Resource information processing method and device, storage medium and program product

Abstract

The embodiment of the invention provides a resource information processing method and device, a storage medium and a program product. After a terminal request is received, a first target object triggered by a preset behavior and a second target object of which the historical interest is higher than a threshold value are included in a conversion range, and a target resource mapping matrix is constructed in real time in combination with resource information updated by a platform in real time and a use rule; converting the matrix into a target resource allocation matrix capable of representing an optimal allocation relationship between resource information and objects by taking maximization of the quantity of resources allocated to all target objects as a target, and completing global assignment at one time; and generating resource description information according to the target resource allocation matrix and displaying the information back to the terminal to guide the user to complete the target operation. According to the scheme, normal form upgrading from static superposition to dynamic matrix global optimal allocation is realized, continuous, accurate and maximized resource saving experience is provided for a user in a scene that multiple target objects, resource information and use rule information thereof change in real time, and the platform object conversion rate is remarkably improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a resource information processing method, device, storage medium, and program product. Background Technology

[0002] With the rapid development of e-commerce, users' exchange behavior has changed significantly, and online exchange has become the mainstream method. The main reasons users choose online exchange can be summarized in two aspects: the convenience of the exchange process and the advantage in resource availability. The advantage in resource availability is mainly reflected in the various resource information released regularly or irregularly by the exchange platform or resource provider. After obtaining the relevant resource information, users can use it when exchanging for a target object, using the resource amount corresponding to the resource information to offset a portion of the target object's total resource amount, thus completing the exchange with fewer actual resources.

[0003] In typical network replacement scenarios, users often add multiple objects to a pre-selection set and submit a combined replacement request. To meet this need, e-commerce platforms have established corresponding resource allocation schemes. Traditional resource allocation schemes mainly rely on static rules, such as allocating resources based on object category or the total resource amount of the replacement request. However, such methods have significant limitations: static rules struggle to effectively capture the synergistic effects of combinations between objects, resulting in insufficient or limited resource savings.

[0004] To address this, a machine learning-based resource information processing method was introduced. While this method can achieve personalized resource information allocation to some extent, when processing multi-object combination replacement requests, the generated resource information allocation results often deviate from the user's actual needs, making it difficult to achieve optimal matching results and resulting in insufficient resource savings. This issue not only affects the user's replacement experience but also negatively impacts the platform's object conversion rate. Summary of the Invention

[0005] The embodiments of this application provide a resource information processing method, device, storage medium, and program product to upgrade the paradigm from static superposition to dynamic matrix global optimal allocation. In scenarios where multiple target objects, resource information, and their usage rules change in real time, it provides users with a continuous, accurate, and maximized resource saving experience, significantly improving the platform object conversion rate.

[0006] This application provides a resource information processing method, including: receiving an information processing request sent by a user terminal, wherein the information processing request is sent by the user terminal when a preset behavior is detected during browsing a platform page, and the information processing request includes attribute information of multiple target objects, including a first target object associated with the preset behavior, and a second target object whose user interest level is greater than a set threshold based on the user's historical behavior; invoking a target resource mapping model, and constructing a target resource mapping matrix between multiple target objects and multiple resource information based on the attribute information of multiple target objects, the latest resource information provided by the platform, and their usage rules, wherein multiple row elements in the target resource mapping matrix represent multiple target objects, and multiple column elements represent multiple target resource information. Target resource information refers to resource information that can be used by at least one target object, determined based on usage rule information. With the goal of maximizing the amount of resources allocated to multiple target objects, resource allocation is performed on multiple target objects based on a target resource mapping matrix to obtain a target resource allocation matrix. In this matrix, row and column elements intersect to form matrix elements, which represent the matching degree between the target resource information corresponding to their column element and the target object corresponding to their row element. Based on the resource allocation matrix, resource description information is generated for multiple target objects, describing the resource allocation status of each target object. This resource description information is then returned to the user terminal, allowing the user terminal to display it on the platform page, guiding the user to perform the target operation.

[0007] This application also provides an electronic device, including: a memory and a processor; the memory for storing a computer program; and the processor, coupled to the memory, for executing the computer program to implement the steps in the above method.

[0008] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps in the above-described method.

[0009] This application also provides a computer program product, which includes a computer program / instructions that, when executed by a processor, enable the processor to perform the steps in the above-described method.

[0010] In this embodiment, upon receiving a user terminal request, the first target object associated with the user's preset behavior during browsing the platform page and the second target object defined based on historical interest are both included in the object conversion scope. A target resource mapping matrix is ​​constructed by combining various resource information and usage rules updated in real time by the platform. Subsequently, with the goal of maximizing the amount of resources allocated to multiple target objects, the target resource mapping matrix is ​​quickly converted into a target resource allocation matrix that represents the allocation relationship of various resource information to each target object. This completes the optimal assignment of various resource information to multiple target objects. Furthermore, a description of the resource allocation status of multiple target objects can be generated based on the target resource allocation matrix and displayed on the user terminal to guide the user to perform the target operation. This process overcomes the bottleneck of optimal resource saving for local target objects caused by sequential matching of resource information, realizing a transformation from "sequential static superposition" to "global optimal matrix allocation." It can quickly complete global optimization, significantly shorten system response latency, and provide users with a shorter response time and a more continuous user experience, even when the various resource information and usage rules provided by the platform change in real time, thereby improving the platform's object conversion efficiency. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1a A flowchart illustrating a resource information processing method provided for an exemplary embodiment of this application; Figure 1b A flowchart illustrating another resource information processing method provided as another exemplary embodiment of this application; Figure 1c A flowchart illustrating yet another exemplary embodiment of the resource information processing method provided in this application; Figure 1d A flowchart illustrating yet another exemplary embodiment of the resource information processing method provided in this application; Figure 2 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of this application. Detailed Implementation

[0012] Following the background technology and addressing the technical problems existing in the background technology, in this embodiment of the application, after receiving a user terminal request, the first target object associated with the user's preset behavior during browsing the platform page and the second target object defined based on historical interest can be included in the object conversion scope. Combined with various resource information and their usage rules updated in real time by the platform, a target resource mapping matrix is ​​constructed. Subsequently, with the goal of maximizing the amount of resources allocated to multiple target objects, the target resource mapping matrix is ​​quickly converted into a target resource allocation matrix that can represent the allocation relationship of various resource information to each target object. This completes the optimal assignment of various resource information to multiple target objects. Furthermore, a description of the resource allocation status of multiple target objects can be generated based on the target resource allocation matrix and displayed on the user terminal to guide the user to perform the target operation. The above process overcomes the bottleneck of optimal local target object resource saving caused by the sequential matching of resource information, realizing the transformation from "sequential static superposition" to "global optimal matrix allocation." It can quickly complete global optimization, significantly shorten system response latency, and provide users with a shorter response time and a more continuous user experience when the various resource information and their usage rules provided by the platform change in real time, thereby improving the platform's object conversion efficiency.

[0013] It should be noted that, in the cases involving user information in the embodiments of this application, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) comply with relevant laws and standards.

[0014] Additionally, it should be noted that when user interaction operations or triggering operations are involved in the embodiments of this application, these operations include, but are not limited to, various interaction methods such as touch operations, gesture operations, voice operations, head movement operations, and eye movement operations. Touch operations include, but are not limited to, click operations, double-click operations, long-press operations, swipe operations, pinch operations, or mouse hover operations. Swipe operations include, but are not limited to, straight-line swipes and curved-line swipes.

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] Figure 1a This is a flowchart illustrating a resource information processing method provided for an exemplary embodiment of this application. Figure 1a As shown, the method includes: 101. Receive information processing requests sent by user terminals. The information processing requests are sent by users of user terminals when they detect preset behaviors while browsing platform pages. The information processing requests include attribute information of multiple target objects. The multiple target objects include a first target object associated with the preset behavior and a second target object whose user interest is greater than a set threshold based on the user's historical behavior. 102. Call the target resource mapping model. Based on the attribute information of multiple target objects, the latest resource information provided by the platform and their usage rules, construct a target resource mapping matrix between multiple target objects and multiple resource information. The multiple row elements in the target resource mapping matrix represent multiple target objects, and the multiple column elements represent multiple target resource information. The target resource information is the resource information that can be used by at least one target object, as determined by the usage rules. 103. With the goal of maximizing the amount of resources allocated to multiple target objects, resource allocation is performed on multiple target objects based on the target resource mapping matrix to obtain the target resource allocation matrix. In the target resource allocation matrix, the row elements and column elements intersect to form matrix elements. The matrix elements represent the matching degree between the target resource information corresponding to the column element and the target object corresponding to the row element. 104. Based on the target resource allocation matrix, generate resource description information corresponding to multiple target objects. The resource description information is used to describe the resource allocation of multiple target objects. 105. Return the resource description information to the user terminal so that the user terminal can display the resource description information on the platform page. The resource description information is used to guide the user to perform the target operation.

[0017] In this embodiment, each platform provides multiple object categories, each containing multiple objects. An object category refers to a set of categories divided based on objects sharing common attributes or functional characteristics; such categories might include housekeeping services, clothing, food, furniture, or electronic products. An object is a replaceable entity; for example, an object could be a housekeeping service, a piece of clothing, a home furnishing product, or an electronic product. When a user has a replacement need for a target object, they can replace it from the target platform. If there are multiple target objects, the user can simultaneously combine and replace multiple target objects. After the replacement is completed, all target objects belong to the user.

[0018] In this embodiment, different objects correspond to the same or different amounts of resources, and the amount of resources can represent the value of each object. Typically, to improve the conversion rate of platform objects, platforms provide various types of resource information or full-category resource information. This resource information is used to adjust the amount of resources for each object. Resource information refers to the electronic carrier used to adjust the amount of resources for an object. Different types of resource information correspond to their own resource adjustment ranges; for example, one type of resource information corresponds to a resource adjustment range of 20 resources. Each type of resource information also has corresponding usage rules. For example, the usage rule for a certain type of resource information is: 20 resources minus 20 resources for every 200 resources purchased, and the scope of application is limited to the food category. Thus, the use of resource information can reduce the amount of resources a user needs to exchange for a target object, thereby improving the conversion rate of the target platform object.

[0019] In this embodiment, during the user's browsing of the target platform's page displayed on the user terminal, a series of user behaviors on the platform page are monitored. These behaviors can be preset behaviors, and the user terminal can send an information processing request to the target platform upon detecting the user's preset behaviors, so that the target platform can receive the information processing request. Referring to Figure 1, in step 101, the information processing request sent by the user terminal is received. The information processing request is sent by the user terminal when the preset behaviors are detected during the browsing of the platform page. The information processing request includes attribute information of multiple target objects. The multiple target objects include a first target object associated with the preset behaviors and a second target object whose user interest is greater than a set threshold based on the user's historical behaviors.

[0020] In this embodiment, the user's browsing of the platform page involves multiple replacement stages, including but not limited to: browsing stage, pre-selection stage, submission stage, and confirmation stage. The browsing stage refers to the process where the user enters the full category page, target category page, or target object details page to browse, but does not perform any active operation (such as clicking, adding to favorites, or adding to the pre-selection set). This stage can record the browsing trajectory and the duration of each dwell time (such as the browsing time for a specific object). The pre-selection stage refers to the process from when the user performs any active action such as adding the target object to the pre-selection set, comparing multiple times, or enabling resource reduction reminders, until the final submission operation is triggered. The submission stage refers to the process of issuing a next-step combination verification confirmation instruction for multiple target objects selected in the pre-selection set, but before the final resource verification is completed. The confirmation phase refers to the process from when a user triggers an object combination replacement confirmation event (e.g., clicking the "Confirm" button for the first time) until the final replacement operation is completed. It's important to note that the completion of the object combination replacement confirmation event indicates successful replacement of the target object; conversely, its failure to do so indicates unsuccessful replacement. In other words, the confirmation phase involves the user using appropriate resources to replace the target object, but simply executing the confirmation event does not equate to replacement completion; replacement may still fail. Successful replacement of the target object is only indicated after the object combination replacement confirmation event is completed. The pre-selection set refers to the set of candidate objects created by the user through actions such as clicking and saving before formally submitting a combination replacement request for multiple target objects. The target objects in this pre-selection set are those for which the user has potential replacement intentions and are used for subsequent matching calculations.

[0021] In this embodiment, the preset behavior can involve various stages of the above-mentioned replacement process. Based on this, the preset behavior includes, but is not limited to: the behavior of browsing the target object for more than a preset browsing time, the behavior of adding the target object to a pre-selection set, the behavior of submitting a combination of multiple target objects, the behavior of submitting multiple combinations of multiple target objects multiple times but not confirming any of them, the behavior of confirming the combination request submitted for the combination of target objects (or executing the object combination replacement confirmation event), the object combination replacement confirmation event not being completed or the object combination replacement confirmation event being completed. Based on this, the multiple target objects included in the information processing request in step 101 above are the first target objects triggered by the user through any of the above-mentioned preset behaviors during the browsing of the platform page. The historical behavior for determining the second target object can be the preset behavior before the currently triggered preset behavior during the current browsing of the platform page, or the historical behavior can be the historical behavior triggered before the current browsing of the platform page. This embodiment does not limit this, as long as the user's interest is greater than a set threshold.

[0022] In addition, the target object's attribute information is a collection of static and dynamic attribute information describing the target object itself and its current environmental state. Static attribute information includes, but is not limited to, object category, specifications, base resource quantity, whitelist of applicable resource information, and inventory limit. Dynamic attribute information includes, but is not limited to, current inventory, platform incentive activity identifier, and resource quantity fluctuation range within a specified time period. Platform incentive activities refer to the platform increasing the resource quantity discount for certain object categories (e.g., changing "200 resources minus 20 resources" to "200 resources minus 30 resources"), issuing more resource information (i.e., increasing discount quotas), or lowering the usage threshold of resource information to encourage users to prioritize objects that can use these resource information (which can be considered as adjusting the usage rules of resource information). Environmental attribute information includes, but is not limited to: the real-time access popularity of the page where the object is located, and its ranking compared with similar objects in terms of resource quantity (i.e., the object's value). The target platform generates corresponding resource information for each type of resource information's usage rules. This resource information is dynamically distributed by the target platform and serves to carry and execute all usage rules and parameter sets that modify the resource quantity of the target object. The usage rules and parameter sets include, but are not limited to: resource information identifiers, trigger conditions, adjustment parameters, and status fields. Specifically, resource information identifiers include, but are not limited to: resource information ID, effective period, and applicable object category range; trigger conditions include, but are not limited to: quantity tiers, membership levels, and geographical restrictions; quantity tiers refer to a pre-set sequence of "object quantity" or "cumulative resource quantity" tiers on the target platform. For each tier that the actual quantity of combined objects and the total resource quantity crosses, the target platform will increase the corresponding resource quantity reduction by one level. For example, if the quantity of combined objects is greater than 2, the total resource quantity will be reduced by 5 resources; if the quantity of combined objects is greater than 4, the total resource quantity will be reduced by 10 resources; and if the quantity of combined objects is greater than 6, the total resource quantity will be reduced by 15 resources. Adjustment parameters include, but are not limited to: resource quantity reduction magnitude, stackable markers (i.e., stackable usage), and the maximum upper limit for resource quantity savings. Status fields include, but are not limited to: the stock of each resource information, the quantity issued / received, and the real-time receipt rate.

[0023] In this embodiment, the target platform provides various resource information and their usage rules. This information is not static and can be dynamically adjusted or updated based on the specific circumstances of the platform. Therefore, upon receiving an information processing request, the platform can obtain the latest usage rules for various resource information based on the request. Referring to step 102 in Figure 1, the target mapping model is invoked. Based on the attribute information of multiple target objects, the various resource information, and their usage rules, a target resource mapping matrix is ​​constructed between multiple target objects and various resource information. The multiple rows in the target resource mapping matrix represent multiple target objects, and the multiple columns represent multiple target resource information. Target resource information is resource information that can be used by at least one target object, determined by the usage rules. The target resource mapping matrix is ​​constructed in real-time based on the dynamic multiple target objects and the latest dynamic resource information and their usage rules. This ensures dynamic real-time matching between the dynamically changing multiple target objects and the latest resource information and their usage rules provided by the platform, laying the foundation for maximizing subsequent resource savings.

[0024] Furthermore, after constructing the target resource mapping model, referring to step 103 in Figure 1, with the goal of maximizing the amount of resources allocated to multiple target objects, resource allocation processing is performed on multiple target objects based on the target resource mapping matrix to obtain the target resource allocation matrix. The target resource allocation matrix is ​​a structured data representation that shows the resource information and quantity allocated to different target objects. For example, in a platform incentive activity, this target resource allocation matrix defines the resource information of different amounts of resources allocated to user A, user B, etc.

[0025] In this embodiment, the row and column elements of the target resource allocation matrix intersect to form matrix elements, and each matrix element represents the matching degree between the target resource information corresponding to its column element and the target object corresponding to its row element. Furthermore, multiple target objects included in the target resource mapping matrix can simultaneously share the same target resource information, or each can independently use a separate target resource information. Accordingly, the matching degree has two meanings: if multiple target objects can share the same target resource information, the higher the matching degree, the greater the proportion of resources allocated to that target resource information for the corresponding target object; if multiple target objects each use a separate target resource information, the higher the matching degree, the higher the priority for each target object to obtain that target resource information. In the embodiments described above and below in this application, the focus is on the scenario where multiple target objects each independently use a separate target resource information.

[0026] Furthermore, after obtaining the target resource allocation matrix, referring to step 104 in Figure 1, resource description information corresponding to multiple target objects can be generated based on the target resource allocation matrix. The resource description information can be used to express the allocation of resource information, the total amount of resources consumed by multiple target objects, and the degree of resource saving. For example, the resource allocation situation can be expressed through specific information formats, including but not limited to: the quantity and value of allocated resource information, such as "a total of 3 resource information were issued, saving a total of 30 resources"; at the same time, the actual amount of resources paid after saving can be explained, such as "the original resource amount was 180, and the actual resource amount after saving was 150", thus clearly reflecting the resource allocation result of "saving 30 resources and ultimately consuming 150 resources".

[0027] Furthermore, after obtaining the resource description information, the resource description information can be returned to the user terminal so that the user terminal can display the resource description information on the platform page. The resource description information is used to guide the user to perform the target operation, such as a replacement operation. After the replacement is completed, multiple target objects belong to the user.

[0028] The specific implementation methods of each step involved in 101-105 above can be found in the relevant descriptions of the following embodiments, which will not be repeated here.

[0029] The technical solutions provided in the above embodiments of this application, upon receiving a user terminal request, can include a first target object associated with the user's preset behavior during browsing the platform page and a second target object defined based on historical interest into the object conversion scope. Combined with various resource information and usage rules updated in real time by the platform, a target resource mapping matrix is ​​constructed in real time. Subsequently, with the goal of maximizing the amount of resources allocated to multiple target objects, the target resource mapping matrix is ​​quickly converted into a target resource allocation matrix that can represent the allocation relationship of various resource information to each target object. This completes the optimal assignment of various resource information to multiple target objects. Furthermore, it can generate a description of the resource allocation status of multiple target objects based on the target resource allocation matrix and display it on the user terminal to guide the user to perform the target operation. This process overcomes the bottleneck of optimal resource saving for local target objects caused by sequential matching of resource information, realizing a transformation from "sequential static superposition" to "dynamic matrix global optimal allocation." It can quickly complete global optimization, significantly shorten system response latency, and provide users with a shorter response time and a more continuous experience of maximizing resource saving, even when multiple target resource objects and various resource information and usage rules provided by the platform change in real time, thereby improving the platform's object conversion efficiency.

[0030] In some embodiments, a resource mapping model is invoked to construct a resource mapping matrix between multiple target objects and various resource information based on the attribute information of multiple target objects, the latest resource information provided by the platform, and their usage rules. This includes: invoking the resource mapping model to perform semantic refinement and / or structural processing on the attribute information of multiple target objects and the usage rules information of the latest resource information provided by the platform, obtaining structured attribute information of multiple target objects and structured usage rules information of the latest resource information provided by the platform; calculating the basic matching degree between multiple target objects and various resource information based on the structured attribute information of multiple target objects and the structured usage rules information of various resource information, wherein the basic matching degree is used to characterize the static conformity between the target objects and the usage rules information of resource information, which is determined by the inherent attributes of the target objects and the usage rules information, and does not change with the user state or scenario; and constructing a resource mapping matrix between multiple target objects and various resource information based on the basic matching degree between multiple target objects and various resource information. This embodiment transforms unstructured text into a computable vector of relationships between target objects and resource information usage rules through semantic refinement and structuring. This eliminates ambiguity in natural language and greatly improves the accuracy of matrix elements in the subsequently constructed target resource mapping matrix. This lays the foundation for providing users with a user experience that maximizes resource savings with shorter response times and more continuous operation, thereby improving the platform's object conversion rate.

[0031] Optionally, by utilizing a target resource mapping model, semantic refinement and / or structuring processing is performed on the attribute information of multiple target objects and the usage rule information of various resource information to obtain multiple structured object information and usage rule information of various structured resource information. This includes: performing word segmentation and entity recognition on object titles, extracting category, sub-object, and seasonal attributes, and mapping them into dense vectors to form structured attribute information of target objects; and performing dependency parsing analysis on the usage rule information of various resource information to extract trigger types, threshold values, adjustment ranges, and applicable categories, and mapping them into structured fields to obtain structured usage rule information of various resource information. Thus, the originally unstructured natural language text is transformed into directly readable and computable structured information, providing a precise and unified data foundation for the subsequent construction of the target resource mapping matrix and target resource allocation matrix. Based on this, the data computation speed can be improved, and the experience of precise, real-time, and imperceptible resource saving can be enhanced, thereby increasing the conversion rate of platform objects.

[0032] For example, before information structuring, the received raw information is unstructured natural language text, such as an object title saying "Summer Cool Set including T-shirt + Shorts," and resource information usage rules stating "Buy 199 resources and get 20 resources off, limited to home appliances category." While this text is understandable to humans, computers struggle to directly parse its semantics and cannot use it for subsequent matrix calculations. After semantic refinement, the natural language text can be further segmented, entity extracted, and vectorized to obtain structured information. Specifically, for the target object's attribute information, the category "Set," the sub-object list containing "T-shirt" and "Shorts," and the seasonal attribute "Summer" can be extracted, and a multi-dimensional (e.g., 128-dimensional) dense vector can be generated as the target object's structured attribute information. For the resource information usage rules, dependency parsing can be used to extract trigger type "discount," threshold value of 199 resources, adjustment range of 20 resources, and applicable category "home appliances," and these fields can be encapsulated into structured usage rule information.

[0033] Optionally, based on the structured attribute information of multiple target objects and the structured usage rule information of various resource information, the basic matching degree between multiple target objects and various resource information is calculated. This includes: performing cosine similarity calculation on the object category vector in the structured attribute information of multiple target objects and the applicable category vector in the structured usage rule information of various resource information to obtain a category fit score; performing negative correlation normalization on the remaining stock of target objects and the corresponding stock threshold of resource information to obtain a stock fit score; and fusing the category fit score and the stock fit score to obtain the basic matching degree. In this embodiment, through two-level calculation of vector cosine and stock negative correlation, the system can quantify the degree of "target object-resource information" fit in milliseconds, avoiding semantic deviations caused by traditional regular expression matching, making the basic matching degree error less than 5%, and providing a high-precision input basis for subsequent priority compensation and global resource minimization processing. On this basis, it can ensure accurate, real-time, and imperceptible resource saving experience for users, thereby improving the conversion rate of platform objects.

[0034] Optionally, based on the basic matching degree between multiple target objects and various resource information, a target resource mapping matrix is ​​constructed, including: constructing an initial resource mapping matrix based on the basic matching degree between multiple target objects and various resource information; and performing sparsity prediction and clustering dimensionality reduction processing on the initial resource mapping matrix to obtain the target resource mapping matrix. In this embodiment, without losing basic matching information, a "complement first, then cut" strategy is used to compress the volume and enhance the quality of the initial resource mapping matrix, enabling the platform to output a high-confidence, low-latency target resource mapping matrix in real time. This lays the groundwork for subsequently outputting a high-confidence, low-latency target resource allocation matrix, providing users with a user experience that maximizes resource savings with shorter response times and more continuous processing, and improving the platform's object conversion efficiency.

[0035] Optionally, the initial resource mapping matrix is ​​subjected to sparsity prediction and clustering dimensionality reduction to obtain the target resource mapping matrix. This includes: predicting matrix elements in the initial resource mapping matrix whose confidence is lower than a set threshold, and replacing these matrix elements with historically optimal weighted matching scores to obtain the initial resource mapping matrix. The historically optimal weighted matching score is a reference value formed by weighting the best matching score of the same pair of "target object-resource information" in past successful allocations after time decay and success rate. This reference value is used to fill in the matrix elements with insufficient confidence, ensuring that the replacement data reflects the historical best performance without becoming outdated or distorted. Furthermore, a dual clustering strategy based on target object category and resource information type is adopted to cluster the rows and columns of the initial resource mapping matrix separately, and the initial resource mapping matrix is ​​divided into multiple sub-matrices according to the clustering results. The row dimension and column dimension of each sub-matrix are both less than or equal to a preset threshold. The multiple sub-matrices are merged according to the original row and column coordinates to obtain the target resource mapping matrix. Specifically, for matrix element positions with confidence levels below a set threshold, historical best-match weighted values ​​are used for filling to complete sparsity prediction processing. Then, a dual clustering method of "object category + strategy type" is applied to divide the original matrix into sub-matrices with both row and column counts not exceeding a preset threshold, providing structured input for maximizing resource allocation to multiple target objects. This embodiment compresses the high-dimensional sparse matrix into multiple small and dense sub-matrices through low-confidence backfilling and dual clustering, increasing density while significantly reducing memory usage and construction time. The merged sub-matrices retain global information distribution, providing a high-confidence, fast-loading mapping foundation for subsequent resource allocation.

[0036] One possible implementation involves predicting matrix elements in the initial resource mapping matrix whose confidence level is lower than a set threshold and replacing these matrix elements with historical best weighted matching scores. This includes: extracting features from three dimensions for each matrix element: basic matching score, historical success rate, and real-time consistency; inputting these three-dimensional features into a lightweight Logistic regression model that has been trained offline and updated incrementally online, and outputting the confidence level in the [0,1] interval; marking the positions of elements with confidence levels lower than a preset confidence threshold (e.g., 0.3) as low-confidence states and entering the historical best weighted value filling process.

[0037] The method employs a dual clustering strategy based on target object categories and resource information types to divide the initial adjustment decision matrix into multiple sub-matrices. One optional implementation includes: initial grouping based on the condition that the cosine similarity of the object category vectors is greater than or equal to a preset cosine similarity (e.g., greater than or equal to 0.8); and then secondary grouping according to the rule that the resource information type labels are completely identical, thereby forming a set of sub-matrices. Sub-matrices can be matrices with a row dimension less than or equal to 50 and a column dimension less than or equal to 50.

[0038] For example, the target resource mapping matrix K can be a two-dimensional numerical matrix, such as a target object set. Its matrix row index Correspondingly, a set of rules for the use of resource information. Its matrix column index ,but Its elements , For target object With resources The basic matching score is calculated as follows: (semantic similarity ± inventory fit, normalized to [0,1]). Specifically, the basic matching score is calculated as follows: First, the title, category, and attribute tags of the target object are vectorized, and the cosine similarity between them and the text vector of the usage rule information of the resource information is calculated. At the same time, the current inventory of the target object is read, and it is evaluated whether it meets the minimum usage threshold specified by the usage rule information to obtain the inventory fit. Then, the cosine similarity and inventory fit are weighted, fused, and normalized to form a basic matching score between 0 and 1. The higher the score, the more the strategy matches the target object in the "category-inventory" dimension, and the more qualified it is to participate in the subsequent resource quantity deduction.

[0039] In this embodiment, in order to further improve resource saving and enhance platform object conversion efficiency, real-time dynamic information is introduced. The real-time dynamic information includes: user real-time dynamic information and platform real-time dynamic information. The platform real-time dynamic information includes: first platform real-time dynamic information and second platform real-time dynamic information.

[0040] The real-time dynamic information of users refers to the instantaneous behavioral signals generated by users during the browsing of platform pages, reflecting their preferences and urgency. Real-time behavioral information includes, but is not limited to: browsing trajectory, interaction events, replacement signals, and preference intensity. Browsing trajectory includes, but is not limited to: the duration of time spent on the details page of the target object and the click sequence in the comparison list. The click sequence in the comparison list refers to the order in which the user clicks on the object identifiers added to the "comparison list" during the browsing of the platform page, along with the corresponding timestamps, used to depict the horizontal comparison path. Interaction events include, but are not limited to: adding to the pre-selection set, enabling resource value reduction reminders, and multiple triggers of confirmation without completion. Multiple triggers of confirmation without completion refers to the user's multiple attempts to perform the confirmation event without success. Replacement signals include, but are not limited to: the number of replacement interruptions, replacement page backtracking, and re-entering the comparison. The number of replacement interruptions refers to the cumulative number of times the user has triggered the execution of the confirmation event but the execution was terminated due to active cancellation, network abnormalities, or timeouts. Preference intensity includes, but is not limited to: the number of consecutive views of similar objects and the click interval of the resource value reduction reminder control.

[0041] The first platform's real-time dynamic information refers to instantaneous status signals generated by the platform in real time during a user's browsing of the platform's pages. These signals can influence the effectiveness or attractiveness of resource value adjustment strategies. The first platform's real-time dynamic information for the target platform includes at least the rate of change in the target object's inventory and the rate of resource information retrieval. The rate of change in inventory refers to the rate at which the remaining available quantity of the target object changes over time, i.e., the speed at which the platform's inventory decreases or increases. This is used to measure the scarcity of the object and adjust matching weights in real time. For example, when "the inventory decline rate for the home appliance category is greater than 20 units / minute" is detected, it is considered a piece of first platform real-time dynamic information and used to incrementally adjust the matrix weights.

[0042] The second platform's real-time dynamic information refers to the target platform's real-time event information. This real-time event information refers to discrete events generated instantaneously on the platform during user browsing, which can alter the availability or attractiveness of resource information. These include, but are not limited to: inventory reduction alarms, sudden announcement keywords, and temporary changes to resource information usage rules and rule text change events. Sudden announcement keywords refer to the set of keywords strongly related to the current matching scenario in the announcement text released instantaneously by the platform during platform activity incentives, replenishment, sales restrictions, or system anomalies. These keywords, after semantic extraction, are directly used as external trigger features for incrementally correcting the priority compensation weight of the target adjustment decision matrix, ensuring that resource information weights are synchronized with the platform's latest operational status within seconds. Rule text change events refer to operations where the platform adds, modifies, or abolishes rule descriptions such as triggering conditions, applicable categories, superimposed attributes, or effective periods for resource information. Once such a rule text change event occurs, the changed content is immediately parsed, and the structured fields of the corresponding resource information's usage rules are updated, ensuring that subsequent matrix construction and global optimal solutions are based on the latest rules, avoiding mismatches or failures due to rule lag. An example of an emergency announcement is: the platform pushes an announcement about the "Home Appliance Festival." The system extracts the keyword "Home Appliance Festival" as an external trigger and immediately increases the priority compensation weight of home appliance-related resource information. An example of a declining inventory alert is: when the inventory decline rate of the air conditioner category exceeds a set threshold, an inventory alert event is triggered. The platform automatically lowers the row weight of this category to avoid continuing to recommend already scarce products. An example of a temporary change to the usage rules of resource information is: the engine suddenly releases "resource information that reduces resource quantity by 50 for every 200 resources" and expands the stackable markers. After parsing, the system dynamically inserts the new resource information into the target resource mapping matrix, generating a new column dimension for subsequent global optimal assignment.

[0043] Based on this, in an optional embodiment, such as Figure 1b As shown, after obtaining the target resource mapping matrix, the following operations can also be performed: 201. Call the target resource mapping model to perform multi-dimensional demand analysis on real-time dynamic information of users and obtain multi-dimensional demand information; 202. Perform semantic refinement and structuring processing on the real-time dynamic information of the first platform to obtain structured real-time dynamic information of the first platform; 203. Based on multi-dimensional demand information and structured real-time dynamic information of the first platform, calculate the priority compensation matching degree between each target object and various resource information; 204. Integrate the basic matching degree and priority compensation matching degree between each target object and various resource information to update the target resource mapping matrix.

[0044] The multi-dimensional demand information in steps 201-203 above refers to a low-dimensional feature vector or factor set obtained by collecting and integrating real-time user behavior information and platform event information, and after multi-source sequence modeling and factor mapping. This information is used to quantitatively describe the urgency and acceptable range of user resource adjustment. This multi-dimensional demand information directly participates in the priority compensation matching degree calculation of the resource mapping matrix and is the core basis for deciding whether to lower the triggering conditions, add strategies, or increase the adjustment range. In this embodiment, through semantic refinement and structuring processing, unstructured text is transformed into a computable vector of relationships between objects and strategies, eliminating ambiguity in natural language and greatly improving the accuracy of the subsequently constructed resource mapping matrix elements. At the same time, real-time user behavior is encoded into structured demand factors such as willingness intensity and resource sensitivity, which directly participate in the priority compensation matching degree calculation, achieving zero-lag alignment of "demand-intensity". The overall response time is shortened to the level of hundreds of milliseconds, providing users with a more continuous and accurate resource adjustment experience. Moreover, accurate demand factors map high conversion willingness to high-saving potential strategies in advance, resulting in an additional increase in the platform's object conversion rate.

[0045] Optionally, based on real-time user dynamic information, a multi-dimensional demand analysis is performed to obtain multi-dimensional demand information, including: inputting the user's most recent N operation sequences into the target resource mapping model recursively into the neural network layer to obtain behavioral latent vectors, where N is positive integer data; mapping the latent vectors into three-dimensional influence factors: willingness intensity, resource quantity sensitivity, and trigger condition adjustability; and weighting and summing the three-dimensional influence factors with the platform-side stock change rate and announcement keyword weights to obtain multi-dimensional demand information.

[0046] Optionally, based on multi-dimensional demand information and real-time dynamic information of the platform, the priority compensation matching degree between each target object and various resource information is calculated, including: determining at least one influencing factor between each target object and various resource information based on multi-dimensional demand information and real-time dynamic information of the first platform, the influencing factor includes at least the user's willingness to replace the target object, resource quantity sensitivity, and the adjustability of the triggering conditions of the usage rules information of various resource information; weighting and summing at least one influencing factor between each target object and various resource information according to preset weights to obtain the compensation coefficient; obtaining the basic matching degree between each target object and various resource information; and calculating the priority compensation matching degree between each target object and various resource information based on the basic matching degree and compensation coefficient between each target object and various resource information. Wherein, the user's willingness to replace the target object refers to quantifying the urgency of the user to complete the replacement within the current session, which is mapped to a continuous value of [0,1] after being encoded by GRU from behavioral sequences such as browsing depth, comparison times, and price reduction reminder clicks. The higher the value, the stronger the replacement willingness. Resource sensitivity refers to the quantification of a user's sensitivity to changes in resource quantity (decreased or increased). It is calculated by weighting the number of interrupted historical replacements, the frequency of price reduction alerts, and rollback behavior related to resource quantity changes. A higher value indicates a more sensitive user to changes in resource quantity. The adjustability of the triggering conditions for resource information usage rules is a Boolean indicator. True indicates that the current strategy trigger threshold has exceeded the user's historical tolerance range, allowing the system to lower the threshold or generate a new strategy; it is determined by the difference between the user's historical tolerance threshold and the current threshold through threshold judgment. In this embodiment, by mapping real-time dynamic information of the platform to external incentive weights and integrating them with the basic matching degree, the system can reduce the weight of the corresponding strategy element by 20%-30% at the moment of a sudden drop in stock or a sudden announcement, achieving second-level synchronization of "platform event - matrix element", avoiding the delay of traditional manual configuration, and ensuring that the adjustment results obtained by the user at any time are consistent with the current state of the platform, improving the real-time performance and accuracy of resource quantity minimization processing. On this basis, it can guarantee the user's accurate, real-time, and imperceptible resource quantity saving experience, thereby improving the conversion rate of platform objects.

[0047] Further optionally, based on multi-dimensional demand information and real-time dynamic information of the platform, at least one influencing factor affecting priority compensation is determined. One optional implementation includes: concatenating the willingness intensity, resource sensitivity, trigger condition adjustability, stock decrease rate, resource information retrieval rate, and announcement keyword weight into a unified feature vector; mapping the feature vector to the willingness intensity factor, resource sensitivity factor, and trigger condition adjustability factor through a gated fully connected network, with the gating coefficient dynamically controlled by the stock rate and keyword weight to achieve second-level amplification of factors by platform events. In this embodiment, through the gated fully connected network, the platform event signal can be directly amplified to the influencing factor at the moment of sudden stock decrease or announcement, causing the priority compensation weight of the corresponding resource information to be increased, achieving millisecond-level linkage of "platform event-factor-matrix", avoiding the delay of traditional manual configuration, and ensuring that the global resource minimization processing is always synchronized with the user's real-time needs and the platform's instantaneous state. On this basis, it can guarantee the user's accurate, real-time, and imperceptible resource saving experience, thereby improving the conversion rate of platform objects.

[0048] In the above optional embodiments, when dynamic information is introduced, the target resource mapping matrix is ​​updated based on the priority step size matching degree. In another optional embodiment, when dynamic information is introduced, the target resource mapping matrix can be regenerated. Specifically, the target resource mapping model is invoked, and a target resource mapping matrix between multiple target objects and multiple resource information is constructed based on the attribute information of multiple target objects, the usage rule information of multiple resource information, and real-time dynamic information. This includes: using the target resource mapping model to perform semantic refinement and / or structuring processing on the attribute information of multiple target objects and the usage rule information of multiple resource information to obtain multiple structured object information and usage rule information of multiple structured resource information; and performing multi-dimensional demand analysis based on real-time behavior information to obtain multi-dimensional demand information; and constructing a resource mapping matrix between multiple target objects and usage rule information of multiple resource information based on the usage rule information of multiple structured object information, multiple structured resource information, and multi-dimensional demand information.

[0049] For example, the target resource mapping matrix K can be a two-dimensional numerical matrix, such as a set of target objects. Its matrix row index Correspondingly, a set of rules for the use of resource information. Its matrix column index ,but Its elements .in, , Let be the weighting coefficient, satisfying , , The specific data is determined by dynamic system optimization or offline training; For target object With resource information The basic matching degree (semantic similarity ± stock fit, normalized to [0,1]); Priority compensation matching degree (weighted summation of influencing factors such as user willingness intensity, resource quantity sensitivity, and trigger condition adjustability, normalized to [0,1]). Among them, the basic matching degree is used to characterize the static consistency between the target object and the usage rule information of resource information. It is determined by the inherent attributes of the target object and the usage rule information, and does not change with the user state or scenario.

[0050] Specifically, the basic matching degree calculation process is as follows: First, the title, category, and attribute tags of the target object are vectorized, and the cosine similarity between them and the text vector of the usage rule information is calculated; at the same time, the current inventory of the target object is read, and it is assessed whether it meets the minimum usage threshold specified by the usage rule information to obtain the inventory fit; then, the cosine similarity and inventory fit are weighted, fused, and normalized to form a basic matching degree score between 0 and 1. The higher the score, the more the strategy matches the target object in the "category-inventory" dimension, and the more qualified it is to participate in the subsequent resource quantity deduction.

[0051] Furthermore, the priority compensation matching degree, based on the basic matching, incorporates real-time dynamic information from both the user and the platform to dynamically adjust the matching score. Specifically, the calculation process for the priority compensation matching degree is as follows: First, the user's willingness to complete the replacement for the current target object is extracted from their recent behavior sequence, such as the number of consecutive browsing sessions or multiple entries into the pre-selection set. Second, this historical data is combined to estimate the user's sensitivity to resource quantity. Resource quantity sensitivity quantifies the gap between the upper limit of resources the user is willing to accept to complete the replacement and the current strategy trigger threshold. A larger gap indicates a more price-sensitive user, and the system is more inclined to lower the usage threshold of resource information or provide a higher resource quantity discount (i.e., adjust the usage rules of the corresponding resource information). Simultaneously, it detects whether the platform allows temporary lowering of trigger conditions or generation of new resource information and its usage rules, quantifying the adjustability of trigger conditions. Adjustability of triggering conditions refers to the flexibility of the platform in temporarily lowering the original trigger threshold, increasing the resource amount deduction, or generating new resource information and its usage rules based on the target object's inventory pressure, activity surplus (the amount of resource information that has not yet been claimed or locked in the platform's incentive activities in the current year), or user level. The higher the adjustability, the more capable the system is of relaxing the conditions in a timely manner to meet user needs. Furthermore, after weighting and summing factors such as user willingness intensity, resource amount sensitivity, and triggering condition adjustability and normalizing them, a priority compensation matching degree of 0 to 1 is obtained. The higher this score, the more urgent the user's demand for the resource amount deduction is, and the more inclined the platform is to release the corresponding resource information immediately, thereby obtaining a higher ranking weight among resource information of the same level.

[0052] Based on this, the construction process of the target resource mapping matrix K is as follows: First, perform basic matching degree calculation, that is, through the target mapping model, semantically vectorize the attribute information of the target object and the usage rule information of the target resource information, and calculate the cosine similarity between them; on this basis, the stock surplus and category adaptability indicators are superimposed, and after normalization, the basic matching degree indicator BaseMatch is finally obtained. Second, perform priority compensation matching degree calculation, that is, map the real-time dynamic information of users and the real-time dynamic information of the platform into three key influencing factors: willingness intensity, resource quantity sensitivity, and trigger condition adjustability. Through weighted summation and normalization, the priority compensation indicator PriorityComp is formed. Finally, matrix fusion and sparsity correction are implemented. First, an initial decision matrix is ​​generated based on the aforementioned linear weighting method. For matrix elements with confidence levels below a set threshold, historical best-matching weighted values ​​are used to fill in the positions, completing the sparsity prediction process. Then, a dual clustering method of "object category + strategy type" is used to divide the original matrix into sub-matrices with no more than a preset threshold in both row and column counts, providing structured input for subsequent independent solutions that minimize resource usage. Matrix fusion refers to combining the basic matching degree and priority compensation matching degree into a unified cost matrix according to weights, allowing information on objects, strategies, users, and platforms to be aggregated in the same numerical table at once. Sparsity correction involves filling elements with confidence levels below the threshold in the synthesized matrix with historical best-matching weighted values ​​and using dual clustering of object category and strategy type to divide the sub-matrices. This fills in the zero-value gaps caused by data sparsity and reduces the row and column size, ensuring that the subsequent global optimal algorithm completes the solution in milliseconds. The historical best weighted value refers to the matching score of the same object and strategy combination that was actually used in the past and achieved the greatest total resource saving. After time decay, it is stored in the cache and used to fill in the confidence of the current sparse positions, so that the matrix can immediately obtain a computable effective cost value. Confidence is a quantitative indicator of the system's credibility of the current element value in the matrix. The higher the confidence value, the more complete and closer the cost data at that position is to reality; conversely, it is considered sparse or noisy and needs to be filled in using the historical best weighted value.

[0053] Therefore, the resource mapping matrix, after sparse and cluster preprocessing, maps the target object, resource information, user intentions, and platform events across the entire domain into a unified cost matrix. The smaller the element value, the greater the potential for resource saving. This matrix directly drives row-column optimal assignment with the goal of "maximizing the amount of resources allocated to multiple target objects," also known as the global optimal algorithm. This algorithm traverses all target objects and all resource information at the same time to find a complete pairing scheme with no conflicts and the lowest total cost in one go, rather than the local suboptimal caused by sequential accumulation. Relying on sparsity prediction and dual clustering, regardless of the expansion of the row and column scale, the above-mentioned one-to-one optimal assignment of "target object-resource information" can still be completed in milliseconds, completely eliminating the limitations of local optima caused by sequential accumulation. At the same time, the model pre-fills and prunes the initial matrix, significantly reducing the number of iterations and ensuring that it remains fast and robust even in ultra-large-scale scenarios.

[0054] Optionally, the platform also includes real-time dynamic information from a second platform. Based on this, such as... Figure 1c As shown, the following operations can also be performed: 301. Call the resource mapping model to perform semantic refinement on the real-time dynamic information of the second platform to obtain external causal features related to multiple target objects, which serve as influencing factors for priority compensation. 302. Based on the influence factor of priority compensation, the priority compensation matching degree of the corresponding matrix elements in the target resource mapping matrix is ​​incrementally corrected to update the target resource mapping matrix.

[0055] In some embodiments, after obtaining the target resource mapping matrix, with the goal of maximizing the amount of resources allocated to multiple target objects, resource allocation processing is performed on multiple target objects based on the target resource mapping matrix to obtain a resource allocation matrix. This includes: applying a positive inventory correlation correction coefficient to the matrix elements in the target resource mapping matrix to obtain a corrected target resource mapping matrix; performing a maximum weight matching algorithm on the corrected target resource mapping matrix to obtain an initial resource allocation matrix; if a conflict occurs in the initial resource allocation matrix where a single resource information is repeatedly assigned to multiple target objects, then the conflicting resource information is released in reverse order of the target object submission time, and the maximum weight matching algorithm is iteratively executed until the conflict is eliminated to obtain a conflict-free target resource allocation matrix.

[0056] Optionally, a stock-related positive correlation correction coefficient is applied to the matrix elements in the target resource mapping matrix to obtain a corrected target resource mapping matrix. This includes: obtaining the stock balance and stock change rate of multiple target objects at the current time, and calculating the stock scarcity degree of each target object based on the stock balance and stock change rate; generating the stock sufficiency degree of the corresponding target object based on the stock scarcity degree, wherein the correction coefficient is positively correlated with the stock sufficiency degree and is limited to a preset lower limit; and performing row-by-row multiplication correction on all elements in the row corresponding to the target object in the target resource mapping matrix using the stock correction coefficient to obtain the corrected target resource mapping matrix for use in subsequent processing steps.

[0057] Specifically, in the constructed target resource mapping matrix that is about to be used for KM minimization, if an object is in a state of scarce stock but still participates in assignment with a high weight, it will cause a waste of resources with "high occupancy and low fulfillment". Therefore, a "stock adjustment coefficient" is applied to the matrix element K[i,j], so that the weights of all strategies of objects with lower stock levels are reduced synchronously, thereby reducing the probability of the object being assigned. In this process, the coefficient generation logic involved is as follows: real-time extraction of target objects Current inventory balance and the rate of decline of stock within the past Δt ; Calculate the scarcity of existing stock ,in The maximum historical stock for this category is given by λ, where λ is the rate penalty factor. Mapping to the [β,1] interval yields the correction coefficients. =max(β,1- ), β∈(0,1) is the lower bound to prevent the weights from returning to zero; all elements in the i-th row of the matrix are uniformly multiplied by . ,Right now The revised matrix K′ retains the original semantics and priority information while also reflecting inventory pressure in a timely manner. This allows the KM algorithm to allocate resources to objects with ample inventory during subsequent minimization, reducing resource failures due to shortages and improving the platform's fulfillment rate. Based on this, the conversion rate of platform objects can be improved.

[0058] Optionally, when resource sensitivity indicates difficulty for users to meet the triggering conditions of resource information, the triggering conditions of the corresponding resource information are dynamically lowered using the target resource mapping model, and the adaptability of the lowered triggering conditions is used as a weighting factor for the priority compensation matching degree; based on the weighting factor, the priority compensation matching degree of the corresponding element in the target resource mapping matrix is ​​updated. Alternatively, the target resource mapping model can be used to generate new resource information and its usage rules in real time; a column dimension corresponding to the new resource information is added to the target resource mapping matrix, and the matching degree of the intersection elements of each target object row and the new resource information column is filled, so that the added column participates in the subsequent resource maximization processing allocated to multiple target objects. In this embodiment, a gated fully connected network is used to concatenate the intensity of intent, sensitivity to resource availability, adjustability of triggering conditions, stock reduction rate, strategy retrieval rate, and announcement keyword weights into a unified feature vector. The channel coefficients corresponding to the stock reduction rate and keyword weights are dynamically amplified. When the platform experiences a sudden drop in stock availability or an unexpected announcement, the corresponding influencing factors are increased by 20%-40% within milliseconds, directly improving the priority compensation matching degree. This causes the weights of relevant strategy elements in the matrix to increase instantly, ensuring that the maximization of resource allocation to multiple target objects is always synchronized with users' real-time needs and the platform's instantaneous state. This avoids the delays of traditional manual settings and achieves zero-lag linkage between "event-factor-matrix". Based on this, accurate, real-time, and imperceptible resource saving experiences for users can be guaranteed, thereby improving the conversion rate of platform targets.

[0059] Optionally, using a target mapping model, the triggering conditions for the corresponding resource information are dynamically adjusted downwards, and the adjusted triggering condition fit is used as a weighting factor for the priority compensation matching degree. This includes: parsing the current threshold value of the resource information and the user's resource quantity sensitivity factor; if the sensitivity factor is greater than a preset threshold and the current threshold exceeds the user's historical tolerance range, a new threshold is calculated using a piecewise linear function, for example, new threshold = max(original threshold × 0.8, user's historical median tolerance); the new threshold is negatively correlated with the user's current acceptable value and normalized to obtain the triggering condition fit, which is then used as a weighting coefficient and superimposed on the priority compensation matching degree to update the corresponding elements in the target resource mapping matrix.

[0060] Optionally, based on the weighting factor, the priority compensation matching degree of the corresponding element in the target resource mapping matrix is ​​updated, including: using the trigger condition fit degree as the weighting factor and adopting a linear superposition formula, for example, the updated priority compensation matching degree = the original priority compensation matching degree × (1 + α × weighting factor), where α is the system dynamic coefficient and α∈[0,0.2]; the updated priority compensation matching degree is atomically written in the sparse matrix storage structure, which can quickly complete the element-level update and trigger the recalculation of the subsequent allocation of resources to each target object to maximize the processing.

[0061] Optionally, using a target mapping model, new resource information and its usage rules are generated in real time, including: activating the target resource mapping model when the resource sensitivity exceeds the threshold sensitivity threshold (e.g., 0.7) and the current strategy cannot meet the user's tolerance range; obtaining a new threshold value using Bayesian regression sampling with the user's historical median tolerance as the mean and the stock decline rate as the variance; using a lightweight Transformer to predict the adjustment range within the range of [-20%, -5%], and extending the applicable categories of the original strategy using synonyms; encapsulating the new threshold, adjustment range, applicable categories, and overlayable tags into a platform standard strategy JSON, and calling the strategy creation API within 10ms to return the new strategy ID; dynamically adding a column to the target resource mapping matrix, filling in the cross-matching degree between each object and the new strategy, so that it can immediately participate in the subsequent processing of maximizing the allocation of resources to multiple target objects.

[0062] Further optional, such as Figure 1d As shown, after adjusting the resource quantities of multiple target objects based on the combination of usage rules information for the target resource information, the following operations can also be performed: 401. Continuously monitor changes in various real-time dynamic information; 402. When the monitoring results meet the preset trigger conditions, dynamically update the target resource mapping matrix, and re-allocate resources to multiple target objects based on the target resource mapping matrix with the goal of maximizing the amount of resources allocated to multiple target objects. 403. Based on the updated resource allocation matrix, generate resource description information corresponding to multiple target objects.

[0063] The preset triggering condition in step 402 refers to a set of configurable rules used by the backend to determine whether to initiate the "incremental update of resource mapping matrix + recalculation of resource allocation to each target object" process during continuous monitoring of real-time dynamic information. This set of rules consists of the following dimensions (supporting single items or weighted combinations): object-side thresholds (such as inventory balance change rate, inventory safety lower limit), resource-side thresholds (such as adding / revoking / changing rules, adjusting overlayable attributes), user behavior thresholds (such as confirmation page dwell time, resource quantity reduction reminders upon re-clicking, number of confirmation interruptions), platform time thresholds (such as sudden platform incentive activity announcements, category daily switching, competitor resource quantity reduction information), and time thresholds (the period since the last strategy assignment exceeding a fixed time). When any single indicator is triggered or the weighted comprehensive score exceeds the preset threshold, the system determines that the preset triggering condition is met, immediately executes the conflict-free optimal strategy re-solution, and synchronously pushes the updated results and natural language explanations to the user terminal. Typical trigger scenarios include, but are not limited to: abnormal change rate of inventory balance for any target object or inventory falling below the safety threshold; usage rules information for newly added, invalidated, or changed resource information on the platform; user stays on the confirmation page for more than a preset threshold, repeatedly clicks "Resource Quantity Decrease Reminder," or returns to the comparison page; the system detects sudden platform incentive activity announcements, category day switching, or competitor resource quantity reduction; and timed fallback refresh triggered when the maximum calculation of the resource quantity allocated to multiple target objects exceeds a preset time threshold. This solution constructs a closed-loop mechanism of "continuous monitoring - dynamic recalculation - real-time push," which can capture real-time dynamic changes in stock status, resource information, and user and platform data within a critical window of several minutes after a user submits a replacement request. Once the monitored data meets the preset trigger conditions, the system immediately updates the target resource mapping matrix incrementally and resolves the problem of maximizing the amount of resources allocated to multiple target objects. This mechanism controls the response latency to the order of hundreds of milliseconds, allowing users to achieve "seamless resource replacement and explicit resource savings" without manually refreshing or supplementing target object information, effectively reducing the order abandonment rate in the replacement process. At the same time, it proactively provides quantitative feedback on the "extra savings of ×× resources compared to the original plan" through natural language explanation, forming positive psychological guidance and significantly improving the replacement conversion rate and user satisfaction. For platform operators, this mechanism can effectively avoid asset losses caused by over-issuance of resource adjustment vouchers and failure to fulfill obligations by dynamically reducing the weight of stock shortage or invalid strategies, thereby optimizing object conversion efficiency. Based on this, the system further improves the conversion efficiency of platform objects while ensuring that users obtain accurate, real-time, and seamless resource savings.

[0064] Optionally, continuous monitoring of changes in various real-time dynamic information includes: using the target resource mapping model to perform multi-source time-series analysis on the user's browsing trajectory, the acquisition trend of various resource information, and the rate of change of the stock of target objects, predicting the candidate target objects that the user may add later and their candidate probabilities; and writing the candidate target objects into the target resource mapping matrix in advance with a pre-matched matching degree, so as to complete the matching degree update before the actual replacement behavior occurs. In this embodiment, multi-source time-series analysis upgrades the "post-event response" to "pre-event pre-setting." The probability prediction and matrix matching degree of candidate categories are completed before the user actually adds them to the pre-selection set. Therefore, when the user clicks or swipes again, the system can provide optimized matching results in milliseconds, significantly shortening decision-making waiting time. The pre-updated matching degree ensures that newly added target objects are instantly integrated into the global resource allocation to multiple target objects, maximizing the amount of resources allocated to them. This avoids the discount gaps caused by the traditional "add to pre-selection set first, then recalculate" approach, improving the continuity of target object additions and the user's perception of cost savings. Adding target objects refers to new objects temporarily added by the user to the existing candidate set. The system immediately incorporates these into the target resource mapping matrix and re-executes the globally optimal assignment to trigger a higher-level resource reduction. For the platform, the pre-loading mechanism reduces redundant calculations during peak concurrency, lowering CPU usage. Simultaneously, incorporating the rate of change in stock into the prediction model allows for the priority push of candidate categories with ample stock to users, reducing resource information failure and fulfillment risks caused by stockouts, achieving a dual improvement in object conversion efficiency and user experience. Based on this, while ensuring users can accurately and instantly access resources without any loss of experience, the system further improves the conversion efficiency of platform objects.

[0065] Optionally, when a change in the usage rules of any resource information is detected, the target mapping model is used to perform semantic analysis and element prediction on the usage rules of any resource information to obtain the predicted new triggering conditions, superimposed attributes, and applicable object category range, which are then used as the prediction results. Semantic analysis involves breaking down the rule text suddenly released by the platform into computable units to extract fields such as "trigger threshold, reduction range, superimposed markers, and applicable categories," and converting them into structured data. Element prediction involves using the model to calculate, based on existing fields, content that has not yet been announced but is highly likely to be adjusted (such as new threshold values, superimposed ranges, and effective periods), generating a complete set of strategy elements in advance for immediate addition or weight adjustment to the matrix. Furthermore, based on the prediction results, the matching degree of the target resource mapping matrix is ​​updated using the corresponding matrix elements. In this embodiment, the moment a change in the usage rules of resource information is detected, the new threshold, superimposed attributes, and applicable categories are immediately output through semantic analysis and element prediction, and the prediction results are directly mapped to update the matching degree of matrix elements, achieving a second-level closed loop of "strategy change → matrix refresh." Its technical advantages are reflected in three aspects: First, no manual intervention is required for rule configuration; platform activity launches or rule adjustments can be quickly synchronized to the decision matrix, greatly improving operational efficiency. Second, after real-time matching correction, the system's subsequent minimization solution is always based on the latest rules, preventing users from being unable to use resources due to old thresholds or category restrictions, significantly reducing complaints and order abandonment rates. Third, by refreshing stackable attributes in advance, the system ensures that stacked resource combinations are correctly included in the global resource minimization, improving the overall resource utilization rate of the platform and reducing financial losses caused by mismatches. Ultimately, users can instantly enjoy the best offers that are perfectly aligned with the latest activities at any stage, experiencing continuous and "seamless updates." Based on this, while ensuring users accurately and instantly access resources without any sense of loss, the system further improves the conversion efficiency of platform targets.

[0066] Optionally, at least a portion of the user's real-time dynamic information indicates that the user is in the replacement phase, which includes at least a browsing phase, a pre-selection phase, and a confirmation phase. Based on this, using a target resource mapping model, the resource quantity adjustment range of the resource information usage rule information is progressively increased according to the advancing order of the replacement phases. In this way, while ensuring users accurately and instantly access resources without any noticeable resource savings, the system further improves the conversion efficiency of platform objects.

[0067] Optionally, when the replacement phase is the confirmation phase, the system monitors the waiting time after the user triggers the object confirmation event. If the waiting time exceeds a preset time threshold, new resource information with different adjustment ranges is created immediately according to preset or real-time generated resource quantity adjustment rules. The target resource mapping matrix is ​​then dynamically updated based on the new resource information to present the user with resource information, usage rules, and resource quantity adjustment ranges that match the current waiting time. Based on this, the system ensures users receive accurate, real-time, and seamless resource saving experiences while further improving the conversion efficiency of platform objects.

[0068] For ease of understanding, the above-mentioned technical solutions of this application will be described in their entirety below.

[0069] 1. Triggering and Data Acquisition: The platform continuously monitors user actions such as "adding to a pre-selection set, submitting a combination request, repeatedly clicking to confirm incomplete actions, and re-entering the details page." Once such behavior occurs, three types of anonymized data are immediately retrieved via the event bus: First, the static attributes (category, specifications, baseline value) and dynamic attributes (stock balance, platform incentive activity identifier, price comparison index) of the target object; second, the usage rule information of currently effective resource information (strategy ID, trigger condition expression, adjustment range, overlayable markers, applicable categories); third, real-time dynamic information from the first platform (at least including the target object's inventory change rate and resource information retrieval rate); fourth, real-time dynamic information from the second platform (at least including inventory decline alarms, sudden announcement keywords, temporary changes to resource information usage rule information, and rule text change events); and fifth, real-time dynamic information of the user (at least including browsing history, interaction events, replacement signals, and preference intensity). This data is aggregated into the target mapping model input buffer for subsequent unified processing.

[0070] 2. Semantic-driven matrix construction: This model is implemented by cascading a pre-trained large model with a lightweight decision network. The specific steps are as follows: Step A: Semantic Refinement. The object title, description, and strategy text are segmented, entity extracted, and vectorized to output structured object information and structured strategies.

[0071] Step B: Multi-dimensional Demands. User behavior sequences are mapped to three-dimensional factors: willingness intensity, price sensitivity, and threshold adjustability. These factors are then integrated with existing user growth rate and announcement weight to form multi-dimensional demand information.

[0072] Step C: Matrix Generation. Using object ID as the row and strategy ID as the column, calculate the basic row-column matching degree (semantic similarity ± existing resource fit). Then, superimpose multi-dimensional requirements to obtain the priority compensation matching degree. The two are weighted and merged into a resource mapping matrix; the smaller the element value, the greater the potential for resource saving.

[0073] 3. Sparse clustering and resource minimization: (1) Sparsity prediction: For elements with confidence scores below the threshold, fill them with historical best weighting values ​​to improve matrix density.

[0074] (2) Clustering dimensionality reduction: The matrix is ​​divided into multiple sub-matrices with rows × columns ≤ 50 by adopting a dual clustering method of "object category + strategy type".

[0075] (3) Maximizing the amount of resources allocated to multiple target objects: Each submatrix runs the improved KM algorithm in parallel to solve the row-column one-to-one minimum cost assignment. After merging, the initial strategy set is obtained, ensuring that each object is regulated by only one strategy, the sum of the overall resource consumption of multiple target objects is minimized, and there is no conflict between resource information.

[0076] 4. Dynamic correction and immediate effect: If a user's resource sensitivity indicator shows difficulty in reaching the original threshold, the system will adjust the trigger value of the corresponding resource information and update the matrix weights in real time. Simultaneously, it can instantly generate new resource information and its usage rules, add matrix columns, and populate cross-matching degrees, allowing the new strategy to immediately participate in the next round of assignment. When any strategy rule change is detected, the large model parses the new threshold, stackable attributes, and applicable categories within seconds, incrementally correcting the corresponding weights without manual configuration.

[0077] 5. Predictive preloading and facilitating the triggering of composite requests: Continuously monitor browsing history, strategy claim rate, and inventory change rate. Through multi-source time-series analysis, predict the categories that users may add to the pre-selection set. Write the predicted categories into the matrix in advance with a pre-matching degree to achieve zero-lag updates before the actual addition to the pre-selection set. When the user is in the confirmation stage and the waiting time exceeds the preset threshold, create new strategies with different strengths in real time according to preset or real-time generated discount rules, add a new matrix column and push a countdown reminder to complete the triggering of the promotion combination request.

[0078] 6. Closed-loop update and explanation push: After the assignment is completed, the system continues to monitor the real-time dynamics. Once preset conditions such as "excessive change rate of stock, change of strategy rules, user rollback comparison or timeout after replacement" are triggered, the matrix is ​​immediately updated incrementally and the solution is re-executed (e.g., "This adjustment saved a total of XX resources") and pushed to the user terminal, forming a closed loop of "monitoring-recalculation-push".

[0079] Therefore, this solution transforms from "static rule overlay" to "semantic-driven global row and column resource volume adjustment range," providing users with greater discounts, shorter response times, and a more continuous user experience in complex and ever-changing platform incentive activity scenarios, while improving the conversion efficiency of platform objects.

[0080] Figure 2 A schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this application. For example... Figure 2 As shown, the electronic device includes: a memory 20a and a processor 20b; the memory 20a is used to store computer programs; the processor 20b is coupled to the memory 20a and is used to execute the computer programs to implement the steps in the above method.

[0081] The detailed implementation methods and beneficial effects of each step in this embodiment have been described in detail in the foregoing embodiments, and will not be elaborated here.

[0082] Furthermore, such as Figure 2 As shown, the electronic device also includes other components such as a communication component 20c, a display 20d, a power supply component 20e, and an audio component 20f. Figure 2 The diagram only shows some components and does not mean that the electronic device includes only these components. Figure 2 The components shown.

[0083] in addition, Figure 2 The components within the dashed box are optional, not mandatory, and their specific requirements depend on the product form of the electronic device. The electronic device in this embodiment can be a terminal device such as a desktop computer, laptop computer, smartphone, or IoT device, or a server-side device such as a conventional server, cloud server, or server array. If the electronic device in this embodiment is a terminal device such as a desktop computer, laptop computer, or smartphone, it may include... Figure 2 The components within the dashed box; if the electronic device in this embodiment is implemented as a conventional server, cloud server, or server array, etc., it may be omitted. Figure 2 The component within the dashed box.

[0084] The aforementioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0085] The aforementioned communication components are configured to facilitate wired or wireless communication between the device containing the communication components and other devices. The device containing the communication components can access wireless networks based on communication standards, such as WiFi, 2G, 6G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication components receive broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication components also include a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wide Band (UWB), Bluetooth (BT), and other technologies.

[0086] The aforementioned display includes a screen, which may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a Touch Panel, the screen can be implemented as a touchscreen to receive input signals from the user. The Touch Panel includes one or more touch sensors to sense touches, swipes, and gestures on the Touch Panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation.

[0087] The aforementioned power supply components provide power to various components within the device in which they reside. These power supply components may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to the device in which they reside.

[0088] The aforementioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0089] Accordingly, exemplary embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement the steps in the above-described method.

[0090] Accordingly, an exemplary embodiment of this application also provides a computer program product comprising a computer program / instructions that, when executed by a processor, cause the processor to perform the steps in the above-described method.

[0091] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, compact disc read-only memory (CD-ROM), optical storage, etc.) containing computer-usable program code.

[0092] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0095] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interfaces, network interfaces, and memory.

[0096] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0097] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, Digital Video Disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0098] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, object, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, object, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, object, or apparatus that includes that element.

[0099] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A resource information processing method, characterized in that, include: The system receives an information processing request sent by a user terminal. The information processing request is sent by the user of the user terminal when a preset behavior is detected during the browsing of the platform page. The information processing request includes attribute information of multiple target objects. The multiple target objects include a first target object associated with the preset behavior and a second target object whose user interest is determined to be greater than a set threshold based on the user's historical behavior. The target resource mapping model is invoked, and a target resource mapping matrix is ​​constructed between the multiple target objects and the multiple resource information based on the attribute information of the multiple target objects, the latest resource information provided by the platform, and their usage rules. The multiple row elements in the target resource mapping matrix represent multiple target objects, and the multiple column elements represent multiple target resource information. The target resource information is resource information that can be used by at least one target object, as determined by the usage rules. With the goal of maximizing the amount of resources allocated to the multiple target objects, resource allocation processing is performed on the multiple target objects based on the target resource mapping matrix to obtain a target resource allocation matrix. In the target resource allocation matrix, row elements and column elements intersect to form matrix elements, and the matrix elements represent the matching degree between the target resource information corresponding to its column element and the target object corresponding to its row element. Based on the target resource allocation matrix, resource description information corresponding to the multiple target objects is generated, and the resource description information is used to describe the resource allocation of the multiple target objects; The resource description information is returned to the user terminal so that the user terminal can display the resource description information on the platform page. The resource description information is used to guide the user to perform the target operation.

2. The method according to claim 1, characterized in that, The resource mapping model is invoked, and based on the attribute information of the multiple target objects, the latest resource information provided by the platform, and their usage rules, a target resource mapping matrix is ​​constructed between the multiple target objects and the multiple resource information, including: The resource mapping model is invoked to perform semantic refinement and / or structural processing on the attribute information of the multiple target objects and the usage rule information of the various resource information recently provided by the platform, so as to obtain the structured attribute information of the multiple target objects and the structured usage rule information of the various resource information recently provided by the platform; Based on the structured attribute information of the multiple target objects and the structured usage rule information of the multiple resource information, calculate the basic matching degree between the multiple target objects and the multiple resource information; Based on the basic matching degree between the multiple target objects and the multiple resource information, a target resource mapping matrix is ​​constructed between the multiple target objects and the multiple resource information.

3. The method according to claim 2, characterized in that, Based on the basic matching degree between the multiple target objects and the multiple resource information, a target resource mapping matrix is ​​constructed between the multiple target objects and the multiple resource information, including: Based on the basic matching degree between the multiple target objects and the multiple resource information, an initial resource mapping matrix is ​​constructed; The initial resource mapping matrix is ​​subjected to sparsity prediction and clustering dimensionality reduction to obtain the target resource mapping matrix.

4. The method according to claim 3, characterized in that, The initial resource mapping matrix is ​​subjected to sparsity prediction and clustering dimensionality reduction processing to obtain the target resource mapping matrix, including: Predict the matrix elements in the initial resource mapping matrix whose confidence level is lower than a set threshold, and replace the matrix elements with the historical best weighted matching degree to obtain the initial resource mapping matrix; A dual clustering strategy based on target object category and resource information type is adopted to cluster the rows and columns of the initial resource mapping matrix respectively, and the initial resource mapping matrix is ​​divided into multiple sub-matrices according to the clustering results. The row dimension and column dimension of each sub-matrix are less than or equal to a preset threshold. The multiple sub-matrices are merged according to their original row and column coordinates to obtain the target resource mapping matrix.

5. The method according to claim 1, characterized in that, The user corresponds to real-time dynamic information of the user, and the platform corresponds to real-time dynamic information of the first platform; After obtaining the target resource mapping matrix, the following is also included: The target resource mapping model is invoked to perform multi-dimensional demand analysis on the real-time dynamic information of the user, thereby obtaining multi-dimensional demand information. Based on the multi-dimensional demand information and the real-time dynamic information of the first platform, the priority compensation matching degree between each target object and the various resource information is calculated. The target resource mapping matrix is ​​updated by integrating the basic matching degree and priority compensation matching degree between each target object and the various resource information.

6. The method according to claim 5, characterized in that, Based on the multi-dimensional demand information and the platform's real-time dynamic information, the priority compensation matching degree between each target object and the various resource information is calculated, including: Based on the multi-dimensional demand information and the real-time dynamic information of the first platform, at least one influence factor for priority compensation between each target object and the various resource information is determined. The influence factor includes at least the user's willingness to replace the target object, the sensitivity to resource quantity, and the adjustability of the triggering conditions of the usage rules information of various resource information. The compensation coefficient is obtained by weighting and summing at least one influencing factor between each target object and the various resource information according to a preset weight. Obtain the basic matching degree between each target object and the various resource information; Based on the basic matching degree and compensation coefficient between each target object and the various resource information, the priority compensation matching degree between each target object and the various resource information is calculated.

7. The method according to claim 1, characterized in that, The platform also corresponds to real-time dynamic information from a second platform; the method further includes: The resource mapping model is invoked to perform semantic refinement on the real-time dynamic information of the second platform, thereby obtaining external causal features related to the multiple target objects, which serve as influencing factors for priority compensation. Based on the influence factor of the priority compensation, the priority compensation matching degree of the corresponding matrix element in the target resource mapping matrix is ​​incrementally corrected to update the target resource mapping matrix.

8. The method according to claim 1, characterized in that, With the objective of maximizing the amount of resources allocated to the multiple target objects, resource allocation processing is performed on the multiple target objects based on the target resource mapping matrix to obtain a resource allocation matrix, including: The inventory positive correlation correction coefficient is applied to the matrix elements in the target resource mapping matrix to obtain the corrected target resource mapping matrix; The maximum weight matching algorithm is applied to the modified target resource mapping matrix to obtain the initial resource allocation matrix; If a conflict occurs in the initial resource allocation matrix where a single resource is repeatedly assigned to multiple target objects, the conflicting resource information is released in reverse order of the target object submission time, and the maximum weight matching algorithm is iteratively executed until the conflict is eliminated, so as to obtain a conflict-free target resource allocation matrix.

9. The method according to claim 1, characterized in that, Also includes: Continuously monitor changes in various real-time dynamic information; When the monitoring results meet the preset trigger conditions, the target resource mapping matrix is ​​dynamically updated, and the resource allocation is re-performed based on the target resource mapping matrix to maximize the amount of resources allocated to the multiple target objects, resulting in an updated resource allocation matrix. Based on the updated resource allocation matrix, resource description information corresponding to the multiple target objects is generated.

10. The method according to claim 9, characterized in that, Continuously monitor changes in various real-time dynamic information, including: Multi-source time-series analysis is performed on the user's browsing trajectory, the acquisition trend of various resource information, and the inventory change rate of target objects to predict the candidate target objects that the user may add in the future and their candidate probabilities. The candidate target objects are written into the target resource mapping matrix in advance with a pre-matched matching degree, so as to complete the matching degree update before the actual replacement behavior occurs.

11. The method according to claim 10, characterized in that, Also includes: When any resource information and its usage rule information are detected to change, semantic parsing and element prediction are performed on the usage rule information of the resource information to obtain new triggering conditions, overlayable attributes and applicable object category range, which are used as prediction results. Based on the prediction results, update the matching degree of the target resource mapping matrix that matches the corresponding matrix element.

12. The method according to any one of claims 1-11, characterized in that, At least a portion of the real-time user dynamic information indicates that the user is in the replacement phase, the replacement phase including at least a browsing phase, a pre-selection phase, and a confirmation phase; the method further includes: According to the progressive order of the replacement stages, the resource quantity adjustment range of the various resource information is increased step by step.

13. The method according to claim 12, characterized in that, When the replacement phase is a confirmation phase, the waiting time after the user triggers the object confirmation event is monitored; the method further includes: If the waiting time exceeds a preset time threshold, new resource information with different adjustment ranges is created immediately according to preset or real-time generated resource quantity adjustment rules, and the target resource allocation matrix is ​​updated based on the new resource information.

14. An electronic device, characterized in that, include: Memory and processor; The memory is used to store a computer program; the processor, coupled to the memory, is used to execute the computer program to implement the steps of any one of claims 1-13.

15. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it causes the processor to perform the steps of any one of the methods of claims 1-13.

16. A computer program product, characterized in that, The computer program product includes a computer program / instruction that, when executed by a processor, causes the processor to perform the steps of any one of the methods of claims 1-13.

Images