Intelligent inquiry method and device, electronic equipment and storage medium

Abstract

The application provides an intelligent inquiry method and device, an electronic device and a storage medium, and belongs to the technical field of intelligent inquiry. The method comprises the following steps: receiving a query request input by a user, wherein the query request comprises a query condition of a target material, and the query condition comprises an attribute item value corresponding to an attribute item to be queried; acquiring a price item value of the target material matched with the query condition from a target ticket database; the target ticket database comprises a plurality of query results, each query result comprises a plurality of material attribute items and respective corresponding attribute item values; the plurality of material attribute items at least comprise a price item; and the plurality of material attribute items comprise at least one structured feature item and at least one unstructured feature item. The intelligent inquiry method and device, the electronic device and the storage medium provided by the application can solve the cost control problem caused by the complexity of price influencing factors and inaccurate data.

Description

Technical Field

[0001] This application belongs to the field of intelligent price inquiry technology, and more specifically, relates to an intelligent price inquiry method and apparatus, electronic equipment, and storage medium. Background Technology

[0002] Currently, when conducting material procurement inquiries, enterprises generally rely on historical procurement transaction data as a core reference. Most companies collect and organize historical electronic invoices, purchase contracts, and other documents generated during the procurement process. Some companies also introduce basic artificial intelligence (AI) tools to assist data processing. For example, they use AI-driven optical character recognition (OCR) technology to extract key information such as material names, specifications, and unit prices from invoices, or use simple AI search algorithms to match similar historical records in a database. Other companies build basic databases to store this data, typically containing basic information such as material names, specifications, unit prices, purchase dates, and supplier names. When users need to query the price of a specific target material, some scenarios utilize AI tools to initially filter records, then extract corresponding historical price data to assist in assessing the current procurement cost range, or combine the procurement personnel's experience with the AI's preliminary estimation results to determine the difference between historical prices and the current market environment.

[0003] However, existing methods for managing and querying historical data related to material inquiries still have significant limitations: On the one hand, existing databases primarily focus on storing basic material identification information and prices. Even with the introduction of AI tools, they fail to utilize AI technology to perform correlation analysis and feature mining on multi-dimensional attributes that directly affect prices, such as transportation distance, process requirements, and payment methods. This results in an inability to fully trace the reasons for different prices. On the other hand, data updates often rely on manual operations. Even with AI-assisted data entry, it is difficult to cover the in-depth processing of unstructured data. This not only makes it difficult to guarantee the real-time nature of the data but may also lead to incomplete data due to insufficient AI extraction accuracy or human oversight. Furthermore, AI matching algorithms can only perform preliminary screening based on single fields such as material name and specifications. They cannot accurately correlate price differences under different times and influencing conditions through semantic understanding, multi-feature fusion, and other AI capabilities. This makes it difficult for users to obtain historical price references that highly match their current procurement needs, thus hindering their ability to accurately determine the reasonable costs to be invested in current procurement and making it difficult to control the cost of material procurement. Summary of the Invention

[0004] The purpose of this application is to provide an intelligent price inquiry method, device, electronic device, and storage medium to solve the cost control problem caused by complex price influencing factors and inaccurate data.

[0005] A first aspect of this application provides an intelligent price inquiry method, including:

[0006] Receive query requests input by users. The query requests contain query conditions for the target material. The query conditions include the attribute values ​​corresponding to the attribute items to be queried.

[0007] Retrieve the price value of the target material that matches the query criteria from the target invoice database; the target material that matches the query criteria includes multiple query results, each query result including: multiple material attribute items and their corresponding attribute values; the multiple material attribute items include at least one price item; the multiple material attribute items include at least one structured feature item and at least one unstructured feature item;

[0008] The target invoice database is updated using the following methods:

[0009] Retrieve historical electronic tickets to be updated at preset intervals;

[0010] For each historical electronic ticket, perform the following operations:

[0011] Feature extraction is performed on the historical electronic ticket to obtain the historical ticket data corresponding to the historical electronic ticket. The historical ticket data includes the feature values ​​corresponding to each feature item.

[0012] Candidate bill data are obtained by matching the structured feature values ​​in historical bill data with the structured feature values ​​of each bill stored in the current target bill database.

[0013] The matching results are obtained by matching the unstructured feature values ​​in historical invoice data and the unstructured feature values ​​in candidate invoice data;

[0014] If there are candidate invoices with a matching degree greater than the first preset matching degree threshold, the historical invoice data will be stored in the target invoice database.

[0015] A second aspect of this application provides an intelligent price inquiry device, comprising:

[0016] The query request receiving module is used to receive query requests input by users. The query request contains the query conditions for the target material. The query conditions include the attribute values ​​corresponding to the attribute items to be queried.

[0017] The inquiry module is used to retrieve the price item values ​​of target materials that match the query conditions from the target invoice database. The target materials that match the query conditions include multiple query results. Each query result includes: multiple material attribute items and their corresponding attribute item values. The multiple material attribute items include at least one price item. The multiple material attribute items include at least one structured feature item and at least one unstructured feature item.

[0018] The target invoice database is updated using the following methods:

[0019] Retrieve historical electronic tickets to be updated at preset intervals;

[0020] For each historical electronic ticket, perform the following operations:

[0021] Feature extraction is performed on the historical electronic ticket to obtain the historical ticket data corresponding to the historical electronic ticket. The historical ticket data includes the feature values ​​corresponding to each feature item.

[0022] Candidate bill data are obtained by matching the structured feature values ​​in historical bill data with the structured feature values ​​of each bill stored in the current target bill database.

[0023] The matching results are obtained by matching the unstructured feature values ​​in historical invoice data and the unstructured feature values ​​in candidate invoice data;

[0024] If there are candidate invoices with a matching degree greater than the first preset matching degree threshold, the historical invoice data will be stored in the target invoice database.

[0025] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent inquiry method.

[0026] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the intelligent price inquiry method described above.

[0027] The beneficial effects of the intelligent price inquiry method, device, electronic device, and storage medium provided in this application embodiment are as follows: Firstly, this application embodiment requires receiving a query request containing the value of the attribute item to be queried, which can accurately locate the target material required by the user and avoid result deviations caused by fuzzy queries. The target invoice database contains multiple query results, each containing structured / unstructured feature items and price items, which can completely retain material price data under different times and different influencing conditions, providing a basis for cost comparison. Simultaneously, the database uses a dual matching mechanism—regularly acquiring historical electronic invoices, filtering candidate data through structured feature matching, and accurately verifying data through unstructured feature matching—to ensure both data timeliness and accuracy. Based on this, when users query, they can obtain multi-dimensional historical price data matching their current needs, clearly understand the price differences under different times and conditions, and then combine this with the current procurement scenario to determine the reasonable cost to be invested, completely solving the cost control problem caused by complex price influencing factors and inaccurate data. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating an intelligent price inquiry method provided in an embodiment of this application;

[0030] Figure 2 A schematic diagram illustrating the process of updating the target invoice database according to an embodiment of this application;

[0031] Figure 3 A structural block diagram of an intelligent price inquiry device provided in an embodiment of this application;

[0032] Figure 4 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0033] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0034] It is understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0035] It should be noted that the terms "first," "second," etc., used in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0036] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0037] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the intelligent price inquiry method provided in this application. The intelligent price inquiry method provided in this application embodiment can be executed by an electronic device, and the method may include:

[0038] S101: Receive a query request input by the user. The query request contains the query conditions for the target material. The query conditions include the attribute values ​​corresponding to the attribute items to be queried.

[0039] In this embodiment, a query request is a user-initiated instruction to obtain the price of a target material, including query conditions. The target material is the specific material whose price the user wishes to query, such as HRB400E steel bars or PVC drainage pipes; it is the object of the price inquiry. The query conditions are the filtering criteria set by the user to locate the target material, consisting of the attribute items to be queried and their corresponding attribute item values, used to limit the query scope. Attribute items are field names describing material characteristics, such as material name, transportation distance, price, and free text description; they are the basic units organizing material information in the target invoice database. Attribute item values ​​are the specific content of the corresponding attribute item; for example, the attribute item value for the attribute item "Material Name" is "HRB400E Steel Bars," and the attribute item value for the attribute item "Price" is "4800 RMB / ton."

[0040] In this embodiment, the user can input a query request through the client. The request includes the query conditions for the target material, that is, the user needs to specify the attribute items of the material to be queried and the corresponding attribute item values.

[0041] For example, query the price of "Attribute item: Material name, Attribute item value: HRB400E steel bar; Attribute item: Specification model, Attribute item value: Diameter 20mm".

[0042] S102: Retrieve the price item value of the target material that matches the query conditions from the target invoice database; the target material that matches the query conditions includes multiple query results, each query result including: multiple material attribute items and their respective attribute item values; the multiple material attribute items include at least one price item; the multiple material attribute items include at least one structured feature item and at least one unstructured feature item.

[0043] In this embodiment, the target invoice database is a database storing electronic invoice data for the company's historical procurement transactions, supporting price inquiry queries and periodic updates. When a user initiates a query, this embodiment can perform batch comparison and screening of a large number of historical transaction records in the database based on the query conditions, quickly locating the target material that matches the query conditions, corresponding to multiple query results. Each query result corresponds to the invoice data of one historical transaction, and each query result includes multiple material attribute items and their corresponding attribute values. In addition, this embodiment can also perform batch queries. Users can input multiple query conditions at once. These conditions can correspond to different types of materials or to a batch of materials of the same type but different models. This embodiment can leverage an AI-driven multi-task parallel processing engine to simultaneously perform batch screening of massive historical transaction records in the database. Through precise field-level matching of structured feature items and semantic parsing of unstructured feature items, multiple query results and price values ​​are independently filtered for each query condition, and output synchronously in batches. This eliminates the need to repeatedly initiate a single query, significantly improving the efficiency of centralized price inquiries for multiple materials, perfectly adapting to the cost research needs of enterprises before bulk procurement. The price field is a specific field in the material attributes section of each query result that records the transaction price of the material. For example, "unit price: 5780 yuan / ton" in a query result is the target field for the user's price inquiry. Structured attribute fields are fields in the material attributes section of each query result that have a fixed format and clear meaning, such as transportation distance, supplier name, purchase quantity, and payment method. Their attribute values ​​have a uniform format, such as transportation distance: 80km, payment method: monthly settlement for 30 days, and can be directly used for field-level matching. Unstructured attribute fields are fields in the material attributes section of each query result that do not have a fixed format and exist in free text form, such as descriptions in the invoice remarks column such as "cold-resistant cables need to be fitted with protective sleeves."

[0044] In this embodiment, the target material matching the query conditions corresponds to multiple query results, each containing complete material attribute information: including both the price item the user needs to query and structured and unstructured feature items used for matching. During the query, the user-input query conditions are compared with the corresponding attribute item values ​​of each query result in the database, filtering out multiple query results that are completely or highly matched, and then extracting the price item value of the target material from these results to complete the price query.

[0045] For example, suppose a power engineering company needs to purchase "YJV22 cross-linked polyethylene insulated power cable" for substation construction. Because the price of this cable is significantly affected by the purchase time (prices increase during peak season due to tight supply and demand, and decrease during off-season) and the delivery location (transportation distance affects logistics costs), the historical purchase price fluctuates greatly (e.g., the off-season price in XX month of XX year was 48 yuan / meter, and the peak season price in April was 52 yuan / meter), manual cost estimation is prone to errors.

[0046] The procurement specialist initiated a query request through the system client, setting the query criteria to: "Material Name" value "YJV22 Cross-linked Polyethylene Insulated Power Cable" and "Specification Model" value "3×120". The attribute item "Purchase Time" value is "June 2025", and the attribute item "Delivery Location" value is "Substation in the eastern suburbs of a certain city".

[0047] The system matches data from the target invoice database, which stores structured features (transportation distance, purchase time, unit price) and unstructured features (note: "anti-aging packaging required for summer high temperatures"). It then filters invoices for the same specifications and similar transportation distances in year YY and month YY, extracting the currently matched price value of "51 yuan / meter". Based on this, the purchasing specialist calculates the total cost of purchasing 10,000 meters of cable to be 510,000 yuan, avoiding a 30,000 yuan cost gap caused by estimating based on the historical low price of 48 yuan / meter, thus achieving precise cost control.

[0048] Figure 2 The following is a schematic diagram illustrating the process of updating a target invoice database according to an embodiment of this application, wherein the target invoice database is updated in the following manner:

[0049] Retrieve historical electronic tickets to be updated at preset intervals;

[0050] In this embodiment, historical electronic invoices are electronic vouchers generated by the enterprise in past procurement processes, such as e-invoices. They contain the latest material information and price data and serve as the data source for updating the target invoice database. The preset time can be set by the enterprise according to business needs, such as every Wednesday or every Monday, to ensure regular synchronization of the latest transaction data. The purpose of retrieving historical electronic invoices to be updated at preset intervals is to supplement the database with new transaction records and avoid using outdated data during queries.

[0051] For each historical electronic ticket, perform the following operations:

[0052] S201: Extract features from the historical electronic ticket to obtain the historical ticket data corresponding to the historical electronic ticket. The historical ticket data includes the feature values ​​corresponding to each feature item.

[0053] In this embodiment, historical invoice data is a structured dataset obtained by extracting features from historical electronic invoices. It contains various feature items and their feature values, and can be directly used for database matching. Feature items are specific fields in the historical invoice data that describe material characteristics, corresponding one-to-one with material attribute items. For example, the feature item "transportation distance" corresponds to the structured feature item "transportation distance" in the material attribute items. Feature values ​​are the specific content of the corresponding feature item; for example, the feature value of the feature item "transportation distance" is 80km.

[0054] In this embodiment, AI-driven recognition technology and field matching models are used to accurately extract feature values ​​of structured feature items from fixed-format fields such as material name and purchase quantity in the invoice header. For text content without a fixed format, an AI semantic parsing model is used for text mining and key information extraction to obtain feature values ​​of unstructured feature items, which are then integrated into historical invoice data. This approach improves the efficiency and accuracy of feature extraction through AI technology, effectively eliminates format differences in the original invoices, and provides a unified and reliable data foundation for subsequent matching.

[0055] S202: Match the structured feature values ​​in the historical invoice data with the structured feature values ​​of each invoice stored in the current target invoice database to obtain candidate invoice data; match the unstructured feature values ​​in the historical invoice data with the unstructured feature values ​​in the candidate invoice data.

[0056] In this embodiment, the candidate invoice data is a set of existing invoices that may belong to the same material as the historical invoice to be updated, selected from the database by structured feature value matching, and used for subsequent unstructured feature matching.

[0057] In this embodiment, the matching range can be quickly narrowed down using structured features. Structured feature values ​​refer to the numerical values ​​of fixed-format fields in historical invoices, such as transportation distance: 80km, supplier name: Huagang Group, etc. This embodiment utilizes an AI-driven field-level precise comparison algorithm to intelligently match these values ​​with the structured feature values ​​of each existing invoice in the database, quickly filtering out existing invoices with highly similar structured features to form candidate invoice data (i.e., a set of invoices that may belong to the same material as the historical invoice to be updated), significantly reducing the computational load of subsequent unstructured matching. Unstructured feature values ​​refer to the free text content in historical invoices, such as "high-toughness PVC drainage pipe" and "cold-resistant, suitable for outdoor installation in northern regions." This embodiment uses an AI semantic vectorization model to convert the text into semantic vectors, and then uses an AI similarity matching algorithm to calculate its semantic similarity with the unstructured feature values ​​in the candidate invoice data, obtaining precise matching results. This embodiment enhances the efficiency of structured comparison and the ability to understand unstructured semantics through AI technology, verifying whether historical invoices and candidate invoices belong to the same material, avoiding misjudgments caused by relying solely on structured feature matching.

[0058] For example, suppose a building materials company updates its historical electronic invoices for PVC drainage pipes. The structured feature value of this invoice is: "Specification: DN110mm; Supplier: Shunda Building Materials; Transportation distance: 50km". The system, using an AI-driven field-level precise comparison algorithm, intelligently matches these values ​​with the structured feature values ​​of existing invoices in the target invoice database, filtering out two candidate invoices (both with "DN110mm specification, supplied by Shunda Building Materials, transportation distance 45-55km").

[0059] The unstructured feature value of the historical invoice is "cold-resistant type, suitable for outdoor installation in northern regions," while the unstructured feature value of the candidate invoice is "freeze-resistant type, for outdoor use in northern regions." To accurately calculate the semantic similarity between the two, this embodiment employs the following specific technical means:

[0060] A lightweight pre-trained model, all-MiniLM-L6-v2 (which can be loaded via the sentence-transformers library), specifically optimized for sentence semantic matching, can be used. This model maps text to 384-dimensional normalized semantic vectors.

[0061] Input the two texts corresponding to the unstructured feature values ​​of historical tickets and the unstructured feature values ​​of candidate tickets into the above model. The model outputs a normalized semantic vector. For illustration, the exemplary values ​​and calculation process of the first 6 dimensions of the vector are shown below:

[0062] Example values ​​for the historical ticket remarks vector V1: [0.12, -0.45, 0.08, 0.67, -0.23, 0.51, ...];

[0063] Example values ​​for candidate ticket memo vector V2: [0.15, -0.38, 0.10, 0.72, -0.19, 0.48, ...];

[0064] The values ​​here represent examples of a single, actual output from the model for the given text, illustrating the numerical characteristics and computational process of the vectors. In practical applications, because the model weights are fixed, each inference for the same text will produce the same vector; for different but semantically similar texts, vectors with different numerical values ​​but high cosine similarity will be produced.

[0065] Since vectors V1 and V2 have been normalized (with a magnitude approximately of 1), the cosine similarity is simplified to a dot product. Based on the first 6 dimensions of the example values ​​above, the dot product calculation principle is as follows:

[0066] The dot product V1·V2 = (0.12×0.15) + (-0.45×(-0.38)) + (0.08×0.10) + (0.67×0.72) + (-0.23×(-0.19)) + (0.51×0.48) + … is summed over all 384 dimensions.

[0067] After calculating all 384 dimensions of numerical values, the final dot product (i.e., cosine similarity) result is 0.85.

[0068] The result of 0.85 exceeds the system's preset first matching threshold of 0.8, thus confirming that the two tickets are semantically matched successfully. This process effectively avoids misjudgments caused by relying solely on structured feature matching while ignoring the semantic equivalence of material characteristics.

[0069] S203: If there are candidate invoice data with a matching degree greater than the first preset matching degree threshold, then the historical invoice data will be stored in the target invoice database.

[0070] In this embodiment, the first preset matching threshold is an intelligent similarity standard preset based on the AI ​​algorithm's learning and optimization of historical matching data, used to determine whether historical invoices and candidate invoices belong to the same material. If at least one candidate invoice and a historical invoice have a matching degree exceeding the threshold calculated by the AI ​​similarity calculation model, it indicates that the AI ​​intelligently determines that the historical invoice belongs to a material type already existing in the database and can be safely stored in the database to supplement the historical transaction records of that material; if the threshold is not reached, further manual verification is required. By using AI to assist in setting the threshold and determining the matching results, the integrity and accuracy of the database are precisely balanced, preventing redundant or erroneous data from entering the system and improving the intelligence level of data filtering.

[0071] As can be seen from the above, this embodiment first requires receiving query requests containing the values ​​of the attributes to be queried, which can accurately locate the target material required by the user and avoid the result deviation caused by fuzzy queries. The target invoice database contains multiple query results, each containing structured / unstructured feature items and price items, which can completely retain material price data under different times and different influencing conditions, providing a basis for cost comparison. At the same time, the database uses a dual matching mechanism of periodically acquiring historical electronic invoices, filtering candidate data through structured feature matching and accurately verifying data through unstructured feature matching, to ensure both data timeliness and data accuracy. Based on this, when users query, they can obtain multi-dimensional historical price data that matches their current needs, clearly understand the price differences under different times and conditions, and then combine this with the current procurement scenario to judge the cost-effectiveness of the investment, completely solving the cost control problem caused by the complexity of price influencing factors and inaccurate data.

[0072] In one embodiment of this application, the method further includes:

[0073] Based on the query conditions, retrieve multiple material attribute items and their corresponding attribute values ​​for the target material that match the query conditions from the target invoice database;

[0074] Extract at least one material attribute from multiple material attribute items as an influencing factor affecting the price of the target material; for each influencing factor, determine the rate of change corresponding to the influencing factor based on the attribute item value corresponding to the influencing factor;

[0075] The weight of the corresponding impact factor is determined based on the rate of change of the impact factor.

[0076] The predicted price of the target material is calculated based on the benchmark price of the target material, the rate of change of the influencing factors, and the corresponding weights; the benchmark price is the latest market price of the target material.

[0077] In this embodiment, the influencing factor refers to the feature items extracted from multiple material attributes of the target material that have a significant impact on its price, such as transportation distance, purchase quantity, process requirements, and supplier cooperation level. These are key factors causing price fluctuations. The rate of change is the deviation ratio of the current attribute value of a certain influencing factor from the historical reference value of that factor, such as the deviation ratio from the historical average, reflecting the magnitude of change of the influencing factor. The benchmark price is the latest market price of the target material, such as the real-time price obtained through the market API interface, which serves as the basic reference value for calculating the predicted price and reflects the current basic pricing level of the market.

[0078] In this embodiment, based on the user-input query conditions, multiple material attribute items and their corresponding attribute values ​​for the target material matching the query conditions are filtered from the target invoice database. This process is not limited to obtaining the price item; it also needs to extract all structured and unstructured feature items and their attribute values ​​related to the material, providing a data foundation for subsequent analysis of which factors affect the price.

[0079] From the complete attribute information obtained, identify the features that have a significant impact on the price of the target material, namely the influencing factors, such as transportation distance or supplier cooperation level; at the same time, calculate the rate of change of each influencing factor, that is, the deviation ratio of the current attribute value of the influencing factor from the historical reference value. For example, if the current transportation distance is 100km and the historical average is 80km, then the rate of change is (100-80) / 80=25%, reflecting the magnitude of change of each influencing factor.

[0080] In this embodiment, the weight of each factor can be determined based on the rate of change of the influencing factors. The weight is a numerical value that measures the degree of influence of a factor on the price; the greater the fluctuation (i.e., the higher the rate of change), the more significant the influence of that factor on the price, and the higher its weight; conversely, the lower the weight, the lower the weight. This ensures that factors with a greater impact on the price have a more significant weight in subsequent calculations. Starting from the benchmark price, and combining the rate of change and weight of each influencing factor, the predicted price of the target material is calculated using a preset formula.

[0081] As can be seen from the above, this embodiment obtains complete material attribute information matching the query conditions, accurately extracts the factors affecting the price and their rate of change, determines the factor weights based on the volatility of the rate of change, and calculates the predicted price in conjunction with the latest market benchmark price. This approach quantifies the impact of each factor on the price, making the price more relevant to the current procurement scenario, solving the problem of inaccurate cost control caused by unclear influencing conditions, and improving the accuracy of procurement cost assessment.

[0082] In one embodiment of this application, a predicted price quote for the target material is calculated based on the benchmark price of the target material, the rate of change of influencing factors, and the corresponding weights, including:

[0083] The predicted price of the target material is obtained by calculating based on the first formula;

[0084] The first formula is:

[0085]

[0086] in, Indicates the predicted price. Indicates the benchmark price. This represents the rate of change of the i-th influencing factor. This represents the weight corresponding to the rate of change of the i-th influencing factor.

[0087] In this embodiment, the benchmark price of the target material and the rate of change of each influencing factor extracted earlier are first obtained. and weight For each influencing factor, through Calculate its single-factor adjustment contribution to the price. The higher the weight, the greater the normalized rate of change, and the more significant the adjustment of the price by that factor. =0.3、 When =0.5, =0.3×(0.5 / (1+0.5))=0.1, meaning this influencing factor pushes up the price by 10%; by summing up the adjustment contributions of all individual influencing factors, the total adjustment magnitude is obtained. For example, if the adjustment contributions of the three factors are 0.1, -0.05, and 0.03 respectively, the total adjustment magnitude is 0.08; using the benchmark price Multiply by "1 + total adjustment range" to get the predicted price. If the total adjustment is positive, the quoted price is higher than the benchmark price; if it is negative, the quoted price is lower than the benchmark price. The final result is consistent with the actual situation of each influencing factor, avoiding the quotation deviation caused by extreme fluctuations of a single factor, and achieving accurate cost prediction.

[0088] in, This represents the normalization term for the rate of change of the i-th influencing factor, which is used to normalize the change rate of the i-th influencing factor. Limited to a reasonable range of [-0.5, 0.5] (e.g.) When =1, this term =0.5; At that time, the item This avoids distortion of predicted prices due to extreme changes in a single factor, ensuring computational stability.

[0089] For example, suppose a power company needs to purchase "YJV22-3×120 "Cross-linked polyethylene cable", the system calculates and predicts the price: benchmark price The latest market price is 50 yuan / meter (obtained via API). Three influencing factors were selected:

[0090] Transportation distance: Current 120km, historical average 100km, rate of change =(120-100) / 100=0.2; Due to large fluctuations in transportation distance, the weight... =0.4;

[0091] Purchase quantity: Current 8000 meters, historical average 10000 meters, rate of change =(8000-10000) / 10000=-0.2; moderate fluctuation, weight. =0.3;

[0092] Supplier rating: Currently A (historically mostly B), rate of change =0.5; small fluctuations, weight =0.3.

[0093] Calculated using the formula: the single-factor adjusted contribution values ​​are approximately 0.4 × (0.2 / 1.2) ≈ 0.067 and 0.3 × (-0.2 / 1.2), respectively. -0.05, 0.3 × (0.5 / 1.5) = 0.1; Total adjustment range ≈ 0.117. Forecast price. =50×(1+0.117)=55.85 yuan / meter.

[0094] This result takes into account factors such as transportation, quantity, and suppliers, avoiding the extreme impact of a single factor, and provides an accurate reference for purchasing 10,000 meters of cable (estimated total cost of 558,500 yuan), solving the problem of deviation from estimation based solely on historical prices.

[0095] From the above, it can be concluded that this embodiment achieves its effect through the formula... Normalize the rate of change of influencing factors to avoid price distortion caused by extreme fluctuations of a single factor; combine with weights By overlaying the influence of multiple factors and accurately quantifying the role of each factor in price, and then calculating and predicting quotations based on benchmark prices, the accuracy of quotations is effectively improved, providing a reliable basis for precise control of procurement costs.

[0096] In one embodiment of this application, extracting at least one material attribute from multiple material attribute items as an influencing factor affecting the price of a target material includes:

[0097] Feature processing is performed on multiple material attribute items and their corresponding attribute values ​​to obtain target feature data;

[0098] Calculate the average of the price values ​​across multiple query results;

[0099] Calculate the price deviation between the price value of each query result and the average price;

[0100] The target feature data and price deviation are clustered to obtain multiple clusters; each cluster contains the target feature data corresponding to multiple query results that match the target material and the price deviation corresponding to each query result;

[0101] For each cluster, the correlation coefficient between the target feature data within the cluster and all price deviations within the cluster is calculated. The material attribute items corresponding to the target feature data whose absolute value of the correlation coefficient is greater than the preset correlation coefficient threshold are determined as the influencing factors affecting the price of the target material.

[0102] The target feature data includes encoded features and unstructured vector features; the feature processing includes:

[0103] Extract the structured feature values ​​corresponding to the structured feature items from multiple material attribute items, and numerically encode the structured feature values ​​to obtain the coded features;

[0104] Extract the unstructured feature values ​​corresponding to the unstructured feature items from multiple material attribute items, and perform semantic vectorization processing on the unstructured feature values ​​to obtain unstructured feature vectors.

[0105] In this embodiment, the target feature data is a set of computable data obtained after feature processing of material attribute items and attribute values, including two categories: coded features and unstructured feature vectors. Coded features are numerical features obtained by numerically encoding the attribute values ​​of structured feature items, used to solve the problem that categorical attributes cannot participate in mathematical calculations. Unstructured feature vectors are multi-dimensional numerical vectors obtained by semantically vectorizing the attribute values ​​of unstructured feature items, used to convert text information into a computer-analyzable format. Price deviation is the relative difference between the price value of a single query result and the average price of multiple query results, used to measure the degree of fluctuation of a single record's price relative to a benchmark level, and serves as a label for the correlation between features and price fluctuations. The preset correlation coefficient threshold is a preset critical value used to determine whether a feature is strongly correlated with the price deviation. Attribute items corresponding to features with correlation coefficients exceeding this threshold are determined to be key influencing factors affecting prices.

[0106] In this embodiment, the multiple material attribute items and attribute values ​​of the target material are first standardized to solve the problem that non-numerical attributes cannot participate in the calculation.

[0107] For structured feature items, after extracting their attribute values, the classification type values, such as prepayment or 30-day monthly settlement, are numerically encoded to obtain coded features, such as prepayment=0, 30-day monthly settlement=1, converting fixed-format non-numerical attributes into quantifiable values; for unstructured feature items, after extracting their attribute values, semantic vectorization techniques, such as Word2Vec, are used to convert the text into unstructured feature vectors, converting text information without a fixed format into computer-computable vector data;

[0108] Finally, the two types of results are integrated to form target feature data, providing a unified data format for subsequent statistical analysis.

[0109] In this embodiment, since the price of a single query result may be affected by accidental factors (such as temporary discounts), it is necessary to first determine the price benchmark and then measure the price fluctuation of each record.

[0110] Calculate the average price value of multiple query results. Using the prices of all matching query results for the target material as a sample, the average value is taken as the price benchmark to eliminate the randomness of individual data. Calculate the price deviation. For each query result, the price deviation is represented by the difference between its price value and the above average value (e.g., deviation = (single price - average value) / average value). This value directly reflects the degree of price fluctuation of a single record relative to the benchmark.

[0111] In this embodiment, the target feature data and price deviation are input into a clustering algorithm, such as the k-means clustering algorithm (K-Means), with the aim of grouping query results with similar features and consistent price fluctuation trends into one category.

[0112] For example, records with a transportation distance > 100km, payment method = monthly settlement for 60 days and a positive price deviation are grouped into one cluster, while records with a transportation distance < 50km, payment method = prepayment and a negative price deviation are grouped into another cluster;

[0113] Clustering serves to avoid correlation distortion caused by feature mixing. Records within the same cluster have similar characteristics and price fluctuation patterns, so subsequent analysis only needs to be performed on the correlation between features and prices within the cluster, which greatly improves the screening accuracy.

[0114] For each cluster, the correlation coefficient (such as the Pearson correlation coefficient) between the target feature data within the cluster and the price deviation is calculated to quantify the strength of the linear association between the feature and price fluctuations.

[0115] A preset correlation coefficient threshold is set, and the original material attribute items corresponding to features whose absolute correlation coefficient exceeds the threshold are identified as influencing factors. The higher the correlation coefficient, the more synchronous the change of the feature is with the change of price deviation, that is, the more obvious the impact of the feature on the price, and finally achieve accurate screening from all attributes to influencing factors.

[0116] For example, suppose a company queries YJV22-3×120 To determine the influencing factors of power cable prices, after obtaining 10 matching query results from the target invoice database, proceed as follows:

[0117] First, feature processing is performed: In the structured feature items, the transportation distance is directly taken as a numerical value, and the payment method is encoded as coded features with prepayment=0 and monthly settlement for 30 days=1; the unstructured feature item notes are converted into unstructured feature vectors through Word2Vec and integrated into the target feature data.

[0118] Next, calculate the price benchmark and deviation: the average price value of the 10 results is 50 yuan / meter. The deviation of a certain price of 55 yuan / meter is (55-50) / 50=10%, and the deviation of another price of 48 yuan / meter is -4%.

[0119] K-Means clustering was then used to cluster the four results with a transportation distance > 80km, payment method = 1 and positive deviation into cluster 1, and the five results with a transportation distance < 50km, payment method = 0 and negative deviation into cluster 2.

[0120] Finally, the correlation coefficients were calculated: in cluster 1, the correlation coefficients between transportation distance and deviation were 0.78 and 0.66, both exceeding the preset threshold of 0.6. Therefore, transportation distance and payment method were determined to be the price influencing factors of the cable.

[0121] As can be seen from the above, this embodiment transforms non-numerical attributes into computable data through feature processing, establishes a benchmark based on the price mean, calculates deviations to quantify fluctuations, then uses clustering to avoid correlation distortion caused by feature confounding, and finally selects strongly correlated attributes as influencing factors based on correlation coefficients. This can accurately identify key price influencing factors, providing a reliable basis for subsequent accurate pricing and cost control, and solving the problem of traditional methods struggling to determine core factors.

[0122] In one embodiment of this application, candidate bill data is obtained by matching the structured feature values ​​in historical bill data with the structured feature values ​​of each bill data stored in the current target bill database, including:

[0123] Calculate the confidence level of structured feature values ​​in historical invoice data, and use structured feature values ​​with a confidence level greater than a preset confidence level as target feature values; the confidence level is a score of the accuracy of structured feature value recognition during the feature extraction process of historical electronic invoices;

[0124] Candidate invoice data are obtained by matching the target feature values ​​with the structured feature values ​​of each invoice stored in the current target invoice database.

[0125] In this embodiment, confidence level refers to a quantitative score of the accuracy of structured feature value identification during the feature extraction process of historical electronic tickets. A higher score indicates a more reliable extraction result for the structured feature value and less susceptibility to identification errors. The preset confidence level is a pre-defined threshold used to filter reliable structured feature values. Only structured feature values ​​with a confidence level exceeding this threshold are identified as target feature values ​​and used as the basis for subsequent matching. Its function is to eliminate interference from low-reliability features in the matching results. The target feature value refers to a structured feature value with a confidence level greater than the preset confidence level. It is core structured feature data that has undergone reliability verification and can be used for matching, ensuring that subsequent matching with the structured feature values ​​of tickets in the database is based on accurate and reliable information.

[0126] In this embodiment, when extracting features from historical electronic invoices, the extracted structured feature values ​​are simultaneously scored for recognition accuracy, i.e., confidence level. This score quantifies the reliability of the structured feature value. For example, the purchase quantity extracted from a clear table field on an electronic invoice has a higher confidence level than the payment method extracted from a blurred OCR recognition area. A preset confidence level (e.g., 60 points out of 100) is set as a reliability threshold, and structured feature values ​​with a confidence level greater than this threshold are identified as target feature values. The purpose is to exclude inaccurate or unreliable structured features, such as misidentifying "transportation distance: 180km" due to blurred invoices when it should actually be "80km," ensuring the validity of subsequent matching criteria.

[0127] Based on the selected target feature values, a matching process is performed with the structured feature values ​​of each invoice in the current target invoice database. Invoices with highly consistent structured features are selected to form candidate invoice data. By focusing on reliable structured features, the matching range is significantly narrowed, improving the efficiency and accuracy of subsequent unstructured feature matching.

[0128] As can be seen from the above, this embodiment calculates the confidence level of structured feature values ​​and selects target feature values ​​with higher than the preset confidence level for matching. This can eliminate interference from low-reliability features that are not accurately identified, ensure the reliability of the matching basis, thereby improving the accuracy of candidate ticket data, reducing the error of subsequent unstructured matching, and providing effective support for accurate database updates.

[0129] In one embodiment of this application, the target invoice database includes multiple historical matching invoice records;

[0130] The method also includes:

[0131] In response to the absence of a matching degree greater than the first preset matching degree threshold in the candidate bill data, the historical bill data is matched with the feature values ​​corresponding to each feature item in at least one historical matching bill data.

[0132] If the number of matches between at least one historical matching invoice and the corresponding feature values ​​of each feature item in the historical invoice data is greater than the preset number, then the historical invoice data will be stored in the manual matching database.

[0133] In this embodiment, historical matching document data refers to document data that has been successfully entered into the target document database through matching verification in the past. It is valid historical data recognized by the system. The preset quantity is a preset threshold (e.g., 3) used to determine whether there are enough common feature values ​​matching between historical document data and historical matching document data. Exceeding this threshold indicates that the two have a high degree of feature overlap and further manual verification is required. The manual matching database is a database specifically used to store historical document data that does not meet the automatic entry standard but has certain common features. The data in this database needs to be manually verified before deciding whether to transfer it to the target document database, playing a role of manual backup and balancing data integrity and accuracy.

[0134] In this embodiment, when the matching degree between all tickets in the candidate ticket data and historical ticket data does not exceed the first preset matching degree threshold, it indicates that there are currently no highly similar tickets in the database. However, the historical ticket data may still be valuable, so a secondary matching mechanism is initiated. The historical ticket data is compared with multiple historical matching ticket data already stored in the target ticket database, and the feature values ​​corresponding to each feature item in both are matched one by one. The number of successfully matched feature values ​​is counted. A preset number is set as a judgment threshold. If the number of matching feature values ​​between at least one historical matching ticket data and the historical ticket data exceeds the threshold, it indicates that there is some commonality between the two, but it does not meet the automatic entry standard and requires manual verification. Therefore, the historical ticket data is stored in the manual matching database. If the threshold is not reached, it may be invalid data or extremely special data, and it is not entered into the database for the time being. This approach avoids omitting potentially valid data and reduces the risk of incorrect entry through manual verification.

[0135] As can be seen from the above, this embodiment addresses scenarios where candidate invoices lack high matching degrees. It uses historically matched invoices already verified in the database as a benchmark and performs secondary feature value matching on the historical invoice data. By filtering out invoices with commonalities through a preset number, it stores them in a manual database. This avoids omitting potentially valid data and uses manual verification as a fallback to prevent erroneous data entry, ensuring both the completeness and accuracy of database updates and laying a solid data foundation for subsequent accurate price inquiries.

[0136] In one embodiment of this application, the method further includes:

[0137] Adjust the first preset matching degree threshold;

[0138] The adjustment process for the first preset matching threshold includes:

[0139] Retrieve all historical matching records and calculate the matching success rate at a first preset matching threshold.

[0140] Compare the matching success rate with the preset matching success rate range;

[0141] If the matching success rate is higher than the upper limit of the preset matching success rate range within the preset time period, the first preset matching degree threshold will be reduced by the first adjustment step.

[0142] If the matching success rate is lower than the lower limit of the preset matching success rate range within the preset time period, the first preset matching degree threshold will be increased according to the second adjustment step.

[0143] In this embodiment, historical matching records refer to all result records generated by past document matching operations, including information such as the matching degree and whether the matching was successful for each match. The matching success rate is the ratio of the number of successfully matched records to the total number of matched records under the current first preset matching degree threshold, reflecting the actual matching effect of the current threshold. The preset matching success rate range is a preset success rate interval reflecting a reasonable matching effect. The upper limit ensures that the matching is not too lenient, avoiding the entry of low-quality data into the database, while the lower limit ensures that the matching is not too strict, avoiding the omission of valid data. It serves as the benchmark for determining whether the threshold needs adjustment. The preset time period is a fixed time period used to observe the trend of the matching success rate, avoiding erroneous threshold adjustments due to short-term accidental fluctuations and ensuring that adjustments are based on stable business patterns. The first adjustment step size is the amount by which the first preset matching degree threshold is reduced when the matching success rate is higher than the upper limit of the preset range. The second adjustment step size is the amount by which the first preset matching degree threshold is reduced when the matching success rate is lower than the lower limit of the preset range. It is usually set to a different magnitude than the first adjustment step size to ensure that the threshold is fine-tuned to a reasonable range, balancing adjustment efficiency and stability.

[0144] In this embodiment,

[0145] First, all historical matching records in the system are collected. Based on the current first preset matching threshold, the matching success rate is calculated, which is the proportion of successfully matched records to the total number of matched records. This serves as the core indicator for evaluating the reasonableness of the current threshold. The calculated matching success rate is compared with a preset matching success rate range (e.g., 80%-90%, a preset reasonable range that balances matching accuracy and data coverage) to determine if the current threshold needs adjustment: if the success rate is too high, it indicates that the threshold may be too low, leading to too many low-quality matches; if the success rate is too low, it indicates that the threshold may be too high, leading to missed valid matches. A preset time period (e.g., 1 week) is set as the observation period to ensure that adjustments are based on stable trends rather than random fluctuations: if the matching success rate is consistently higher than the upper limit of the preset range during this period, it indicates that the current threshold is too lenient and needs to be lowered by the first adjustment step (e.g., decreasing by 5% each time) to reduce low-quality matches; if it is consistently lower than the lower limit of the preset range (e.g., below 80%), it indicates that the current threshold is too strict and needs to be increased by the second adjustment step (e.g., increasing by 3% each time) to reduce the omission of valid data.

[0146] Through the above dynamic adjustments, the first preset matching threshold is always adapted to the actual data characteristics, balancing the accuracy and completeness of the matching.

[0147] As can be seen from the above, this embodiment calculates the matching success rate under the current threshold by obtaining historical matching records, compares it with the preset success rate range, and dynamically adjusts the first preset matching degree threshold by step size within a preset time period. This avoids the problem of matching being too lenient or too strict due to a fixed threshold, ensuring that the threshold continuously adapts to the actual business situation, stabilizing the matching effect, guaranteeing the quality of database data, and laying a solid data foundation for subsequent accurate price inquiries.

[0148] Based on the same inventive concept, this application also provides an intelligent price inquiry device for implementing the intelligent price inquiry method described above. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more intelligent price inquiry device embodiments provided below can be found in the limitations of the intelligent price inquiry method described above, and will not be repeated here.

[0149] This application provides an intelligent price inquiry device, such as... Figure 3 As shown, the intelligent inquiry device 20 includes: an inquiry request receiving module 21 and an inquiry module 22;

[0150] The query request receiving module 21 is used to receive query requests input by users. The query request includes query conditions for the target material. The query conditions include the attribute values ​​corresponding to the attribute items to be queried.

[0151] The inquiry module 22 is used to retrieve the price item value of the target material that matches the query conditions from the target invoice database. The target material that matches the query conditions includes multiple query results. Each query result includes: multiple material attribute items and their corresponding attribute item values. The multiple material attribute items include at least one price item. The multiple material attribute items include at least one structured feature item and at least one unstructured feature item.

[0152] The target invoice database is updated using the following methods:

[0153] Retrieve historical electronic tickets to be updated at preset intervals;

[0154] For each historical electronic ticket, perform the following operations:

[0155] Feature extraction is performed on the historical electronic ticket to obtain the historical ticket data corresponding to the historical electronic ticket. The historical ticket data includes the feature values ​​corresponding to each feature item.

[0156] Candidate bill data are obtained by matching the structured feature values ​​in historical bill data with the structured feature values ​​of each bill stored in the current target bill database.

[0157] The matching results are obtained by matching the unstructured feature values ​​in historical invoice data and the unstructured feature values ​​in candidate invoice data;

[0158] If there are candidate invoices with a matching degree greater than the first preset matching degree threshold, the historical invoice data will be stored in the target invoice database.

[0159] In one embodiment of this application, the intelligent price inquiry device 20 further includes: a price prediction module; specifically used for:

[0160] Based on the query conditions, retrieve multiple material attribute items and their corresponding attribute values ​​for the target material that match the query conditions from the target invoice database;

[0161] Extract at least one material attribute from multiple material attribute items as an influencing factor affecting the price of the target material; for each influencing factor, determine the rate of change corresponding to the influencing factor based on the attribute item value corresponding to the influencing factor;

[0162] The weight of the corresponding impact factor is determined based on the rate of change of the impact factor.

[0163] The predicted price of the target material is calculated based on the benchmark price of the target material, the rate of change of the influencing factors, and the corresponding weights; the benchmark price is the latest market price of the target material.

[0164] In one embodiment of this application, when extracting at least one material attribute from multiple material attribute items as an influencing factor affecting the price of a target material, the price prediction module is further configured to:

[0165] Feature processing is performed on multiple material attribute items and their corresponding attribute values ​​to obtain target feature data;

[0166] Calculate the average of the price values ​​across multiple query results;

[0167] Calculate the price deviation between the price value of each query result and the average price;

[0168] The target feature data and price deviation are clustered to obtain multiple clusters; each cluster contains the target feature data corresponding to multiple query results that match the target material and the price deviation corresponding to each query result;

[0169] For each cluster, the correlation coefficient between the target feature data within the cluster and the price deviation of all data within the cluster is calculated. The material attribute items corresponding to the target feature data whose absolute value of the correlation coefficient is greater than the preset correlation coefficient threshold are identified as the influencing factors affecting the price of the target material.

[0170] The target feature data includes encoded features and unstructured vector features; the feature processing includes:

[0171] Extract the attribute values ​​corresponding to the structured feature items from multiple material attribute items, and numerically encode the attribute values ​​to obtain the encoded features;

[0172] Extract the attribute values ​​corresponding to unstructured feature items from multiple material attribute items, and perform semantic vectorization processing on the attribute values ​​to obtain unstructured feature vectors.

[0173] In one embodiment of this application, when calculating the predicted price of the target material based on its benchmark price, the rate of change of influencing factors, and their corresponding weights, the price prediction module is further used for:

[0174] The predicted price of the target material is obtained by calculating based on the first formula;

[0175] The first formula is:

[0176]

[0177] in, Indicates the predicted price. Indicates the benchmark price. This represents the rate of change of the i-th influencing factor. This represents the weight corresponding to the rate of change of the i-th influencing factor.

[0178] In one embodiment of this application, candidate bill data is obtained by matching the structured feature values ​​in historical bill data with the structured feature values ​​of each bill data stored in the current target bill database, including:

[0179] Calculate the confidence level of structured feature values ​​in historical invoice data, and use structured feature values ​​with a confidence level greater than a preset confidence level as target feature values; the confidence level is a score of the accuracy of structured feature value recognition during the feature extraction process of historical electronic invoices;

[0180] Candidate invoice data are obtained by matching the target feature values ​​with the structured feature values ​​of each invoice stored in the current target invoice database.

[0181] In one embodiment of this application, the target invoice database includes multiple historical matching invoice records;

[0182] Also includes:

[0183] In response to the absence of a matching degree greater than the first preset matching degree threshold in the candidate bill data, the historical bill data is matched with the feature values ​​corresponding to each feature item in at least one historical matching bill data.

[0184] If the number of matches between at least one historical matching invoice and the corresponding feature values ​​of each feature item in the historical invoice data is greater than the preset number, then the historical invoice data will be stored in the manual matching database.

[0185] In one embodiment of this application, it further includes:

[0186] Adjust the first preset matching degree threshold;

[0187] The adjustment process for the first preset matching threshold includes:

[0188] Retrieve all historical matching records and calculate the matching success rate at a first preset matching threshold.

[0189] Compare the matching success rate with the preset matching success rate range;

[0190] If the matching success rate is higher than the upper limit of the preset matching success rate range within the preset time period, the first preset matching degree threshold will be reduced by the first adjustment step.

[0191] If the matching success rate is lower than the lower limit of the preset matching success rate range within the preset time period, the first preset matching degree threshold will be increased according to the second adjustment step.

[0192] See Figure 4 , Figure 4This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 4 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 3 The functions of the query request receiving module 21 and the inquiry module 22 are shown.

[0193] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0194] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0195] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store information such as a target ticket database.

[0196] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the intelligent inquiry method provided in the embodiments of this application, or they can execute the implementation method of the electronic device described in the embodiments of this application, which will not be repeated here.

[0197] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0198] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0199] Those skilled in the art will recognize that the modules / units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0200] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0201] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules, units, or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or modules / units, or it may be an electrical, mechanical, or other form of connection.

[0202] The modules / units described as separate components may or may not be physically separate. Similarly, the components shown as modules / units may or may not be physical modules / units; they may be located in one place or distributed across multiple network modules / units. Some or all of the modules / units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0203] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0204] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent bidding method, characterized in that, The method comprises the following steps: receiving a query request input by a user, the query request comprising a query condition of a target material, the query condition comprising an attribute item value corresponding to an attribute item to be queried; obtaining a price item value of the target material matching the query condition from a target ticket database; the target material matching the query condition comprises a plurality of query results, each query result comprising a plurality of material attribute items and respective attribute item values; the plurality of material attribute items at least comprise a price item; the plurality of material attribute items comprise at least one structured feature item and at least one unstructured feature item; wherein the target ticket database is a database updated by the following method: obtaining a historical electronic ticket to be updated at a preset time interval; for each historical electronic ticket, the following operations are performed: performing feature extraction on the historical electronic ticket to obtain historical ticket data corresponding to the historical electronic ticket, the historical ticket data comprising respective feature values corresponding to each feature item; performing corresponding matching between the structured feature item values in the historical ticket data and the structured feature values of each ticket data stored in the current target ticket database to obtain candidate ticket data; performing matching between the unstructured feature item values in the historical ticket data and the unstructured feature values in the candidate ticket data to obtain a matching result; if there is ticket data in the candidate ticket data with a matching degree greater than a first preset matching degree threshold, then storing the historical ticket data in the target ticket database; The method further comprises the following steps: based on the query condition, obtaining a plurality of material attribute items and respective attribute item values of the target material matching the query condition from the target ticket database; extracting at least one material attribute item from the plurality of material attribute items as an influencing factor affecting the price of the target material; for each influencing factor, determining a change rate corresponding to the influencing factor based on the attribute item value corresponding to the influencing factor; determining the weight of the corresponding influencing factor based on the change rate corresponding to the influencing factor; based on the benchmark price of the target material, the change rate of the influencing factor and the corresponding weight, calculating a predicted price of the target material; the benchmark price is the latest market price of the target material.

2. The intelligent inquiry method of claim 1, wherein the step of extracting at least one material attribute item from the plurality of material attribute items as an influencing factor affecting the price of the target material comprises: performing feature processing on the plurality of material attribute items and corresponding attribute item values to obtain target feature data; calculating the average value between the price item values of the plurality of query results; calculating the price deviation between the price item value of each query result and the average value; clustering the target feature data and the price deviation to obtain a plurality of clustering clusters; each clustering cluster comprises target feature data corresponding to a plurality of query results matching the target material and a price deviation corresponding to each query result. For each cluster, the correlation coefficient of the target feature data in the cluster and all price deviation degrees in the cluster is calculated, and the material attribute item corresponding to the target feature data with an absolute value of the correlation coefficient greater than a preset correlation coefficient threshold is determined as an influencing factor affecting the target material price; The target feature data includes coded features and unstructured vector features; the feature processing process includes: Extracting the structured feature item value corresponding to the structured feature item in the plurality of material attribute items, and numerically encoding the structured feature item value to obtain coded features; Extracting the unstructured feature item value corresponding to the unstructured feature item in the plurality of material attribute items, and performing semantic vectorization processing on the unstructured feature item value to obtain an unstructured feature vector.

3. The intelligent price inquiry method of claim 1, wherein, The calculation based on the benchmark price of the target material, the change rate of the influencing factor, and the corresponding weight to obtain the predicted price of the target material includes: Calculating based on the first formula to obtain the predicted price of the target material; The first formula is: wherein, represents a predicted offer, represents a reference price, represents a rate of change of the ith influencing factor, represents a weight corresponding to the rate of change of the ith influencing factor.

4. The intelligent price inquiry method of claim 1, wherein, The corresponding matching between the structured feature item value in the historical ticket data and the structured feature value of each ticket data stored in the current target ticket database to obtain candidate ticket data includes: Calculating the confidence of the structured feature item value in the historical ticket data, and taking the structured feature item value with a confidence greater than a preset confidence as a target feature item value; the confidence is the recognition accuracy score of the structured feature item value in the feature extraction process of the historical electronic ticket; According to the corresponding matching between the target feature item value and the structured feature value of each ticket data stored in the current target ticket database, candidate ticket data is obtained.

5. The intelligent price inquiry method of claim 1, wherein, The target ticket database includes a plurality of historical matching ticket data; Further comprising: In response to the absence of ticket data with a matching degree greater than a first preset matching degree threshold in the candidate ticket data, matching the feature values corresponding to each feature item in the historical ticket data and at least one historical matching ticket data; If the matching number of feature values corresponding to each feature item in the historical ticket data and at least one historical matching ticket data is greater than a preset number, the historical ticket data is stored in the artificial matching library.

6. The intelligent price inquiry method of claim 5, wherein, Further comprising: Adjusting the first preset matching degree threshold; The adjustment process of the first preset matching degree threshold includes: Obtaining all historical matching records, calculating the matching success rate under the first preset matching degree threshold, Comparing the matching success rate with a preset matching success rate range; If the matching success rate is higher than the upper limit of the preset matching success rate range within a preset time period, the first preset matching degree threshold is reduced by a first adjustment step; If the matching success rate is lower than the lower limit of the preset matching success rate range within a preset time period, the first preset matching degree threshold is increased by a second adjustment step.

7. An intelligent price inquiry device, characterized by It includes: The query request receiving module is configured to receive a query request input by a user, wherein the query request includes a query condition of a target material, and the query condition includes an attribute item value corresponding to an attribute item to be queried. an inquiry module, configured to acquire a price item value of a target material matching the query condition from a target ticket database; the target material matching the query condition comprises a plurality of query results, each query result comprising a plurality of material attribute items and respective attribute item values; the plurality of material attribute items at least comprises a price item; the plurality of material attribute items comprises at least one structured feature item and at least one unstructured feature item; wherein the target ticket database is a database updated by the following manner: acquiring a historical electronic ticket to be updated every preset time; for each historical electronic ticket, performing the following operations: performing feature extraction on the historical electronic ticket to obtain historical ticket data corresponding to the historical electronic ticket, wherein the historical ticket data comprises respective feature values corresponding to each feature item; performing corresponding matching between the structured feature item value in the historical ticket data and the structured feature value of each ticket data stored in the current target ticket database to obtain candidate ticket data; performing matching between the unstructured feature item value in the historical ticket data and the unstructured feature value in the candidate ticket data to obtain a matching result; if there is a ticket data in the candidate ticket data with a matching degree greater than a first preset matching degree threshold, storing the historical ticket data into the target ticket database; a price prediction module, specifically configured to: acquire a plurality of material attribute items and respective attribute item values of a target material matching the query condition from a target ticket database based on the query condition; extract at least one material attribute item from the plurality of material attribute items as an influence factor affecting the price of the target material; for each influence factor, determine a change rate corresponding to the influence factor based on the attribute item value corresponding to the influence factor; determine a weight of the corresponding influence factor based on the change rate corresponding to the influence factor; calculate a predicted offer price of the target material based on a benchmark price of the target material, the change rate of the influence factor, and the corresponding weight; the benchmark price is the latest market price of the target material.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, the processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. the computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images