Hardware supply chain intelligent matching recommendation system and method

Abstract

The invention discloses an intelligent matching recommendation system and method for a hardware supply chain, and particularly relates to the technical field of intelligent matching recommendation. The method comprises the steps that original demand information submitted by a purchaser is analyzed, key fields are extracted, and feature vectors are generated; by comparing current and historical purchase records, a difference field is identified, a two-dimensional vector is constructed according to a change amplitude and a service weight, and difference risks are divided into four levels by using a clustering algorithm. And the system further calculates function dependence, historical co-occurrence and purchase quantity correlation among the materials, and quantifies a strong association relationship. In combination with the difference level and the strong association index, the system calculates a comprehensive urgency degree, and a supplier selection strategy is determined according to the comprehensive urgency degree: when the urgency degree is low, suppliers are globally screened through similarity matching; and when the urgency degree is high, the historical cooperative suppliers are preferentially recommended, so that fine grading of purchase difference risks and intelligent supplier decision making are realized, and the supply chain management efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent matching recommendation technology, and more specifically, to a hardware supply chain intelligent matching recommendation system and method. Background Art

[0002] With the advancement of intelligent manufacturing and digital supply chains, hardware product production processes are placing higher demands on supply chain responsiveness, accuracy, and flexibility. Existing supply chain management systems (SCMs) mostly focus on static supplier information management or simple material matching-based recommendation methods, lacking the ability to understand and quantify the multi-dimensional semantics between procurement requirements and supplier capabilities. Furthermore, existing technologies generally struggle to identify discrepancy risks between procurement requirements and historical procurement data. They also lack in-depth analysis of the relationships between procurement materials, making it difficult to effectively determine whether materials should be bundled or to implement tiered management for discrepancy risks. Therefore, when procurement requirements change or potential risks arise, manual judgment often relies on experience, resulting in inaccurate supplier recommendations and difficulty in responding to sudden market or production demand changes, impacting the continuity of production plans and overall supply chain efficiency. To address these issues, a method and system are urgently needed that can intelligently analyze procurement requirements, quantitatively rank discrepancy risks, and intelligently recommend suppliers based on historical data to enhance the intelligence and decision-making efficiency of supply chain management. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a hardware supply chain intelligent matching recommendation system and method to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions: The hardware supply chain intelligent matching recommendation method includes the following steps: Parse the original purchase demand information submitted by the purchaser, extract and structure key fields, and generate a purchase demand feature vector; compare the current purchase list with the historical purchase record list, identify the difference fields, and construct a two-dimensional vector based on the change range and business weight of the difference fields; use clustering algorithms to map the difference fields into different difference field levels; Strong correlation indicators between materials are generated based on the support, promotion and procurement volume time series correlation coefficients between purchased materials. The comprehensive urgency is calculated based on the strong correlation indicators between the difference field level and the materials, and different supplier selection strategies are adopted based on the comprehensive urgency: when the comprehensive urgency is lower than the system threshold, the best supplier is selected from all suppliers through similarity matching; when the comprehensive urgency is higher than the system threshold, suppliers are selected from suppliers with whom historical cooperation has been carried out.

[0005] In a preferred embodiment, the identification of difference fields combines the historical purchase record list with the current purchase list, first extracts all material items from the current purchase list, and aligns and matches them with the historical purchase records based on their unique identifiers; after the pairing is completed, semantic mapping and format standardization are performed through the field dictionary to unify the field naming, data type and unit of measurement, and then, each pair of current and historical records is compared field by field to identify the difference fields.

[0006] In a preferred embodiment, the business weight is obtained through offline learning of logistic regression based on historical material shortages and rework losses.

[0007] In a preferred embodiment, a K-Means clustering algorithm is used to cluster the two-dimensional vector consisting of the change amplitude of the difference field and the business weight, and the difference field is mapped into different difference field levels.

[0008] In a preferred embodiment, the frequency ratio of material A and material B appearing in the purchase order at the same time is counted to obtain the support between the purchase materials. The degree of improvement ; where n(A∩B) is the number of times material A and material B co-occur, n(A) and n(B) are their respective purchase frequencies, and N is the total number of purchase orders; the correlation coefficient of the purchase quantity time series is ; where qA(t) and qB(t) represent the purchase quantity sequences of material A and material B at time t, respectively, capturing the consistency of the demand changes between the two.

[0009] In a preferred embodiment, the support degree, promotion degree and purchase quantity time series correlation coefficients between the comprehensive purchase materials are weighted summed to obtain the strong correlation index between the materials.

[0010] In a preferred embodiment, the comprehensive urgency is calculated by weighted summation based on the strong correlation index between the difference field level and the material.

[0011] In a preferred embodiment, when the comprehensive urgency is higher than the system threshold, suppliers are preferentially selected from suppliers with whom historical cooperation has been conducted. The comprehensive coverage, capability score and supplier proportion are weighted and summed to calculate the supplier's comprehensive score. Candidate suppliers with high comprehensive scores are selected as procurement suppliers. The candidate suppliers are the set of all suppliers with whom historical cooperation has been conducted.

[0012] In a preferred embodiment, the supplier proportion is the ratio of the number of supply chains that include the supplier divided by the total number of supply chains; the coverage is 1 if all coupling parts can be supplied at one time; if only p% of them are covered, the value is taken according to the coverage rate; the capability score is obtained by weighted summation after multi-dimensional normalization of technical similarity, delivery reliability, and historical performance.

[0013] In a preferred embodiment, the following modules are included: The procurement demand parsing and vectorization module is used to parse the original procurement demand information submitted by the purchaser, extract and structure key fields, and generate a procurement demand feature vector; The difference analysis and risk assessment module compares the current purchase list with the historical purchase record list, identifies the difference fields, and constructs a two-dimensional vector based on the change range and business weight of the difference fields. It also uses a clustering algorithm to map the difference fields into different difference field levels. The comprehensive urgency and supplier strategy selection module is used to generate strong correlation indicators between purchased materials based on the support, promotion, and purchase volume time series correlation coefficients. The comprehensive urgency is calculated based on the strong correlation indicators between the difference field level and the materials. Different supplier selection strategies are adopted based on the comprehensive urgency: when the comprehensive urgency is lower than the system threshold, the best supplier is selected from all suppliers through similarity matching; when the comprehensive urgency is higher than the system threshold, suppliers with historical cooperation are preferentially selected. The supplier comprehensive scoring and recommendation module is used to select suppliers from suppliers with whom we have cooperated in the past when the comprehensive urgency is higher than the system threshold. The supplier comprehensive score is calculated by weighted summation of comprehensive coverage, capability score and supplier proportion. Candidate suppliers with high comprehensive scores are selected as procurement suppliers. The candidate suppliers are the set of all suppliers with whom we have cooperated in the past.

[0014] The technical effects and advantages of the present invention are as follows: The present invention achieves accurate recognition of complex free text and structured data through intelligent parsing and vectorization of procurement demand information, significantly improving the system's ability to understand diverse procurement needs. By combining historical procurement records, using the magnitude of change in difference fields and business weights to construct a two-dimensional vector, and adopting a clustering algorithm for risk grading, it effectively achieves quantitative management of procurement difference risks, overcoming the problem of insufficient understanding of multi-dimensional differences and supply chain coupling relationships in existing technologies. At the same time, the proposed Must-Buy Index (MBI) accurately depicts the strong correlation between materials through multi-dimensional indicators such as support, lift, and procurement volume time series correlation, providing a scientific basis for bundled procurement and improving the intelligence level of procurement decisions.

[0015] This invention also adaptively selects suppliers based on the overall urgency. When the urgency is low, the optimal supplier is globally screened through similarity matching. When the urgency is high, historical suppliers are prioritized to ensure supply chain stability and rapid response. By calculating the supplier's coverage, capability score, and historical share, a multi-dimensional comprehensive scoring model is formed, enabling scientific supplier selection. Overall, this invention significantly improves the intelligence and automation of the procurement process, enhances the flexibility and decision-making efficiency of the supply chain, and has important practical application value and broad promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 Schematic diagram of the process of the hardware supply chain intelligent matching recommendation method of the present invention; Figure 2 Schematic diagram of the process for quantifying and modeling the urgency of differential risks; Figure 3 This is a structural diagram of the hardware supply chain intelligent matching recommendation system of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] Example 1: The hardware supply chain intelligent matching recommendation method of the present invention, such as Figure 1 As shown, the following steps are included: Step S1: Procurement demand information analysis This step is used to semantically understand and structure the original procurement demand information submitted by the purchaser, laying a data foundation for subsequent matching and recommendation.

[0019] First, the system receives the original purchase requirement information submitted by the purchaser. This information may come from a standardized form, a free-text description, or an attached document containing technical specifications. To ensure universality, the system converts all input content into a processable text stream and uses this as the basis for demand analysis.

[0020] During the text processing phase, the system first pre-processes the natural language content, employing techniques such as Chinese word segmentation, part-of-speech tagging, and named entity recognition to identify key fields such as material names, technical processes, specifications, quantity units, delivery time requirements, and relevant certification requirements. For content involving numerical information, the system automatically identifies and standardizes the units. If the original text contains references, ambiguities, or missing expressions, such as "delivery as soon as possible," the system will standardize based on the contextual semantics or platform default rules. For example, "as soon as possible" defaults to a 14-day delivery period.

[0021] After completing keyword extraction and numerical standardization, the system maps the original procurement demand information into structured procurement fields, mainly including: product category, target application scenario, material type, process flow, dimensional accuracy, functional indicators, delivery time, batch size, cost constraints, quality certification requirements and other contents.

[0022] Step S2: Constructing the purchasing demand feature vector The purpose of this step is to convert the structured procurement fields (i.e., the original procurement demand information) into numerical feature vector representations as input for subsequent steps.

[0023] First, based on the structured procurement fields obtained in step S1, each field is encoded and quantified. For categorical fields (such as target product category, industry, and certification requirements), the system uses one-hot encoding or embedded vector encoding to represent them, converting each category value into a sparse or dense vector of fixed dimension.

[0024] The system normalizes numeric fields (such as batch quantity, lead time, wall thickness, and budgeted unit price). Common methods include maximum-minimum normalization, Z-score normalization, or segmented mapping to eliminate scale differences between numerical dimensions. For example, a field such as "delivery time is 15 days" can be normalized to the [0, 1] interval for quantitative comparison with the supplier's deliverability field.

[0025] For textual requirements extracted from natural language descriptions (such as "high requirements for appearance" and "intention for long-term cooperation"), the system uses a word vector model or semantic embedding model (such as Word2Vec, GloVe, or BERT) to encode them into fixed-dimensional semantic vectors. Semantic vectors typically represent the semantic information of a sentence in a continuous space of 100 to 768 dimensions, allowing subsequent sentence-level or phrase-level similarity calculations. For hierarchical material descriptions (such as "Plastics > Engineering Plastics > Polycarbonate (PC)"), the system uses a hierarchical encoding method to simultaneously retain multiple levels of semantic labels in the vector to enhance the discriminative power of the representation.

[0026] After all fields are encoded, the system concatenates these numerical, categorical, and semantic vectors into a complete feature vector, which can be directly used to calculate the matching score between the supplier capability vector. The dimension of the feature vector is fixed, and the arrangement order is organized according to the field category to ensure the consistency of the input structure. In order to improve the distinguishing ability of the vector, this step can also introduce a feature weight initialization strategy, that is, the relative importance of certain fields is set according to historical matching data or the experience of business experts, and recorded as the initial weight parameters for subsequent scoring functions. The final output procurement demand feature vector represents the "position" of the current task in the multidimensional feature space, and can be used to support various recommendation mechanisms such as content matching, vector search, and collaborative filtering.

[0027] Step S3: Standardization of supplier capability information This step structures and standardizes candidate supplier information to facilitate efficient matching and scoring with the procurement requirement's feature vector. Supplier capability data typically comes from platform registration information, company profiles, third-party qualification certification databases, or historical transaction records. This data structure is complex, encompassing structured fields (such as registered capital, core industry, and geographic location), semi-structured fields (such as product catalogs and technical lists), and unstructured fields (such as company profiles and customer feedback). To ensure consistency in the matching process, the system must convert supplier information into a representation consistent with the procurement requirement's feature vector.

[0028] Step S4: Differential risk quantification and urgency modeling like Figure 2As shown, the historical purchase record list is combined with the current purchase list. The historical purchase record list integrates the purchase lists in the past period of time (for example, one month, three months); first, all material items are extracted from the current purchase list, and aligned and matched with the historical purchase records based on their unique identifiers (such as item_id, spec_code, or the composite key item_id+version); the matching logic prioritizes the most recent historical orders with a status of "completed" or "in stock" as the comparison baseline to ensure that the comparison basis has business comparability.

[0029] After matching, the system uses a field dictionary to perform semantic mapping and format standardization, unifying field names, data types, and units of measurement. For example, price fields are converted to a unified currency, quantity fields are standardized to standard numbers of pieces or standard packaging quantities, and delivery dates are converted to ISO dates. Enumerated fields are mapped and synonyms merged according to the current version of the standard dictionary to eliminate false discrepancies caused by inconsistent representations in historical data. Subsequently, each pair of current and historical records is compared field by field.

[0030] After identifying the difference fields, two normalization indicators are calculated for each field. One of them is the change magnitude of the difference field: ; where index i corresponds to the i-th difference field in the difference field, This is the value of this field in the historical purchase record list. is the value of the current purchase list of this field, and are the maximum and minimum values of the same field in the past six months (or any selected historical window), and their difference gives the historical lower and upper limits. In this way, no matter whether the field is price, threshold or embedded enumeration, it is linearly converted into a relative amplitude. If it is close to 0, it means that the change is almost negligible, while if it is close to 1, it means that the fluctuation is far greater than in the past.

[0031] The second is business weight , based on historical material shortages and rework losses, it is obtained through offline learning through logistic regression.

[0032] The change magnitude of each difference field and business weight Combined into a two-dimensional vector: ; Forming a sample set Used for cluster modeling. Sets the number of risk levels, K, to be classified. Typically, K=4, corresponding to {Trivial, Minor, Major, Critical}. Optional adaptive strategies include using the Elbow Method or Silhouette Coefficient to determine the optimal K.

[0033] Using the standard K-Means algorithm for modeling, K clusters are ultimately obtained. Each cluster is assigned a risk level label and sorted based on the location of its centroid (mean vector). The corresponding difference field levels are named Critical, Major, Minor, and Trivial. Each difference field is then mapped to four levels. This example uses four levels as an example. Finally, the difference field level is obtained.

[0034] A continuously computable strong correlation indicator, the Must-Buy Index (MBI), is constructed. Its core concept is to quantify multiple dimensions that influence procurement coupling, including functional dependency, historical co-occurrence, cost synergy, and supply risk. These factors are then normalized and weighted, resulting in a real-number score in the [0,1] interval that serves as the basis for determining whether procurement should be bundled.

[0035] Specifically, in the historical co-occurrence layer, the frequency ratio of material A and material B appearing in the purchase order at the same time is counted to obtain the support ; and calculate the lift Where n(A∩B) is the number of times material A and material B co-occur, n(A) and n(B) are their respective purchase frequencies, and N is the total number of purchase orders. If Lift > 1, it means that material A and material B are more related than by chance.

[0036] Further introduce the correlation coefficient of the purchase volume time series ; where qA(t) and qB(t) represent the purchase quantity sequences of material A and material B at time t, respectively, capturing the consistency of the demand changes between the two.

[0037] The strong correlation index is obtained by combining the correlation coefficient of promotion, support and purchase volume time series. According to the following formula: ; Where QG represents the strong correlation index; 、 、 To adjust the parameters, it can be designed as needed.

[0038] The comprehensive urgency is calculated by combining the incremental level and the weighted sum of the strong correlation indicators. First, numerical values are assigned to the difference field levels Critical, Major, Minor, and Trivial in sequence; for example, Critical is assigned a value of 1, Major is assigned a value of 0.75, Minor is assigned a value of 0.5, and Trivial is assigned a value of 0.25.

[0039] By the following formula: ; Where Z represents the comprehensive urgency, and DG represents the value after the incremental level is assigned; 、 The weight coefficients of the values assigned to the strong correlation index and incremental level respectively.

[0040] Step S5: Comprehensive supplier scoring and selection decision If the comprehensive urgency is less than the system threshold, it means that the difference field is significantly different from the historical purchase record list and is relatively unimportant. The supplier is determined according to the similarity matching method in the existing technology. The following is a brief description.

[0041] A similarity calculation relationship is established between the difference field feature vector and the supplier capability vector, which is processed in the same way as the procurement requirement feature vector. The matching degree is evaluated based on their proximity in the multidimensional feature space, providing a baseline score for subsequent recommendation ranking. The difference field feature vector and the supplier capability vector set are first input into the matching engine, entering a one-to-many matching calculation process. The core of the matching process is to construct a scoring function to measure the similarity between any supplier capability vector and the difference field feature vector. This scoring function typically incorporates multiple calculation strategies based on different feature types. Suppliers with the highest similarity are selected.

[0042] If the comprehensive urgency is greater than the system threshold, it means that the difference field has a strong connection with the historical purchase record list and is relatively important; selection is made from historical suppliers; and it is more stable.

[0043] For each candidate supplier, the candidate supplier represents a supplier with which the supplier has cooperated in the past; There are multiple existing supply chains, and the proportion of suppliers in each supply chain is counted; specifically, the supplier proportion is the number of supply chains that include the supplier divided by the total number of supply chains.

[0044] Define two items that can be directly queried from your existing vector library: Coverage cs∈[0,1]: If all coupling parts can be supplied at one time, then take 1; if only p% of them are covered, take the value according to the coverage rate, such as 10% coverage, then take 0.1.

[0045] Capability score as∈[0,1]: It is the weighted sum of normalized multi-dimensional factors such as technical similarity, delivery reliability, historical performance, and unit price.

[0046] The comprehensive score of suppliers is calculated based on comprehensive coverage, capability score and supplier proportion. Candidate suppliers with high comprehensive scores are selected as procurement suppliers.

[0047] The specific formula is as follows: ; ZH represents the supplier's comprehensive score; zb represents the supplier's share; the higher the supplier's share, the more popular the supplier is, the more preferred the purchaser is, and the higher the supplier's comprehensive score is; 、 、 They are the weight coefficients of coverage, capability score and supplier proportion respectively.

[0048] Example 2: The design of the hardware supply chain intelligent matching recommendation system of the present invention is based on the method in Example 1, specifically as follows Figure 3 The following modules are shown: procurement demand analysis and quantification module, supplier information standardization module, difference analysis and risk assessment module, comprehensive urgency and supplier matching module, supplier comprehensive scoring and recommendation module; Procurement demand parsing and vectorization module: used to receive and parse the original procurement demand information submitted by the purchaser, extract key fields such as material name, technical parameters, specifications, delivery cycle, quantity, certification requirements, etc. through natural language processing technology, and convert the parsed information into a multi-dimensional procurement demand feature vector through unique hot encoding, normalization or semantic embedding, which serves as the basic data for subsequent matching.

[0049] Supplier information standardization module: used to collect and process multi-dimensional data such as supplier registration information, product catalogs, technical capabilities, historical transaction records, etc., and unify and structure and standardize them into supplier capability vectors consistent with the procurement demand vector format to achieve vectorized input for subsequent matching calculations.

[0050] Variance Analysis and Risk Assessment Module: This module compares current procurement needs with historical procurement records, identifies discrepancies, and categorizes these discrepancies into four risk levels: Critical, Major, Minor, and Trivial. This module calculates the magnitude of the change and uses business weights derived from historical material shortages or rework losses. It also constructs a Must-Buy Index (MBI) to analyze the functional dependencies, historical co-occurrences, and time-series correlations between materials, forming a strong correlation indicator to assist in deciding whether to bundle purchases.

[0051] Comprehensive Urgency and Supplier Matching Module: Based on risk levels and strong correlation indicators, a comprehensive urgency index is calculated to determine the urgency of procurement needs and historical dependencies. If the comprehensive urgency is low, a similarity matching algorithm is used to select the most compatible supplier from all candidate suppliers. If the urgency is high, prioritizing recommendations from suppliers with whom we have previously collaborated to ensure stable and reliable supply.

[0052] Supplier comprehensive scoring and recommendation module: used to comprehensively evaluate the coverage, capability scores and proportion of candidate suppliers in the historical supply chain, form a comprehensive supplier score through weighted calculation, rank the candidate suppliers, and recommend the supplier with the highest comprehensive score as the optimal cooperation partner. The recommendation results are fed back to the user to support subsequent model optimization and continuous learning.

[0053] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0055] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0056] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0057] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. Hardware supply chain intelligent matching recommendation method, characterized by: The following steps are involved: Parse the original purchase demand information submitted by the purchaser, extract and structure key fields, and generate a purchase demand feature vector; compare the current purchase list with the historical purchase record list, identify the difference fields, and construct a two-dimensional vector based on the change range and business weight of the difference fields; use clustering algorithms to map the difference fields into different difference field levels; Generate strong correlation indicators between materials based on the support, promotion and purchase quantity time series correlation coefficients between purchased materials; The comprehensive urgency is calculated based on the strong correlation index between the difference field level and the material, and different supplier selection strategies are adopted based on the comprehensive urgency: when the comprehensive urgency is lower than the system threshold, the best supplier is selected from all suppliers through similarity matching; when the comprehensive urgency is higher than the system threshold, suppliers are selected from suppliers with whom historical cooperation has been carried out.

2. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: The identification difference field combines the historical purchase record list with the current purchase list, first extracts all material items from the current purchase list, and aligns and matches them with the historical purchase record based on their unique identifiers; After the pairing is completed, semantic mapping and format standardization are performed through the field dictionary to unify field naming, data types, and units of measurement. Subsequently, each pair of current and historical records is compared field by field to identify the difference fields.

3. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: The business weights are obtained through offline logistic regression learning based on historical material shortages and rework losses.

4. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: The K-Means clustering algorithm is used to cluster the two-dimensional vector consisting of the change amplitude of the difference field and the business weight, and map the difference field into different difference field levels.

5. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: Count the frequency ratio of material A and material B appearing in the purchase order at the same time, and get the support between the purchase materials The degree of improvement ; where n(A∩B) is the number of times material A and material B co-occur, n(A) and n(B) are their respective purchase frequencies, and N is the total number of purchase orders; the correlation coefficient of the purchase quantity time series is ; where qA(t) and qB(t) represent the purchase quantity sequences of material A and material B at time t, respectively, capturing the consistency of the demand changes between the two.

6. The hardware supply chain intelligent matching recommendation method according to claim 5, characterized in that: The support, promotion and purchase quantity time series correlation coefficients between comprehensive procurement materials are weighted summed to obtain the strong correlation index between materials.

7. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: The comprehensive urgency is calculated by weighted summation based on the strong correlation indicators between the difference field level and the material.

8. The hardware supply chain intelligent matching recommendation method according to claim 1, characterized in that: When the comprehensive urgency is higher than the system threshold, suppliers are selected from suppliers with whom we have cooperated in the past. The comprehensive score of the supplier is calculated by weighted summation of the comprehensive coverage, capability score and supplier proportion. Candidate suppliers with high comprehensive scores are selected as procurement suppliers. The candidate suppliers are the set of all suppliers with whom we have cooperated in the past.

9. The hardware supply chain intelligent matching recommendation method according to claim 8, characterized in that: The supplier proportion is the ratio of the number of supply chains that include this supplier to the total number of supply chains; the coverage is 1 if all coupling parts can be supplied at one time; if only p% of them are covered, the value is taken according to the coverage rate; the capability score is obtained by weighted summation of multi-dimensional normalization of technical similarity, delivery reliability, and historical performance.

10. Hardware supply chain intelligent matching recommendation system, characterized by: The recommendation system is based on the method according to any one of claims 1 to 9 and includes the following modules: The procurement demand parsing and vectorization module is used to parse the original procurement demand information submitted by the purchaser, extract and structure key fields, and generate a procurement demand feature vector; The difference analysis and risk assessment module compares the current purchase list with the historical purchase record list, identifies the difference fields, and constructs a two-dimensional vector based on the change range and business weight of the difference fields. It also uses a clustering algorithm to map the difference fields into different difference field levels. The comprehensive urgency and supplier strategy selection module is used to generate strong correlation indicators between materials based on the support, promotion and purchase quantity time series correlation coefficients between purchased materials; Based on the strong correlation index between the difference field level and the material, the comprehensive urgency is calculated and different supplier selection strategies are adopted based on the comprehensive urgency: when the comprehensive urgency is lower than the system threshold, the best supplier is selected from all suppliers through similarity matching; when the comprehensive urgency is higher than the system threshold, suppliers with historical cooperation are selected first; The supplier comprehensive scoring and recommendation module is used to select suppliers from suppliers with whom we have cooperated in the past when the comprehensive urgency is higher than the system threshold. The supplier comprehensive score is calculated by weighted summation of comprehensive coverage, capability score and supplier proportion. Candidate suppliers with high comprehensive scores are selected as procurement suppliers. The candidate suppliers are the set of all suppliers with whom we have cooperated in the past.

Images