A Two-Dimensional Matching Supplier Intelligent Price Query Collaboration Method and System

By standardizing preprocessing and dynamically reorganizing supplier matching in a two-dimensional manner, the problem of materials remaining in the original process after the bill of materials is corrected is solved, and dynamic linkage of supplier matching is achieved, ensuring the synergy and accuracy of price inquiries.

CN122550262BActive Publication Date: 2026-11-17JINRUNFANGZHOU SCI & TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202610716218.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-05-22
Publication Date
2026-11-17
Estimated Expiration
2046-05-22

AI Technical Summary

Technical Problem

In existing technologies, the standardization and correction of the bill of materials and the two-dimensional supplier matching are performed as independent steps. This results in the materials remaining in the original process cluster after the correction, the supplier matching range of the category dimension being outdated, and the delivery constraints of the time sequence dimension being misaligned with the actual construction needs, causing the price query collaboration to fail.

Method used

By acquiring price request data, performing standardized preprocessing, generating a standardized material record set, obtaining process attribution based on project type tags, calculating correlation scores, performing hierarchical clustering to generate material combination clusters, performing two-dimensional supplier matching, receiving error correction feedback, updating material attributes, regenerating the corrected category path, triggering dynamic reorganization of combination clusters, updating the latest arrival date, re-executing supplier matching, and sending change notifications and supplementary quotation requests.

Benefits of technology

It enables dynamic and interconnected reassessment of material properties and dual-dimensional supplier matching, ensuring that quotation requests are consistent with the revised actual needs, avoiding invalid and missing quotations, and improving the coordination and accuracy of price inquiries.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122550262B_ABST
    Figure CN122550262B_ABST
Patent Text Reader

Abstract

This invention relates to the field of engineering procurement price inquiry technology, and discloses a collaborative method and system for intelligent price inquiry of suppliers with dual-dimensional matching. The method includes: acquiring price inquiry request data and performing material standardization preprocessing; performing material process attribution mapping and intra-process material aggregation to generate a combination cluster with the latest arrival date of the materials; performing dual-dimensional supplier matching from the category dimension and the time sequence dimension and pushing quotation requests; receiving quotation data and processing supplier error correction feedback; detecting process attribution changes caused by error correction; performing dynamic reorganization of the combination cluster; re-performing dual-dimensional supplier matching and pushing supplementary quotation requests; and summarizing quotation data and generating a phased procurement quotation recommendation scheme.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of engineering procurement price inquiry technology, specifically to a collaborative method and system for intelligent price inquiry of suppliers with dual-dimensional matching. Background Technology

[0002] In engineering material procurement and price inquiry scenarios, the material list submitted by the purchaser typically involves material requirements for multiple construction stages. Existing processes require simultaneously filtering matching suppliers and sending price requests based on both product category coverage and delivery time.

[0003] In existing technologies, material standardization error correction and two-dimensional supplier matching are executed as independent steps. When a supplier submits error correction feedback during the quotation process, resulting in a change in the material category path, the construction process attribution for that material may also change. Since the error correction does not trigger a re-evaluation of the two-dimensional matching conditions, the corrected material remains in the original process cluster. The supplier matching scope in the category dimension depends on the category path, and the delivery constraints in the time sequence dimension depend on the process time constraints. Changes in the category path cause the basic matching conditions for both dimensions to fail simultaneously. The original process cluster lacks the material, resulting in incomplete category coverage. The newly assigned process also omits the material, leading to a systematic deviation between the submitted quotation request and the corrected actual needs, thus causing the price query coordination to fail. Summary of the Invention

[0004] This invention provides a collaborative method and system for intelligent price query of suppliers with dual-dimensional matching, which solves the technical problems in related technologies such as insufficient standardization of bill of materials, single supplier matching dimension, and inability of error correction feedback during the quotation process to dynamically drive process attribution and combination cluster reorganization.

[0005] This invention discloses a collaborative method for intelligent price query of suppliers with dual-dimensional matching, which includes at least the following steps: obtaining price query request data, performing standardized preprocessing on the bill of materials, generating a standardized set of material records containing category paths, and extracting project type tags, project address information and project start date; Standard construction process data is obtained based on project type tags. The category paths of each material are matched with the material requirement list of each process to determine the process affiliation. The correlation score of materials within the same process is calculated and hierarchical clustering is performed to generate material combination clusters. The latest arrival date of materials is estimated based on process dependencies. For each material combination cluster, perform supplier matching in terms of category and time sequence, take the intersection of the two dimensions to generate a list of candidate suppliers and push the quotation request; Receive supplier quotation data and error correction feedback, group and statistically analyze error correction feedback by attribute dimension, and extract reliable error correction results that meet the consistency threshold; Based on the reliable error correction results, update the material properties and regenerate the corrected category path. Then, rematch the corrected category path with the material requirement list of each process to obtain the corrected process number. Compare the process numbers before and after the correction to identify the materials belonging to the process. Remove the materials with changed ownership from the original cluster and add them to the material set of the new ownership process. Recalculate the relevance score, perform dynamic cluster reorganization, and update the latest arrival date of the materials. For the clusters affected by the reorganization, a two-dimensional supplier matching process is re-executed based on the revised category path and the updated latest material arrival date. Change notices are sent to the original suppliers and supplementary quotation requests are pushed to the newly matched suppliers.

[0006] Furthermore, the correlation score is calculated by weighting the matching relationship strength value, the historical co-occurrence frequency value, and the category proximity value; wherein, the matching relationship strength value indicates the degree to which the two materials are specified as a matching pair in the construction specifications; The historical co-occurrence frequency value represents the frequency with which two materials appear simultaneously in historical purchase orders; the category proximity value represents the normalized reciprocal of the distance between the category paths of the two materials in the category system tree structure. The matching relationship strength value, the historical co-occurrence frequency value, and the category proximity value are processed to the same numerical range using a mean normalization method based on the range before participating in the weighted calculation.

[0007] Furthermore, the determination criteria for the consistency threshold include: the first condition is that the frequency of occurrence of the correction value accounts for a proportion of the total number of all feedbacks in the same attribute dimension that reaches a preset frequency proportion. The second condition is that the number of suppliers submitting the corrected value is not less than the preset minimum number of suppliers. When both conditions are met, the corrected value is considered a reliable error correction result. If there is no corrected value that meets the consistency threshold on the same attribute dimension, the error correction feedback for that attribute dimension is not effective, and the original value from the standardization preprocessing stage is retained.

[0008] Furthermore, the dynamic reorganization of the combined cluster also includes: determining whether the number of remaining materials in the original combined cluster is lower than the minimum aggregation threshold; if so, dissolving the original combined cluster and downgrading the remaining materials to independent material entries. Conversely, the original cluster structure is retained; the correlation score between the material whose ownership has been changed and each material in the existing clusters within the new process is calculated. When the correlation score between the material whose ownership has been changed and at least one material in a cluster exceeds the aggregation threshold, the material whose ownership has been changed is merged into the corresponding cluster; otherwise, the material whose ownership has been changed is included in the new process as an independent material item.

[0009] Furthermore, when multiple attribution change materials in the same batch point to the same new process, the merging operations of these attribution change materials are executed sequentially in descending order of relevance score; After each attribute change material is incorporated, the member list of the combined cluster within the new process is updated so that the relevance score calculation of subsequent attribute change materials is based on the latest combined cluster status.

[0010] Furthermore, in the time-series supplier matching, the method for determining whether the historical delivery cycle meets the latest material arrival date constraint is: to obtain the expected arrival date by adding the candidate supplier's historical average delivery cycle to the current date; When the expected arrival date is not later than the latest arrival date of the materials, the supplier is deemed to meet the timing dimension constraint; wherein, the latest arrival date of the materials is calculated as follows: starting from the project start date, the earliest start date of each process is determined according to the process dependency relationship, and then the lower limit of the process duration range is subtracted from the earliest start date of each process.

[0011] Furthermore, when pushing quotation requests, exploratory quotation requests are simultaneously pushed to candidate suppliers for materials whose confidence level is lower than a preset threshold during the standardized preprocessing stage; The exploratory quotation request indicates that the standardization results of the material are pending confirmation, and the supplier submits error correction feedback simultaneously when submitting the quotation.

[0012] Furthermore, after pushing the supplementary quote request, it also includes: converting the credible error correction results into incremental training samples and storing them in the buffer pool. Each incremental training sample contains the original text features before error correction and the standardized mapping results after error correction. When the number of samples in the buffer pool reaches the batch threshold, the incremental fine-tuning of the standardized mapping model is triggered; wherein, the standardized mapping model takes the text features of the structured material entries as input and outputs the category probability distribution and confidence level corresponding to the category path.

[0013] Furthermore, it also includes summarizing all valid quotation data and calculating a comprehensive score for each material combination cluster and individual material item within each process based on the normalized value of total cost, delivery coordination risk coefficient, and normalized value of time margin. Among them, the value of the delivery coordination risk coefficient is positively correlated with the number of suppliers involved in the quotation scheme; The normalized value of the time margin is obtained by normalizing the average of the difference between the delivery time commitment of each supplier and the latest arrival date of the materials; the quotation schemes in each process are sorted according to the comprehensive score, and the phased procurement quotation recommendation scheme is generated by arranging them according to the construction process sequence.

[0014] This invention provides a dual-dimensional matching supplier intelligent price query collaborative system, comprising: The standardized preprocessing module is used to acquire price inquiry request data, perform standardized preprocessing on the bill of materials to generate a standardized material record set containing category paths, and extract project type tags, project address information and project start date; The process attribution mapping and aggregation module is used to obtain standard construction process flow data based on project type labels, match the category path of each material with the material requirement list of each process to determine the process attribution, calculate the correlation score of materials within the same process and perform hierarchical clustering to generate material combination clusters, and calculate the latest arrival date of materials based on process dependencies. The dual-dimensional supplier matching module is used to perform supplier matching in terms of category and time sequence for each material combination cluster, and generate a candidate supplier list by taking the intersection of the two dimensions of candidate suppliers and pushing quotation requests. The error correction feedback processing module is used to receive supplier quotation data and error correction feedback, group and statistically analyze error correction feedback by attribute dimension, and extract reliable error correction results that meet the consistency threshold. The process attribution change detection module is used to update material properties based on reliable error correction results and regenerate the corrected category path. The corrected category path is then matched with the material requirement list of each process to obtain the corrected process number. The process numbers before and after the correction are compared to identify materials whose process attribution has changed. The dynamic reorganization module of the composite cluster is used to remove materials with changed ownership from the original composite cluster and add them to the material set of the new ownership process, recalculate the relevance score, perform dynamic reorganization of the composite cluster, and update the latest arrival date of the materials. The supplementary matching and push module is used to re-perform two-dimensional supplier matching for the clusters affected by the reorganization based on the revised category path and the updated latest arrival date of materials, send change notifications to the original suppliers and push supplementary quotation requests to the newly matched suppliers.

[0015] This invention adds a process attribution change detection step after the standardized error correction process. By comparing the construction process numbers before and after correction, when a change in the category path is detected that causes a change in the process number, it automatically triggers dynamic reorganization of the combination cluster and a joint reassessment of the two-dimensional supplier matching. The materials with changed attribution are removed from the original combination cluster and added to the material set of the new process. The supplier selection scope in the category dimension is recalculated based on the corrected category path, and the delivery constraints in the time sequence dimension are recalibrated based on the latest arrival date of the materials in the new process. As a result, the corrected materials no longer remain in the original process's combination cluster, and the supplier coverage in the category dimension and the delivery constraints in the time sequence dimension are updated to be consistent with the corrected material attributes. This overcomes the problems of outdated matching results and misaligned delivery times when standardized error correction and two-dimensional supplier matching are executed independently. By sending change notifications to the original suppliers and pushing supplementary quotation requests to the newly matched suppliers, the content of the quotation requests is consistent with the actual needs after correction, avoiding invalid and omitted quotations, and realizing collaborative correction of price queries. Attached Figure Description

[0016] Figure 1 This is a flowchart of the dual-dimensional matching supplier intelligent price query collaborative method provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the confidence distribution of the material standardization mapping provided in an embodiment of the present invention; Figure 3 This is a schematic diagram showing the distribution of material process attribution and latest arrival date provided in the embodiments of the present invention; Figure 4 This is a schematic diagram of the consistency statistics of supplier error correction feedback provided in an embodiment of the present invention; Figure 5 This is a schematic diagram showing the comparison between material correlation score and aggregation threshold provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the two-dimensional supplier matching and screening results provided in an embodiment of the present invention; Figure 7 This is a schematic diagram comparing the comprehensive scoring of the material 03 quotation scheme in the embodiments of the present invention across three dimensions; Figure 8 This is a schematic diagram summarizing the comprehensive scoring of the material 03 quotation scheme provided in the embodiment of the present invention; Figure 9 This is a schematic diagram illustrating the changes in process attribution and arrival date before and after the category path change provided in this embodiment of the invention. Detailed Implementation

[0017] Example 1; In engineering material procurement and price inquiry scenarios, the material list submitted by the procuring party typically involves material requirements across multiple construction stages. It's necessary to simultaneously filter matching suppliers and submit price requests based on both category coverage and delivery time dimensions. When a supplier submits error correction feedback on the material description during the quotation process, causing a change in the material category path, the construction stage attribution for that material may also change. Since supplier matching at the category level depends on the category path, and supplier matching at the time level depends on the process time constraints, changes in the category path will cause the fundamental conditions for both dimensions of matching to simultaneously fail. In the existing process, standardized error correction and two-dimensional supplier matching are executed independently. After error correction, the linkage and re-evaluation of the two-dimensional matching conditions are not triggered. The corrected materials remain within the original process cluster, the supplier matching scope at the category level is outdated, the delivery constraints at the time level are misaligned with actual construction needs, the original process cluster has incomplete category coverage due to the absence of this material, the newly assigned process omits this material, and the submitted price requests create a systematic deviation from the corrected actual needs, resulting in the failure of price query coordination.

[0018] A collaborative method for dual-dimensional matching supplier intelligent price query according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain price request data and perform material standardization preprocessing; The system acquires user-submitted price request data, performs natural language processing on the material list text within the request data to identify material names, specifications, and quantities, and generates structured material entries. It then performs standardization preprocessing on each structured material entry, mapping material names and specifications to a predefined category system to generate a set of standardized material records. Each standardized material record contains a standardized material keyword, an attribute list, and a category path. Finally, it extracts project type tags, project address information, and project start date from the price request data.

[0019] It should be noted that the category path refers to the complete hierarchical path of a material in the category system from the top-level category node to the bottom-level category node. For example, if the category path generated after standardization preprocessing of a material is "cement / silicate cement / ordinary silicate cement", it means that the material belongs to the bottom-level category of ordinary silicate cement in the silicate cement subcategory under the cement category.

[0020] The above standardization preprocessing is performed using a standardization mapping model. This model takes the textual features of structured material entries as input, outputs a fully connected classification layer, and outputs the category probability distribution and confidence score corresponding to the category path. The standardization mapping model is trained in a supervised manner, using historical standardization mapping results as labeled data, with cross-entropy loss as the optimization objective, and employing the Adam optimization algorithm for parameter updates.

[0021] Step 2: Perform material process attribution mapping and intra-process material aggregation; Step 2 may specifically include: Step 201: Retrieve the corresponding standard construction procedure flow data from the construction procedure template library based on the project type tag. The standard construction procedure flow data includes the procedure number, procedure name, material requirement list, and procedure dependencies for each procedure.

[0022] Step 202: Match the category path of each material in the standardized material record set with the material requirement list of each process to determine the construction process number of each material, and generate a material process attribution mapping table. Each record in the material process attribution mapping table contains a material identifier, category path, and construction process number.

[0023] Step 203: Calculate the pairwise correlation score between materials within the same construction process. The correlation score is calculated based on a weighted average of the matching relationship strength value, historical co-occurrence frequency value, and category proximity value. The calculation formula is as follows:

[0024] in, The score is based on relevance. The matching strength value indicates the degree to which two materials are specified as compatible for use in the construction specifications. The historical co-occurrence frequency value represents the frequency at which two materials appear simultaneously in historical purchase orders; The category proximity value represents the normalized reciprocal of the distance between two materials' category paths within the category hierarchy tree structure. In other words, the smaller the number of nodes in the shortest path between two materials in the category hierarchy tree, the better. The larger the value; , , For the preset weighting coefficients, and To eliminate the influence of dimensional differences on weighted calculations, , , Before being used in the calculation, the values ​​were processed to the same range using a mean normalization method based on the range.

[0025] Step 204: Perform hierarchical clustering on materials whose correlation scores exceed the aggregation threshold to generate a set of material combination clusters within the process. Hierarchical clustering takes the correlation scores between each pair of materials as input and the set of material combination clusters as output. Materials whose correlation scores do not exceed the aggregation threshold and are not clustered into any combination cluster are retained as independent material entries.

[0026] Step 205: Based on the process dependencies and the duration of each process, calculate the latest arrival date of materials for each process in a forward direction from the project start date, and mark the latest arrival date of materials to the corresponding combination cluster and independent material entries.

[0027] It should be noted that the calculation method for the latest arrival date of materials for each process is as follows: Starting from the project commencement date, determine the earliest start date of each process according to the process dependencies. Then, subtract the lower limit of the timeframe range for each process from its earliest start date to obtain the latest arrival date of the materials for that process. The process dependencies define the constraints between processes, meaning that a process can only begin after all its preceding processes are completed. The timeframe range is the standard timeframe interval for each process pre-stored in the construction process template library. The lower limit of the timeframe range is used during calculation to ensure that the materials arrive no later than the earliest start time of the process.

[0028] Step 3: Perform two-dimensional supplier matching and push quotation requests; Step 3 may specifically include: Step 301: Perform category-based matching on the material combination clusters within each process. Obtain the main category coverage data of candidate suppliers, filter candidate suppliers whose main category coverage clusters exceed a preset proportion of the category path set, and generate a category-based candidate supplier list.

[0029] Step 302: Perform time-series matching on the material combination clusters within each process. Filter candidate suppliers whose service areas cover the project address and whose historical delivery cycles meet the latest material arrival date constraint, and generate a time-series candidate supplier list.

[0030] It should be noted that the method for determining whether the historical delivery cycle meets the latest material arrival date constraint is as follows: add the candidate supplier's historical average delivery cycle to the current date to obtain the expected arrival date; when the expected arrival date is not later than the latest material arrival date, the supplier is deemed to meet the timing dimension constraint. The historical average delivery cycle is taken from the average delivery cycle of similar materials in the supplier's historical order records.

[0031] Step 303: Obtain the intersection of the candidate supplier list based on the product category dimension and the candidate supplier list based on the time series dimension to generate a combined candidate supplier list. Perform the same two-dimensional matching logic on each individual material item to generate a single-item candidate supplier list.

[0032] Step 304: Send quotation requests, including material combination details and delivery time requirements, to each candidate supplier.

[0033] In this embodiment, to obtain market price references for materials with low confidence levels in category matching, in step 304, exploratory quote requests are simultaneously pushed to candidate suppliers for materials with confidence levels below a preset threshold during the standardization preprocessing stage. The exploratory quote request indicates that the standardization result for the material is pending confirmation, and suppliers can submit error correction feedback simultaneously when submitting their quotes.

[0034] Step 4: Receive quotation data and process supplier error correction feedback; Step 4 may specifically include: Step 401: Receive the quotation data and feedback information submitted by the supplier, and identify the error correction feedback records from the feedback information. Each error correction feedback record includes the target material identifier, the error correction attribute dimension identifier, and the corrected value after correction.

[0035] Step 402: Perform structured parsing on each error correction feedback record to extract the original value before correction, the corrected value after correction, and the error correction attribute dimension identifier. The error correction attribute dimension identifier includes the material subject dimension, specification parameter dimension, and category affiliation dimension.

[0036] Step 403: For multiple error correction feedbacks for the same material item, group them by attribute dimension and count the frequency of each correction value. Select the correction value whose frequency exceeds the consistency threshold as the reliable error correction result.

[0037] It should be noted that the consistency threshold is determined by two conditions: first, the frequency of the corrected value must be a predetermined proportion of the total number of feedbacks for the same attribute dimension; second, the number of suppliers submitting the corrected value must be no less than a predetermined minimum number of suppliers. When both conditions are met, the corrected value is considered a reliable error correction result. If no corrected value meets the consistency threshold for the same attribute dimension, the error correction feedback for that attribute dimension is ineffective, and the original value from the standardization preprocessing stage is retained.

[0038] Step 5: Detect and correct any changes in process attribution caused by errors; For reliable error correction results that have passed step 4, step 5 can be further performed. Step 5 may specifically include: Step 501: For each material involved in the reliable error correction result, update the corrected value to the standardized material main term and attribute list, and regenerate the corrected category path based on the updated material attributes.

[0039] Step 502: Re-match the revised category path with the material requirement list of each process in the construction process template to obtain the revised construction process number.

[0040] Step 503: Compare the construction process numbers before and after the correction. When the construction process numbers before and after the correction are the same, the process affiliation of the material has not changed, and the material's position in the original cluster is retained. When the construction process numbers before and after the correction are inconsistent, the material is marked as a material with a changed affiliation, the original process number, the new process number, and the identifier of the original cluster are recorded, and it is added to the set of materials with changed process affiliation.

[0041] Step 6: Perform dynamic recombination of the combined clusters; For the set of materials whose process attribution has changed as identified in step 5, step 6 can be further performed. Step 6 may specifically include: Step 601: For each attribute change material, remove it from the original cluster. Recalculate the pairwise correlation scores between the remaining materials in the original cluster.

[0042] Step 602: Determine whether the number of remaining materials in the original cluster is lower than the minimum aggregation threshold. If yes, dissolve the original cluster and downgrade the remaining materials to independent material entries; otherwise, if the number of remaining materials is not lower than the minimum aggregation threshold, retain the original cluster structure.

[0043] It should be noted that the minimum aggregation threshold is the lower limit for the number of materials required to maintain the effectiveness of a cluster. When the number of materials in a cluster falls below the minimum aggregation threshold, the category combination pricing advantage of the cluster no longer holds, and therefore the remaining materials are downgraded to independent material items for independent supplier matching.

[0044] Step 603: Add the material with the changed attribution to the material set of the new process. Calculate the correlation score between the material with the changed attribution and each material in each existing cluster within the new process. When the correlation score between the material with the changed attribution and at least one material in a cluster exceeds the aggregation threshold, the material with the changed attribution is merged into the corresponding cluster. When the correlation score between the material with the changed attribution and none of the materials in any existing cluster within the new process exceeds the aggregation threshold, the material with the changed attribution is included as an independent material item in the new process.

[0045] Step 604: For the combined clusters and independent material entries affected by the reorganization, recalculate the latest arrival date of the materials based on the process dependencies and duration of the new process, and replace the original latest arrival date of the materials.

[0046] In this embodiment, when multiple attribution change materials in the same batch point to the same new process, step 603 performs the merging operation of these attribution change materials sequentially in descending order of relevance score. After each attribution change material is merged, the member list of the combined cluster within the new process is updated so that the relevance score calculation of subsequent attribution change materials is based on the latest combined cluster state.

[0047] Step 7: Re-execute the two-dimensional supplier matching and push a supplementary quotation request; Step 7 may specifically include: Step 701: Re-perform two-dimensional supplier matching for the combined clusters and individual material entries affected by the reorganization. For the category dimension, re-select candidate suppliers whose main category coverage meets the preset ratio based on the revised category path. For the time-series dimension, re-select candidate suppliers whose historical delivery cycles meet the constraints based on the updated latest material arrival date. Take the intersection of the two dimension selection results to generate an updated candidate supplier list.

[0048] Step 702: Send a notification of the change in quotation range to the suppliers associated with the original material cluster. The notification should include the removed material identifiers and the revised material combination details. Push supplementary quotation requests, including material details and updated delivery time requirements, to the newly added candidate suppliers in the updated candidate supplier list.

[0049] In this embodiment, to reduce the error rate of the same type of standardized mapping in subsequent price lookup, after step 702, the reliable error correction results are converted into incremental training samples and stored in a buffer pool. Each incremental training sample contains the original text features before error correction and the standardized mapping results after error correction. When the number of samples in the buffer pool reaches the batch threshold, incremental fine-tuning of the standardized mapping model is triggered. Using the standardized mapping results after error correction as the annotation target and the cross-entropy loss function as the optimization objective, the Adam optimization algorithm is used to incrementally update the category path mapping parameters.

[0050] Step 8: Summarize the quotation data and generate a phased procurement quotation recommendation plan; Step 801: Summarize all valid quotes, including initial quotes and supplementary quotes.

[0051] Step 802: Calculate the overall score for each material combination cluster and individual material item within each process. The formula for calculating the overall score is:

[0052] in, For comprehensive scoring; The total cost is normalized by normalizing the total cost of each bid scheme using the mean normalization method based on the range. Weighted by total cost; The risk factor for delivery coordination is positively correlated with the number of suppliers involved in the pricing proposal; the more suppliers, the higher the risk. The larger the value, A mean normalization method based on the range is used to process the values ​​to and from the mean. , Same numerical range; To coordinate risk weights; The normalized value of the time margin is obtained by taking the mean value after processing the time margin between each supplier's delivery time commitment and the latest arrival date of the materials using the mean normalization method based on the range. As a time margin weight; , , For the preset weighting coefficients, and .

[0053] It should be noted that, The time margin in a quote refers to the difference in days between the delivery date promised by the supplier and the latest arrival date of the materials for that process. A larger difference indicates a more ample delivery time margin. The higher the value, the higher the overall score. The higher the margin, the better. By incorporating time margin into the overall score, the ranking results in step 802 simultaneously reflect the advantages and disadvantages in three dimensions: cost, coordination risk, and delivery time margin.

[0054] Step 803: Sort the quotation schemes within each process according to the comprehensive score, arrange them according to the construction process sequence to generate a phased procurement quotation recommendation scheme, and feed it back to the user terminal.

[0055] This implementation adds a process attribution change detection step after the standardized error correction process, specifically comparing the construction process numbers before and after correction in step 5. When a change in material category path is detected, causing a change in the construction process number, the dynamic reorganization of the combination cluster in step 6 and the linked re-evaluation of the two-dimensional supplier matching in step 7 are automatically triggered. Since the material with the changed attribution is removed from the original combination cluster and added to the material set of the new process, the members of the original combination cluster and the combination cluster of the new process are updated synchronously. The supplier screening range in the category dimension is recalculated based on the corrected category path, and the delivery constraints in the time sequence dimension are recalibrated based on the latest arrival date of the material in the new attribution process. Therefore, the corrected material no longer remains in the combination cluster of the original process, and the supplier coverage in the category dimension and the delivery constraints in the time sequence dimension are updated to a state consistent with the corrected material attributes, overcoming the problems of outdated matching results and misaligned delivery times when standardized error correction and two-dimensional supplier matching are executed independently. By sending change notifications to the original suppliers and pushing supplementary quotation requests to the newly matched suppliers, the content of the quotation requests is consistent with the actual needs after correction, avoiding invalid and omitted quotations, and realizing collaborative correction of price queries.

[0056] The following is an example of an application of the present invention, such as Figure 2-9 As shown, the implementation process is as follows: In a scenario involving material procurement price inquiries, Project A, the procuring entity, submitted a batch of material procurement price inquiries for the structural construction phase through a price inquiry platform in Month B of 20XX. This batch of material lists included various types of materials such as concrete, reinforcing steel, and waterproofing coatings, covering three construction phases: foundation engineering, main structure, and waterproofing. Among these, the waterproofing coating materials were marked as pending confirmation due to vague original descriptions and the standardized mapping model's output confidence level being below a preset threshold, and a exploratory price request was simultaneously pushed to the platform. After the exploratory price request was issued, the platform received feedback from suppliers correcting errors related to the material, indicating discrepancies between the material's category classification and the standardized results.

[0057] In Step 1, the original description of the waterproof coating in the price request submitted by the purchaser is "JS waterproof coating film, thickness 1.5mm, usage 300㎡". The natural language processing module identifies the material name "JS waterproof coating film", the specification parameter "thickness 1.5mm" and the quantity "300㎡", generating a structured material entry. The standardized mapping model maps this entry to the category path "coating / waterproof coating / polymer cement waterproof coating", with an output confidence score of 0.61, which is lower than the preset threshold of 0.75. Therefore, the material is marked as pending confirmation. At the same time, the project type label "residential building", project address "City Z", and project start date "20XX Year B Month 08" are extracted from the price request.

[0058] Table 1. Standardization Preprocessing Results

[0059] In step 2, based on the project type tag "Residential Building", the corresponding standard construction process data is obtained from the construction process template library. The material process attribution mapping table is generated through category path matching: materials 01 and 04 belong to process W01 (foundation engineering), materials 02 belong to process W02 (main structure), and the category path of materials 03, "coating / waterproof coating / polymer cement waterproof coating", is matched to process W03 (waterproofing engineering).

[0060] Within process W01, the correlation score between material 01 (ready-mixed concrete C30) and material 04 (ordinary Portland cement) is calculated using the following formula:

[0061] in, For relevance scoring, To score the category matching score, Score for historical co-occurrence. Scoring is based on adjacent processes. , , These are the corresponding weighting coefficients. Substituting the normalized scores of the three items into the formula, we get... The aggregation threshold of 0.65 is exceeded, and the two materials aggregate into a composite cluster C01. Process W02 contains only one material, 02, which is retained as an independent material entry. Process W03 contains only one material, 03, which is also retained as an independent material entry. Based on the process dependencies (W01→W02→W03) and the project duration, the latest arrival date of each material for each process is calculated forward from the start date of B month 08, 20XX.

[0062] Table 2 Material Process Attribution Mapping and Arrival Date

[0063] In step 3, a two-dimensional supplier matching process is performed on cluster C01. The category dimension filters candidate suppliers whose main product categories simultaneously cover "C30 concrete" and "ordinary Portland cement," generating a category-based candidate supplier list (including suppliers SP01 and SP02). The time-series dimension filters suppliers whose service areas cover City Z and whose historical average delivery cycle meets the arrival constraint of B month 10, 20XX. The intersection of these suppliers' data is used to generate a combined candidate supplier list (SP01). The same two-dimensional matching logic is performed on material 03, generating a single-item candidate supplier list (SP04 and SP05). Quotation requests are pushed to each candidate supplier, with the exploratory quotation request for material 03 marked as "Standardization result pending confirmation."

[0064] In step 4, the platform received error correction feedback from suppliers SP04, SP05, and SP06 regarding material 03 between February 18th and February 22nd, 20XX. All three suppliers submitted feedback at the category classification level, with the corrected value pointing to "waterproofing materials / waterproof rolls / SBS modified bitumen waterproof rolls," which differs from the "coatings / waterproof coatings / polymer cement waterproof coatings" generated during the standardization preprocessing stage. Grouped by attribute dimension, a total of 3 feedbacks were received at the category classification level. The corrected value "SBS modified bitumen waterproof rolls" appeared 3 times, accounting for 100%, and was submitted by 3 suppliers. This meets both the frequency threshold (60%) and the minimum number of suppliers (2), thus the corrected value is considered a reliable error correction result.

[0065] Table 3 Summary of Supplier Error Correction Feedback

[0066] In step 5, based on the reliable error correction results, the category path of material 03 is updated to "Waterproofing Materials / Rolled Waterproofing / SBS Modified Bituminous Waterproofing Roll". The corrected category path is then re-matched with the construction process template. The material requirement list for "SBS Modified Bituminous Waterproofing Roll" corresponds to process W02 (main structure), and the corrected construction process number is W02. Comparing the process number W03 before and after the correction with the corrected process number W02, they are inconsistent. Material 03 is marked as a material with a changed affiliation. The original process number W03, the new process number W02, and the identifier of the original independent entry are recorded, and the material is added to the set of materials with changed affiliation for the process.

[0067] In step 6, since material 03 exists as an independent material entry in the original process W03, there is no remaining material in process W03 after removal. The independent entry of process W03 is directly cleared, and there is no need to perform the minimum aggregation threshold judgment.

[0068] Add material 03 to the material set of process W02, and calculate the correlation score between material 03 and the existing material 02 (threaded steel bar) in process W02. After normalization and substituting into the formula, we get... Since the polymerization threshold of 0.65 was not exceeded, material 03 was included in process W02 as an independent material item. Based on the process dependencies and timeframe of process W02, the latest arrival date of material 03 was recalculated, updated from the original date of C month 15, 20XX corresponding to W03 to B month 28, 20XX corresponding to W02.

[0069] Table 4 Comparison of Cluster Dynamic Recombination Before and After Table 4

[0070] In step 7, a two-dimensional supplier matching process is re-executed for material 03. The category dimension is re-screened based on the revised category path "SBS modified bitumen waterproof membrane." Original single-item candidate suppliers SP04 and SP05 have main product categories covering this final-level category, and the newly added SP07 also covers this category. The time dimension is re-screened based on the updated latest delivery date of the material, B / 28, 20XX. SP04 and SP07 meet the historical average delivery cycle constraint, while SP05's historical average delivery cycle is too long and does not meet the constraint. The candidate supplier list is updated to SP04 and SP07 after taking the intersection. A notification of the change in quotation range is sent to SP04, including information on the change in the category path for material 03 and the updated delivery time requirement. A supplementary quotation request containing details of material 03 and the delivery requirement of B / 28, 20XX is sent to the newly added candidate supplier SP07. The original exploratory quotation recipient SP05 is no longer included in the candidate list because it does not meet the time dimension constraint.

[0071] Table 5. Updated Results of Two-Dimensional Supplier Matching

[0072] In step 8, after summarizing all valid quotation data, a comprehensive score is calculated for materials 02 and 03 within process W02. Taking the two valid quotations received for material 03 (quotation SP04 and quotation SP07) as an example, the comprehensive score is calculated using the following formula:

[0073] in, For comprehensive scoring, This is the normalized value of the total cost. To coordinate risk factors, This is the normalized value of the time margin. , , These are the corresponding weighting coefficients. SP04 pricing scheme. , , Substituting into the formula, we get SP07 pricing scheme , , Substituting into the formula, we get SP04 received a slightly higher overall score, ranking first among the recommended pricing schemes for Material 03. Arranged by process sequence W01 and W02, a phased procurement pricing recommendation scheme was ultimately generated and fed back to the purchasing party.

[0074] Table 6. Calculation Results of Comprehensive Score for Material 03

[0075] The entire data flow begins with a vague material description submitted by the purchaser. Step 1 involves standardized preprocessing to generate a category path with low confidence. Steps 2 and 3 use this path to complete initial process attribution mapping and two-dimensional supplier matching. Step 4 aggregates consistent error correction feedback from multiple suppliers and identifies reliable correction values. Step 5 detects a change in the category path causing the construction process number to change from W03 to W02. Step 6 removes material 03 from the original process W03 and incorporates it into an independent entry for process W02, while simultaneously updating the latest arrival date of the material. Step 7 re-matches candidate suppliers based on the corrected category path and updated delivery constraints and pushes supplementary quotation requests. Step 8 summarizes valid quotations and generates a phased recommendation scheme with a comprehensive score ranking. Data status is transmitted unidirectionally between steps, and the correction signal, triggered in Step 5, propagates downstream step by step, ensuring that the failure of the two-dimensional matching conditions caused by the category path change is completely corrected within the same batch.

[0076] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A two-dimension matching supplier intelligent price inquiry collaboration method, characterized in that, At least the following steps are included: Obtain price inquiry request data, perform standardized preprocessing on the bill of materials, generate a standardized material record set containing category paths, and extract project type tags, project address information, and project start date; Standard construction procedure data is obtained based on project type tags. The category paths of each material are matched with the material requirement lists of each procedure to determine the procedure affiliation. For materials within the same procedure, a correlation score is calculated, and hierarchical clustering is performed to generate material combination clusters. The latest arrival date of materials is estimated based on the procedure dependencies. The correlation score is calculated by weighting the matching relationship strength value, historical co-occurrence frequency value, and category proximity value, using the following formula: ; The score is based on relevance. The matching strength value indicates the degree to which two materials are specified as compatible for use in the construction specifications. The historical co-occurrence frequency value represents the frequency at which two materials appear simultaneously in historical purchase orders; The category proximity value represents the normalized reciprocal of the distance between two materials' category paths within the category hierarchy tree structure. In other words, the smaller the number of nodes in the shortest path between two materials in the category hierarchy tree, the better. The larger the value; , , For the preset weighting coefficients, and To eliminate the influence of dimensional differences on weighted calculations, , , Before being included in the calculation, each value was processed to the same numerical range using a mean normalization method based on the range. For each material combination cluster, perform supplier matching in terms of category and time sequence, take the intersection of the two dimensions to generate a list of candidate suppliers and push the quotation request; Receive supplier quotation data and error correction feedback, group and statistically analyze error correction feedback by attribute dimension, and extract reliable error correction results that meet the consistency threshold; Based on the reliable error correction results, update the material properties and regenerate the corrected category path. Then, rematch the corrected category path with the material requirement list of each process to obtain the corrected process number. Compare the process numbers before and after the correction to identify the material belonging to the process. Remove the materials with changed ownership from the original cluster and add them to the material set of the new ownership process. Recalculate the relevance score, perform dynamic cluster reorganization, and update the latest arrival date of the materials. For the clusters affected by the reorganization, supplier matching is re-executed based on the revised category path and the updated latest material arrival date, along with the category and time dimensions. Change notices are sent to the original suppliers, and supplementary quotation requests are pushed to the newly matched suppliers.

2. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, The criteria for determining the consistency threshold include: the first condition is that the frequency of the correction value in the error correction feedback accounts for a proportion of the total number of all feedbacks in the same attribute dimension that reaches a preset frequency proportion. The second condition is that the number of suppliers submitting the corrected value is not less than the preset minimum number of suppliers. When both conditions are met, the corrected value is considered a reliable error correction result. If there is no corrected value that meets the consistency threshold on the same attribute dimension, the error correction feedback for that attribute dimension is not effective, and the original value from the standardization preprocessing stage is retained.

3. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, The dynamic reorganization of the cluster also includes: determining whether the number of remaining materials in the original cluster is lower than the minimum aggregation threshold; if so, disbanding the original cluster and downgrading the remaining materials to independent material entries. Conversely, the original cluster structure is retained; the correlation score between the material whose ownership has been changed and each material in the existing clusters within the new process is calculated. When the correlation score between the material whose ownership has been changed and at least one material in a cluster exceeds the aggregation threshold, the material whose ownership has been changed is merged into the corresponding cluster; otherwise, the material whose ownership has been changed is included in the new process as an independent material item.

4. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 3, characterized in that, When multiple attribution change materials in the same batch point to the same new process, the merging operations of these attribution change materials are executed sequentially in descending order of relevance score; After each attribute change material is incorporated, the member list of the combined cluster within the new process is updated so that the relevance score calculation of subsequent attribute change materials is based on the latest combined cluster status.

5. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, In the time-series supplier matching, the method for determining whether the historical delivery cycle meets the latest material arrival date constraint is: to add the candidate supplier's historical average delivery cycle to obtain the expected arrival date; When the expected arrival date is not later than the latest arrival date of the materials, the supplier is deemed to meet the timing dimension constraint; wherein, the latest arrival date of the materials is calculated as follows: starting from the project start date, the earliest start date of each process is determined according to the process dependency relationship, and then the earliest start date of each process is subtracted from the lower limit of the process's duration range.

6. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, When sending a quote request, a probe quote request is simultaneously sent to candidate suppliers for materials whose confidence level is lower than a preset threshold in the standardized preprocessing stage. The probing quote request indicates that the standardization result of the material is pending confirmation, and the supplier submits error correction feedback simultaneously when submitting the quote.

7. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, After pushing the supplementary quote request, the process also includes: converting the reliable error correction results into incremental training samples and storing them in a buffer pool. Each incremental training sample contains the original text features before error correction and the standardized mapping results after error correction. When the number of samples in the buffer pool reaches the batch threshold, the incremental fine-tuning of the standardized mapping model is triggered; wherein, the standardized mapping model takes the text features of the structured material entries as input and outputs the category probability distribution and confidence level corresponding to the category path.

8. The dual-dimensional matching supplier intelligent price query collaborative method according to claim 1, characterized in that, It also includes summarizing all valid quotation data and calculating a comprehensive score for each material combination cluster and individual material item within each process based on the normalized value of total cost, delivery coordination risk coefficient, and normalized value of time margin; Among them, the value of the delivery coordination risk coefficient is positively correlated with the number of suppliers involved in the quotation scheme; The normalized value of the time margin is obtained by normalizing the average of the difference between the delivery time commitment of each supplier and the latest arrival date of the materials; the quotation schemes in each process are sorted according to the comprehensive score, and the phased procurement quotation recommendation scheme is generated by arranging them according to the construction process sequence.

9. A dual-dimensional matching supplier intelligent price query collaborative system, used to execute the dual-dimensional matching supplier intelligent price query collaborative method according to any one of claims 1 to 8, characterized in that, include: The standardized preprocessing module is used to acquire price inquiry request data, perform standardized preprocessing on the bill of materials to generate a standardized material record set containing category paths, and extract project type tags, project address information and project start date; The process attribution mapping and aggregation module is used to obtain standard construction process flow data based on project type labels, match the category path of each material with the material requirement list of each process to determine the process attribution, calculate the correlation score of materials within the same process and perform hierarchical clustering to generate material combination clusters, and calculate the latest arrival date of materials based on process dependencies. The dual-dimensional supplier matching module is used to perform supplier matching in terms of category and time sequence for each material combination cluster, and to generate a candidate supplier list by taking the intersection of the two dimensions of candidate suppliers and pushing quotation requests. The error correction feedback processing module is used to receive supplier quotation data and error correction feedback, group and statistically analyze error correction feedback by attribute dimension, and extract reliable error correction results that meet the consistency threshold. The process attribution change detection module is used to update material properties based on reliable error correction results and regenerate the corrected category path. The corrected category path is then matched with the material requirement list of each process to obtain the corrected process number. The process numbers before and after the correction are compared to identify materials whose process attribution has changed. The dynamic reorganization module of the composite cluster is used to remove materials with changed ownership from the original composite cluster and add them to the material set of the new ownership process, recalculate the relevance score, perform dynamic reorganization of the composite cluster, and update the latest arrival date of the materials. The supplementary matching and push module is used to re-execute supplier matching in terms of category and time sequence for the clusters affected by the reorganization, based on the revised category path and the updated latest arrival date of materials. It sends change notifications to the original suppliers and pushes supplementary quotation requests to the newly matched suppliers.

Citation Information

Patent Citations

  • Purchase order generation method and device

    CN120563041A

  • Knowledge graph-based factory process parameter intelligent matching method and system

    CN120832531A