross-project citation count of the corresponding technical achievement for the current node in the statistics table. All nodes in the network are processed in the same way to complete the cross-project citation count for all nodes.

[0071] After obtaining the cross-project citation counts for each node, a preset threshold is set as the standard for judging shared elements. This threshold is configured according to the enterprise's R&D collaboration model and project scale, typically set to 2 or 3, indicating that a technological achievement is considered shared only if it is cited by at least two or three different projects. All records in the statistics table are traversed, and nodes with cross-project citation counts exceeding the preset threshold are selected. The identifiers of these nodes are collected into a candidate shared node list. For each node in the candidate shared node list, the original R&D element data is traced back, and the specific R&D element that generated the technological achievement is located based on the mapping relationship between the node and the technological achievement. R&D elements may include R&D personnel, technical equipment, software tools, or experimental materials, etc. A precise match is performed between the element identifier field in the R&D element data and the association relationship between the technological achievement. All matched R&D elements are extracted, and a shared identifier field is added to these elements in the R&D element data. Setting the field value to true indicates that the element is reused across multiple R&D projects. Simultaneously, the set of projects served by each shared element is recorded, providing basic data for subsequent cost allocation calculations.

[0072] In identifying shared elements, it's necessary to address the version evolution of technological achievements. When a technological achievement is updated after being referenced by other projects, the inheritance relationship between versions should be tracked, treating different versions of the technological achievement as different states of the same node. Specifically, this involves maintaining a version chain of technological achievements in the R&D element data, recording the generation time of each version and the identifier of its predecessor version. When constructing a cross-project achievement transfer network, all versions of the same technological achievement are mapped to the same node, but the specific version and reference time are distinguished in the edge's time attribute. This approach ensures that version evolution does not lead to duplicate identification of shared elements, while preserving the complete timeline of technological achievement transfer.

[0073] Furthermore, loop structures may exist in cross-project deliverable networks, indicating that multiple projects have formed a closed loop of mutual referencing technology. When counting cross-project citations, nodes in the loop may receive higher citation counts, but these citations are partly due to the circular transfer of technological achievements rather than genuine cross-project reuse. To avoid misjudgment, loop detection is performed on the cross-project deliverable network before screening shared elements, using depth-first search or topology sorting algorithms to identify strongly connected components in the network. For nodes inside the loop, only citations from outside the loop are counted when counting cross-project citations, excluding mutual references between nodes within the loop. This process ensures that the identification results of shared elements more accurately reflect the true reuse of technological achievements between independent projects, avoiding statistical bias caused by technical exchanges between projects.

[0074] In a cross-project deliverables network, the delivery path of technological achievements generated by shared elements is traced. Based on the delivery path, the direct and indirect contributions of shared elements to each R&D project are calculated, generating an initial contribution matrix including:

[0075] In the cross-project results delivery network, locate the source node corresponding to the technical results generated by the shared elements, traverse from the source node to each target node, record the traversal path, trace the delivery path of the technical results generated by the shared elements, and generate a set of delivery paths.

[0076] Extract the technical feature description text corresponding to each node on each transmission path in the transmission path set, construct the semantic evolution graph between the technical feature description texts of adjacent nodes on each transmission path, and identify the technical fault nodes on the transmission path through the semantic evolution graph.

[0077] Remove transmission paths containing technology gap nodes from the transmission path set, and use the number of remaining single-hop transmission paths as the direct contribution value of the shared elements to the R&D project to which the target node belongs, and use the number of remaining multi-hop transmission paths as the indirect contribution value of the shared elements to the R&D project to which the target node belongs.

[0078] The matrix with the row index of CCB as the shared element identifier and the column index as the R&D project identifier is filled with the direct contribution value and the indirect contribution value to generate the initial contribution matrix.

[0079] After constructing a cross-project deliverables network, it is necessary to conduct in-depth contribution measurement of the shared elements within the network. The cross-project deliverables network is organized in a node-edge format, where nodes represent technological deliverables generated in various R&D projects, and edges represent the citation relationships between deliverables. Each node carries metadata information, including a unique identifier for the deliverable, its project number, the identifier of the element that generated the deliverable, and a technical feature description text. The technical feature description text typically contains structured or semi-structured information such as the technical principles, implementation methods, and application scope of the deliverable, providing a data foundation for subsequent semantic analysis.

[0080] For shared elements reused across multiple R&D projects, the first step is to locate the source node corresponding to the generated technological achievement within the cross-project deliverable network. The source node is determined by precisely matching the element identifier with the element identifier in the node's metadata. A shared element may generate multiple technological achievements, thus potentially corresponding to multiple source nodes. Starting from each source node, a breadth-first search or depth-first search algorithm is used to expand outwards, visiting all target nodes reachable through reference relationships. During the traversal, the complete path sequence from the source node to each target node is recorded; this path sequence consists of sequentially connected node identifiers. All recorded path sequences are then aggregated to form the delivery path set for the shared element. Each path in the delivery path set reflects the specific trajectory of the technological achievement's propagation from the source project to other projects; the path length, i.e., the number of hops, reflects the degree of indirectness in the delivery.

[0081] To accurately assess contribution values, it is necessary to identify potential technical gaps in the transmission path. A technical gap refers to a situation where, although there are reference relationships between adjacent nodes in a transmission path, the technical content lacks substantial semantic continuity; this could be a formal reference or a weakly related reference. Identifying technical gaps requires extracting the technical feature description text of each node in each path of the transmission path set. For each transmission path, the technical feature description text between adjacent node pairs is extracted in node order, constructing a set of text pairs. Natural language processing is performed on each text pair, including word segmentation, part-of-speech tagging, entity recognition, and key term extraction. The extracted key terms are mapped to a predefined technical technology ontology or knowledge graph to obtain semantic vector representations of the terms. The semantic similarity between the technical feature description texts of adjacent nodes is calculated using metrics such as cosine similarity, word-shift distance, or sentence embedding similarity based on a pre-trained language model.

[0082] When constructing a semantic evolution graph, each node along the transmission path is treated as a node in the graph, and the edge weights between adjacent nodes are set as the semantic similarity of the corresponding text pairs. The semantic evolution graph visually displays the evolution trend of technical content along the transmission path. A semantic similarity threshold is set; when the semantic similarity between adjacent nodes is lower than this threshold, a technological gap is identified, and subsequent nodes in the adjacent nodes are marked as technological gap nodes. The semantic similarity threshold can be adjusted according to the characteristics of specific technical fields; the threshold can be appropriately lowered for fields with rapid technological evolution, and raised for fields with strong technological accumulation. By traversing the semantic evolution graphs of all transmission paths, all technological gap nodes are identified, and the transmission paths in which these nodes reside are recorded.

[0083] The transfer paths containing technology discontinuity nodes are removed from the set of transfer paths to ensure genuine semantic continuity of technology transfer in the remaining paths. This removal operation is performed by checking each path for marked technology discontinuity nodes; if a node is found, the path is removed from the set. The remaining set of transfer paths is further categorized into single-hop transfer paths and multi-hop transfer paths. A single-hop transfer path is a path directly from the source node to the target node, with a path length of 1, indicating that the technological achievements generated by the shared elements are directly referenced by the target project. A multi-hop transfer path is a path with a path length greater than 1, indicating that the technological achievements are referenced by the target project after being transformed or recreated by intermediate projects, reflecting an indirect transfer relationship.

[0084] When calculating the direct contribution of shared elements to each R&D project, the number of paths in the remaining single-hop transmission paths that use nodes of each R&D project as target nodes is counted. For a specific R&D project, its direct contribution value equals the total number of single-hop paths ending at nodes of that project. This value reflects the frequency and intensity of the direct adoption of the shared element's technological achievements by that project. When calculating the indirect contribution value, the number of paths in the remaining multi-hop transmission paths that use nodes of each R&D project as target nodes is counted. The indirect contribution value reflects the indirect impact of shared elements on the target project through the technology transfer of other projects. The calculation of the indirect contribution value can further consider the impact of path length; longer paths can be assigned smaller weighting coefficients to reflect the attenuation of influence during transmission.

[0085] After obtaining the direct and indirect contributions of all shared elements to each R&D project, an initial contribution matrix is ​​constructed. The initial contribution matrix adopts a two-dimensional matrix structure, with row indices corresponding to shared element identifiers and column indices corresponding to R&D project identifiers. The matrix's dimensions are determined by the number of shared elements and the number of R&D projects. The contribution value of each shared element to each R&D project is filled into the corresponding position in the matrix. This can be done by storing direct and indirect contribution values ​​in separate matrices, or by combining them into a single contribution value through weighted summation. If a weighted summation method is used, weight coefficients are set for direct and indirect contributions, typically with the direct contribution weight coefficient being greater than the indirect contribution weight coefficient to reflect the priority of direct impact. Each element in the initial contribution matrix reflects the overall contribution intensity of the corresponding shared element to the corresponding R&D project. This matrix provides a quantitative basis for subsequent cost allocation calculations and lays the data foundation for identifying incremental collaborative contributions and performing nonlinear corrections.

[0086] Construct a semantic evolution graph of the technical feature description texts of adjacent nodes on each transmission path, and identify technical discontinuity nodes on the transmission path through the semantic evolution graph, including:

[0087] The text describing technical features is semantically vectorized and encoded to generate a set of node semantic vectors.

[0088] Based on the set of semantic vectors of nodes, construct the semantic vector sequence of nodes on each transmission path, calculate the vector angle between adjacent semantic vectors in the semantic vector sequence, use the vector angle as the edge weight between adjacent nodes, use the nodes in the set of transmission paths as graph nodes, connect adjacent graph nodes with edge weights, and construct a semantic evolution graph.

[0089] Nodes with edge weights exceeding a preset angle threshold in the semantic evolution graph are identified as candidate fault nodes. The research and development time interval and overlap of participants in the technological achievements corresponding to the preceding and following adjacent nodes of the candidate fault nodes are extracted.

[0090] When the research and development time interval exceeds a preset time threshold and the overlap of participants is lower than a preset overlap threshold, the candidate fault node is marked as a technology fault node.

[0091] In the process of cross-project transfer of technological achievements, the semantic changes of technical feature description texts can reflect the continuity and integrity of knowledge transfer. To deeply analyze the quality of knowledge transfer along the transfer path, it is necessary to conduct semantic evolution analysis on the technical feature description texts of adjacent nodes in the transfer path.

[0092] For each node in the cross-project deliverable network, its technical feature description text records the core technical content of the technological achievement generated by that element. These technical feature description texts are used as objects of semantic analysis, and semantic vectorization encoding is performed using a pre-trained language model. Specifically, the technical feature description texts are input into semantic encoding models such as BERT or RoBERTa, and the deep semantic representation of the text is extracted through the model's encoding layer, thereby converting the technical feature description text of each node into a high-dimensional dense vector. This encoding process can capture the semantic features of domain terminology associations, similarity of technical principles, and key innovations in the technical description. The encoded vector dimension is typically set to 768 or 1024 dimensions to ensure sufficient expression of semantic information. By encoding the technical feature description texts of all nodes, a set of node semantic vectors is generated, where each vector corresponds to an element node in the deliverable network.

[0093] For each delivery path identified in the cross-project deliverable network, the nodes along the path are arranged in chronological order of the technological achievements' citation. Based on the node arrangement, corresponding semantic vectors are extracted from the node semantic vector set to construct a semantic vector sequence for that delivery path. This sequence reflects the semantic evolution trajectory of the technological achievements along the delivery path. Within the semantic vector sequence, the angle between adjacent semantic vectors is calculated to quantify the degree of semantic difference between adjacent nodes. Let the semantic vectors of two adjacent nodes on the delivery path be... and The vector angle between the two The cosine similarity was calculated using the cosine similarity function. Then, the angle between the vectors is obtained through the inverse cosine function. The larger the vector angle, the more significant the semantic differences in the technical features of adjacent nodes, potentially indicating a break or jump in the transfer of technical knowledge. The calculated vector angle is used as the edge weight between adjacent nodes in the transfer path. Using all nodes in the transfer path set as graph nodes, adjacent graph nodes on each transfer path are connected by edge weights to construct a complete semantic evolution graph. This graph expresses the semantic evolution process of technological achievements in the form of a directed graph, and the edge weights intuitively reflect the semantic coherence of technical knowledge during the transfer process. The semantic evolution graph not only contains the semantic evolution information of a single transfer path but also integrates the cross-relationships between multiple transfer paths, comprehensively showcasing the pattern of technical knowledge transfer among R&D elements.

[0094] In a semantic evolution graph, nodes that may indicate technological discontinuities are identified. All edges in the graph are traversed, and successor nodes connected to edges whose weights exceed a preset angle threshold are extracted and marked as candidate discontinuity nodes. The preset angle threshold is typically set between 60 and 90 degrees, and the specific value can be adjusted according to the semantic change characteristics of the technological field. When the vector angle between adjacent nodes exceeds this threshold, it indicates a significant deviation in the semantic descriptions of the technological features of the two nodes, potentially signifying a significant change in technological direction or a knowledge gap during the transfer of technological knowledge.

[0095] For each identified candidate fault node, the R&D time information of the technological achievements corresponding to its preceding and following adjacent nodes is extracted. The time interval between the completion time of the technological achievement of the preceding adjacent node and the start time of the technological achievement of the candidate fault node is calculated. This time interval reflects the time delay in the transfer of technological achievements. A longer time interval may lead to the forgetting of technical knowledge, changes in the technical environment, or the loss of technical personnel, thus affecting the continuous transfer of technical knowledge. Simultaneously, the R&D participant lists for the technological achievements of the candidate fault node and the technological participants lists for the technological achievements of the preceding adjacent node are extracted, and the overlap between the two lists is calculated. The overlap of participants is obtained by calculating the ratio of the intersection to the union of the two sets of participants, i.e. ,in This represents the set of participants from the preceding adjacent node. This represents the set of participants in the candidate fault node. Low overlap among participants indicates significant changes in the R&D team of the technological achievement, which may lead to the interruption of tacit knowledge transfer.

[0096] A comprehensive assessment is conducted to determine whether candidate fault nodes constitute genuine technology fault nodes. When the R&D time interval exceeds a preset time threshold and the overlap of participants is below a preset overlap threshold, the candidate fault node is considered to have a significant fault effect in the transfer of technical knowledge and is marked as a technology fault node. The preset time threshold can be set to 180 to 365 days, and the preset overlap threshold can be set to 0.2 to 0.3. Identifying technology fault nodes is crucial for accurately assessing the contribution value of shared elements, as the existence of technology faults weakens the actual impact of shared element technological achievements in subsequent projects. When calculating the indirect contribution value of shared elements to each R&D project, the contribution attenuation coefficient needs to be adjusted according to the number and location of technology fault nodes in the transfer path to ensure that the contribution value calculation accurately reflects the transfer effect of technical knowledge.

[0097] By constructing a semantic evolution graph and identifying technological discontinuity nodes, obstacles to knowledge flow at the semantic level of technological achievement transfer can be revealed, providing a quantitative assessment basis for the quality of technological knowledge transfer in R&D management. This analysis process combines natural language processing technology with graph analysis methods, enabling in-depth analysis of the cross-project technological achievement transfer process. This helps identify key links that require strengthened knowledge management and optimize the allocation strategy of R&D resources.

[0098] like Figure 2 The diagram shows the cost allocation flowchart for R&D projects based on collaborative contribution correction in this embodiment.

[0099] This involves identifying multiple R&D elements within the same R&D project that have bidirectional referencing relationships from a cross-project deliverable network, calculating the synergistic contribution increment of these elements, performing a nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocating the total cost to each R&D project based on the corrected contribution matrix.

[0100] Traverse the nodes in the cross-project deliverable network, extract the project identifier of the corresponding technical deliverable of each node, and filter the nodes with the same project identifier to form a set of nodes within the project.

[0101] Detect bidirectional edge connections between nodes in the node set within the project, identify R&D element pairs with bidirectional reference relationships, mark them as collaborative element combinations, and generate a combination list;

[0102] Extract the bidirectional reference time window of each collaborative element combination in the combination list, calculate the overlapping area of ​​the bidirectional reference time window as the collaborative coefficient, and multiply the collaborative coefficient with the contribution value of the collaborative element combination at the corresponding position in the initial contribution matrix to generate the collaborative contribution increment.

[0103] The incremental contribution of collaboration is nonlinearly and recursively amplified based on the dependency depth between R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution of collaboration to generate a corrected contribution matrix.

[0104] The sum of the weights corresponding to each R&D project column in the modified contribution matrix is ​​used as the project weight. The proportion of the project weight to the total weight of all projects is used as the allocation ratio. The total cost is then allocated to each R&D project according to the allocation ratio.

[0105] After generating the initial contribution matrix, it is necessary to further consider the potential synergistic effects between different R&D elements within the same R&D project. These synergies often manifest as mutual referencing and support between the technological achievements of two or more R&D elements, with their overall contribution to the project exceeding the simple sum of their individual contributions. To accurately quantify these synergies, in-depth analysis is required within the cross-project deliverables network.

[0106] The process iterates through all nodes in the cross-project deliverable network, with each node representing a specific technological achievement. For each node, its associated metadata information is extracted, particularly the project identifier to which the technological achievement belongs. By comparing project identifiers, all nodes belonging to the same R&D project are grouped together to form a project-specific node set. For example, project A might contain nodes N1, N2, N5, N... 12 The technological achievements corresponding to these nodes all originated from the research and development process of Project A. The same node aggregation operation is performed on each independent project in the network to ensure that no internal nodes of any project are missed.

[0107] After obtaining the internal node set for each project, an internal connectivity analysis is performed on the node set for each project. This involves checking if there is a directed edge connecting any two nodes within the set, paying particular attention to whether there is a corresponding slave node N. i Pointing to node N j The edges and from node N j Pointing to node N i The edges. This bidirectional edge connection implies a mutual reference relationship between the two technological achievements, i.e., node N. i The corresponding technological achievement referenced node N during its generation process. j The corresponding technological achievements, and node N j The corresponding technological achievement also cited node N. i The corresponding results. This bidirectional referencing typically occurs during iterative R&D processes, where early results provide the foundation for later research, while later improvements, in turn, optimize the application of earlier solutions. When such bidirectional connections are detected, the R&D elements associated with the corresponding two nodes are marked as a pair of collaborative elements, and this element pair is recorded in a combination list. The combination list includes information such as element identifier pairs, their respective projects, and the time of reference.

[0108] For each pair of synergistic element combinations in the combination list, extract the time information of their bidirectional references. Each reference relationship is accompanied by a timestamp, recording the start and end times of the reference; this time period constitutes the time window of the reference. For a synergistic element combination, there are references in two directions, thus two time windows. Calculate the overlap between these two time windows on the time axis; the ratio of the overlap duration to a certain baseline duration can be used as the synergy coefficient φ. The longer the overlap duration, the closer the synergy between the two elements in the R&D process, and the more significant the synergistic effect. In the initial contribution matrix, each R&D element has a corresponding contribution value for different projects. For the two elements in the synergistic element combination, find their corresponding contribution values ​​C in the matrix. i and C jThe synergy coefficient φ is then calculated with these two contribution values ​​to generate the synergy contribution increment ΔC. This increment reflects the additional contribution due to the synergy effect.

[0109] When calculating the incremental contribution of collaborative elements, the dependency depth within the combination of collaborative elements also needs to be considered. Dependency depth measures the degree of interdependence between two elements, which can be determined by analyzing factors such as the connection strength, citation frequency, and complexity of the technological achievements they produce in the network. The greater the dependency depth, the deeper the synergistic effect produced by the combination of the two elements. This effect is not linearly additive but exhibits a non-linear growth characteristic. A non-linear recursive amplification mechanism is introduced to adjust the incremental contribution of collaborative elements based on the dependency depth d. When the dependency depth is small, the amplification coefficient is close to 1, and the incremental contribution of collaborative elements basically remains the same. When the dependency depth increases, the amplification coefficient grows exponentially or exponentially, resulting in a significant amplification of the incremental contribution of collaborative elements. The recursive amplification process can be carried out layer by layer. Each layer calculates the amplification factor based on the current dependency depth and passes the result to the next layer until the preset recursion depth limit is reached or the dependency relationship terminates.

[0110] After calculating and amplifying the synergistic contribution increments, these increments are applied to correct the initial contribution matrix. For each pair of synergistic elements in the combination list, their corresponding row and column positions are located in the initial contribution matrix, and the calculated amplified synergistic contribution increments are superimposed on the original contribution values ​​at the corresponding positions. This correction process may involve addition operations on matrix elements or more complex matrix operations to ensure that the corrected values ​​accurately reflect the contribution of the synergistic effect. After the correction operation is completed, a corrected contribution matrix is ​​generated. This matrix contains more comprehensive contribution information than the initial contribution matrix, considering both the direct and indirect contributions of individual elements, as well as the additional contributions generated by synergy between elements.

[0111] The total cost is allocated based on a modified contribution matrix. Each column of the modified contribution matrix corresponds to a research and development project, and each row corresponds to a research and development element. For a given research and development project, the total weight value of the project is obtained by summing all elements in its columns. This weighting value comprehensively reflects the overall contribution of all R&D elements to the project. The sum of the weighting values ​​for all R&D projects is calculated. Then calculate the proportion of each item's weight value to the total weight value, i.e., the weight of each item. Proportion of apportionment This ratio represents the relative importance and contribution of the project within the overall R&D system. It also represents the company's total R&D costs. Allocated according to the sharing ratio, project The cost to be allocated is This cost-sharing method ensures that the cost allocation matches the actual contribution of each project, avoiding the unfairness that may result from simply allocating costs based on the number of people or time.

[0112] Through the above processing steps, synergistic effects are incorporated into the cost allocation calculation system, resulting in more accurate and reasonable allocation results. The synergistic effects among R&D elements within the project are quantified and identified, and their enhancing effects on the overall project contribution are accurately reflected through a non-linear amplification mechanism. The revised contribution matrix serves as the basis for allocation, providing a more comprehensive perspective on contribution assessment and offering more reliable data support for management's resource allocation decisions. In practical applications, the calculation method of the synergy coefficient, the quantification method based on the depth of dependence, and the specific form of the non-linear amplification function can be flexibly adjusted according to the R&D characteristics of different industries and enterprises to adapt to diverse R&D management needs. The entire process remains transparent and traceable, facilitating subsequent auditing and optimization adjustments.

[0113] The incremental contribution of collaboration is non-linearly recursively amplified based on the dependency depth among R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution, resulting in a corrected contribution matrix including:

[0114] Extract the nodes corresponding to the technological achievements generated by each R&D element within the collaborative element combination in the cross-project achievement delivery network, traverse the delivery path of each node in the cross-project achievement delivery network, and count the number of intermediate nodes in the delivery path from the starting node to the ending node as the dependency depth value.

[0115] A non-linear amplification function is constructed by using the dependency depth value as an exponential parameter. The collaborative contribution increment is then recursively amplified by exponential powers through the non-linear amplification function to generate the amplified collaborative contribution increment.

[0116] Extract the row and column coordinates of the collaborative element combination in the initial contribution matrix, and superimpose the amplified collaborative contribution increment onto the original contribution value at the corresponding position in the initial contribution matrix according to the row and column coordinates. Update the contribution allocation value of the R&D project to which the collaborative element combination belongs in the initial contribution matrix to generate the corrected contribution matrix.

[0117] In cross-project deliverable networks, once multiple R&D elements within the same R&D project are identified as forming collaborative relationships, a deep quantitative analysis of the dependencies between these collaborative elements is required. The core of this quantitative analysis lies in assessing the degree of technological dependence between collaborative elements and transforming this dependence into a non-linear correction parameter for the contribution value. Specifically, for identified combinations of collaborative elements, it is necessary to locate the network node corresponding to the technological deliverable generated by each R&D element within the topology of the cross-project deliverable network. These nodes form specific positional relationships within the network, and by analyzing the connections between nodes, the delivery order and dependence strength of technological deliverables during the R&D process can be revealed.

[0118] When extracting the nodes corresponding to the technological achievements generated by each R&D element within the collaborative element combination, a mapping relationship between nodes and elements is established. Each technological achievement node carries attribute information such as the unique identifier of the achievement, its generation time, and the R&D project to which it belongs. In the cross-project achievement transfer network, nodes are connected by directed edges, the direction of which indicates the direction of reference of the technological achievement. For any R&D element within the collaborative element combination, its generated technological achievement may act as a starting node to transfer to other nodes, or as an intermediate node to receive achievements from other nodes and continue to transfer them, or as a terminating node to only receive and not transfer them outwards. When traversing all transfer paths of each node in the network, a depth-first search strategy is adopted, starting from the node and tracing downwards along the direction of the directed edge, recording all reachable path sequences.

[0119] During path traversal, for each complete transmission path, the number of intermediate nodes traversed from the starting node to the ending node is counted. This number reflects the depth of the technological achievement's transmission hierarchy in the R&D process; a larger value indicates that the achievement is positioned earlier in the R&D chain and has a wider impact. For each R&D element within the collaborative element combination, the average number of intermediate nodes for its generated technological achievements across all transmission paths is calculated, and this average is used as the dependency depth value for that R&D element. When a R&D element's technological achievements have multiple transmission paths of different lengths, a weighted average of the number of intermediate nodes across all paths is required. The weighting coefficient can be set according to the importance of the path, for example, using the investment scale of downstream projects or the frequency of achievement application as weighting factors.

[0120] After obtaining the dependency depth value of each R&D element, this value is used as a key parameter of the nonlinear amplification function. The design goal of the nonlinear amplification function is to establish an exponential mapping relationship between dependency depth and contribution increment, so that R&D elements with greater dependency depth obtain a more significant contribution amplification effect. In specific construction, the dependency depth value is combined with a preset basic amplification coefficient to form an expression for calculating the amplification factor. This function adopts an exponential form to ensure that the amplification effect grows at an accelerating rate with increasing dependency depth. For the synergistic contribution increment already calculated for the combination of synergistic elements, an exponential operation is performed through the nonlinear amplification function. During the operation, the dependency depth value is directly used as the exponent, so that the amplification factor grows exponentially with the dependency depth.

[0121] When performing exponential recursive amplification, the incremental collaborative contribution is multiplied by an amplification coefficient built based on dependency depth. This amplification process is not a simple linear scaling, but a non-linear enhancement achieved through an exponential function. When the dependency depth is small, the amplification effect is relatively mild, ensuring that the correction of the contribution value remains within a reasonable range; when the dependency depth is large, the amplification effect is significantly enhanced, fully reflecting the cumulative impact of deep dependencies on project contributions. During the recursive amplification process, for each R&D element within the collaborative element combination, the amplified contribution increment is calculated separately based on its individual dependency depth value. Then, these individual amplified values ​​are summed to generate the overall amplified collaborative contribution increment for the collaborative element combination. This recursive processing method ensures that the amplification effect considers both the deep influence of individual elements and the overall effect generated by the collaboration of multiple elements.

[0122] After nonlinearly amplifying the synergistic contribution increment, the amplification result needs to be accurately mapped to the initial contribution matrix. The rows of the initial contribution matrix represent each R&D element, the columns represent each R&D project, and the matrix elements represent the contribution value of a specific element to a specific project. For a combination of synergistic elements, the row indices of all R&D elements within that combination in the initial contribution matrix, as well as the column indices of the R&D projects to which these elements belong, are extracted to form a set of row and column coordinate positions. These coordinate positions identify the matrix cells where contribution value correction is needed. For the amplified synergistic contribution increment generated by the combination of synergistic elements, it is decomposed into the matrix cells corresponding to each coordinate position according to a preset allocation rule. The allocation rule can be based on the proportional allocation of each element's importance weight in the synergistic relationship, or it can be differentiated based on the magnitude of each element's original contribution value.

[0123] When performing the contribution value aggregation operation, the amplified collaborative contribution increment is added to the original contribution value at the corresponding position in the initial contribution matrix according to the calculated allocation ratio. This aggregation process is an incremental correction based on the original contribution, rather than replacing the original value. For multiple collaborative elements in the same R&D project, the aggregation of their contribution increments will lead to an increase in the overall contribution value of the project in the revised matrix. During the matrix update process, it is necessary to simultaneously adjust the contribution values ​​of other non-collaborative elements to ensure that the total contribution value of the entire matrix remains constant. Specifically, the sum of the amplified collaborative contribution increments is proportionally deducted from the contribution values ​​of other elements, or the global contribution ratio is adjusted through normalization to ensure that the revised matrix still meets the total cost constraint.

[0124] When updating the contribution allocation values ​​of R&D projects belonging to collaborative element combinations in the initial contribution matrix, the sum of all elements in the corresponding matrix column for each project is recalculated to confirm the total contribution value obtained by the project from all elements. Due to the superposition of collaborative contribution increments, the total contribution value of the project will increase significantly, which truly reflects the additional value brought by the synergy effect. At the same time, for each R&D element within the collaborative element combination, its numerical distribution in the corresponding row of the matrix is ​​updated to more accurately reflect the impact of deep dependencies in the contribution allocation to different projects. After completing the correction of the contribution values ​​of all collaborative element combinations, the generated corrected contribution matrix maintains the same structure as the initial contribution matrix, but the numerical distribution has been non-linearly adjusted to more accurately reflect the actual impact of cross-project synergy on cost allocation.

[0125] After the modified contribution matrix is ​​generated, subsequent cost allocation calculations are performed based on it. The value of each element in the matrix represents the modified contribution weight of the corresponding element to the corresponding project. The summation and normalization of all element values ​​in each project column yields the cost allocation ratio for each project. This cost allocation method based on the modified contribution matrix fully considers the deep dependencies and non-linear amplification effects of collaborative elements, resulting in a fairer and more reasonable cost allocation and avoiding the underestimation of collaborative value by traditional linear allocation methods. Through this non-linear recursive amplification and precise matrix modification technique, the scientific allocation of costs for complex R&D projects is achieved, providing reliable technical support for cost management in multi-project parallel R&D environments.

[0126] When R&D element data changes, the affected delivery path is located in the cross-project outcome delivery network. The contribution matrix is ​​locally updated and corrected along the affected delivery path, and the updated allocated costs for each R&D project are output, including:

[0127] The R&D elements that have changed in the R&D element data are identified as changed elements. All transmission paths of the changed elements in the cross-project result transmission network are traced, and a set of affected transmission paths is generated.

[0128] The impact attenuation coefficient is calculated based on the length of each transmission path in the set of affected transmission paths, and the impact attenuation coefficient is applied to the original contribution value corresponding to the affected transmission path for recalculation.

[0129] The recalculated contribution values ​​are updated to the corresponding positions in the corrected contribution matrix to generate a partially updated corrected contribution matrix. The sum of the weight values ​​corresponding to each R&D project column in the partially updated corrected contribution matrix is ​​calculated as the weight of the updated project. The total cost is redistributed to each R&D project according to the proportion of the weight of the updated project to the total weight of all updated projects.

[0130] In real-world R&D management scenarios, the data for R&D elements is not static. For example, the cost of a certain R&D element may change due to personnel adjustments, or the citation relationships of a technological achievement may be added or deleted as the project progresses. To maintain the real-time accuracy of cost allocation results, a dynamic update mechanism needs to be established, rather than recalculating the entire network with each change.

[0131] When the system detects changes in R&D element data, it first needs to identify which specific R&D elements have changed. These changes may include adjustments to element costs, modifications to the technological achievements generated by the element, changes to the project to which the element belongs, or additions or deletions to the citation relationships between technological achievements. These changed R&D elements are marked as changed elements, and their element identifiers and the specific details of the changes are recorded. Suppose that a set of changed elements is detected at a certain point in time, containing several element identifiers.

[0132] For each changing element, a backtracking analysis needs to be performed in the cross-project deliverable network. Starting with a given changing element, the process traces downstream along the reference relationships of the technological deliverables generated by that element, identifying all potentially affected paths. Specifically, starting from the node corresponding to the changing element, a depth-first or breadth-first traversal is performed along the directed edges, recording all node sequences reachable from that changing element. Each node sequence constitutes a delivery path. All these delivery paths are then aggregated to form a set of affected delivery paths. Each path in this set contains a complete delivery chain starting from the changing element, passing through several intermediate technological deliverable nodes, and finally reaching a target project.

[0133] After identifying the set of affected transmission paths, not all contribution values ​​on these paths are updated in the same way. The degree of influence of the changing element on different paths varies, and this degree of influence is closely related to the path length. Path length refers to the number of edges traversed between the changing element node and the target project node. The longer the path, the more technological achievement transformation stages it passes through, and the influence of the changing element will gradually attenuate during transmission. Therefore, it is necessary to calculate an influence attenuation coefficient for each affected transmission path.

[0134] The attenuation coefficient can be calculated using an exponential attenuation model. Let the length of a transmission path be the path length value, and the attenuation factor be a positive number less than 1, for example, a value between 0.8 and 0.95. The attenuation coefficient can be expressed as the attenuation factor raised to the power of the path length value. Thus, when the path length is 1, the attenuation coefficient is close to 1, indicating almost complete influence; when the path length is 2, the attenuation coefficient decreases; the longer the path length, the smaller the attenuation coefficient, indicating a weaker influence. This method allows for the quantification of the impact of changing factors on downstream projects at different distances.

[0135] After obtaining the impact attenuation coefficient for each affected transmission path, the original contribution values ​​corresponding to these paths need to be recalculated. The original contribution value refers to the direct or indirect contribution value corresponding to that transmission path in the revised contribution matrix. Taking a transmission path from a changed element to the target project as an example, assume that the original contribution value corresponding to this path in the revised contribution matrix is ​​a certain value. If the cost of the changed element changes, for example, from the original cost value to a new cost value, then the new contribution value of this path should be recalculated based on the new cost value.

[0136] The specific recalculation process needs to consider the type of change. If it's a change in factor costs, the new contribution value equals the product of the new cost value and the weights of each transmission along the path, multiplied by the impact attenuation coefficient. If it's a change in technological achievement citation relationships, such as adding a new citation relationship, a new path needs to be added to the transmission path set, and an initial contribution value and impact attenuation coefficient need to be calculated for this new path. If a citation relationship is deleted, the corresponding transmission path needs to be removed from the affected transmission path set, and the contribution value corresponding to that path needs to be set to zero in the matrix. If it's a change in collaborative relationships, the incremental collaborative contribution of the affected factors needs to be recalculated, and the corresponding nonlinear corrections need to be updated.

[0137] After recalculating the contribution values ​​of all affected transmission paths, these newly calculated contribution values ​​are updated to the corresponding positions in the corrected contribution matrix. The corrected contribution matrix is ​​a two-dimensional matrix, where rows correspond to R&D elements and columns correspond to R&D projects. For each path in the set of affected transmission paths, its position in the matrix is ​​determined based on its starting element and ending project, and the original value at that position is replaced with the recalculated contribution value. If a matrix position corresponds to multiple transmission paths, the contribution values ​​of these paths are summed before updating. For unaffected transmission paths, their corresponding contribution values ​​in the matrix remain unchanged, thus achieving local updates rather than global recalculation, significantly improving computational efficiency.

[0138] In the locally updated corrected contribution matrix, each column represents a research and development project, and the sum of all elements in that column represents the total contribution of all research and development elements to that project. This total contribution value is used as the updated project weight for that project. The updated project weight is calculated for each project by iterating through all columns of all research and development projects, and then the updated project weights of all projects are summed to obtain the total updated project weight.

[0139] After obtaining the updated project weights for each project and the sum of all updated project weights, the cost can be reallocated. The cost allocated to each R&D project equals the total cost multiplied by the proportion of that project's updated project weight to the sum of all updated project weights. Assuming the total cost is a fixed value, the updated project weight for a project is its weight value, and the sum of all updated project weights is the total weight value, then the allocated cost for that project = total cost × project weight value / total weight value. This calculation is performed on all R&D projects to obtain the updated allocated cost for each project.

[0140] The updated cost allocation for each R&D project is output for management decision-making. The output can be a list or a report, including the identifier of each project, the allocated cost before the update, the allocated cost after the update, and the changed amount and percentage. This partial update mechanism allows the system to quickly respond to data changes, accurately reflecting the impact of changing factors on the cost allocation of each project without recalculating the entire network. This ensures the real-time nature and accuracy of the cost allocation results, while significantly reducing computational resource consumption. It is particularly suitable for real-world applications with a large number of R&D factors and frequent changes.

[0141] A second aspect of this invention provides a project cost allocation and calculation system based on R&D element data, comprising:

[0142] The data acquisition unit is used to acquire R&D element data of multiple R&D projects, wherein the R&D element data includes element identifier, project to which it belongs, and the resulting technological achievements.

[0143] The shared identification unit is used to build a cross-project results transfer network based on the citation relationship of technological achievements in different R&D projects, and to identify shared elements reused in multiple R&D projects.

[0144] The contribution calculation unit is used to trace the transmission path of technological achievements generated by shared elements in a cross-project results transmission network, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the transmission path, and generate an initial contribution matrix.

[0145] The collaborative correction unit is used to identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project result delivery network, calculate the collaborative contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the collaborative contribution increment, generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix.

[0146] The dynamic update unit is used to locate the affected transmission path in the cross-project result transmission network when the R&D element data changes, locally update and correct the contribution matrix along the affected transmission path, and output the updated cost allocation for each R&D project.

[0147] A third aspect of the present invention provides an electronic device, comprising:

[0148] processor;

[0149] Memory used to store processor-executable instructions;

[0150] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0151] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0152] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for project cost allocation and calculation based on R&D element data, characterized in that, include: Acquire research and development element data for multiple research and development projects, including element identifiers, associated projects, and resulting technological achievements; Based on the citation relationships of technological achievements across different R&D projects, a cross-project achievement transfer network is constructed to identify shared elements reused across multiple R&D projects. In the cross-project results delivery network, trace the delivery path of the technological achievements generated by shared elements, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the delivery path, and generate an initial contribution matrix. Identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project deliverable network, calculate the synergistic contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix. When R&D element data changes, the affected transmission path is located in the cross-project result transmission network, the contribution matrix is ​​locally updated and corrected along the affected transmission path, and the updated cost allocation for each R&D project is output.

2. The method according to claim 1, characterized in that, Based on the citation relationships of technological achievements across different R&D projects, a cross-project deliverable network is constructed, identifying shared elements reused across multiple R&D projects, including: Extract the citation relationships of technological achievements in R&D element data among different R&D projects. The citation relationships include the source project identifier, the target project identifier, and the citation time, and generate a set of citation relationships. A directed graph is constructed based on the source project identifier and the target project identifier in the reference relationship set. The nodes of the directed graph represent technical achievements, and the edges represent reference relationships. The reference time in the reference relationship set is appended to the edges of the directed graph as a time attribute to form a cross-project achievement transfer network. Traverse the nodes in the cross-project technology achievement transfer network, count the number of edges pointing to each node that contain different source project identifiers, and generate the cross-project citation count of the technology achievement corresponding to each node. Nodes that are referenced across projects more than a preset threshold are filtered out. The R&D elements corresponding to the filtered nodes are extracted from the R&D element data, and the extracted R&D elements are marked as shared elements that can be reused across multiple R&D projects.

3. The method according to claim 1, characterized in that, In a cross-project deliverables network, the delivery path of technological achievements generated by shared elements is traced. Based on the delivery path, the direct and indirect contributions of shared elements to each R&D project are calculated, generating an initial contribution matrix including: In the cross-project results delivery network, locate the source node corresponding to the technical results generated by the shared elements, traverse from the source node to each target node, record the traversal path, trace the delivery path of the technical results generated by the shared elements, and generate a set of delivery paths. Extract the technical feature description text corresponding to each node on each transmission path in the transmission path set, construct the semantic evolution graph between the technical feature description texts of adjacent nodes on each transmission path, and identify the technical fault nodes on the transmission path through the semantic evolution graph. Remove transmission paths containing technology gap nodes from the transmission path set, and use the number of remaining single-hop transmission paths as the direct contribution value of the shared elements to the R&D project to which the target node belongs, and use the number of remaining multi-hop transmission paths as the indirect contribution value of the shared elements to the R&D project to which the target node belongs. The matrix with the row index of CCB as the shared element identifier and the column index as the R&D project identifier is filled with the direct contribution value and the indirect contribution value to generate the initial contribution matrix.

4. The method according to claim 3, characterized in that, Construct a semantic evolution graph of the technical feature description texts of adjacent nodes on each transmission path, and identify technical discontinuity nodes on the transmission path through the semantic evolution graph, including: The text describing technical features is semantically vectorized and encoded to generate a set of node semantic vectors. Based on the set of semantic vectors of nodes, construct the semantic vector sequence of nodes on each transmission path, calculate the vector angle between adjacent semantic vectors in the semantic vector sequence, use the vector angle as the edge weight between adjacent nodes, use the nodes in the set of transmission paths as graph nodes, connect adjacent graph nodes with edge weights, and construct a semantic evolution graph. Nodes with edge weights exceeding a preset angle threshold in the semantic evolution graph are identified as candidate fault nodes. The research and development time interval and overlap of participants in the technological achievements corresponding to the preceding and following adjacent nodes of the candidate fault nodes are extracted. When the research and development time interval exceeds a preset time threshold and the overlap of participants is lower than a preset overlap threshold, the candidate fault node is marked as a technology fault node.

5. The method according to claim 1, characterized in that, This involves identifying multiple R&D elements within the same R&D project that have bidirectional referencing relationships from a cross-project deliverable network, calculating the synergistic contribution increment of these elements, performing a nonlinear correction on the initial contribution matrix based on the synergistic contribution increment to generate a corrected contribution matrix, and allocating the total cost to each R&D project based on the corrected contribution matrix. Traverse the nodes in the cross-project deliverable network, extract the project identifier of the corresponding technical deliverable of each node, and filter the nodes with the same project identifier to form a set of nodes within the project. Detect bidirectional edge connections between nodes in the node set within the project, identify R&D element pairs with bidirectional reference relationships, mark them as collaborative element combinations, and generate a combination list; Extract the bidirectional reference time window of each collaborative element combination in the combination list, calculate the overlapping area of ​​the bidirectional reference time window as the collaborative coefficient, and multiply the collaborative coefficient with the contribution value of the collaborative element combination at the corresponding position in the initial contribution matrix to generate the collaborative contribution increment. The incremental contribution of collaboration is nonlinearly and recursively amplified based on the dependency depth between R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution of collaboration to generate a corrected contribution matrix. The sum of the weights corresponding to each R&D project column in the modified contribution matrix is ​​used as the project weight. The proportion of the project weight to the total weight of all projects is used as the allocation ratio. The total cost is then allocated to each R&D project according to the allocation ratio.

6. The method according to claim 5, characterized in that, The incremental contribution of collaboration is non-linearly recursively amplified based on the dependency depth among R&D elements within the collaborative element combination. The initial contribution matrix is ​​then corrected based on the amplified incremental contribution, resulting in a corrected contribution matrix including: Extract the nodes corresponding to the technological achievements generated by each R&D element within the collaborative element combination in the cross-project achievement delivery network, traverse the delivery path of each node in the cross-project achievement delivery network, and count the number of intermediate nodes in the delivery path from the starting node to the ending node as the dependency depth value. A non-linear amplification function is constructed by using the dependency depth value as an exponential parameter. The collaborative contribution increment is then recursively amplified by exponential powers through the non-linear amplification function to generate the amplified collaborative contribution increment. Extract the row and column coordinates of the collaborative element combination in the initial contribution matrix, and superimpose the amplified collaborative contribution increment onto the original contribution value at the corresponding position in the initial contribution matrix according to the row and column coordinates. Update the contribution allocation value of the R&D project to which the collaborative element combination belongs in the initial contribution matrix to generate the corrected contribution matrix.

7. The method according to claim 1, characterized in that, When R&D element data changes, the affected delivery path is located in the cross-project outcome delivery network. The contribution matrix is ​​locally updated and corrected along the affected delivery path, and the updated allocated costs for each R&D project are output, including: The R&D elements that have changed in the R&D element data are identified as changed elements. All transmission paths of the changed elements in the cross-project result transmission network are traced, and a set of affected transmission paths is generated. The impact attenuation coefficient is calculated based on the length of each transmission path in the set of affected transmission paths, and the impact attenuation coefficient is applied to the original contribution value corresponding to the affected transmission path for recalculation. The recalculated contribution values ​​are updated to the corresponding positions in the corrected contribution matrix to generate a partially updated corrected contribution matrix. The sum of the weight values ​​corresponding to each R&D project column in the partially updated corrected contribution matrix is ​​calculated as the weight of the updated project. The total cost is redistributed to each R&D project according to the proportion of the weight of the updated project to the total weight of all updated projects.

8. A project cost allocation and calculation system based on R&D element data, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire R&D element data of multiple R&D projects, wherein the R&D element data includes element identifier, project to which it belongs, and the resulting technological achievements. The shared identification unit is used to build a cross-project results transfer network based on the citation relationship of technological achievements in different R&D projects, and to identify shared elements reused in multiple R&D projects. The contribution calculation unit is used to trace the transmission path of technological achievements generated by shared elements in a cross-project results transmission network, calculate the direct and indirect contribution values ​​of shared elements to each R&D project based on the transmission path, and generate an initial contribution matrix. The collaborative correction unit is used to identify multiple R&D elements with bidirectional referencing relationships within the same R&D project from the cross-project result delivery network, calculate the collaborative contribution increment of multiple R&D elements, perform nonlinear correction on the initial contribution matrix based on the collaborative contribution increment, generate a corrected contribution matrix, and allocate the total cost to each R&D project based on the corrected contribution matrix. The dynamic update unit is used to locate the affected transmission path in the cross-project result transmission network when the R&D element data changes, locally update and correct the contribution matrix along the affected transmission path, and output the updated cost allocation for each R&D project.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images