A multi-platform-based adaptive recharge scheduling method and system

Abstract

The application is suitable for the technical field of mobile terminal advertisement scheduling and resource adaptive control, and provides a kind of adaptive recharge scheduling method and system based on multi-platform, the method comprises: under the premise that current GPU load is greater than critical GPU load, adjustment coefficient is generated according to the difference between the two, and the content complexity corresponding to the initial advertisement content is adjusted based on the adjustment coefficient;Select the advertisement content that best matches the adjusted content complexity from several candidate platforms, replace the initial advertisement content with it, and push it to the recharge page of the target user.The application introduces terminal device GPU load as a key technical parameter for advertisement scheduling in the scenario of charging, dynamically determines the critical GPU load based on the correlation between user interaction behavior and GPU load change in historical samples, and generates an adjustment coefficient based on the degree of excess when the current GPU load exceeds the critical GPU load, to adaptively adjust the complexity of the advertisement content.

Description

Technical Field

[0001] This invention belongs to the field of mobile terminal advertising scheduling and resource adaptive control technology, and particularly relates to an adaptive recharge scheduling method and system based on multiple platforms. Background Technology

[0002] With the development of mobile internet technology, mobile phone top-up has become one of the most common and frequently used basic services on mobile apps. To improve conversion rates and user value, existing mobile phone top-up pages or top-up plan selection pages typically display advertising content related to the operating system, such as mobile phone top-up promotions, package recommendations, or related product promotions. At the same time, advertising models are gradually shifting from single-source to multi-platform parallel development. Different advertising platforms provide diverse advertising content to the app and obtain display opportunities through polling or sequential delivery mechanisms to achieve a relatively fair distribution of exposure.

[0003] During the aforementioned advertising campaigns, different ad content exhibits significant differences in display format, rendering method, and interactive structure, resulting in varying degrees of resource consumption on terminal devices. When users load ad content during the recharge phase, ad rendering itself increases the GPU load on the terminal device, especially when the ad content includes animation effects, video displays, or multi-layered interactive components, where the impact on terminal performance is more pronounced. For terminal devices that have been used for extended periods, have many background processes running, or have low hardware performance, displaying ads during the recharge phase can easily lead to page lag, loading delays, or incomplete ad display, thereby affecting user experience and the actual effectiveness of ad display.

[0004] To address the issue of ad delivery under high terminal load, existing technologies typically employ methods such as delaying page loading or directly replacing ads with lower-complexity ones. However, delaying loading increases user waiting time and affects the smoothness of the recharge process; while simply replacing ads with lower-complexity ones lacks precise judgment on the degree of complexity matching, easily leading to frequent exposure of some lower-complexity ads and undermining the fairness of multi-platform delivery mechanisms. Furthermore, existing solutions generally do not fully consider the correlation between changes in terminal GPU load and user ad interaction behavior, making it difficult to balance terminal performance and platform interests while ensuring ad effectiveness. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive recharge scheduling method and system based on multiple platforms, aiming to solve the problems mentioned in the background art.

[0006] This invention is implemented as follows: an adaptive recharge scheduling method based on multiple platforms, the method comprising:

[0007] When identifying a target user's behavior of topping up mobile phone credit using a terminal device with a high GPU load, determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the topping-up page.

[0008] Call the preset database and filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one. Each sample includes the user's interaction behavior with the historical advertisements during the recharge process.

[0009] Calculate the level of interaction behavior for each sample and analyze whether the level of interaction behavior increases as the GPU load decreases and remains stable after reaching a turning point; if so, set the GPU load corresponding to the turning point as the critical GPU load.

[0010] Given that the current GPU load is greater than the critical GPU load, an adjustment coefficient is generated based on the difference between the two, and the content complexity of the initial ad content is adjusted based on the adjustment coefficient.

[0011] Select the ad content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial ad content, and push it to the target user's recharge page.

[0012] As a further limitation of the technical solution of the present invention, the current GPU load of the terminal device is determined by: obtaining real-time data on the core load utilization, video memory usage and computing task queue length of the terminal device's GPU, and combining this data with hardware monitoring information provided by the GPU driver to calculate the current GPU load.

[0013] As a further limitation of the technical solution of the present invention, the phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

[0014] As a further limitation of the technical solution of the embodiment of the present invention, the calculation process of the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and performing weighted calculation on each interactive data to obtain a comprehensive interaction score.

[0015] As a further limitation of the technical solution of this invention embodiment, under the premise that the current GPU load is greater than the critical GPU load, the step of generating an adjustment coefficient based on the difference between the two, and adjusting the content complexity corresponding to the initial advertisement content based on the adjustment coefficient includes:

[0016] Determine if the current GPU load exceeds the critical GPU load. If the current GPU load exceeds the critical GPU load, generate an adjustment factor based on the degree to which the current GPU load exceeds the critical GPU load.

[0017] The preset correction model is invoked, and the content complexity of the initial advertisement content is adjusted by combining the adjustment coefficients to obtain the adjusted content complexity.

[0018] As a further limitation of the technical solution of this embodiment of the invention, the preset correction model is:

[0019] ;

[0020] in, This refers to the adjusted complexity of the content. This refers to the complexity of the initial advertisement content. This refers to the minimum allowable value for content complexity. This refers to the current GPU load. This refers to critical GPU load. This refers to the degree to which the current GPU load exceeds the critical GPU load. This refers to the preset control correction strength coefficient, and it satisfies... Greater than 0, This refers to the adjustment factor.

[0021] An adaptive recharge scheduling system based on multiple platforms, the system comprising:

[0022] The load detection module is used to determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the recharge page when the target user is identified to be using a terminal device with a high GPU load to recharge mobile phone credit.

[0023] The sample filtering module is used to call a preset database to filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one, and each sample includes the user's interaction behavior with the historical advertisements during the recharge process.

[0024] The interactive behavior analysis module is used to calculate the level of interactive behavior for each sample and analyze whether the level of interactive behavior increases as the GPU load decreases and remains stable after reaching a turning point; if so, the GPU load corresponding to the turning point is set as the critical GPU load.

[0025] The complexity adjustment module is used to generate an adjustment coefficient based on the difference between the current GPU load and the critical GPU load, and then adjust the content complexity of the initial advertisement content based on the adjustment coefficient.

[0026] The advertising platform selection module is used to select the advertising content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial advertising content, and push it to the target user's recharge page.

[0027] As a further limitation of the technical solution of the present invention, the current GPU load of the terminal device is determined by: obtaining real-time data on the core load utilization, video memory usage and computing task queue length of the terminal device's GPU, and combining this data with hardware monitoring information provided by the GPU driver to calculate the current GPU load.

[0028] As a further limitation of the technical solution of the present invention, the phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

[0029] As a further limitation of the technical solution of the embodiment of the present invention, the calculation process of the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and performing weighted calculation on each interactive data to obtain a comprehensive interaction score.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] This invention introduces the GPU load of the terminal device as a key technical parameter for advertising scheduling in the context of mobile phone top-up. Based on the correlation between user interaction behavior and GPU load changes in historical samples, it dynamically determines the critical GPU load. When the current GPU load exceeds the critical GPU load, it generates an adjustment coefficient based on the degree of exceedance to adaptively adjust the complexity of the advertising content.

[0032] This solution avoids the user experience degradation and unfair multi-platform placement issues caused by delayed page loading or simple replacement of low-complexity ads in existing technologies. It achieves a balance between ad display effectiveness and terminal performance, even without changing the objective load conditions of the terminal. Furthermore, by using adjusted content complexity as an abstract matching parameter, it enables refined and fair ad content selection across multiple advertising platforms. This improves the actual interaction effect for users with ad-interaction habits while protecting the placement rights of different advertising platforms, demonstrating significant practical value and promising prospects for promotion. Attached Figure Description

[0033] Figure 1 A flowchart of the method provided in the embodiments of the present invention;

[0034] Figure 2 This is a flowchart illustrating the method for adjusting the content complexity of the initial advertisement content in the embodiments of the present invention.

[0035] Figure 3 The application architecture diagram of the system provided in the embodiments of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] Figure 1 A flowchart of the method provided by an embodiment of the present invention is shown.

[0038] Specifically, an adaptive recharge scheduling method based on multiple platforms includes the following steps:

[0039] Step S100: When identifying a target user using a terminal device with a high GPU load to recharge phone credit, determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the recharge page.

[0040] The current GPU load of the terminal device is determined by acquiring real-time data on the core load utilization, memory usage, and computation task queue length of the terminal device's GPU, combined with hardware monitoring information provided by the GPU driver.

[0041] In this embodiment of the invention, the invention is applied to a mobile phone top-up system on an app. In existing mobile phone top-up services, the top-up page or top-up plan selection page typically displays advertising content related to the operating system, such as promotional activities, package recommendations, service binding guidance, or promotional information for other related products. This type of advertising has become a common component of the top-up process, used to promote services and reach users without adding extra steps for the user.

[0042] As advertising business models evolve, the number of advertising platforms partnering with mobile apps continues to increase. These platforms differ in their ad content types, display formats, and delivery strategies. Typically, to ensure display opportunities across platforms, existing systems employ polling or sequential delivery mechanisms, pushing ad content from different platforms to different users' recharge pages in turn, thus achieving a relatively fair distribution of exposure.

[0043] During the aforementioned advertising campaign, different ad content exhibits significant differences in display format, rendering method, and interactive structure, thus forming the so-called content complexity. Content complexity is an abstract technical metric used to characterize the degree of graphics rendering resource consumption when ad content is displayed and interacted with on a terminal device. Specifically, it can be understood as a comprehensive reflection of factors such as the number of animation effects included in the ad content, the hierarchy of dynamic elements, the need for video or 3D image rendering, the number of real-time interactive components, and the page refresh rate. Higher content complexity means greater GPU computing and rendering resource requirements on the terminal device.

[0044] During the process of users topping up their mobile phone credit, loading advertising content onto the recharge page increases the GPU processing load on the terminal device, especially when the advertising content contains complex animation effects, dynamic video displays, or multi-layered interactive components. When the terminal device is already under high GPU load—for example, when the device has been running for a long time, has many background programs, low hardware performance, or is an older model—continuing to load highly complex advertising content can easily lead to screen lag, loading delays, or incomplete ad display on the recharge page, thus affecting the user experience and the actual display effect of the advertisement.

[0045] To address the aforementioned issues, existing technologies typically incorporate load identification mechanisms. When high load is detected on the terminal device, two main processing methods are employed. One method is to delay the loading time of the recharge page or advertising content to allow terminal resources to be released. However, this method directly increases user waiting time, affecting the smoothness of the recharge process. The other method is to directly replace the content with less complex advertising content. However, this method often lacks precise judgment on the degree of complexity matching, easily leading to the frequent selection of low-complexity ads. This results in some advertising platforms gaining higher exposure rates, undermining fairness in multi-platform advertising scenarios.

[0046] After analyzing a large amount of historical data and user behavior, those skilled in the art discovered a clear correlation between the display effect of advertising content and the user's interaction with advertising before, during, or after recharging, for users with a habit of interacting with ads. Specifically, the smoothness, completeness, and responsiveness of the advertising content directly affect the actual level of user interaction with the ad, which is precisely the core effect that advertising aims to achieve. Furthermore, under high device load, reducing GPU load can significantly improve the display effect of advertising content, indicating a quantifiable correlation between GPU load and ad display effectiveness.

[0047] Further research revealed that in high-load scenarios, the improvement in ad display performance does not increase infinitely with decreasing GPU load, but rather reaches a relative equilibrium. When the GPU load decreases to the level corresponding to this equilibrium, the ad display performance reaches its optimal or most balanced state under high-load conditions. This load level represents the critical GPU load that balances ad display performance and system performance while maintaining a high-load terminal. Based on this, those skilled in the art have proposed that by comparing the current GPU load with this critical GPU load, and without changing the objective load conditions of the terminal, the complexity of the ad content can be adjusted accordingly. This reduces the complexity of the ad content while indirectly reducing the demand for GPU resources, thereby improving the display performance. Furthermore, this adjustment process is based on data patterns and critical point determination, and will not compromise the fairness of ad delivery across multiple platforms.

[0048] In step S100, the high GPU load state is not determined subjectively, but rather based on pre-set judgment criteria according to different terminal device types, hardware performance, and historical operating data. By statistically analyzing the GPU load data of different devices under various usage scenarios, it is determined that when the GPU load of a terminal device exceeds a certain preset load level, it is prone to lag or unsmooth display during the loading of advertising content on the recharge page. Thus, the state exceeding this load level is judged as a high GPU load state.

[0049] In the specific implementation process, real-time operating data such as the core load utilization, video memory usage, and computation task queue length of the terminal device's GPU are acquired and combined with hardware monitoring information provided by the GPU driver to comprehensively calculate the current GPU load. When the calculated current GPU load exceeds the preset load standard for the corresponding device type, the terminal device is identified as being in a high GPU load state, thus providing a basis for determining and adjusting the complexity of the initial advertising content.

[0050] Furthermore, the multi-platform-based adaptive recharge scheduling method also includes the following steps:

[0051] Step S200: Call the preset database and filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one. Each sample includes the user's interaction behavior with the historical advertisements during the recharge stage.

[0052] The phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

[0053] In this embodiment of the invention, the preset database is not a newly created data source, but rather a data set naturally accumulated over a long period of operation by existing mobile phone top-up systems, advertising delivery systems, and terminal operation monitoring systems. The data acquisition method itself is a prior art. This invention does not change the data collection process of the existing system, but rather structures and utilizes the existing data. Specifically, the preset database includes at least terminal device operation data, advertising delivery record data, and user interaction behavior data during the top-up phase. The terminal device operation data reflects the GPU load of different devices under different usage states; the advertising delivery record data characterizes the advertising content's delivery platform, display format, and corresponding content complexity parameters; and the user interaction behavior data depicts the user's actual response to the advertising content during the top-up phase.

[0054] In step S200, the preset database is invoked to filter out historical samples that meet the criteria for subsequent interactive behavior analysis and determination of critical GPU load. The historical samples in this database are all derived from the behavioral data of real users in actual recharge scenarios. The data source is reliable and can truly reflect the objective relationship between terminal load status, advertising content complexity, and user interactive behavior.

[0055] During the historical sample screening process, ensuring that the content complexity of the historical advertisements received by users in each sample is consistent with the content complexity of the currently matched initial advertisement content is crucial to guaranteeing a clear benchmark for comparison during the analysis. Under a multi-platform advertising mechanism, the advertisement content that should have been displayed in the current recharge process is the initial advertisement content. Therefore, using the content complexity of this initial advertisement content as the analytical benchmark avoids introducing additional variables between advertisements of different complexities, making the impact of GPU load changes on user interaction behavior more controllable and comparable. Simultaneously, this approach ensures that the necessity for adjusting content complexity can be determined without affecting the original delivery order or platform fairness, thus meeting the fairness requirements of multi-platform advertising scenarios.

[0056] Furthermore, the requirement that each historical sample include user interaction with historical advertising content during the recharge phase is based on a targeted analysis of user behavior characteristics. This invention does not focus on all recharge users, but rather on user groups with a history of advertising interaction. For this type of user, there is a significant correlation between advertising display effectiveness and their interaction behavior, and their behavioral data is of higher reference value for analyzing the relationship between GPU load and advertising effectiveness. Compared to users who do not interact with ads at all, this user sample can more realistically reflect the actual impact of changes in advertising content display quality.

[0057] When screening historical samples, in addition to device type, operating system, network connection quality, and recharge type, further limitations can be made by considering factors such as terminal hardware performance level, screen resolution range, recharge page version number, or advertisement display area location. These conditions will all affect the rendering method of advertisement content and the user's visual experience to varying degrees. Therefore, taking them into account during the sample screening process helps to reduce the interference of non-GPU load factors on the analysis results.

[0058] Employing relatively strict screening criteria ensures sufficient comparability between historical samples and current recharge scenarios, avoiding the introduction of noisy data due to differences in equipment, network fluctuations, or business processes. This improves the accuracy of subsequent interactive behavior trend analysis and critical GPU load determination. Meanwhile, allowing for tolerance ranges in some conditions takes into account unavoidable minor differences in actual operating environments, such as different models of the same device type, different versions of the same operating system, or short-term fluctuations in network quality. By setting reasonable tolerance ranges, we can avoid statistical distortion caused by insufficient sample size while increasing the coverage of historical samples while maintaining overall consistency, making the analysis results more stable and reliable.

[0059] Furthermore, the multi-platform-based adaptive recharge scheduling method also includes the following steps:

[0060] Step S300: Calculate the level of interaction behavior for each sample and analyze whether the level of interaction behavior increases as the GPU load decreases and remains stable after reaching the inflection point; if so, set the GPU load corresponding to the inflection point as the critical GPU load.

[0061] The calculation process for the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and weighting the various interaction data to obtain a comprehensive interaction score.

[0062] In this embodiment of the invention, step S300 is used to further analyze the screened historical samples to quantify user advertising interaction under different GPU load conditions and verify whether changes in GPU load have a stable and predictable impact on advertising interaction behavior under specific user groups and terminal device conditions. The core purpose of this step is to determine whether there is a representative trend, thereby providing an objective basis for whether to adjust the complexity of advertising content in the future.

[0063] The level of interactive behavior is obtained by quantifying user interactions during the historical ad content delivery process. In practice, from the start of historical ad content delivery, various types of user interaction data related to the ad are continuously collected during the recharge phase, and the collected click count, dwell time, swipe depth, and swipe frequency are used as basic behavioral indicators. These various interaction data reflect the user's active attention to the ad content, the duration of sustained attention, and the depth and frequency of interaction, thus depicting the user's true response to the ad content from different dimensions.

[0064] When calculating the level of interactive behavior, various interactive data are weighted, and the weighted results are summarized to obtain a comprehensive interaction score. The weight of different interactive behaviors in the comprehensive interaction score can be preset according to business needs or historical statistical results. For example, in some scenarios, the weight of click behavior can be increased, and in other scenarios, the weight of dwell time or scrolling depth can be increased, so that the comprehensive interaction score can more accurately reflect the user's actual willingness to interact with the advertisement.

[0065] Besides the weighted calculation method mentioned above, the degree of interactive behavior can also be quantified in other ways. For example, an interaction scoring model can be built based on normalized multi-dimensional interaction indicators, or a statistical distribution model can be established based on historical behavior data to compare and analyze the average interaction level within different GPU load ranges. Regardless of the specific calculation method used, the essence is to transform user interaction behavior during ad display into comparable numerical indicators for subsequent trend analysis.

[0066] In step S300, the analysis of the level of interactive behavior is not conducted in isolation, but rather in conjunction with the GPU load of corresponding historical samples. The overall interaction score is compared and analyzed in descending order of GPU load. This analysis determines whether the level of interactive behavior shows an overall upward trend as the GPU load gradually decreases when the target user and their terminal device are under high load. This verifies the technical hypothesis proposed in step S100 that there is a correlation between ad display effectiveness and GPU load. Only when this trend is verified in historical samples will the subsequent content complexity-based adjustment process continue, to avoid ineffective or unnecessary ad adjustments without supporting data.

[0067] In further trend analysis, when it was found that the level of interactive behavior increased as GPU load decreased, and then stabilized after reaching a certain load level, the GPU load corresponding to this inflection point was set as the critical GPU load. This critical GPU load is used to characterize the state in which the ad display effect and terminal resource consumption are relatively balanced under high load conditions. The corresponding GPU load level can ensure the ad display effect while avoiding excessive impact on terminal performance.

[0068] The aforementioned trends align with the general patterns of terminal graphics rendering and user behavior. On one hand, when GPU load is high, rendering resources for ad content are limited, and screen stuttering or slow response directly reduces users' willingness to interact. As GPU load decreases, the smoothness and completeness of ad content improve, making it easier for users to engage effectively. On the other hand, once GPU load decreases to a certain level, the ad display effect has reached a relatively optimal state perceptible to users, and further reducing the load will not significantly improve user interaction, thus forming a trend of relatively stable interaction levels. By identifying this inflection point and using it as the critical GPU load, this invention can provide a stable and practically significant reference benchmark for adjusting the complexity of subsequent ad content in high-load scenarios.

[0069] Furthermore, the multi-platform-based adaptive recharge scheduling method also includes the following steps:

[0070] Step S400: Given that the current GPU load is greater than the critical GPU load, generate an adjustment coefficient based on the difference between the two, and adjust the content complexity corresponding to the initial advertisement content based on the adjustment coefficient.

[0071] Step S500: Select the advertising content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial advertising content, and push it to the target user's recharge page.

[0072] Specifically, Figure 2 A flowchart is shown to adjust the content complexity corresponding to the initial advertisement content.

[0073] Specifically, under the premise that the current GPU load is greater than the critical GPU load, an adjustment coefficient is generated based on the difference between the two, and the content complexity corresponding to the initial advertisement content is adjusted based on the adjustment coefficient. This includes the following steps:

[0074] Step S401: Determine whether the current GPU load is greater than the critical GPU load. If the current GPU load is greater than the critical GPU load, generate an adjustment coefficient based on the degree to which the current GPU load exceeds the critical GPU load.

[0075] Step S402: Call the preset correction model and adjust the content complexity of the initial advertisement content by combining the adjustment coefficient to obtain the adjusted content complexity.

[0076] The preset correction model is:

[0077] ;

[0078] in, This refers to the adjusted complexity of the content. This refers to the complexity of the initial advertisement content. This refers to the minimum allowable value for content complexity. This refers to the current GPU load. This refers to critical GPU load. This refers to the degree to which the current GPU load exceeds the critical GPU load. This refers to the preset control correction strength coefficient, and it satisfies... Greater than 0, This refers to the adjustment factor.

[0079] In this embodiment of the invention, step S400 is the core processing step of the entire technical solution. Its purpose is to make a basis-based adaptive adjustment of the content complexity of the advertisement through a quantitative method under the premise that the current objective operating state of the terminal device cannot be changed, so as to balance the advertising display effect, terminal operation stability and the fairness of multi-platform delivery in high GPU load scenarios.

[0080] When the current GPU load exceeds the critical GPU load, it indicates that the terminal device is already in a data-verified high-load range. Continuing to serve advertisements with the original content complexity within this range can easily lead to a decline in ad display effectiveness and negatively impact the user's interactive experience during the recharge phase. Based on this premise, this invention does not employ a fixed-amplitude or discrete-level adjustment method, but rather generates a corresponding adjustment coefficient based on the degree to which the current GPU load exceeds the critical GPU load. This adjustment coefficient reflects the degree of deviation of the current load state from the critical state. The greater the deviation, the weaker the terminal's capacity to handle the complexity of the advertisement content, and the greater the need to reduce the content complexity; a smaller deviation requires only minor adjustments. In this way, continuous, controllable complexity adjustments strongly correlated with the load state can be achieved, rather than simply and crudely replacing or blocking content.

[0081] The adjustment mechanism is based on the interactive behavior change pattern obtained in step S300. Specifically, within the high-load range, the advertising interaction effect improves as the GPU load decreases, and then stabilizes near the critical GPU load. Therefore, using the difference between the current GPU load and the critical GPU load as the basis for adjustment ensures that the direction of complexity adjustment aligns with the actual user behavior patterns, thereby improving the effectiveness and interpretability of the adjustment results.

[0082] In step S402, the content complexity corresponding to the initial advertisement content is adjusted by calling a preset correction model and combining it with the generated adjustment coefficients, resulting in the adjusted content complexity. It should be noted that this adjusted content complexity does not directly correspond to any specific advertisement content, nor does it constitute an independent substantive object. Essentially, it is an abstract parameter or level identifier used for subsequent matching judgments. The purpose of this parameter is to provide a clear and unified reference standard for the selection of the advertisement platform in step S500, enabling the system to compare and filter multiple candidate advertisement platforms based on the same complexity dimension.

[0083] By adopting the above approach, we can avoid the problem of simply reducing the complexity of ad content to the minimum under high load conditions, which would lead to frequent exposure of low-complexity ads. At the same time, we can reasonably select the most suitable ad content while still maintaining differentiation in complexity adjustment. This protects the interests of platforms that have invested more in ad content design and production, and maintains the overall fairness of the multi-platform placement mechanism.

[0084] When the current GPU load does not exceed the critical GPU load, it means that the terminal device is in a state where it can handle the current initial advertising content well. In this case, there is no need to adjust the complexity of the initial advertising content. The system directly keeps the initial advertising content unchanged and enters the subsequent advertising display process, thereby avoiding unnecessary adjustment operations.

[0085] In step S500, the system selects the advertising content that best matches the adjusted content complexity from several candidate advertising platforms. Specifically, each candidate advertising platform pre-labels the advertising content it provides with corresponding content complexity parameters. When advertising replacement is required, the system compares the matching degree between the content complexity of the advertising content on each platform and the adjusted content complexity, selects the advertising content with the smallest difference or that falls within the corresponding complexity range, and pushes it to the target user's recharge page. In this way, refined advertising scheduling based on terminal load status can be achieved without changing the original multi-platform polling or fair delivery mechanism.

[0086] The adjustment model employed in this invention is intuitive and effective. It normalizes the difference between the current GPU load and the critical GPU load, and combines this with preset control strength parameters to achieve a linear adjustment of the initial content complexity. This approach is easy to understand, implement, and debug. In practical applications, other calculation methods can also be used to achieve similar effects. For example, a piecewise function can be used to set different adjustment ranges in different load intervals, or a historical statistical model can be used to map load differences to corresponding complexity adjustment ratios. None of these methods affect the core idea of ​​this invention.

[0087] The overall implementation process of this invention is described below with specific numerical examples. Assume that a terminal device detects a current GPU load of 80 during the recharge phase, and the critical GPU load obtained based on historical sample analysis is 60, indicating that the current GPU load exceeds the critical GPU load by 20 units. Assume the initial ad content complexity is 100, the minimum allowable content complexity is 40, and the preset control strength parameter is 0.5. Then, based on the ratio between the current load exceeding the critical load, the initial content complexity is reduced proportionally, resulting in an adjusted content complexity of approximately 83. The system then searches for ad content with a content complexity close to 83.3 in the candidate ad platform for delivery, rather than directly reducing it to the minimum complexity, thereby maintaining fairness in ad delivery while ensuring improved ad display effects.

[0088] Through the above steps, this invention effectively addresses the core research problem raised in step S100: how to improve ad display effectiveness and increase the actual interaction level of users with ad-interaction habits under high GPU load scenarios without prolonging the recharge process or compromising the fairness of multi-platform ad delivery. This solution can dynamically adjust the ad content delivery strategy according to the terminal's operating status, exhibiting good adaptability and scalability. It is applicable to various terminal devices and ad delivery scenarios, possessing high practical application value and promising prospects for promotion.

[0089] Furthermore, Figure 3 An application architecture diagram of the system provided in an embodiment of the present invention is shown.

[0090] In another preferred embodiment of the present invention, an adaptive recharge scheduling system based on multiple platforms includes:

[0091] The load detection module 100 is used to determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the recharge page when the target user is identified to be using a terminal device with a high GPU load to recharge mobile phone credit.

[0092] The current GPU load of the terminal device is determined by acquiring real-time data on the core load utilization, memory usage, and computation task queue length of the terminal device's GPU, combined with hardware monitoring information provided by the GPU driver.

[0093] Furthermore, the multi-platform adaptive recharge scheduling system also includes:

[0094] The sample filtering module 200 is used to call a preset database to filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one, and each sample includes the user's interaction behavior with the historical advertisements during the recharge stage.

[0095] The phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

[0096] Furthermore, the multi-platform adaptive recharge scheduling system also includes:

[0097] The interactive behavior analysis module 300 is used to calculate the degree of interactive behavior for each sample and analyze whether the degree of interactive behavior increases as the GPU load decreases and remains stable after reaching a turning point; if so, the GPU load corresponding to the turning point is set as the critical GPU load.

[0098] The calculation process for the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and weighting the various interaction data to obtain a comprehensive interaction score.

[0099] Furthermore, the multi-platform adaptive recharge scheduling system also includes:

[0100] The complexity adjustment module 400 is used to generate an adjustment coefficient based on the difference between the current GPU load and the critical GPU load, and adjust the content complexity corresponding to the initial advertisement content based on the adjustment coefficient.

[0101] Furthermore, the multi-platform adaptive recharge scheduling system also includes:

[0102] The advertising platform selection module 500 is used to select the advertising content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial advertising content, and push it to the target user's recharge page.

[0103] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0104] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0105] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0107] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-platform adaptive recharge scheduling method, characterized in that, The method includes: When identifying a target user's behavior of topping up mobile phone credit using a terminal device with a high GPU load, determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the topping-up page. Call the preset database and filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one. Each sample includes the user's interaction behavior with the historical advertisements during the recharge process. Calculate the level of interaction behavior for each sample and analyze whether the level of interaction behavior increases as the GPU load decreases and remains stable after reaching a turning point; if so, set the GPU load corresponding to the turning point as the critical GPU load. Given that the current GPU load is greater than the critical GPU load, an adjustment coefficient is generated based on the difference between the two, and the content complexity of the initial ad content is adjusted based on the adjustment coefficient. Select the ad content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial ad content, and push it to the target user's recharge page.

2. The adaptive recharge scheduling method based on multiple platforms according to claim 1, characterized in that, The current GPU load of the terminal device is determined by acquiring real-time data on the core load utilization, memory usage, and computation task queue length of the terminal device's GPU, combined with hardware monitoring information provided by the GPU driver.

3. The adaptive recharge scheduling method based on multiple platforms according to claim 1, characterized in that, The phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

4. The multi-platform adaptive recharge scheduling method according to claim 1, characterized in that, The calculation process for the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and weighting the various interaction data to obtain a comprehensive interaction score.

5. The adaptive recharge scheduling method based on multiple platforms according to claim 1, characterized in that, Given that the current GPU load is greater than the critical GPU load, the steps to generate an adjustment coefficient based on the difference between the two and adjust the content complexity of the initial ad content based on the adjustment coefficient include: Determine if the current GPU load exceeds the critical GPU load. If the current GPU load exceeds the critical GPU load, generate an adjustment factor based on the degree to which the current GPU load exceeds the critical GPU load. The preset correction model is invoked, and the content complexity of the initial advertisement content is adjusted by combining the adjustment coefficients to obtain the adjusted content complexity.

6. The adaptive recharge scheduling method based on multiple platforms according to claim 5, characterized in that, The preset correction model is: ； in, This refers to the adjusted complexity of the content. This refers to the complexity of the initial advertisement content. This refers to the minimum allowable value for content complexity. This refers to the current GPU load. This refers to critical GPU load. This refers to the degree to which the current GPU load exceeds the critical GPU load. This refers to the preset control correction strength coefficient, and it satisfies... Greater than 0, This refers to the adjustment factor.

7. A multi-platform adaptive recharge scheduling system, characterized in that, The system includes: The load detection module is used to determine the current GPU load of the terminal device and the content complexity of the initial advertisement content matched for the recharge page when the target user is identified to be using a terminal device with a high GPU load to recharge mobile phone credit. The sample filtering module is used to call a preset database to filter out several historical samples that are consistent with the current recharge background conditions. The GPU load corresponding to the samples is different. The content complexity of the historical advertisements received by the user in each sample is consistent with the current one, and each sample includes the user's interaction behavior with the historical advertisements during the recharge process. The interactive behavior analysis module is used to calculate the level of interactive behavior for each sample and analyze whether the level of interactive behavior increases as the GPU load decreases and remains stable after reaching a turning point; if so, the GPU load corresponding to the turning point is set as the critical GPU load. The complexity adjustment module is used to generate an adjustment coefficient based on the difference between the current GPU load and the critical GPU load, and then adjust the content complexity of the initial advertisement content based on the adjustment coefficient. The advertising platform selection module is used to select the advertising content that best matches the adjusted content complexity from several candidate platforms, replace it with the initial advertising content, and push it to the target user's recharge page.

8. The multi-platform adaptive recharge scheduling system according to claim 7, characterized in that, The current GPU load of the terminal device is determined by acquiring real-time data on the core load utilization, memory usage, and computation task queue length of the terminal device's GPU, combined with hardware monitoring information provided by the GPU driver.

9. The multi-platform adaptive recharge scheduling system according to claim 7, characterized in that, The phrase "consistent with the current recharge background conditions" specifically means that the sample and the terminal device used by the current target user have the same device type, operating system, network connection quality, and recharge type within the same or allowable tolerance range.

10. The multi-platform adaptive recharge scheduling system according to claim 7, characterized in that, The calculation process for the degree of interactive behavior includes: collecting user interaction data with the advertisement from the beginning of the historical advertisement content delivery, including the number of clicks, dwell time, swipe depth and swipe frequency; and weighting the various interaction data to obtain a comprehensive interaction score.

Images