An intelligent order dispatching method and system based on multi-product combination price difference identification

Abstract

This invention discloses an intelligent order dispatching method and system based on multi-product combination price difference identification, belonging to the field of computer technology. The method includes: acquiring multi-product order data; generating a combination price difference representation vector based on a price difference identification rule base; acquiring a processing party capability profile; performing multi-dimensional matching degree calculation and dynamically adjusting weights; selecting the optimal processing party for order dispatch; and updating the rule base and profile based on feedback data. The system includes corresponding functional modules, supporting price difference identification, capability profile construction, dynamic matching, and online learning. This invention aims to solve the problems of unreasonable order allocation, low efficiency, and reduced revenue caused by the inability of existing order dispatching mechanisms to accurately identify multi-product combination price differences. Through value-driven intelligent matching, it significantly improves order dispatching efficiency, resource utilization, and operational revenue.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an intelligent order dispatching method and system based on multi-product combination price difference identification. Background Technology

[0002] With the rapid development of the air transport industry and the widespread adoption of online travel service platforms, airfare orders have evolved from single-segment products to composite orders that include basic tickets, baggage allowance, seat selection services, meals, insurance, and membership benefits. While these multi-product bundled orders enhance the personalized user experience, they also significantly increase the complexity of order processing and revenue management. Current mainstream order dispatch mechanisms are generally based on static rules or historical experience, and their core logic typically only considers superficial indicators such as total order volume, processor load, or unit price of a single product, lacking the ability to deeply analyze the internal structure and economic value of multi-product bundles.

[0003] Among these factors, identifying price differences in multi-product combinations is a crucial step affecting the quality of order dispatch decisions. Different combinations of ancillary services result in varying total prices and marginal revenues. However, existing systems, upon receiving an order, often directly assign the entire order to a pre-defined processing unit, failing to quantitatively analyze the synergistic effects, cost allocation, and potential premium space among the products within the combination. This coarse-grained processing model leads to high-value combinations being misclassified as ordinary orders, or inefficient processing units undertaking complex, high-return orders that should be handled by high-capacity units, resulting in revenue leakage and resource misallocation.

[0004] Existing technologies generally suffer from problems when dealing with multi-product combination orders, such as a lack of ability to identify price differences, rigid order dispatch strategies, and insufficient matching accuracy of processing parties. Specifically, these problems manifest as: the inability to dynamically analyze the split pricing capabilities of additional products, making it difficult to accurately calculate the net profit difference under different combination schemes; the order dispatch logic does not integrate multi-dimensional capability parameters such as the processing party's historical fulfillment efficiency, service costs, and customer satisfaction on specific product combinations; and in scenarios with a surge in order volume or frequent adjustments to product rules, the system lacks an adaptive optimization mechanism based on feedback data, causing the order dispatch results to deviate from the optimal solution. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent order dispatching method and system based on multi-product combination price difference identification, so as to solve the problems of unreasonable order dispatching, low processing efficiency and overall loss of operating revenue caused by the inability to accurately identify the price difference generated by multi-product combinations in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A smart order dispatching method based on multi-product combination price difference identification includes:

[0008] Step S1: Obtain the multi-product order data to be processed. The multi-product order data includes main product information, supplementary product information, product combination identifier, product unit price, product cost, user level, service timeliness requirements, and total order amount.

[0009] Step S2: Based on the preset product combination price difference identification rule library, extract the price difference features from the multi-product order data and generate the combination price difference representation vector of the order. The combination price difference representation vector includes the combination gross profit value, the unit time revenue density, the additional service premium coefficient, and the combination complexity score.

[0010] Step S3: Obtain real-time capability profile data for all available processors. The capability profile data includes historical processing success rate, average processing time, processing experience index for specific product combinations, current load rate, service cost coefficient, and regional service capability tags.

[0011] Step S4: Perform multi-dimensional matching degree calculation on the combined price difference representation vector and the capability profile data of each processing party to generate a comprehensive matching score for each processing party for the order. The comprehensive matching score is determined by a weighted linear combination formula, and its weight coefficients are dynamically adjusted through an online learning mechanism based on historical order feedback data.

[0012] Step S5: Based on the comprehensive matching score, select the processing party with the highest score as the target processing party, and generate a dispatch instruction to allocate the multi-product order to the target processing party;

[0013] Step S6: After the order is processed, collect the actual processing result data, including the actual processing time, actual cost, user satisfaction score and revenue achievement rate, and use the actual processing result data to update the product combination price difference identification rule base and the processing party capability profile data.

[0014] As one embodiment of the present invention, the construction process of the product combination price difference identification rule base in step S2 includes: extracting completed multi-product order records from the historical order database, parsing the product combination structure of each record, and identifying the combination pattern of the main product and the supplementary product; calculating the difference between the actual gross profit value and the benchmark gross profit value under each combination pattern, as the price difference benchmark for that combination pattern; establishing a combination price difference mapping function based on the price difference benchmark, combination complexity, and service timeliness requirements to form an initial rule base; and continuously integrating the price difference data of newly completed orders into the rule base through an incremental learning mechanism to achieve dynamic evolution of the rule base.

[0015] As one embodiment of the present invention, the process of generating the combined price difference representation vector in step S2 specifically includes: calculating the total gross profit of the order, i.e., the total order amount minus the sum of the costs of all products; calculating the revenue density per unit time, i.e., the total gross profit of the order divided by the estimated processing time, wherein the estimated processing time is determined based on the product combination type and the average processing time of similar orders in history; calculating the additional service premium coefficient, i.e., the proportion of additional service revenue to the total order revenue; and calculating the combination complexity score, which is determined by the number of product types, the number of service dependency layers, and the cross-regional service indicator, and is quantified using a preset complexity score matrix.

[0016] As one embodiment of the present invention, the process of constructing the processing capability profile data in step S3 includes: statistically analyzing the historical order processing records of each processing party to calculate its processing success rate and average processing time under various product combinations; generating regional service capability tags based on the processing party's region, personnel configuration, and system support capabilities; calculating the current load rate based on the ratio of the number of orders currently being processed to the maximum processing capacity; determining its service cost coefficient under different product combinations through a cost accounting module; integrating the above indicators into a structured capability profile vector, and updating it in real time through a sliding time window mechanism.

[0017] As one embodiment of the present invention, the specific steps of multi-dimensional matching degree calculation in step S4 include: normalizing the combined price difference representation vector and the processing party capability profile vector so that the values ​​of each dimension are on the same scale; setting an initial weight vector, including revenue weight, timeliness weight, cost weight and matching experience weight; calculating the weighted inner product to obtain the preliminary matching score; introducing a penalty term to deduct the score of processing parties whose current load rate exceeds the threshold; and finally, the matching score is obtained by subtracting the penalty term from the preliminary matching score.

[0018] As one embodiment of the present invention, the online learning mechanism in step S4 uses gradient descent to update the weight coefficients, and its loss function is defined as the mean square error between the actual revenue achievement rate and the predicted matching score; after each order is processed, if the actual revenue achievement rate is lower than the preset threshold, the weight of the corresponding dimension is adjusted in reverse to reduce the influence of that dimension in subsequent matching.

[0019] As one embodiment of the present invention, the dispatch instruction in step S5 includes a unique order identifier, a target processor identifier, product combination details, expected processing time limit, revenue target value, and service quality requirements; the dispatch instruction is asynchronously sent to the order receiving interface of the target processor through a message queue, and a dispatch log is recorded for subsequent auditing and optimization.

[0020] According to another aspect of the present invention, an intelligent order dispatch system based on multi-product combination price difference identification is provided, comprising:

[0021] The multi-product order data acquisition module is used to acquire multi-product order data to be processed. The multi-product order data includes main product information, supplementary product information, product combination identifier, product unit price, product cost, user level, service timeliness requirements, and total order amount.

[0022] The combined price difference representation generation module is used to extract price difference features from the multi-product order data based on a preset product combined price difference identification rule library, and generate a combined price difference representation vector for the order. The combined price difference representation vector includes the combined gross profit value, the unit time revenue density, the additional service premium coefficient, and the combined complexity score.

[0023] The processing provider capability profile acquisition module is used to acquire real-time capability profile data of all available processing providers. The capability profile data includes historical processing success rate, average processing time, processing experience index of specific product combination, current load rate, service cost coefficient, and regional service capability tag.

[0024] The multidimensional matching degree calculation module is used to perform multidimensional matching degree calculation on the combined price difference representation vector and the capability profile data of each processing party, and generate a comprehensive matching score for each processing party for the order. The comprehensive matching score is determined by a weighted linear combination formula, and its weight coefficients are dynamically adjusted through an online learning mechanism based on historical order dispatch feedback data.

[0025] The target processing party determination module is used to select the processing party with the highest score as the target processing party based on the comprehensive matching score, and generate a dispatch instruction to allocate the multi-product order to the target processing party;

[0026] The feedback data collection and update module is used to collect actual processing result data after the order is processed, including actual processing time, actual cost, user satisfaction score and revenue achievement rate, and use the actual processing result data to update the product combination price difference identification rule base and the processing party capability profile data.

[0027] As one embodiment of the present invention, the system further includes a product combination price difference identification rule base construction unit, which is used to extract completed multi-product order records from the historical order database, perform product combination structure parsing on each record, identify the combination pattern of main products and supplementary products; calculate the difference between the actual gross profit value and the benchmark gross profit value under each combination pattern, as the price difference benchmark for that combination pattern; establish a combination price difference mapping function based on the price difference benchmark, combination complexity and service timeliness requirements to form an initial rule base; and continuously integrate the price difference data of newly completed orders into the rule base through an incremental learning mechanism.

[0028] As one embodiment of the present invention, the system further includes a dynamic update unit for the processing provider's capability profile, which is used to perform statistical analysis on the historical order processing records of each processing provider, calculate its processing success rate and average processing time under various product combinations; generate regional service capability tags based on the processing provider's region, personnel configuration, and system support capabilities; calculate the current load rate based on the ratio of the number of orders currently being processed to the maximum processing capacity; determine its service cost coefficient under different product combinations through a cost accounting module; and integrate the above indicators into a structured capability profile vector, which is updated in real time through a sliding time window mechanism.

[0029] In one embodiment of the present invention, the multidimensional matching degree calculation module integrates a normalization processing submodule, a weight management submodule, a weighted inner product calculation submodule, and a load penalty submodule; the normalization processing submodule performs minimum-maximum scaling on the input vector; the weight management submodule stores and dynamically updates the weight vector; the weighted inner product calculation submodule performs vector dot product operation; and the load penalty submodule applies a fixed percentage deduction to the score of the overload processing party according to a preset load threshold.

[0030] As one embodiment of the present invention, the online learning mechanism is implemented by a weight update engine, which adopts a gradient descent algorithm and uses the deviation between the actual profit achievement rate and the predicted matching score as a loss signal to periodically adjust the weight vector to ensure that the matching model continuously approaches the optimal order dispatch strategy.

[0031] As one embodiment of the present invention, the order dispatch instruction generation unit encapsulates the unique order identifier, target processor identifier, product combination details, expected processing time limit, revenue target value and service quality requirements into a structured data packet, pushes it to the order receiving interface of the target processor through a reliable message transmission protocol, and simultaneously writes it into the distributed log system.

[0032] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0033] This invention constructs a multi-product combination price difference identification model, which can accurately quantify the revenue differences brought about by different product combinations, overcoming the limitations of traditional order dispatching systems that allocate based solely on a single price or processing volume; by introducing a processing party capability profile, it achieves multi-dimensional dynamic matching between order characteristics and processing party capabilities, avoiding revenue loss caused by high-value orders being handled by inefficient processing parties.

[0034] This invention employs an online learning mechanism, enabling the matching weights to be continuously optimized based on actual processing results. This ensures that the dispatch strategy always adapts to changes in the business environment. By combining complexity scoring with timeliness requirements, the system effectively balances the contradiction between maximizing revenue and ensuring service experience. This significantly improves customer satisfaction while enhancing overall operational efficiency.

[0035] The technical solution of this invention, without increasing additional hardware costs, achieves a fundamental shift in order allocation from "experience-driven" to "data-driven" and "value-driven" through intelligent algorithm reconstruction at the software level, demonstrating significant technological advancement and industrial application value. Attached Figure Description

[0036] Figure 1 This is a schematic diagram illustrating the core principle of an intelligent order dispatching method based on multi-product combination price difference identification in this invention.

[0037] Figure 2 This is a schematic diagram of the overall technical framework of an intelligent order dispatching system based on multi-product combination price difference identification proposed in this invention. Detailed Implementation

[0038] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present invention and not to limit the present invention. For those skilled in the art, the present invention can be practiced without some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present invention by illustrating examples of the invention.

[0039] Example 1

[0040] This invention provides an intelligent order dispatching method and system based on multi-product combination price difference identification, aiming to solve the problems of unreasonable order dispatching, low processing efficiency, and overall operational losses caused by the inability to accurately identify price differences arising from multi-product combinations in existing technologies. By constructing a multi-product combination price difference identification model, combined with processor capability profiling and a dynamic matching mechanism, it achieves accurate identification of high-value orders and optimal order dispatching path planning, thereby improving order processing efficiency, resource utilization efficiency, and overall operational profits.

[0041] In this embodiment, see Figure 1 , Figure 1 This is a schematic diagram illustrating the core principle of an intelligent order dispatching method based on multi-product combination price difference identification in this invention. The intelligent order dispatching method based on multi-product combination price difference identification includes the following steps:

[0042] Step S1: Obtain the multi-product order data to be processed. This data includes main product information, supplementary product information, product combination identifier, product unit price, product cost, user level, service timeliness requirements, and total order amount. In the scenario of processing multi-product airfare orders, a complete order typically consists of a main product (such as a basic ticket) and several supplementary products (such as baggage allowance, priority boarding, meal service, insurance, etc.). These products are stored in a structured format in the order database, with each product containing a unique identifier, price, cost, service attributes, and dependencies. Before entering the order dispatch process, the order data has undergone legality verification, payment status confirmation, and user identity binding through the front-end business system. Main product information includes at least the flight number, departure point, destination, cabin class, flight date, and time period; supplementary product information includes service type, quantity, applicable conditions, and whether binding is mandatory. The product combination identifier uniquely identifies the product packaging strategy used in this order; for example, "economy class + 20 kg baggage + priority boarding" corresponds to the combination identifier C-1024. Product unit price and product cost are recorded separately for the sales price and internal accounting cost of each product, used for subsequent gross profit calculation. User level reflects a user's historical consumption behavior, credit score, and membership level on the platform, divided into four tiers: Ordinary, Silver, Gold, and Platinum, affecting service priority and timeliness tolerance. Service timeliness requirements refer to users' explicit constraints on order processing completion time, such as "ticketing must be completed within two hours" or "confirmation must be made 24 hours before flight departure." The total order amount is the sum of the unit prices of all products and is the original basis for revenue calculation. The above data is retrieved in real time from the order management system through a standardized interface to ensure data integrity and timeliness.

[0043] Step S2: Based on a pre-defined product combination price difference identification rule library, price difference features are extracted from multi-product order data to generate a combination price difference representation vector for the order. This vector includes the combination gross profit value, unit time revenue density, additional service premium coefficient, and combination complexity score. It should be understood that traditional order dispatch logic only focuses on the total order amount or the price of the main product, ignoring the marginal revenue differences brought by additional services. This invention maps different product combination patterns to quantifiable revenue features by constructing a product combination price difference identification rule library. This rule library is stored in a central rule engine and organized in key-value pairs, where the key is the product combination identifier and the value is the corresponding price difference mapping function parameter. The price difference feature extraction process first parses the product combination structure in the order, identifying the association topology between the main product and additional products; then, it calls the mapping function corresponding to the combination identifier in the rule library, inputting the product unit price, cost, and service attributes, and outputting a four-dimensional representation vector. The combination gross profit value equals the total order amount minus the sum of all product costs, directly reflecting the net revenue potential of the order. The unit time revenue density is defined as the gross profit of the portfolio divided by the estimated processing time, where the estimated processing time is weighted by the product portfolio type and the average processing time of similar historical orders. For example, the estimated processing time for orders including cross-airline connecting services is 1.8 times the benchmark value. The ancillary service premium coefficient is calculated as the proportion of ancillary service revenue to total order revenue, used to measure the contribution of non-core products to overall revenue. The portfolio complexity score is determined by the number of product types, the number of service dependency layers, and cross-regional service indicators, and is quantified using a preset complexity scoring matrix: 1 point is added for each additional product type, 0.5 points are added for each layer of service dependency (such as baggage allowance depending on cabin class), and 2 points are added for cross-regional services (such as international segment + domestic segment), with a total maximum score of 10 points. The above four dimensions together constitute the portfolio price difference representation vector, fully characterizing the value density and processing difficulty of the order.

[0044] Step S3: Obtain real-time capability profile data for all available processors. This capability profile data includes historical processing success rate, average processing time, experience index for specific product combinations, current load rate, service cost coefficient, and regional service capability tags. It should be understood that a processor refers to an operational unit with order processing qualifications, which can be a human agent team, an automated robot process, or a hybrid processing node. Each processor is registered with a unique identifier in the system, and its capability profile is maintained. The historical processing success rate is calculated as the ratio of the number of orders successfully completed by the processor in the past thirty days to the total number of orders received, stored separately by product combination type. The average processing time records the average time taken from receiving to completing similar orders, using a sliding window average after removing outliers. The experience index for specific product combinations is calculated using exponential smoothing, with an initial value of zero. For each successful processing of a certain type of combination order, the experience index for that combination increases by 0.1, with a maximum value of 5.0. The current load rate equals the number of orders currently being processed divided by the processor's maximum concurrent processing capacity, which is set by the system administrator based on resource configuration. The service cost coefficient reflects the internal cost of processing a unit of order by the processor, including expenses such as manpower, system calls, and communication, and is calculated differently according to product combination type. The regional service capability tag describes the geographical range and language capabilities supported by the processor, such as "East China - Chinese - English" or "Global - Multilingual". The above indicators are collected in real time by the processor status monitoring module and cached in the in-memory database in the form of structured vectors to ensure the freshness of data during matching calculation.

[0045] Step S4 involves calculating the multi-dimensional matching degree between the combined price difference representation vector and the capability profile data of each processing party, generating a comprehensive matching score for each processing strategy for the order. The comprehensive matching score is determined by a weighted linear combination formula, and its weight coefficients are dynamically adjusted based on historical order feedback data through an online learning mechanism. It should be understood that the matching degree calculation is the core decision-making step of this invention.

[0046] First, the combined price difference representation vector and the processing capability profile vector are normalized by using a minimum-maximum scaling method to map the values ​​of each dimension to the 0-1 range, eliminating dimensional differences. Then, an initial weight vector is set. These correspond to the revenue weight, timeliness weight, experience weight, load weight, cost weight, and regional matching weight, respectively, with an initial value of 1 / 6 for each. The weighted inner product calculation formula is as follows:

[0047]

[0048] in, to The normalized combined gross profit value, revenue density per unit time, additional service premium coefficient, and combined complexity score are used. to The parameters are: normalized historical processing success rate, reciprocal of average processing time, processing experience index for specific product combinations, 1 minus current load rate, 1 minus service cost coefficient, and regional service capability matching degree. A perfect match is 1, a partial match is 0.5, and a non-match is 0. Preliminary matching score. After correction by the load penalty submodule: if the current load rate of the processor exceeds a preset threshold (e.g., 0.85), the final matching score will be adjusted. ,in, The penalty percentage is fixed at 0.15. This mechanism prevents high-value orders from being assigned to overloaded processing providers, thus ensuring service quality.

[0049] Step S5: Based on the comprehensive matching score, the processor with the highest score is selected as the target processor, and a dispatch instruction is generated to allocate multi-product orders to the target processor. It can be understood that the system iterates through the comprehensive matching scores of all available processors and selects the processor with the highest score as the target. If multiple processors have the same score, they are selected according to their registration priority. The dispatch instruction is encapsulated as a structured data packet, containing a unique order identifier, target processor identifier, product combination details, expected processing time limit, revenue target value, and service quality requirements. The expected processing time limit is derived from the revenue density per unit time, the revenue target value is the combined gross profit value, and the service quality requirements are generated based on user level and service timeliness requirements. The dispatch instruction is asynchronously pushed to the target processor's order receiving interface via a reliable message queue. The interface adopts an idempotent design to ensure that duplicate delivery does not produce side effects. Simultaneously, the dispatch log is written to a distributed log system, recording the timestamp, order ID, target processor, matching score, and various feature values ​​for subsequent auditing and model optimization.

[0050] Step S6: After order processing is completed, collect actual processing result data, including actual processing time, actual cost, user satisfaction score, and revenue achievement rate. This data is then used to update the product mix price difference identification rule base and the processing provider's capability profile data. It should be understood that the actual processing result data is returned by the processing provider when the order is closed. Actual processing time is calculated from the order dispatch time to the completion confirmation time; actual cost is reported by the processing provider's cost accounting module; user satisfaction score comes from the user feedback interface and ranges from 1 to 5; revenue achievement rate equals the ratio of actual gross profit to predicted mix gross profit.

[0051] These data trigger two update processes: first, updating the processing capability profile by replacing old records with a sliding time window and recalculating historical processing success rates, average processing times, and experience indices; second, updating the product portfolio price difference identification rule base by using the deviation between actual and predicted gross profit values ​​as incremental learning signals to adjust the portfolio complexity scoring weights and estimated processing time coefficients. In addition, an online learning mechanism operates concurrently: if the profit achievement rate is lower than a preset threshold (e.g., 0.9), the weight update engine is activated, using gradient descent to adjust the weight vector, with the loss function defined as:

[0052]

[0053] in, The normalized actual profit achievement rate is equal to the ratio of the actual gross profit to the predicted combined gross profit. This is the final match score.

[0054] The weight update formula is:

[0055]

[0056] Where η is the learning rate, taking a value of 0.01, t is the iteration time step, representing the number of weight updates or the time interval, and L is the loss function. This represents the value of the i-th weight at time t. This mechanism ensures that the matching model continuously approximates the optimal policy.

[0057] In this embodiment, the construction process of the product combination price difference identification rule base includes: extracting completed multi-product order records from the historical order database; parsing the product combination structure of each record to identify the combination pattern of the main product and the supplementary product; calculating the difference between the actual gross profit value and the benchmark gross profit value under each combination pattern as the price difference benchmark for that combination pattern; establishing a combination price difference mapping function based on the price difference benchmark, combination complexity, and service timeliness requirements to form an initial rule base; and continuously integrating the price difference data of newly completed orders into the rule base through an incremental learning mechanism to achieve dynamic evolution of the rule base. The benchmark gross profit value is defined as the sum of the standard gross profit of the main product and the standard gross profit of the supplementary product, and the standard gross profit is periodically published by the financial system. The price difference benchmark reflects the excess revenue or cost spillover brought about by the combination synergy effect.

[0058] In this embodiment, the process of constructing the capability profile data of the processing party includes: statistically analyzing the historical order processing records of each processing party to calculate its processing success rate and average processing time under various product combinations; generating regional service capability tags based on the processing party's region, personnel configuration, and system support capabilities; calculating the current load rate based on the ratio of the number of orders currently being processed to the maximum processing capacity; determining its service cost coefficient under different product combinations through a cost accounting module; integrating the above indicators into a structured capability profile vector, and updating it in real time through a sliding time window mechanism. The sliding window length is set to seven days to ensure that the profile reflects the recent capability status.

[0059] In this embodiment, the online learning mechanism uses gradient descent to update the weight coefficients. Its loss function is defined as the mean squared error between the actual revenue achievement rate and the predicted matching score. After each order is processed, if the actual revenue achievement rate is lower than a preset threshold, the weights of the corresponding dimensions are adjusted in reverse to reduce the impact of that dimension in subsequent matching. The updated weight vector needs to be normalized to ensure that the sum of all weights is 1.

[0060] In this embodiment, the dispatch instruction includes a unique order identifier, a target processor identifier, product combination details, expected processing time limit, revenue target value, and service quality requirements. The dispatch instruction is asynchronously sent to the target processor's order receiving interface via a message queue, and a dispatch log is recorded for subsequent auditing and optimization. The message queue uses persistent storage to ensure that instructions are not lost in the event of system failure.

[0061] Corresponding to the above method, the present invention also provides an intelligent order dispatch system based on multi-product combination price difference identification. See also Figure 2 , Figure 2 This is a schematic diagram of the overall technical framework of an intelligent order dispatching system based on multi-product combination price difference identification proposed in this invention. The system includes a multi-product order data acquisition module, a combination price difference characterization generation module, a processor capability profile acquisition module, a multi-dimensional matching degree calculation module, a target processor determination module, and a feedback data collection and update module.

[0062] The multi-product order data acquisition module connects to the data interface of the order management system, periodically polls or listens for new order events, extracts multi-product order data that meets the order dispatch conditions, and performs field validation and format conversion to ensure data integrity.

[0063] The module for generating composite price difference representations has a built-in rule engine that loads the rule library for recognizing composite price differences. After receiving order data, it parses the composite structure, calls the corresponding mapping function, calculates the composite gross profit value, revenue density per unit time, premium coefficient for additional services, and composite complexity score, and outputs a four-dimensional representation vector.

[0064] The module for acquiring the capability profile of the processor maintains the processor registration center, collects status indicators of each processor in real time, constructs and caches capability profile vectors, and supports fast retrieval by product combination type.

[0065] The multidimensional matching degree calculation module integrates a normalization processing submodule, a weight management submodule, a weighted inner product calculation submodule, and a load penalty submodule. The normalization processing submodule performs min-max scaling on the input vector; the weight management submodule stores the current weight vector and receives update instructions from the online learning engine; the weighted inner product calculation submodule performs dot product operations; and the load penalty submodule applies score deductions based on the load rate threshold.

[0066] The target processor determination module iterates through the matching scores of all processors, selects the one with the highest score, generates a structured dispatch instruction, and sends it to the target processor through a message middleware.

[0067] The feedback data collection and update module listens for order completion events, extracts actual processing result data, triggers the rule base update process and the capability profile update process respectively, and drives the online learning engine to adjust weights.

[0068] Through the collaborative work of the above modules, the system achieves closed-loop control from order access to intelligent order dispatch and model self-optimization, significantly improving the rationality of multi-product order dispatch and operational efficiency.

[0069] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An intelligent order dispatching method based on multi-product combination price difference identification, characterized in that, include: Step S1: Obtain the multi-product order data to be processed. The multi-product order data includes main product information, supplementary product information, product combination identifier, product unit price, product cost, user level, service timeliness requirements, and total order amount. Step S2: Based on the preset product combination price difference identification rule library, extract the price difference features from the multi-product order data and generate the combination price difference representation vector of the order. The combination price difference representation vector includes the combination gross profit value, the unit time revenue density, the additional service premium coefficient, and the combination complexity score. Step S3: Obtain real-time capability profile data for all available processors. The capability profile data includes historical processing success rate, average processing time, processing experience index for specific product combinations, current load rate, service cost coefficient, and regional service capability tags. Step S4: Perform multi-dimensional matching degree calculation on the combined price difference representation vector and the capability profile data of each processing party to generate a comprehensive matching score for each processing party for the order. The comprehensive matching score is determined by a weighted linear combination formula, and its weight coefficients are dynamically adjusted through an online learning mechanism based on historical order feedback data. Step S5: Based on the comprehensive matching score, select the processing party with the highest score as the target processing party, and generate a dispatch instruction to allocate the multi-product order to the target processing party; Step S6: After the order is processed, collect the actual processing result data, including the actual processing time, actual cost, user satisfaction score and revenue achievement rate, and use the actual processing result data to update the product combination price difference identification rule base and the processing party capability profile data.

2. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 1, characterized in that, Step S2, based on a preset product combination price difference identification rule base, extracts price difference features from the multi-product order data to generate a combination price difference representation vector for the order, including: Calculate the total gross profit of the order, which is the total order amount minus the sum of the costs of all products; Calculate the revenue density per unit time, which is the total gross profit of the order divided by the estimated processing time, the estimated processing time being determined based on the product mix type and the average processing time of similar orders in history; Calculate the value-added service premium factor, which is the proportion of value-added service revenue to total order revenue; Calculate the combinatorial complexity score, which is determined by the number of product types, the number of service dependency layers, and cross-regional service indicators, and is quantified using a preset complexity score matrix.

3. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 2, characterized in that, Step S3, obtaining real-time capability profile data of all available processors, includes: Statistical analysis is performed on the historical order processing records of each processor to calculate its success rate and average processing time under various product combinations. Based on the region, personnel configuration, and system support capabilities of the processor, generate regional service capability tags; Calculate the current load factor based on the ratio of the number of orders currently being processed to the maximum processing capacity; The service cost coefficients for different product combinations are determined through the cost accounting module. The above indicators are integrated into a structured capability profile vector, and updated in real time through a sliding time window mechanism.

4. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 3, characterized in that, The combined price difference representation vector from step S4 is used to perform a multi-dimensional matching degree calculation with the capability profile data of each processing party to generate a comprehensive matching score for each processing strategy for the order, including: The combined price difference representation vector and the processing capability profile vector are normalized so that the values ​​of each dimension are on the same scale. Set an initial weight vector, including revenue weight, timeliness weight, cost weight, and matching experience weight; Calculate the weighted inner product to obtain the preliminary matching score; Introduce a penalty clause to deduct points from the score of the processor whose current load rate exceeds the threshold; The final match score is obtained by subtracting the penalty from the initial match score.

5. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 4, characterized in that, The online learning mechanism uses gradient descent to update the weight coefficients, and its loss function is defined as the mean square error between the actual revenue achievement rate and the predicted matching score. After each order is processed, if the actual revenue achievement rate is lower than the preset threshold, the weight of the corresponding dimension is adjusted in reverse to reduce the influence of that dimension in subsequent matching.

6. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 5, characterized in that, The dispatch instruction includes a unique order identifier, a target processor identifier, product combination details, expected processing time limit, revenue target value, and service quality requirements. The dispatch instruction is sent asynchronously to the target processor's order receiving interface via a message queue, and dispatch logs are recorded for subsequent auditing and optimization.

7. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 6, characterized in that, The process of constructing the product combination price difference identification rule base includes: Extract completed multi-product order records from the historical order database, analyze the product combination structure of each record, and identify the combination pattern of main products and supplementary products; Calculate the difference between the actual gross profit value and the benchmark gross profit value for each combination model, and use it as the price difference benchmark for that combination model; Based on the price difference benchmark, combination complexity, and service timeliness requirements, a combination price difference mapping function is established to form an initial rule base. Through an incremental learning mechanism, the price difference data of newly completed orders is continuously integrated into the rule base, enabling the dynamic evolution of the rule base.

8. The intelligent order dispatching method based on multi-product combination price difference identification according to claim 7, characterized in that, In the calculation of the combinatorial complexity score, 1 point is added for each additional product type, 0.5 points are added for each layer of service dependencies, and 2 points are added for cross-regional services, with a total maximum score of 10 points.

9. An intelligent order dispatch system based on multi-product combination price difference identification, characterized in that, include: The multi-product order data acquisition module is used to acquire multi-product order data to be processed. The multi-product order data includes main product information, supplementary product information, product combination identifier, product unit price, product cost, user level, service timeliness requirements, and total order amount. The combined price difference representation generation module is used to extract price difference features from the multi-product order data based on a preset product combined price difference identification rule library, and generate a combined price difference representation vector for the order. The combined price difference representation vector includes the combined gross profit value, the unit time revenue density, the additional service premium coefficient, and the combined complexity score. The processing provider capability profile acquisition module is used to acquire real-time capability profile data of all available processing providers. The capability profile data includes historical processing success rate, average processing time, processing experience index of specific product combination, current load rate, service cost coefficient, and regional service capability tag. The multidimensional matching degree calculation module is used to perform multidimensional matching degree calculation on the combined price difference representation vector and the capability profile data of each processing party, and generate a comprehensive matching score for each processing party for the order. The comprehensive matching score is determined by a weighted linear combination formula, and its weight coefficients are dynamically adjusted through an online learning mechanism based on historical order dispatch feedback data. The target processing party determination module is used to select the processing party with the highest score as the target processing party based on the comprehensive matching score, and generate a dispatch instruction to allocate the multi-product order to the target processing party; The feedback data collection and update module is used to collect actual processing result data after the order is processed, including actual processing time, actual cost, user satisfaction score and revenue achievement rate, and use the actual processing result data to update the product combination price difference identification rule base and the processing party capability profile data.

10. The intelligent order dispatching system based on multi-product combination price difference identification according to claim 9, characterized in that, The multidimensional matching degree calculation module integrates a normalization processing submodule, a weight management submodule, a weighted inner product calculation submodule, and a load penalty submodule; the normalization processing submodule performs minimum-maximum scaling on the input vector; The weight management submodule stores and dynamically updates the weight vector; the weighted inner product calculation submodule performs vector dot product operations; and the load penalty submodule applies a fixed percentage deduction to the score of the overload processing party based on a preset load threshold.

Images