Traffic regulation method, system, computer device and storage medium

Abstract

The application discloses a traffic regulation method and system, a computer device and a storage medium, and comprises the following steps: obtaining a search commodity request of a user; recalling, rough sorting and fine sorting the searched commodity based on the search commodity request of the user; performing rearrangement calculation on the fine sorting calculation result, calculating traffic adjustment strength, efficiency factor and regulation factor in the rearrangement calculation process; and determining a traffic pushing target based on the rearrangement calculation result. The application enables the commodity platform to perform rearrangement on the commodity after the commodity experiences recalling, rough sorting and fine sorting based on the search request of the user, adds individualized weight and efficiency factor in the rearrangement calculation process, performs directional traffic pushing and individualized traffic regulation on the brand of the merchant by using the calculation result of the rearrangement, and can take into account the use efficiency of the user and reduce the unpleasant feeling of the user experience platform as much as possible.

Description

Technical Field

[0001] This invention relates to the field of e-commerce traffic distribution technology, and in particular to a traffic control method, system, computer device, and storage medium. Background Technology

[0002] After more than a decade of development, the e-commerce industry has seen many platforms eliminated, and those that survive continue to compete fiercely. For a platform to survive, it must constantly innovate to meet diverse market demands. Search and recommendation are two common scenarios for e-commerce platforms, both aiming to find products that meet users' individual needs. However, the cooperation between e-commerce platforms and various brands means protecting the interests of each brand. Typically, the process from user access to the final sorted product display involves several stages: recall, coarse ranking, and fine ranking. To interfere with the final ranking, a re-ranking is added after fine ranking. This re-ranking is usually a hard rule added to implement traffic allocation policies, but this often harms the user experience. In today's increasingly convenient internet environment, how e-commerce platforms can more accurately target traffic to brands with limited traffic without compromising the user experience has become a pressing technical problem that needs to be solved.

[0003] Therefore, there is a need to provide a traffic control method, system, computer equipment, and storage medium that can re-rank products based on user search requests after they have undergone recall, coarse ranking, and fine ranking. The re-ranking calculation results can be used to push targeted traffic to merchants' brands, while also taking into account user efficiency and minimizing user dissatisfaction with the platform. Summary of the Invention

[0004] This invention proposes a traffic control method, system, computer device, and storage medium, which can solve the technical problem of how to perform targeted traffic push for product brands without harming the user's platform experience.

[0005] To achieve the above objectives, the present invention provides a flow control method, comprising:

[0006] Obtain the user's product search request;

[0007] Based on the user's search request, the searched products are recalled, coarsely ranked, and finely ranked.

[0008] The fine-order calculation results are rearranged. During the rearrangement calculation, the flow control strength, efficiency factor and control factor are calculated. The rearrangement score is calculated based on the calculated flow control strength, efficiency factor and control factor.

[0009] The traffic push target is determined based on the rearrangement calculation results.

[0010] Furthermore, the process of recalling, coarsely ranking, and finely ranking the searched products based on the user's search product request includes:

[0011] Based on the user's search request for products, a recall operation is performed on the requested products;

[0012] Based on the user's search request for products, a coarse sorting operation is performed on the recalled products;

[0013] Based on the user's search request, perform a fine-ranking operation on the coarsely ranked products;

[0014] The fine sorting operation includes obtaining the fine sorting score, the maximum value of the fine sorting score, and the minimum value of the fine sorting score.

[0015] Furthermore, the calculation of the flow adjustment force during the rearrangement calculation process includes:

[0016] The multiple brands that require traffic control are grouped into multiple product pools;

[0017] Each cargo pool has its corresponding traffic target. Based on historical data, future data is predicted, and the overall target is broken down into traffic targets according to time periods.

[0018] Data is collected and statistically analyzed for each time period of the cargo pool to calculate the current demand for the cargo pool's flow, which is then used as the adjustment level.

[0019] The formula for calculating the adjustment level is:

[0020] AdjustLevel=1-actualUV / targetUV;

[0021] Among them, AdjustLevel represents the adjustment level, actualUV represents the actual traffic completed at the current moment, and targetUV represents the traffic target that should be completed at the current moment;

[0022] Get the adjustment level of the products in the three time periods preceding the current point in time;

[0023] The intensity of the flow control is determined based on the adjustment level.

[0024] The formula for calculating the flow regulation intensity is:

[0025] u(k) = u(k-1) + Δu(k);

[0026] △u(k)=Kp(err(k)-err(k-1))+Kierr(k)+Kd(err(k)-2*err(k-1)+err(k-2)));

[0027] Where u represents a function, k represents the current time, k-1 represents the previous time, k-2 represents the time before that, err represents the adjustment level, Kp is the proportional term (Proportion factor), Ki is the integral term (Integral factor), Kd is the differential term (Differential factor), Δu(k) is the increment value calculated by the formula at the current time, err(k) is the adjustment level at the current time, err(k-1) is the adjustment level at the previous time, and err(k-2) is the adjustment level at the time two moments before that.

[0028] The calculation results will be used as the numerical value for adjusting the flow rate.

[0029] Furthermore, during the rearrangement calculation process, the flow adjustment force, efficiency factor, and control factor are calculated, including:

[0030] In response to obtaining the refined score and the minimum value of the refined score;

[0031] The finely sorted score and its minimum value are calculated using the following formula:

[0032] PidadjustFactor=1+β*2*(log 10 ModelScore / ReqModelScoreMin / logFactor-0.5);

[0033] Wherein, PidadjustFactor represents the efficiency factor, ModelScore represents the fine ranking score of the product to be scored, and logFactor represents the control factor;

[0034] The calculation result is used as the value of the efficiency factor.

[0035] Furthermore, the calculation of the regulatory factor includes:

[0036] In response to the obtained maximum and minimum values ​​of the refined ranking score;

[0037] The maximum and minimum values ​​of the refined ranking score are calculated using the following formula:

[0038] logFactor = log 10 ReqModelScoreMax / ReqModelScoreMin ;

[0039] Wherein, logFactor represents the control factor, ReqModelScoreMin represents the minimum value of the refined ranking score, and ReqModelScoreMax represents the maximum value of the refined ranking score;

[0040] The calculation result is used as the numerical value of the regulatory factor.

[0041] Furthermore, determining the traffic push target based on the rearrangement calculation results includes:

[0042] In response to the obtained calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor, the calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor are calculated using the following formula:

[0043] FinalScore = ModelScore * 10 u(k)*logFactor*PidadjustFactor ;

[0044] Where: FinalScore represents the rearrangement score, ModelScore represents the fine-tuning score, logFactor represents the control factor, PidadjustFactor represents the efficiency factor, and 10 u(k)*logFactor*PidadjustFactor This indicates that the ranking results are weighted based on the degree of traffic adjustment, while also incorporating an efficiency factor;

[0045] The calculation result is used as the value for the rearrangement.

[0046] Furthermore, determining the traffic push target based on the rearrangement calculation results includes:

[0047] Sort the rearrangement calculation results in descending order, determine and output the maximum value of the rearrangement calculation results;

[0048] The product brand corresponding to the maximum value of the rearrangement calculation result is used as the traffic push target.

[0049] On the other hand, a system for identifying hot-swappable hard drives is provided, the system comprising:

[0050] The search module is used to obtain users' search requests for products;

[0051] The product filtering module is used to recall, coarsely rank, and finely rank the searched products based on the user's search product request.

[0052] The rearrangement calculation module is used to rearrange the fine-order calculation results and calculate the flow adjustment force, efficiency factor and control factor during the rearrangement calculation process.

[0053] The output decision module is used to determine the traffic push target based on the reordering calculation results.

[0054] In another aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0055] Obtain the user's product search request;

[0056] Based on the user's search request, the searched products are retrieved, coarsely ranked, and finely ranked.

[0057] The fine-order calculation results are rearranged, and the flow adjustment force, efficiency factor and control factor are calculated during the rearrangement calculation process;

[0058] The traffic push target is determined based on the rearrangement calculation results.

[0059] In another aspect, a computer-readable storage medium is provided, the computer-readable storage medium storing a program that, when executed by a processor, causes the processor to perform the following steps:

[0060] Obtain the user's product search request;

[0061] Based on the user's search request, the searched products are recalled, coarsely ranked, and finely ranked.

[0062] The fine-order calculation results are rearranged, and the flow adjustment force, efficiency factor and control factor are calculated during the rearrangement calculation process;

[0063] The traffic push target is determined based on the rearrangement calculation results.

[0064] The flow control method proposed in this application has the following beneficial effects:

[0065] The traffic control method proposed in this application obtains users' search requests for products, analyzes the characteristics of these requests, and then performs recall, coarse ranking, and fine ranking on the required products. After fine ranking, it obtains the fine ranking score, maximum fine ranking score, minimum fine ranking score, traffic target, and actual traffic situation for products that meet the characteristics. It then recalculates the product score for products that meet the characteristics through re-ranking. This re-ranking calculation includes determining the adjustment level for each product based on the traffic target and actual traffic situation, and then calculating the traffic control intensity based on the adjustment level. The overall strength is weighted on the fine ranking results, while efficiency factors are introduced to take into account efficiency. The value of the control factor is calculated by the maximum and minimum fine ranking scores. The value of the efficiency factor is calculated by the fine ranking score, the minimum fine ranking score, and the control factor value. Finally, based on the calculated traffic control strength, control factor value, efficiency factor value, and fine ranking score, the re-ranking score is calculated. The re-ranking score is used as the product score. The product scores are sorted in descending order, and the re-ranking score of the product is output. The product with the highest score is selected as the target of brand traffic push.

[0066] The above settings enable the product platform to re-rank products based on user search requests, after the products have undergone recall, coarse ranking, and fine ranking. Personalized weights and efficiency factors are added during the re-ranking calculation process. The re-ranking calculation results are used to target traffic to merchants' brands and personalize traffic control, while also taking into account user efficiency and minimizing user dissatisfaction with the platform. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is an application environment diagram of the flow control method provided in the embodiments of this application;

[0069] Figure 2 This is a flowchart of the flow control method provided in the embodiments of this application;

[0070] Figure 3 This is a detailed flowchart of the flow control method provided in the embodiments of this application;

[0071] Figure 4 This is a schematic diagram of the flow control system provided in the embodiments of this application;

[0072] Figure 5 This is a schematic diagram of the computer device provided in the embodiments of this application. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. 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.

[0074] Example 1

[0075] The flow control method provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. The user inputs search terms for product features into the server's search box. The server receives the user's search request, analyzes its features, and performs recall, coarse ranking, and fine ranking of the required products. After fine ranking, it obtains the fine ranking score, maximum fine ranking score, minimum fine ranking score, traffic target, and actual traffic situation for products matching the features. Then, it recalculates the product score for products matching the features through re-ranking. This re-ranking calculation includes calculating the adjustment level of the product matching the features based on the traffic target and actual traffic situation, and calculating the traffic control intensity of the product matching the features based on the adjustment level. The ranking results are weighted by the intensity of traffic adjustment, while also incorporating an efficiency factor. The adjustment factor is calculated using the maximum and minimum ranking scores, and the efficiency factor is calculated using the ranking score, minimum ranking score, and adjustment factor value. Finally, a re-ranking score is calculated based on the calculated traffic adjustment intensity, adjustment factor value, efficiency factor value, and ranking score. This re-ranking score is used as the product's rating, and the product ratings are sorted in descending order. The re-ranked product score is then output, and the product with the highest score is selected as the target for brand traffic push. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server 104 can be a standalone server or a server cluster consisting of multiple servers.

[0076] The traffic control method provided in this application enables product platforms to re-rank products based on user search requests after the products have undergone recall, coarse ranking, and fine ranking. The re-ranking calculation results are used to push targeted traffic to merchants' brands, while also taking into account user efficiency and minimizing user dissatisfaction with the platform.

[0077] In one embodiment, such as Figure 2 As shown, a flow control method is provided, including the following steps S1-S4:

[0078] S1: Obtain the user's search product request.

[0079] In this embodiment of the invention, the user inputs product features into the search box on the product platform's search page. Product features include the product's brand, reviews, color, rating, sales volume, and usage effects. Based on the user's search for product features, the platform analyzes the product features and searches for products matching the searched product features in the platform's product resource pool.

[0080] S2: Based on the user's search request, the searched products are recalled, coarsely ranked, and finely ranked.

[0081] In this embodiment of the invention, when a user inputs search product features into the search box, the platform will recall products from the product resource pool based on the user's search product features. The purpose of recall is to find as many relevant products as possible from the product library based on the product search features, thereby expanding the space for subsequent ranking layers. Afterwards, the recalled products will be coarsely ranked. The purpose of the coarse ranking stage is to select the top N compliant, relevant, and high-transaction-efficiency products from approximately tens of thousands of candidates, where N is usually in the thousands. Afterwards, the coarsely ranked products will be finely ranked. The purpose of the fine ranking is to select the top N compliant, relevant, and high-transaction-efficiency products from approximately thousands of candidates, where N is usually in the hundreds.

[0082] S3: Perform rearrangement calculation on the fine-order calculation results. During the rearrangement calculation process, calculate the flow adjustment strength, efficiency factor and control factor, and calculate the rearrangement score based on the calculated flow adjustment strength, efficiency factor and control factor.

[0083] In this embodiment of the invention, after the selected compliant products are finely ranked, in order to score the selected products more accurately, or because more complex models cannot be implemented, or because the fine ranking score cannot meet special overall ranking requirements, the compliant products will be re-ranked. Common approaches to re-ranking calculation include: reducing ranking candidates, using models with higher computational complexity, such as using long sequence modeling, or ranking modeling with complex network structures; increasing link effectiveness, such as real-time modeling, real-time feature modeling, etc.; multi-objective ensemble, manual weighting, or automatic learning of multiple objective weights; in addition, there is personalized weight learning, such as using clicks or transactions as objectives, so that different users or search terms have different ensemble weights. This embodiment uses personalized weight learning. During the calculation process, the traffic control intensity is calculated based on the traffic target and the current actual traffic situation. The control factor is calculated based on the maximum and minimum values ​​of the fine ranking score. The efficiency factor is calculated based on the fine ranking score and the minimum value of the fine ranking score. Finally, the re-ranking score is calculated based on the calculated traffic control intensity, control factor, and efficiency factor.

[0084] S4: Determine the traffic push target based on the rearrangement calculation results.

[0085] In this embodiment of the invention, after the selected compliant products are finely ranked, in order to score the selected products more accurately, the products after fine ranking will be re-ranked and scored. The result of the re-ranking and scoring is the final product score. The platform will sort the product scores in descending order and take the score represented by the maximum product score as the brand traffic target for targeted push by the platform.

[0086] The above-mentioned flow control method, by combining the technical features mentioned above and making reasonable derivations, achieves the beneficial effect of solving the technical problems raised in the background art.

[0087] Example 2

[0088] This embodiment 2 includes the features of the above embodiments, such as Figure 2 , Figure 3 As shown, this embodiment provides a traffic control method, including: step 100 recalling requested products according to the user's search product request; performing coarse ranking on the recalled products according to the user's search product request; and performing fine ranking on the coarsely ranked products according to the user's search product request; the fine ranking includes obtaining a fine ranking score, a maximum value of the fine ranking score, and a minimum value of the fine ranking score.

[0089] It is understood that, in this embodiment of the invention, after the platform server obtains the user's product search request,

[0090] The analysis identifies the characteristics of user search requests and operates on the required products in the product resource pool. The operation steps include: First, based on the user's search request, recalling the requested products. Recall involves finding matching products based on the user's request. The recall architecture often employs a multi-path recall architecture, where the results from multiple paths are merged and truncated to obtain the final recall result, which is then sent downstream. The most basic search recall is also text recall based on an inverted index. Other classic recall methods include structured recall, personalized recall, and vector recall. The purpose of recall is to find as many relevant products as possible from the product database based on product search characteristics, expanding the space for subsequent ranking layers. Second, based on the user's search request, coarsely ranking the recalled products. Coarse ranking involves roughly sorting the found products. The ranking model is relatively simple and can handle a large number of products, up to 20,000. However, due to computational performance considerations, the choice of coarse ranking model is limited; the simplest is the most efficient. Classic ranking methods include static product ranking, linear regression (LR), and the DSSM dual-tower model. There are also many advanced optimization models. The purpose of coarse ranking is to select the top N compliant, relevant, and efficient products from approximately tens of thousands of candidates. Based on the user's search request, fine ranking further sorts the coarse-ranked products. This fine ranking model is more complex and can handle a smaller number of products, approximately 3000 at a time. With the number of candidates decreasing to the thousands or hundreds, more complex solutions can be chosen in the model design. Currently, deep learning models are predominantly used. Commonly used deep learning ranking models include basic DNN ranking, Dot cross-referencing, enhanced memory features, and the introduction of FM cross-referencing. The purpose of fine ranking is to select the top N compliant, relevant, and efficient products from approximately thousands of candidates.

[0091] After the product recall, coarse ranking, and fine ranking processes, the server will receive the fine ranking calculation results for the candidate compliant products. After receiving the fine ranking calculation results for the candidate compliant products, the server will obtain the fine ranking score, the maximum fine ranking score, and the minimum fine ranking score for the candidate compliant products for subsequent re-ranking calculations.

[0092] This embodiment provides a traffic control method, including: step 200 dividing multiple brands that need traffic control into multiple pools; each pool has its corresponding traffic target, predicting future data based on historical data, and breaking down the overall target into traffic targets according to time periods; collecting and collecting data for each time period of the pool, calculating the current traffic demand of the pool, and using it as the adjustment level;

[0093] The formula for calculating the adjustment level is:

[0094] AdjustLevel=1-actualUV / targetUV;

[0095] Among them, AdjustLevel represents the adjustment level, actualUV represents the actual traffic completed at the current moment, and targetUV represents the traffic target that should be completed at the current moment;

[0096] Obtain the adjustment level of the products in the three time periods preceding the current time point; calculate and determine the traffic control intensity based on the adjustment level;

[0097] The formula for calculating the flow regulation intensity is:

[0098] Incremental PID, output is u(k-1) + Δu(k);

[0099] △u(k)=Kp(err(k)-err(k-1))+Kierr(k)+Kd(err(k)-2*err(k-1)+err(k-2)));

[0100] Where u represents a function, k represents the current time, k-1 represents the previous time, k-2 represents the time before that, err represents the adjustment level, Kp is the proportional term (Proportion factor), Ki is the integral term (Integral factor), Kd is the differential term (Differential factor), Δu(k) is the increment value calculated by the formula at the current time, err(k) is the adjustment level at the current time, err(k-1) is the adjustment level at the previous time, and err(k-2) is the adjustment level at the time two moments before that.

[0101] The calculation result is used as the numerical value of the flow rate adjustment force.

[0102] It is understood that, in this embodiment of the invention, after the platform server obtains the user's product search request,

[0103] The analysis reveals the characteristics of user search requests for products and relies on recalling, coarsely ranking, and finely ranking the required products in the product resource pool. After fine ranking, in order to score the selected products more accurately, or because more complex models cannot be implemented, or because the fine ranking score cannot meet special overall ranking requirements, the platform server will recalculate the ranking based on the fine ranking score, maximum fine ranking score, and minimum fine ranking score of the compliant products. In this embodiment, personalized weight learning is used. During the calculation process, the traffic control intensity is calculated based on the traffic target and the current actual traffic situation.

[0104] The platform server groups numerous brands requiring traffic control into individual product pools. Each product pool has its own corresponding traffic target. Based on historical data, future data is predicted, and the overall target is broken down into different traffic targets for different time periods. `targetUV` represents the traffic target for each product pool at different time periods. Data is also collected and statistically analyzed at different time periods, which is called the actual traffic completion status, represented by `actualUV`. Then, using the traffic target and actual traffic completion data, the current traffic demand of the product pool is calculated, which we call the adjustment level. The platform server obtains the adjustment level of the products in the three time periods before the current time point. These three time periods are all 5 minutes. The traffic adjustment level is calculated every 5 minutes at the current time point. Finally, the traffic control intensity is determined based on the adjustment level.

[0105] The formula for adjusting the level is as follows:

[0106] AdjustLevel=1-actualUV / targetUV;

[0107] The formula for calculating the intensity of traffic control is:

[0108] Incremental PID, the output is u(k) = u(k-1) + Δu(k);

[0109] △u(k)=Kp(err(k)-err(k-1))+Kierr(k)+Kd(err(k)-2*err(k-1)+err(k-2)));

[0110] Where u represents a function, k represents the current time, k-1 represents the previous time, k-2 represents the time before that, err represents the adjustment level, Kp is the proportional term (Proportion factor), Ki is the integral term (Integral factor), Kd is the differential term (Differential factor), Δu(k) is the increment value calculated by the formula at the current time, err(k) is the adjustment level at the current time, err(k-1) is the adjustment level at the previous time, and err(k-2) is the adjustment level at the time two moments before that.

[0111] The embodiment provides a flow control method, including: step 300, in response to obtaining the fine-scale score and the minimum value of the fine-scale score, calculating the value of the efficiency factor.

[0112] Efficiency factor calculation formula:

[0113] PidadjustFactor=1+β*2*(log 10 ModelScore / ReqModelScoreMin / logFactor-0.5);

[0114] Wherein, PidadjustFactor represents the efficiency factor, ModelScore represents the refined ranking score of the product to be scored, and logFactor represents the control factor.

[0115] It is understood that, in this embodiment of the invention, after the platform server obtains the user's product search request,

[0116] The analysis identifies the characteristics of user search requests and performs recall, coarse ranking, and fine ranking of the required products in the product resource pool. After fine ranking, to more accurately score the selected products, or because more complex models cannot be implemented, or because the fine ranking score cannot meet specific overall ranking requirements, the platform server will recalculate the ranking based on the fine ranking score, maximum fine ranking score, and minimum fine ranking score of the compliant products. During the recalculation process, an efficiency factor is calculated based on the fine ranking score and the minimum fine ranking score. The purpose of introducing the efficiency factor is to shorten the calculation time during the recalculation process and minimize negative user experiences. The efficiency factor is automatically calculated by the computer recalculation module using the fine ranking score and the minimum fine ranking score obtained from the user's search requests.

[0117] Efficiency factor calculation formula:

[0118] PidadjustFactor=1+β*2*(log 10 ModelScore / ReqModelScoreMin / logFactor-0.5);

[0119] Here, PidadjustFactor represents the efficiency factor, ModelScore represents the refined ranking score of the product to be scored, and logFactor represents the adjustment factor. Hyperparameters can be adjusted using the β value. A larger β value places greater emphasis on efficiency, but may result in the brand's traffic targets not being met. A smaller β value means making the greatest effort to achieve the brand's traffic targets. Generally, a β value of 0.3 is recommended, and then adjustments can be made based on actual conditions.

[0120] The embodiment provides a flow control method, including: step 400, obtaining the maximum value and minimum value of the fine-scale distribution, and calculating the value of the control factor.

[0121] Formula for calculating regulatory factors:

[0122] logFactor = log 10 ReqModelScoreMax / ReqModelScoreMin ;

[0123] Wherein, logFactor represents the control factor, ReqModelScoreMin represents the minimum value of the refined ranking score, and ReqModelScoreMax represents the maximum value of the refined ranking score.

[0124] Understandably, in this embodiment of the invention, after obtaining a user's product search request, the platform server analyzes the characteristics of the user's product search request and performs recall, coarse ranking, and fine ranking operations on the required products in the product resource pool. After fine ranking, in order to score the selected products more accurately, or because more complex models cannot be implemented, or because the fine ranking score cannot meet special overall ranking requirements, the platform server will perform a re-ranking calculation based on the fine ranking score, the maximum fine ranking score, and the minimum fine ranking score of the obtained compliant products. During the re-ranking calculation, the platform server will calculate the value of the adjustment factor based on the maximum and minimum fine ranking scores. The adjustment factor is one of the terms in the formula for weighting the fine ranking score.

[0125] Formula for calculating regulatory factors:

[0126] logFactor = log 10 ReqModelScoreMax / ReqModelScoreMin ;

[0127] Wherein, logFactor represents the control factor, ReqModelScoreMin represents the minimum fine ranking score of compliant products, and ReqModelScoreMax represents the maximum fine ranking score of compliant products. The calculation process is to take the logarithm of the ratio of the maximum fine ranking score to the minimum fine ranking score for each request. The purpose of taking the logarithm is to reduce dimensionality.

[0128] The embodiment provides a traffic control method, including: step 500, in response to the obtained calculated traffic control intensity, efficiency factor value and control factor value, calculating the calculated traffic control intensity, efficiency factor value and control factor value to obtain the re-ranking score of compliant products calculated by fine ranking.

[0129] Rearrangement calculation formula:

[0130] FinalScore = ModelScore * 10 u(k)*logFactor*PidadjustFactor ;

[0131] Where: FinalScore represents the rearrangement score, ModelScore represents the fine-tuning score, logFactor represents the control factor, PidadjustFactor represents the efficiency factor, and 10 u(k)*logFactor*PidadjustFactor This indicates that the ranking results are weighted based on the intensity of traffic control, while also incorporating an efficiency factor.

[0132] It is understood that in this embodiment of the invention, after obtaining a user's product search request, the platform server analyzes the characteristics of the user's product search request and performs recall, coarse ranking, and fine ranking operations on the required products in the product resource pool. After fine ranking, in order to score the selected products more accurately, or because more complex models cannot be implemented, or because the fine ranking score cannot meet special overall ranking requirements, the platform server will recalculate the fine ranking score, the maximum fine ranking score, and the minimum fine ranking score of the obtained compliant products. During the reranking calculation process, the platform server will recalculate the fine ranking score based on the obtained traffic control intensity, control factor value, and efficiency factor value, and perform reranking processing. Reranking is a weighted processing of the fine ranking result, while also taking efficiency into account by introducing PidadjustFactor as an efficiency factor. This increases efficiency factor control, and the impact on user experience varies depending on the traffic requested by different users. At the same time, the fine ranking result is weighted to obtain the reranking result, which can also be used to achieve the purpose of targeted traffic push to merchants' brands.

[0133] Rearrangement calculation formula:

[0134] FinalScore = ModelScore * 10 u(k)*logFactor*PidadjustFactor ;

[0135] Where: FinalScore represents the rearrangement score, ModelScore represents the fine-tuning score, logFactor represents the control factor, PidadjustFactor represents the efficiency factor, and 10 u(k)*logFactor*PidadjustFactor This indicates that the ranking results are weighted based on PID, and an efficiency factor is introduced.

[0136] The embodiment provides a traffic control method, including: step 600 sorting the rearrangement calculation results in descending order, determining and outputting the maximum value of the rearrangement calculation results; and using the product brand corresponding to the maximum value of the rearrangement calculation results as the traffic push target.

[0137] Understandably, in this embodiment of the invention, after obtaining a user's product search request, the platform server analyzes the characteristics of the user's product search request and performs recall, coarse ranking, and fine ranking operations on the required products in the product resource pool. After fine ranking, in order to score the selected products more accurately, or because more complex models cannot be implemented, or because the fine ranking score cannot meet special overall ranking requirements, the platform server will recalculate the fine ranking score, the maximum fine ranking score, and the minimum fine ranking score of the compliant products. During the reranking calculation process, the platform server will recalculate the fine ranking score based on the obtained traffic control intensity, control factor value, and efficiency factor value, and perform reranking processing. The platform server will sort the reranking scores in descending order based on the reranking calculation results and output the maximum value of the reranking results. At the same time, based on the maximum value of the output reranking calculation results, the product brand corresponding to the maximum value of the reranking calculation results will be determined as the traffic push target. Traffic will be pushed to the product brands according to the ranking results of the reranking scores to achieve the purpose of targeted traffic push to the merchant's brand.

[0138] Example 3

[0139] This embodiment includes the features of the above embodiments. This embodiment provides a flow control system, such as... Figure 4 As shown, the system includes: a search module 1, a product filtering module 2, a rearrangement calculation module 3, and an output determination module 4.

[0140] The search module 1 is used to acquire the user's search product request; the product filtering module 2 is used to recall, coarsely rank, and finely rank the searched products based on the user's search product request; the re-ranking calculation module 3 is used to re-rank the fine ranking calculation result, and calculate the traffic control strength, efficiency factor, and control factor during the re-ranking calculation process; the output judgment module 4 is used to determine the traffic push target based on the re-ranking calculation result.

[0141] The above embodiments only illustrate the structural relationships and components between the various unit modules. For practical applications, please refer to the execution methods of each unit module. Figure 3 The method shown will not be elaborated here.

[0142] The search module 1 is used to acquire users' search requests for products. The search module 1 is displayed on the user's screen, where the user can input various characteristics of the products they want to search for. These characteristics include the product's brand, reviews, color, rating, sales volume, and effectiveness. The platform server analyzes the user's input product characteristics and filters them from the platform's product resource pool based on these characteristics.

[0143] The product filtering module 2 is used to recall, coarsely rank, and finely rank the searched products based on the user's search request. The platform server analyzes the product characteristics input by the user and uses a calculation model to filter products that match the characteristics from the platform's product resource pool. The filtering steps include: recalling the requested products based on the user's search request, the purpose of which is to find as many relevant products as possible from the product library based on the product search characteristics, expanding the space for subsequent ranking layers; coarsely ranking the recalled products based on the user's search request, the purpose of which is to filter out the top N compliant, relevant, and high-transaction-efficiency products from approximately tens of thousands of candidates; and finely ranking the coarsely ranked products based on the user's search request, the purpose of which is to filter out the top N compliant, relevant, and high-transaction-efficiency products from approximately thousands of candidates. The effect of the filtering module 2 is to filter out products that match the user's search characteristics based on the user's search request for the user to choose from.

[0144] The re-ranking calculation module 3 is used to re-rank the results of the fine-ranking calculation. During this re-ranking process, it calculates the traffic control intensity, efficiency factor, and control factor. The platform server, based on the analysis of the product characteristics input by the user, uses a calculation model to filter products that meet the characteristics from the platform's product resource pool. The filtering steps include recalling, coarse-ranking, and fine-ranking the requested products according to the user's search request. Simultaneously, a fine-ranking score, a maximum fine-ranking score, and a minimum fine-ranking score are obtained. After fine-ranking, the traffic target and actual traffic situation of compliant products are obtained. The adjustment level of compliant products can be obtained from the traffic target and actual traffic situation. The traffic control intensity of compliant products can be calculated from the adjustment level. Then, the value of the introduced efficiency factor is calculated by obtaining the fine-ranking score and the minimum fine-ranking score of the compliant products. The value of the control factor is also calculated by obtaining the fine-ranking score and the minimum fine-ranking score. Using the obtained traffic control intensity, efficiency factor, and control factor, the platform server can calculate the re-ranking score of compliant products. The re-ranking score of compliant products is a weighted score calculated by introducing the efficiency factor.

[0145] Output determination module 4 is used to determine the traffic push target based on the re-ranking calculation result. The platform server analyzes the product characteristics input by the user and uses a calculation model to filter products that meet the characteristics from the platform's product resource pool. The filtering steps include recalling, coarsely ranking, and finely ranking the requested products based on the user's search request. After fine ranking, the platform server re-ranks the compliant products and uses the re-ranking score as the final score for compliant products, outputting the re-ranking calculation result. The re-ranking scores are then sorted in descending order. Simultaneously, based on the maximum value of the output re-ranking calculation result, the product brand corresponding to the maximum value of the re-ranking calculation result is determined as the traffic push target.

[0146] Specific limitations regarding the flow control system can be found in the method limitations section above, and will not be repeated here. Each module in the aforementioned flow control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0147] It should be understood that, although Figure 2 , Figure 3 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 3 At least some of the steps in the process 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 executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0148] Example 4

[0149] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a method for recognizing hot-swappable hard drives.

[0150] This computer device can be a terminal, and its internal structure diagram can be as follows: Figure 5As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for recognizing hot-swappable hard drives. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0151] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0152] Obtain the user's product search request;

[0153] Based on the user's search request, the searched products are recalled, coarsely ranked, and finely ranked.

[0154] The fine-order calculation results are rearranged. During the rearrangement calculation, the flow control strength, efficiency factor and control factor are calculated. The rearrangement score is calculated based on the calculated flow control strength, efficiency factor and control factor.

[0155] The traffic push target is determined based on the rearrangement calculation results.

[0156] In one implementation, the processor also performs the following steps when executing a computer program:

[0157] Based on the user's search request for products, recall the requested products;

[0158] Based on the user's search request for products, the recalled products are coarsely ranked;

[0159] Based on the user's search request, the coarsely ranked products are then finely ranked.

[0160] The fine ranking includes obtaining the fine ranking score, the maximum value of the fine ranking score, and the minimum value of the fine ranking score.

[0161] In one implementation, the processor also performs the following steps when executing a computer program:

[0162] The multiple brands that require traffic control are grouped into multiple product pools;

[0163] Each cargo pool has its corresponding traffic target. Based on historical data, future data is predicted, and the overall target is broken down into traffic targets according to time periods.

[0164] Data is collected and statistically analyzed for each time period of the cargo pool to calculate the current demand for the cargo pool's flow, which is then used as the adjustment level.

[0165] The formula for calculating the adjustment level is:

[0166] AdjustLevel=1-actualUV / targetUV;

[0167] Among them, AdjustLevel represents the adjustment level, actualUV represents the actual traffic completed at the current moment, and targetUV represents the traffic target that should be completed at the current moment;

[0168] Get the adjustment level of the products in the three time periods preceding the current point in time;

[0169] The intensity of the flow control is determined based on the adjustment level.

[0170] The formula for calculating the flow regulation intensity is:

[0171] Incremental PID, output is u(k-1) + Δu(k);

[0172] △u(k)=Kp(err(k)-err(k-1))+Kierr(k)+Kd(err(k)-2*err(k-1)+err(k-2)));

[0173] Where u represents a function, k represents the current time, k-1 represents the previous time, k-2 represents the time before that, err represents the adjustment level, Kp is the proportional term (Proportion factor), Ki is the integral term (Integral factor), Kd is the differential term (Differential factor), Δu(k) is the increment value calculated by the formula at the current time, err(k) is the adjustment level at the current time, err(k-1) is the adjustment level at the previous time, and err(k-2) is the adjustment level at the time two moments before that.

[0174] The calculation result is used as the numerical value of the flow regulation intensity.

[0175] In one implementation, the processor also performs the following steps when executing a computer program:

[0176] In response to obtaining the refined score and the minimum value of the refined score;

[0177] The finely sorted score and its minimum value are calculated using the following formula:

[0178] PidadjustFactor=1+β*2*(log 10 ModelScore / ReqModelScoreMin / logFactor-0.5);

[0179] Wherein, PidadjustFactor represents the efficiency factor, ModelScore represents the fine ranking score of the product to be scored, and logFactor represents the control factor;

[0180] The calculation result is used as the value of the efficiency factor.

[0181] In one implementation, the processor also performs the following steps when executing a computer program:

[0182] In response to the obtained maximum and minimum values ​​of the refined ranking score;

[0183] The maximum and minimum values ​​of the refined ranking score are calculated using the following formula:

[0184] logFactor = log 10 ReqModelScoreMax / ReqModelScoreMin ;

[0185] Wherein, logFactor represents the control factor, ReqModelScoreMin represents the minimum value of the refined ranking score, and ReqModelScoreMax represents the maximum value of the refined ranking score;

[0186] The calculation result is used as the numerical value of the regulatory factor.

[0187] In one implementation, the processor also performs the following steps when executing a computer program:

[0188] In response to the obtained calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor, the calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor are calculated using the following formula:

[0189] FinalScore = ModelScore * 10 PID*logFactor*PidadjustFactor ;

[0190] Where: FinalScore represents the rearrangement score, ModelScore represents the fine-tuning score, logFactor represents the control factor, PidadjustFactor represents the efficiency factor, and 10 PID*logFactor*PidadjustFactorThis indicates that the ranking results are weighted based on the intensity of traffic control, while also incorporating an efficiency factor;

[0191] The calculation result is used as the value for the rearrangement.

[0192] In one implementation, the processor also performs the following steps when executing a computer program:

[0193] Sort the rearrangement calculation results in descending order, determine and output the maximum value of the rearrangement calculation results;

[0194] The product brand corresponding to the maximum value of the rearrangement calculation result is used as the traffic push target.

[0195] Example 5

[0196] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it performs the following steps:

[0197] Obtain the user's product search request;

[0198] Based on the user's search request, the searched products are recalled, coarsely ranked, and finely ranked.

[0199] The fine-order calculation results are rearranged. During the rearrangement calculation, the flow control strength, efficiency factor and control factor are calculated. The rearrangement score is calculated based on the calculated flow control strength, efficiency factor and control factor.

[0200] The traffic push target is determined based on the rearrangement calculation results.

[0201] The traffic control method proposed in this application obtains users' search requests for products, analyzes the characteristics of these requests, and then performs recall, coarse ranking, and fine ranking on the required products. After fine ranking, it obtains the fine ranking score, maximum fine ranking score, minimum fine ranking score, traffic target, and actual traffic situation for products that meet the characteristics. It then recalculates the product score for products that meet the characteristics through re-ranking. This re-ranking calculation includes determining the adjustment level for each product based on the traffic target and actual traffic situation, and then calculating the traffic control intensity based on the adjustment level. The control intensity is weighted on the fine ranking results, while efficiency is also taken into account by introducing an efficiency factor. The value of the control factor is calculated by the maximum and minimum fine ranking scores. The value of the efficiency factor is calculated by the fine ranking score, the minimum fine ranking score, and the control factor value. Finally, based on the calculated traffic control intensity, control factor value, efficiency factor value, and fine ranking score, the re-ranking score is calculated. The re-ranking score is used as the product score, and the product scores are sorted in descending order. The re-ranking score of the product is output, and the product with the highest score is selected as the target of brand traffic push.

[0202] The above settings enable the product platform to re-rank products based on user search requests, after the products have undergone recall, coarse ranking, and fine ranking. Personalized weights and efficiency factors are added during the re-ranking calculation process. The re-ranking calculation results are used to target traffic to merchants' brands and personalize traffic control, while also taking into account user efficiency and minimizing user dissatisfaction with the platform.

[0203] 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 computer 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.

[0204] 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.

[0205] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A flow control method, characterized in that, The method includes: Obtain the user's product search request; Based on the user's search request, the searched products are sequentially recalled, coarsely ranked, and finely ranked. The fine-ranking results are rearranged, and during the rearrangement calculation, the flow control strength, efficiency factor and control factor are calculated. The rearrangement score is calculated based on the calculated flow control strength, efficiency factor and control factor. Traffic delivery targets are determined based on reordering. The step of recalling, coarsely ranking, and finely ranking the searched products based on the user's search product request includes: Based on the user's search request for products, a recall operation is performed on the requested products; Based on the user's search request for products, a coarse sorting operation is performed on the recalled products; Based on the user's search request, perform a fine-ranking operation on the coarsely ranked products; The fine sorting operation includes obtaining the fine sorting score, the maximum value of the fine sorting score, and the minimum value of the fine sorting score; The calculation of flow regulation intensity during the rearrangement calculation process includes: The multiple brands that require traffic control are grouped into multiple product pools; Each cargo pool has its corresponding traffic target. Based on historical data, future data is predicted, and the overall target is broken down into traffic targets according to time periods. Data is collected and analyzed for each time period in the cargo pool to calculate the adjustment level. The formula for calculating the adjustment level is: AdjustLevel=1 - actualUV / targetUV; Among them, AdjustLevel represents the adjustment level, actualUV represents the actual traffic completed at the current moment, and targetUV represents the traffic target that should be completed at the current moment; Get the adjustment level of the products in the three time periods preceding the current point in time; The intensity of the flow control is determined based on the adjustment level. The formula for calculating the flow regulation intensity is: u(k) = u(k-1) + Δu(k); △u(k)=Kp(err(k)-err(k-1))+Kierr(k)+ Kd(err(k)-2*err(k-1)+ err(k-2))); Where u represents the function, k represents the current time, k-1 represents the previous time, k-2 represents the time before that, err represents the adjustment level, Kp is the proportional term (Proportion factor), Ki is the integral term (Integral factor), Kd is the differential term (Differential factor), Δu(k) is the increment value calculated by the formula at the current time, err(k) is the adjustment level at the current time, err(k-1) is the adjustment level at the previous time, and err(k-2) is the adjustment level at the time before that. The calculation result will be used as the numerical value of the flow regulation intensity; The flow regulation intensity, efficiency factor, and regulation factor are calculated during the rearrangement calculation process, including: In response to obtaining the refined score and the minimum value of the refined score; The finely sorted score and its minimum value are calculated using the following formula: PidadjustFactor=1+β*2*(log 10 ModelScore / Req ModelScoreMin / logFactor-0.5); Among them, PidadjustFactor represents the efficiency factor, ModelScore represents the fine ranking score of the product to be scored, logFactor represents the adjustment factor, and the β value can be adjusted according to the actual situation. The calculation result is used as the value of the efficiency factor; The calculated regulatory factors include: In response to the obtained maximum and minimum values ​​of the refined ranking score; The maximum and minimum values ​​of the refined ranking score are calculated using the following formula: logFactor= log 10 Req ModelScoreMax / Req ModelScoreMin ; Wherein, logFactor represents the control factor, ReqModelScoreMin represents the minimum value of the refined ranking score, and ReqModelScoreMax represents the maximum value of the refined ranking score; The calculation result is used as the numerical value of the regulatory factor.

2. The flow control method according to claim 1, characterized in that, Determining traffic delivery targets based on rearrangement calculation results includes: In response to the obtained calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor, the calculated values ​​of the flow regulation intensity, the efficiency factor, and the regulation factor are calculated using the following formula: FinalScore=ModelScore*10 u(k)*logFactor*PidadjustFactor ; Where: FinalScore represents the rearrangement score, ModelScore represents the fine-tuning score, logFactor represents the control factor, PidadjustFactor represents the efficiency factor, and 10 u(k)*logFactor*PidadjustFactor This indicates that the ranking results are weighted based on the intensity of traffic control, while also incorporating an efficiency factor; The calculation result is used as the value for the rearrangement.

3. The flow control method according to claim 1, characterized in that, Determining traffic delivery targets based on rearrangement calculation results includes: Sort the rearrangement calculation results in descending order, determine and output the maximum value of the rearrangement calculation results; The product brand corresponding to the maximum value of the rearrangement calculation result is used as the traffic push target.

4. A flow control system for implementing the flow control method according to any one of claims 1-3, characterized in that, The system includes: The search module is used to obtain users' search requests for products; The product filtering module is used to recall, coarsely rank, and finely rank the searched products based on the user's search product request. The rearrangement calculation module is used to rearrange the fine-order calculation results and calculate the flow regulation intensity, efficiency factor and regulation factor during the rearrangement calculation process. The output decision module is used to determine the traffic push target based on the reordering calculation results.

5. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

Images