Method and apparatus for processing service data, device, and medium
By obtaining the indicator sampling values and recommendation constraints of the services to be recommended, and weighted summing is performed to determine the traffic allocation ratio, the problems of traffic concentration and low utilization are solved, and more reasonable and efficient traffic allocation is achieved.
Patent Information
- Application Number
- PCT/CN2024/085342
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-22
- Filing Date
- 2024-04-01
- Publication Date
- 2024-11-28
AI Technical Summary
In the business recommendation scenario, it is difficult for the existing technology to effectively allocate traffic, resulting in excessive traffic concentration on certain services, insufficient traffic that is not saturated, resulting in low traffic utilization.
By obtaining the index sampling values of the services to be recommended, weighted summing is performed based on the proportions that meet and do not meet the recommendation constraints, the traffic allocation ratio of each service is determined to ensure reasonable traffic allocation.
The traffic utilization rate of each service to be recommended is improved to ensure the rationality and efficiency of traffic allocation.
Smart Images

Figure CN2024085342_28112024_PF_FP_ABST
Abstract
Description
Business data processing method, device, equipment and medium
[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on May 22, 2023, with application number 202310583288.9 and application name “Business Data Processing Method, Device, Equipment and Medium”, the entire contents of which are incorporated by reference into this application. Technical Field
[0002] The present application relates to the field of Internet technology, and in particular to a business data processing method, apparatus, device, and medium. Background Art
[0003] In the business recommendation scenario, when there are multiple businesses to be recommended on the platform (such as information A, information B, etc.), it is necessary to allocate corresponding traffic to each business to be recommended to ensure that each business to be recommended can be reasonably exposed and converted. At present, business personnel usually allocate corresponding traffic to each business to be recommended based on historical business data (for example, historical exposure rate and historical conversion rate, etc.), and push it to the corresponding object according to the traffic allocation results corresponding to each business to be recommended; for example, the higher the historical conversion rate of the business to be recommended itself or the same type of business, the more traffic will be allocated to the business to be recommended. However, each business to be recommended may face the problem of traffic saturation. When the business historical data of the business to be recommended or the same type of business is used as the basis for traffic allocation, the traffic may be overly concentrated on some businesses to be recommended that have reached or are about to reach traffic saturation, while the businesses to be recommended that have not reached traffic saturation are only allocated very little traffic, resulting in too low traffic utilization.
[0004] Summary of the Invention
[0005] The embodiments of the present application provide a business data processing method, apparatus, device, and medium, which can improve traffic utilization.
[0006] On the one hand, an embodiment of the present application provides a business data processing method, which is executed by a computer device and includes:
[0007] Obtain N services to be recommended, and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1;
[0008] If all the sample values of the indicators corresponding to the services to be recommended satisfy the recommendation constraints, obtain the recommendation probabilities corresponding to the N services to be recommended, and determine the first traffic allocation ratios corresponding to the N services to be recommended based on the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended;
[0009] Performing a weighted summation of the first traffic allocation ratio and the reference traffic allocation ratio to obtain a second traffic allocation ratio corresponding to N services to be recommended, and pushing the N services to be recommended on the service platform according to the second traffic allocation ratio;
[0010] The traffic reference allocation ratio refers to the traffic allocation ratio adopted by N services to be recommended when each service to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
[0011] An embodiment of the present application provides a service data processing device, including:
[0012] An indicator sampling value acquisition module is used to obtain N services to be recommended and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1;
[0013] A first traffic allocation ratio acquisition module is configured to obtain recommendation probabilities corresponding to N services to be recommended if all M indicator sampling values corresponding to the services to be recommended satisfy the recommendation constraint conditions, and determine first traffic allocation ratios corresponding to the N services to be recommended based on the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended;
[0014] A second traffic allocation ratio acquisition module is configured to perform a weighted summation of the first traffic allocation ratio and the reference traffic allocation ratio to obtain second traffic allocation ratios corresponding to N services to be recommended, and push the N services to be recommended on the service platform according to the second traffic allocation ratios;
[0015] The traffic reference allocation ratio refers to the traffic allocation ratio adopted by N services to be recommended when each service to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
[0016] One aspect of an embodiment of the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the method in one aspect of the embodiment of the present application.
[0017] On one hand, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the steps of the method in one aspect of the embodiment of the present application are executed.
[0018] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in various optional embodiments of the above-mentioned aspect.
[0019] In an embodiment of the present application, after obtaining N services to be recommended and M indicator sampling values corresponding to each service to be recommended, the M indicator sampling values associated with the N services to be recommended can be compared with the recommendation constraints. If one or more of the M indicator sampling values corresponding to the N services to be recommended satisfy the recommendation constraints, the recommendation probability of each service to be recommended under the premise of satisfying the recommendation constraints can be obtained, and then the first traffic allocation ratio of each service to be recommended under the premise of satisfying the recommendation constraints can be determined based on the recommendation probability. The final traffic allocation ratio of each service to be recommended is obtained by weighted summing the first traffic allocation ratio and the traffic reference allocation ratio (the traffic allocation ratio of each service to be recommended under the premise of not satisfying the recommendation constraints). In the process of traffic allocation for N services to be recommended, by setting multiple indicators for the N services to be recommended, each indicator can be configured with its corresponding recommendation constraints, and using multiple indicators and the recommendation constraints corresponding to each indicator as comprehensive considerations in traffic allocation can improve the rationality of the final traffic allocation ratio of each service to be recommended, thereby improving the traffic utilization rate for the N services to be recommended. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] FIG1 is a schematic diagram of a network architecture provided in an embodiment of the present application;
[0021] FIG2 is a schematic diagram of a business recommendation scenario provided by an embodiment of the present application;
[0022] FIG3 is a flow chart of a data processing method provided in an embodiment of the present application;
[0023] FIG4 is a schematic diagram of a modified probability distribution provided in an embodiment of the present application;
[0024] FIG5 is a schematic diagram of an interface of a push service provided in an embodiment of the present application;
[0025] FIG6 is a schematic diagram of an interface of another push service provided in an embodiment of the present application;
[0026] FIG7 is a flow chart of another data processing method provided in an embodiment of the present application;
[0027] FIG8 is a schematic diagram of a flow chart of calculating a diversion ratio according to an embodiment of the present application;
[0028] FIG9 is a schematic structural diagram of a business data processing device provided in an embodiment of the present application;
[0029] FIG10 is a schematic structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.
[0031] Please refer to Figure 1, which is a schematic diagram of the structure of a network architecture provided in an embodiment of the present application. The network architecture may include a server 10d and a terminal cluster. The terminal cluster may include one or more terminal devices. There is no limit on the number of terminal devices included in the terminal cluster. As shown in Figure 1, the terminal cluster may specifically include terminal device 10a, terminal device 10b, and terminal device 10c, etc.; all terminal devices in the terminal cluster (for example, terminal device 10a, terminal device 10b, and terminal device 10c, etc.) can be connected to the server 10d through the network connection, so that each terminal device can exchange data with the server 10d through the network connection.
[0032] Among them, the server 10d shown in Figure 1 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. This application does not limit the type of server.
[0033] The terminal devices in the terminal cluster shown in Figure 1 may include but are not limited to: smart phones, tablet computers, laptops, PDAs, mobile internet devices (MIDs), wearable devices (such as smart watches, smart bracelets, etc.), smart voice interaction devices, smart home appliances (such as smart TVs, etc.), vehicle-mounted devices, aircraft and other electronic devices. This application does not limit the type of terminal devices.
[0034] It is understood that each terminal device in the terminal cluster shown in FIG1 can be integrated with a service platform. This service platform can be an application client, webpage, or applet capable of service push. When the service platform runs in each terminal device, it can respectively exchange data with the server 10d shown in FIG1. For example, when the service platform is an application client, the service platform running in each terminal device can be an independent client or an embedded sub-client integrated into a client. This embodiment of the present application does not limit this.
[0035] In the embodiment of the present application, when the service platform is an application client, the service platform may specifically include, but is not limited to: an in-vehicle client, a smart home client, an entertainment client (e.g., a game client), a multimedia client (e.g., a video client, a music client), an interactive client, and an information client (e.g., a news client). If the terminal device included in the terminal cluster is an in-vehicle device, then the in-vehicle device may be an intelligent terminal in a smart transportation scenario, and the application client running on the in-vehicle device may be referred to as an in-vehicle client.
[0036] In an embodiment of the present application, the server 10d and the terminal device (for example, terminal device 10a, terminal device 10b, and terminal device 10c, etc.) can independently implement the business data processing method provided in the embodiment of the present application; optionally, the server 10d and the terminal device (for example, terminal device 10a, terminal device 10b, and terminal device 10c, etc.) can also collaboratively implement the business data processing method provided in the embodiment of the present application, and the present application does not limit this.
[0037] Please refer to Figure 2, which is a schematic diagram of a business recommendation scenario provided by an embodiment of the present application. As shown in Figure 2, in the business recommendation scenario, assuming that multiple businesses need to be pushed in a certain business platform, the multiple businesses that need to be pushed can be called businesses to be recommended, such as Business 1 to be recommended, Business 2 to be recommended, and Business 3 to be recommended as shown in Figure 1. Before pushing the above-mentioned businesses to be recommended in the business platform, traffic allocation can be performed for each business to be recommended; specifically, conversion rate and average transaction cost can be selected as optimization indicators for Business 1 to be recommended, Business 2 to be recommended, and Business 3 to be recommended, and then corresponding recommendation constraints can be designed for the conversion rate indicator and the average transaction cost indicator, respectively. Among them, the conversion rate can refer to the ratio between the number of times the business to be recommended is purchased or subscribed to and the total number of pushes; the average transaction cost can refer to the ratio of the total transaction cost of the business to be recommended to the number of transactions.
[0038] For example, the recommendation constraint for the conversion rate indicator can be designed as "When the indicator sampling value corresponding to the conversion rate indicator is greater than or equal to 0.78, it indicates that the indicator sampling value satisfies the recommendation constraint; otherwise, it does not satisfy the recommendation constraint." The recommendation constraint for the average transaction cost indicator can be designed as "When the indicator sampling value corresponding to the average transaction cost indicator is greater than or equal to 0.6, it indicates that the indicator sampling value satisfies the recommendation constraint; otherwise, it does not satisfy the recommendation constraint." As shown in Figure 2, using the above example as the recommendation constraint for the conversion rate indicator and the average transaction cost indicator, assume that in the first sampling, for business 1 to be recommended, the indicator sampling value for the conversion rate indicator is 0.8, and the indicator sampling value for the average transaction cost indicator is 0.6; for business 2 to be recommended, the indicator sampling value for the conversion rate indicator is 0.7, and the indicator sampling value for the average transaction cost indicator is 0.6; for business 3 to be recommended, the indicator sampling value for the conversion rate indicator is 0.5, and the indicator sampling value for the average transaction cost indicator is 0.4. By comparing the sampled values of the conversion rate indicator for business 1 to be recommended with the corresponding recommended constraint, it can be determined that the sampled values of the conversion rate indicator for business 1 to be recommended satisfy the corresponding recommended constraint. By comparing the sampled values of the average transaction cost indicator for business 1 to be recommended with the corresponding recommended constraint, it can be determined that the sampled values of the average transaction cost indicator for business 1 to be recommended also satisfy the corresponding recommended constraint. Therefore, in the first sampling, among the three businesses to be recommended, the sampled values of the two indicators corresponding to business 1 to be recommended both satisfy the recommended constraint. Therefore, it can be determined that the first sampling passes the constraint, and the constraint pass count can be updated to 1 (its initial value is 0).
[0039] Similarly, in the second sampling, the sampled indicator values for the conversion rate indicator of each business to be recommended are compared with the recommendation constraint corresponding to the conversion rate indicator, and the sampled indicator values for the average transaction cost indicator of each business to be recommended are compared with the recommendation constraint corresponding to the average transaction cost indicator. It can be determined that at least one of the two indicator sampled values corresponding to business to be recommended 1, the two indicator sampled values corresponding to business to be recommended 2, and the two indicator sampled values corresponding to business to be recommended 3 does not satisfy the recommendation constraint. For example, neither business to be recommended 1 nor business to be recommended 2 satisfies the recommendation constraint corresponding to the conversion rate indicator, and business to be recommended 3 satisfies neither the recommendation constraint corresponding to the conversion rate indicator nor the recommendation constraint corresponding to the average transaction cost. It can then be determined that the second sampling failed the constraint. In this case, the constraint failure count can be updated to 1 (its initial value is 0), and the above process is repeated until the preset total number of samplings is reached (for example, the total number of samplings can be set to 10,000).
[0040] After the total number of sampling times is reached, the recommendation probabilities for recommended services 1, 2, and 3 can be calculated based on the number of constraint passes. The sum of the recommendation probabilities for recommended services 1, 2, and 3 is determined as the cumulative recommendation probability value. Furthermore, the first traffic allocation ratio for each recommended service is determined based on the proportion of the recommendation probability corresponding to each recommended service in the cumulative recommendation probability value. As shown in Figure 2, the first traffic allocation ratio 1 for recommended service 1 can be 0.475, the first traffic allocation ratio 2 for recommended service 2 can be 0.312, and the first traffic allocation ratio 3 for recommended service 3 can be 0.213.
[0041] Furthermore, the first traffic allocation ratio and the traffic reference allocation ratio corresponding to each service to be recommended can be weighted and summed to obtain the second traffic allocation ratio corresponding to each service to be recommended. In the weighted summation process here, the weight of the first traffic allocation ratio can be associated with the number of times the constraint is passed and the total number of samples, and the weight of the traffic reference allocation ratio can be associated with the number of times the constraint is not passed and the total number of samples; the traffic reference allocation ratio refers to the traffic allocation ratio adopted by each service to be recommended when there is an indicator sampling value that does not meet the recommendation constraint conditions for each service to be recommended; the method for determining the traffic reference allocation ratio will be described in detail in the subsequent content. As shown in Figure 2, the second traffic allocation ratio 1 finally obtained by service to be recommended 1 can be 0.495, the second traffic allocation ratio 2 finally obtained by service to be recommended 2 can be 0.295, and the second traffic allocation ratio 3 finally obtained by service to be recommended 3 can be 0.210. Then, the recommended traffic can be allocated to each service to be recommended according to the second traffic allocation ratio corresponding to each service to be recommended. For example, the total recommended traffic provided by the business platform for this business push is 1,000 pushes. Then, according to the second traffic distribution ratio, the recommended traffic that can be obtained by the recommended business 1 is 495 pushes, the recommended traffic that can be obtained by the recommended business 2 is 295 pushes, and the recommended traffic that can be obtained by the recommended business 3 is 210 pushes. At this time, the recommended business 1 can be pushed to 495 registered objects in the business platform, the recommended business 2 can be pushed to 295 registered objects in the business platform, and the recommended business 3 can be pushed to 210 registered objects in the business platform.
[0042] In an embodiment of the present application, in the process of allocating traffic for multiple services to be recommended, by setting multiple indicators for the multiple services to be recommended, each indicator can be configured with a corresponding recommendation constraint condition, and multiple indicators and the recommendation constraint conditions corresponding to each indicator are used as comprehensive considerations in traffic allocation. The rationality of the final traffic allocation ratio of each service to be recommended can be improved, and the traffic utilization rate for multiple services to be recommended can be improved.
[0043] Please refer to Figure 3, which is a flow chart of a business data processing method provided in an embodiment of the present application. It can be understood that the data processing method is executed by a computer device, which can be a terminal device (for example, terminal device 10a, terminal device 10b, and terminal device 10c in the embodiment corresponding to Figure 1), or a server (for example, server 10d in the embodiment corresponding to Figure 1). As shown in Figure 3, the business data processing method can include the following steps S101-S103:
[0044] Step S101: Obtain N services to be recommended, and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1.
[0045] Among them, the services to be recommended can be understood as information that needs to be pushed to the business objects with the help of the business platform, and specifically can include advertisements, games, information, videos, props, commodities, etc. to be pushed. The business platform can be understood as an application client, web page or mini-program that can push services, for example, a video client, an information web page or a game mini-program, etc. N represents the number of services to be recommended, N is an integer greater than 1, and the specific value of N can be 2, 3, 4...; in other words, in the embodiment of the present application, the number of services to be recommended is at least two.
[0046] The business to be recommended is provided by the business party, which may refer to an object that provides the business to be recommended, such as a shopping platform, an information platform, an activity display platform, etc.; the business types of each business to be recommended may be the same or different, and the embodiments of the present application do not limit this. For example, taking N businesses to be recommended including two businesses to be recommended (respectively, business to be recommended 1 and business to be recommended 2), if business to be recommended 1 is advertising information for brand A lipstick, and business to be recommended 2 is advertising information for brand B camera, then it can be determined that business to be recommended 1 and business to be recommended 2 have the same business type (both are advertising information); if business to be recommended 1 is social news information, and business to be recommended 2 is advertising information for brand B camera, then it can be determined that business to be recommended 1 and business to be recommended 2 have different business types. The business to be recommended may be information that has been pushed on the business platform, that is, information that has been exposed on the business platform, or it may be new information that has not been pushed on the business platform, that is, information that has not yet been exposed on the business platform. The business to be recommended may be stored in the business database corresponding to the business platform, which may be a local database, a cloud database, or a backend server corresponding to the business platform. The business database can be configured with at least one type of query interface, such as an Oracle (a relational database management system) interface, a SQL Server (a relational database management system) interface, etc. When the recommended business needs to be pushed on the business platform, the corresponding query result can be called using a programming language to obtain the recommended business from the business database.
[0047] The indicator sampling value may refer to the result obtained by performing one or more random samplings on the recommended business for each indicator; for a single business to be recommended, each time sampling (one sampling may include one or more random samplings) one indicator corresponds to an indicator sampling value. If M indicators are set for the business to be recommended, then each time sampling is performed one business to be recommended may correspond to M indicator sampling values, where M represents the number of indicators, and M is an integer greater than 1, such as M may take the value of 2, 3, 4... Since the actual revenue generated by each business to be recommended after being pushed on the business platform is unknown at the current moment, it is necessary to estimate the expected revenue of each business to be recommended by sampling. The indicator sampling value can be used to update the expected revenue estimate of each business to be recommended, and plays an important role in the subsequent calculation of the second traffic distribution ratio corresponding to each business to be recommended. In an embodiment of the present application, a business to be recommended (for example, business to be recommended i) may correspond to M indicator sampling values, and these M indicator sampling values may be recorded as in, It represents the index sampling value of the recommended service i on the jth index, where j is a positive integer less than or equal to M.
[0048] In an embodiment of the present application, M indicators associated with N services to be recommended can be obtained, and then one or more sampling strategies such as the ε-greedy algorithm, the upper confidence bound algorithm (UCB), the Thompson sampling algorithm (Thompson Sampling) or the Monte Carlo sampling algorithm (Monte Carlo Sampling) are used to obtain the indicator sampling values of each service to be recommended on the M indicators. That is, each service to be recommended can correspond to M indicator sampling values in each sampling.
[0049] Among them, the M indicators set for N services to be recommended can be of the same indicator type or of different indicator types, and this application does not limit this; the indicator types involved in the embodiments of this application may include but are not limited to ratio types and mean types. Among them, ratio type indicators can be used to reflect the effect of the service to be recommended, and ratio type indicators may include but are not limited to click-through rate, conversion rate, etc.; wherein, click-through rate may refer to the ratio between the number of times the service to be recommended is clicked and the total number of pushes; mean type indicators can be used to reflect the overall satisfaction of the registered object with the recommended service, or the resources or cost required to push the service to be recommended, and mean type indicators may include average browsing time, number of clicks, average revenue value, average transaction fee, etc.; wherein, average revenue refers to the average value of the ratio of the number of times the service to be recommended is purchased, subscribed or clicked by the registered object to the total number of pushes.
[0050] The M indicators of the business to be recommended may be M ratio-type indicators, for example, the indicators corresponding to the business to be recommended may be click-through rate and conversion rate, etc.; the M indicators of the business to be recommended may also be M mean-type indicators, for example, the indicators corresponding to the business to be recommended may be average transaction cost and average browsing time, etc.; the M indicators of the recommended business may also be indicators that include both ratio type and mean type, for example, the indicators corresponding to the business to be recommended may be conversion rate and average transaction cost, etc. It is understandable that the M indicators involved in the embodiment of the present application may be positively correlated, for example, conversion rate and average browsing time; the M indicators may also be negatively correlated, for example, conversion rate and average transaction cost. It should be noted that each of the N businesses to be recommended has the same indicators.
[0051] In one possible implementation, an ε-greedy algorithm can be used to obtain sample values of M indicators for each of the N recommended services. Assuming that the M indicators corresponding to the N recommended services include the number of clicks and average revenue, the following describes the process of obtaining the sample values of the M indicators corresponding to any one of the N recommended services (e.g., recommended service i) as an example.
[0052] Specifically, the sampled value of the number of clicks for recommended business i can be denoted as N(i), and the sampled value of the average revenue for recommended business i can be denoted as Q(i). The sampled values N(i) and Q(i) corresponding to recommended business i can be initialized. For example, both N(i) and Q(i) can be initialized to 0 (i.e., N(i) = 0, Q(i) = 0). A value ε is then selected as the first random probability, i.e., the probability of exploring unknown information, and a value 1-ε is selected as the second random probability, i.e., the probability of exploiting known information. In other words, the first random probability ε can be used to control the trade-off between exploration and exploitation. ε is a value greater than 0 and less than 1, and its specific value can be determined based on actual conditions. Generally speaking, the smaller the value of ε, the more likely the algorithm is to exploit; the larger the value of ε, the more likely it is to explore.
[0053] Specifically, the recommended business i can be sampled at least once with a first random probability ε, and the probability g(i) of the recommended business i being selected can be calculated, where g(i) = ε / y+[(1-ε) / y]I(i), where y is the total number of sampling processes that have been performed, and I(i) is an indicator function. When the recommended business i is the recommended business with the highest currently known average reward value among the N recommended businesses, the value of I(i) is 1, indicating that the recommended business i is selected for exploitation; when the recommended business i is not the recommended business with the highest currently known average reward value among the N recommended businesses, the value of I(i) is 0, indicating that other recommended businesses are selected for exploration or exploitation.
[0054] After completing a sampling process for the recommended business i, the index sampling value corresponding to the recommended business i can be updated according to the obtained reward value R. Specifically, the number of times the recommended business i is selected can be increased by 1 (N(i) + = 1), and its average reward value can be updated (Q(i) = Q(i) * (N(i) - 1) / N(i) + R / N(i). Then, according to the updated index sampling value, the next round of sampling process for the recommended business i can be carried out. The same processing flow as the above sampling can be adopted to continue updating the index sampling value until the sampling end condition is met, and the sampling is stopped; the index sampling value obtained after the last sampling is used as the index sampling value corresponding to the recommended business i. The sampling end condition can be reaching a specified number of sampling times, etc.
[0055] In one possible implementation, the process of obtaining the sample values of the M indicators corresponding to the service to be recommended i may include: constructing a first probability distribution corresponding to the M indicators; wherein the first probability distribution can be understood as the prior distribution of each indicator of the service to be recommended, and the first probability distribution can specifically be a beta distribution, a normal distribution, etc. In the embodiment of the present application, the types of the first probability distributions corresponding to the various indicators can be the same or different. For example, when the indicator is click-through rate or conversion rate, the corresponding first probability distribution can be a beta distribution; when the indicator is average browsing time, the corresponding first probability distribution can be a normal distribution. For ease of understanding, the embodiment of the present application takes N businesses to be recommended, including business to be recommended 1, business to be recommended 2, and business to be recommended 3, the jth indicator is any one of the M indicators, and the first probability distribution corresponding to the jth indicator is the beta distribution as an example to describe the process of obtaining the indicator sampling value of the jth indicator of the recommended business. The process of obtaining the indicator sampling values corresponding to other indicators can refer to the process of obtaining the jth indicator; wherein, j is a positive integer less than or equal to M, and the specific value of j can be 2, 3, 4...M.
[0056] Specifically, please refer to Figure 4, which is a schematic diagram of a modified probability distribution provided by an embodiment of the present application. As shown in Figure 4, the first probability distribution 1 corresponding to the service to be recommended 1, the first probability distribution 2 corresponding to the service to be recommended 2, and the first probability distribution 3 corresponding to the service to be recommended 3 can all be initialized to beta distribution Beta (α, β). At this time, the first probability distribution 1 can be recorded as Beta1 (α, β), the first probability distribution 2 can be recorded as Beta2 (α, β), and the first probability distribution 3 can be recorded as Beta3 (α, β); wherein the x-axis represents a random value, and the y-axis represents the probability density of the random value. In the case of prior experience, the initial values of the parameters α and β can be adjusted. For example, the initial values of the parameters α and β can both be adjusted to 1.
[0057] Furthermore, the platform collected data related to the j-th indicator of the business to be recommended 1, the business to be recommended 2 and the business to be recommended 3 can be obtained, and then the first probability distribution corresponding to each business to be recommended can be corrected according to the platform collected data to obtain the second probability distribution corresponding to each business to be recommended. Among them, the platform collected data can be understood as a random value obtained by sampling according to the first probability distribution, or it can also be experimental data obtained by online simulation based on information such as the business attributes of the business to be recommended, or it can also be the push feedback results collected after each business to be recommended is pushed on the business platform for a period of time (for example, historical click counts, historical conversion rates, historical average browsing time or historical average transaction fees, etc.). The second probability distribution can be understood as the posterior distribution of the j-th indicator, and the second probability distribution is of the same type as the first probability distribution. For example, if the first probability distribution is a beta distribution, then the second probability distribution is also a beta distribution.
[0058] Taking the jth indicator, click count, as an example, the correction process for the first probability distribution is described. For example, the platform-collected data corresponding to the recommended business 1 is predicted click count 1, the platform-collected data corresponding to the recommended business 2 is predicted click count 2, and the platform-collected data corresponding to the recommended business 3 is predicted click count 3. The selected recommended business can be determined by comparing the predicted click counts 1, 2, and 3. For example, if both predicted click counts 2 and 3 are less than predicted click count 1, it can be determined that recommended business 1 is selected, and recommended business 2 and 3 are not selected. At this time, the parameter α of Beta1(α, β) (first probability distribution 1) can be updated according to the above comparison results. For example, the parameter α of Beta1(α, β) is increased by 1 to obtain the second probability distribution 1 corresponding to the business to be recommended 1. The second probability distribution 1 can be recorded as Beta11(α+1, β); the parameter β of Beta2(α, β) (first probability distribution 2) and Beta3(α, β) (first probability distribution 3) can be updated. For example, the parameter β of Beta2(α, β) and Beta3(α, β) is increased by 1 respectively to obtain the second probability distribution 2 corresponding to the business to be recommended 2 and the second probability distribution 3 corresponding to the business to be recommended 3; the second probability distribution 2 can be recorded as Beta22(α, β+1), and the second probability distribution 3 can be recorded as Beta33(α, β+1). Optionally, when the platform collects data with multiple sets of data, the first probability distribution can be corrected multiple times. Through multiple corrections, the second probability distribution can be made more in line with expectations.
[0059] As shown in Figure 4, after the first probability distribution has been corrected at least once, the shape of the second probability distribution obtained becomes narrower, indicating that the second probability distribution has a higher confidence level than the first probability distribution. Therefore, the index sampling value obtained by subsequent sampling processing based on the second probability distribution has a higher accuracy. After obtaining the second probability distribution corresponding to each business to be recommended, a random probability value can be obtained from the second probability distribution. Assuming that a random probability value A is randomly selected from the second probability distribution 1, a random probability value B is randomly selected from the second probability distribution 2, and a random probability value C is randomly selected from the second probability distribution 3, then the random probability value A can be determined as the index sampling value for the j-th index of the business to be recommended 1, the random probability value B can be determined as the index sampling value for the j-th index of the business to be recommended 2, and the random probability value C can be determined as the index sampling value for the j-th index of the business to be recommended 3.
[0060] In an embodiment of the present application, for N services to be recommended, a first probability distribution (prior distribution) corresponding to each of the M indicators can be constructed to quantify the influence of each indicator; then, the first probability distribution is corrected according to the data collected by the platform, so that the confidence of the second probability distribution is improved. Therefore, the indicator sampling value obtained by sampling according to the second probability distribution has a higher accuracy, thereby ensuring the reliability and accuracy of the subsequent traffic distribution ratio.
[0061] Step S102: If there are M indicator sampling values corresponding to the services to be recommended that all meet the recommendation constraints, then obtain the recommendation probabilities corresponding to the N services to be recommended, and determine the first traffic allocation ratio corresponding to the N services to be recommended based on the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended.
[0062] Among them, the recommendation constraints can be understood as the constraints that the business to be recommended needs to meet on various indicators, or the experimental design rules that the business to be recommended needs to follow when making recommendations. Specifically, the M indicator sampling values corresponding to each of the N businesses to be recommended can be compared with the corresponding recommendation constraints. If there are one or more of the M indicator sampling values corresponding to the N businesses to be recommended that meet the recommendation constraints, then the step of obtaining the recommendation probabilities corresponding to the N businesses to be recommended is executed. For example, the N businesses to be recommended may include business 1 to be recommended and business 2 to be recommended. If the M indicator sampling values corresponding to business 1 to be recommended meet the recommendation constraints, and there are indicator sampling values in business 2 to be recommended that do not meet the recommendation constraints, then the step of obtaining the recommendation probabilities corresponding to the N businesses to be recommended is still executed.
[0063] Optionally, N recommendation constraints associated with the services to be recommended can be pre-designed. When designing recommendation constraints, different recommendation constraints can be designed for different indicators; in other words, different indicators can correspond to different recommendation constraints. For example, the recommendation constraint corresponding to indicator 1 can be designed as "when the indicator sampling value corresponding to indicator 1 is greater than or equal to 0, the indicator sampling value corresponding to indicator 1 meets the recommendation constraint, otherwise it does not meet the recommendation constraint"; the recommendation constraint corresponding to indicator 2 can be designed as "when the indicator sampling value corresponding to indicator 2 is greater than or equal to 5%, the indicator sampling value corresponding to indicator 2 meets the recommendation constraint, otherwise it does not meet the recommendation constraint". For a certain service to be recommended (for example, service to be recommended 1), if the indicator sampling values corresponding to all indicators meet the corresponding recommendation constraints, it can be determined that the indicator sampling value corresponding to service to be recommended 1 meets the recommendation constraint; if there is any indicator sampling value that does not meet the recommendation constraint for service to be recommended 1, it can be determined that the indicator sampling value corresponding to service to be recommended 1 does not meet the recommendation constraint.
[0064] When designing recommendation constraints for indicators corresponding to the business to be recommended, each of the M indicators associated with the N businesses to be recommended can be converted into a sub-constraint; wherein, a is a positive integer, and the specific value of a can be 1, 2, 3, 4... In other words, the recommendation constraint corresponding to an indicator can include one or more sub-constraints, and each of the M indicators associated with a business to be recommended corresponds to at least one sub-constraint. For example, indicator 1 can correspond to sub-constraint C1 and sub-constraint C2, and indicator 2 can correspond to sub-constraint C3. For ease of understanding, the embodiment of the present application takes any indicator (for example, the jth indicator) in the business to be recommended i as an example to describe the implementation process of converting the jth indicator into a sub-constraint. The conversion process of other indicators corresponding to the business to be recommended i can refer to the conversion process of the jth indicator; the business to be recommended i is any one of the N businesses to be recommended, and the determination process of the sub-constraints corresponding to other businesses to be recommended can refer to the description corresponding to the business to be recommended i.
[0065] In one possible implementation, the design method of the recommended constraint condition may include: determining the target value or target range (e.g., minimum value, maximum value, average value, expected value, etc.) corresponding to the j-th indicator, and then converting each target value or target range into a corresponding sub-constraint condition, thereby obtaining a sub-constraint conditions corresponding to the j-th indicator. For example, the sub-constraint condition C1 with the minimum value as the target can be designed as "I j ≥T1", where I j is the index sampling value corresponding to the jth index, T1 is the minimum value; the sub-constraint C2 with the maximum value as the target can be designed as "I j≤T2", where T2 is the maximum value.
[0066] In one possible implementation, the design method of the recommended constraint condition may further include: designing sub-constraint conditions based on the indicator type corresponding to the j-th indicator. As described above, the indicator type corresponding to the j-th indicator may include a ratio type and a mean type, and different sub-constraint conditions may be set for different indicator types. If the indicator type is a ratio type, then a first constraint design strategy corresponding to the ratio type may be obtained, and then the j-th indicator may be converted into a sub-constraint conditions according to the first constraint design strategy; if the indicator type is a mean type, then a second constraint design strategy corresponding to the mean type may be obtained, and then the j-th indicator may be converted into a sub-constraint conditions according to the second constraint design strategy.
[0067] An example of design rules for sub-constraints corresponding to different indicator types provided in an embodiment of the present application may be shown in Table 1 below:
[0068] Table 1
[0069] As shown in Table 1, a reference business can be set for each of the N recommended businesses. The reference business can be the business with the largest revenue value (conversion rate or exposure rate, etc.) among the businesses pushed by the business platform in the past period (for example, the past week or the past month); the index sampling value of the reference business for the jth index can be recorded as The index sampling value of the recommended business i for the jth index can be recorded as
[0070] It is understandable that in an embodiment of the present application, indicators with different indicator types may have different indicator distributions (also referred to as probability distributions). For example, when the indicator type corresponding to the j-th indicator is a ratio type (for example, conversion rate, etc.), the indicator distribution corresponding to the indicator may be a beta distribution. The indicator sampling value obtained by each sampling of the ratio type indicator has only two values of 0 and 1, so the design rules of the sub-constraint conditions corresponding to the ratio type indicator only support the use of ≤ or ≥ logical operators. When the indicator type corresponding to the j-th indicator is a mean type (for example, average transaction fee, etc.), the indicator distribution corresponding to the indicator may be a normal distribution. The indicator sampling value obtained by each sampling of the mean type indicator is a numerical value x, for example, the numerical value x can take values of 0.1 or 1, etc.; the indicator sampling value of the mean type indicator can be a continuous value, so the design rules of the sub-constraint conditions corresponding to the mean type indicator can support the use of logical operators including ≤ or ≥.
[0071] As shown in Table 1, as an example, the first constraint design strategy can be designed as "when the index sampling value of the recommended business Greater than or equal to the index sampling value corresponding to the control business When , it indicates the sampling value of the indicator Satisfy the sub-constraints, otherwise do not satisfy the sub-constraints". Accordingly, the sub-constraints corresponding to the j-th indicator can be designed as " When Satisfy the sub-constraint condition, otherwise do not satisfy the sub-constraint condition"; Optionally, the second constraint design strategy can be designed as "when the index sampling value of the recommended business Not less than the index sampling value corresponding to the control business 1%, indicating the sampling value of the indicator Satisfy the sub-constraints, otherwise do not satisfy the sub-constraints". Accordingly, the sub-constraints corresponding to the j-th indicator can be designed as " When The sub-constraint condition is satisfied, otherwise the sub-constraint condition is not satisfied. It can be understood that 1% is only used as an example, and in actual applications, other appropriate values can be selected according to actual needs, such as 2% or 5%.
[0072] In an embodiment of the present application, by classifying indicator types, adopting different constraint design strategies for indicators of different indicator types, and converting the indicators into multiple sub-constraint conditions in a targeted manner according to different constraint design strategies, the impact of each indicator on the subsequent traffic allocation ratio can be finely controlled, thereby improving the rationality of the final traffic allocation ratio of each recommended business, and thus improving the traffic utilization rate for N recommended businesses.
[0073] An example of a judgment rule for a recommendation constraint provided in an embodiment of the present application may be shown in Table 2 below:
[0074] Table 2
[0075] Table 2 shows the recommended constraint conditions C result Taking the recommendation constraint corresponding to the jth indicator as an example, the judgment process of the recommendation constraint is described. In the embodiment of the present application, after obtaining the jth indicator sampling value corresponding to each service to be recommended, the jth indicator sampling value can be compared with each sub-constraint in the a sub-constraint to determine whether the jth indicator sampling value satisfies a single sub-constraint in the a sub-constraint. If it satisfies, the label corresponding to the sub-constraint is marked as 1. If it does not satisfy, the label corresponding to the sub-constraint is marked as 0. When the labels corresponding to the a sub-constraints are all 1, the recommendation constraint C is selected. result The corresponding label is marked as 1, which means that the sample value of the jth indicator corresponding to the recommended business meets the recommendation constraint C corresponding to the jth indicator. result; If any of the a sub-constraints has a label of 0, then constraint C is recommended result The corresponding label is marked as 0, which means that the sampling value of the jth indicator corresponding to the recommended business does not meet the recommendation constraint C corresponding to the jth indicator. result The j-th indicator sampling value refers to the indicator sampling value of the recommended business for the j-th indicator.
[0076] That is to say, when the jth index sampling values all meet the sub-constraints C1, C2, ..., C q , ..., sub-constraint C a When the sampling value of the jth indicator corresponding to the recommended business satisfies the recommendation constraint C corresponding to the jth indicator result When the jth index sampling value does not meet one or more of the a sub-constraints (for example, the jth index sampling value does not meet the sub-constraint C2, or the jth index sampling value does not meet the sub-constraint C2 and the sub-constraint C q ), the sampling value of the jth indicator corresponding to the recommended business does not meet the recommendation constraint C corresponding to the jth indicator result Among them, the sub-constraint C q Indicates the qth sub-constraint among a sub-constraints, where q is a positive integer less than or equal to a. As shown in Table 2, the jth indicator sampling value corresponding to the recommended business 1 And the jth indicator sampling value corresponding to the recommended business i Satisfy the recommendation constraint C corresponding to the jth indicator result ; The jth index sampling value corresponding to the recommended business 2 The jth indicator sampling value corresponding to the recommended business N The recommendation constraint C corresponding to the jth indicator is not satisfied result .
[0077] In the embodiment of the present application, the indicator sample value corresponding to an indicator is compared with one or more sub-constraints corresponding to the indicator. If the indicator sample value corresponding to the indicator satisfies each of its corresponding sub-constraints, the indicator sample value is considered to meet the recommendation constraint; otherwise, it does not meet the recommendation constraint. This judgment method can accurately determine whether the indicator sample value meets the conditions, which can improve the rationality of the final traffic allocation ratio of each recommended service, and thus can improve the traffic utilization rate for N recommended services.
[0078] The recommendation probability can be understood as the probability that a particular service is the optimal service, given that the recommendation constraints are met. The cumulative recommendation probability value can be understood as the sum of the probabilities of each service being the optimal service, given that the recommendation constraints are met. In other words, after calculating the recommendation probability for each service, the cumulative recommendation probability value is obtained by adding up the corresponding probabilities.
[0079] In an embodiment of the present application, multiple sampling processes can be performed on each business to be recommended. For each business to be recommended, M indicator sampling values can be obtained after each sampling is completed. The M indicator sampling values obtained after each sampling may not be exactly the same. Therefore, for the sampled indicator values corresponding to N businesses to be recommended obtained from two different samplings, it may be possible that in one sampling, there is a situation where at least one of the M indicator sampling values corresponding to the business to be recommended satisfies the recommendation constraint conditions, while in another sampling, there may be a situation where each business to be recommended has an indicator sampling value that does not satisfy the recommendation constraint conditions. In addition, when the sampling batches are different, the businesses to be recommended that satisfy the recommendation constraint conditions may also be different. For example, in the first sampling, the indicator sampling values corresponding to business to be recommended 1 and business to be recommended 2 may satisfy the recommendation constraint conditions, while in the second sampling, the indicator sampling values corresponding to business to be recommended 3 may satisfy the recommendation constraint conditions. Therefore, when multiple sampling is performed, the indicator sampling values obtained after each sampling are required to be compared with the recommendation constraint conditions to determine whether the current sampling needs to calculate the recommendation probability corresponding to the business to be recommended that satisfies the recommendation constraint conditions. The calculation method of the recommendation probability will be described in detail in step S204 in FIG. 7 .
[0080] The first traffic allocation ratio may refer to a weighted ratio determined based on the proportion of the recommendation probability corresponding to each pending service within the cumulative recommendation probability value of the N pending services. The first traffic allocation ratio may serve as the basis for setting push strategies for the subsequent N pending services. Generally speaking, the greater the recommendation probability of a pending service, the greater the first traffic allocation ratio for that service, and accordingly, the greater the recommended traffic allocated to that service.
[0081] Taking any one of the N recommended services (for example, recommended service i) as an example, the calculation method of the first traffic allocation ratio corresponding to the recommended service i is described. After obtaining the recommendation probability corresponding to the recommended service i (which can be recorded as P(Arm i(Best) )) and the cumulative value of recommendation probability corresponding to N recommended businesses (which can be recorded as ) After that, the first traffic allocation ratio U corresponding to the recommended service i can be calculated i(t) ,in, In other words, the recommendation probability Arm corresponding to the recommended business i can be i(Best) The cumulative value of recommendation probability corresponding to N recommended businesses The ratio between them is determined as the first traffic allocation ratio U corresponding to the recommended service i i(t) .
[0082] Step S103: performing weighted summation on the first traffic allocation ratio and the traffic reference allocation ratio to obtain a second traffic allocation ratio corresponding to N services to be recommended, and pushing the N services to be recommended in the service platform according to the second traffic allocation ratio.
[0083] In the embodiment of the present application, the traffic reference allocation ratio can be understood as the traffic allocation ratio adopted by the N services to be recommended when each service to be recommended has an indicator sampling value that does not meet the recommendation constraint condition; the second traffic allocation ratio can be understood as the traffic allocation ratio adopted by the N services to be recommended when the service is finally pushed on the service platform. It can be understood that when the traffic reference allocation ratio is 0, the second traffic allocation ratio is the first traffic allocation ratio. In this case, the N services to be recommended can be pushed on the service platform according to the first traffic allocation ratio. Optionally, the traffic reference allocation ratio can be based on the rules of business experts. For example, the N services to be recommended include the service to be recommended i and the service to be recommended j. According to prior knowledge, it is estimated that the conversion rate of the service to be recommended i on the service platform is 3 times that of the service to be recommended j. Then, the traffic reference allocation ratio corresponding to the service to be recommended i can be set to 0.75, and the traffic reference allocation ratio corresponding to the service to be recommended j can be set to 0.25.
[0084] Furthermore, the first allocation weight corresponding to the first traffic allocation ratio and the second allocation weight corresponding to the traffic reference allocation ratio can be obtained. The first traffic allocation ratio and the traffic reference allocation ratio are weighted and summed according to the first allocation weight and the second allocation weight to obtain the second traffic allocation ratio corresponding to N recommended services. The values of the first allocation weight and the second allocation weight can be set according to the business expert rules; specifically, the allocation weight corresponding to each allocation ratio can be determined based on the expert's evaluation opinions on the first traffic allocation ratio and the traffic reference allocation ratio. For example, based on prior knowledge, it is determined that the importance of the first traffic allocation ratio is 5 times that of the traffic reference allocation ratio, then the first allocation weight can be set to 5 and the second allocation weight can be set to 1.
[0085] Furthermore, the recommended traffic corresponding to each of the N recommended businesses can be determined based on the second traffic allocation ratio, and then the delivery area range and delivery duration corresponding to each recommended business can be determined based on the recommended traffic; and the N recommended businesses can be pushed on the business platform according to the delivery area range and delivery duration.
[0086] The recommended traffic volume is used to measure the target audience, duration, and number of pushes for the recommended service. Generally speaking, the larger the second traffic allocation ratio for the recommended service, the greater the recommended traffic volume for that service. The total recommended traffic volume provided by the service platform for N recommended services can be obtained, and then the recommended traffic volume for each service can be determined based on the second traffic allocation ratio and the total recommended traffic volume for each service. Assuming the total recommended traffic volume provided by the service platform includes 1,000 pushes and a push duration of 100 minutes, and the N recommended services include Service 1, Service 2, and Service 3, with their second traffic allocation ratios of 0.5, 0.3, and 0.2, respectively; then Service 1 can receive 500 pushes and a push duration of 50 minutes, Service 2 can receive 300 pushes and a push duration of 30 minutes, and Service 3 can receive 200 pushes and a push duration of 20 minutes.
[0087] Optionally, a set of registered objects can be obtained in the business platform, and the historical login information and basic object information of each registered object in the set of registered objects can be obtained, and the historical login information and basic object information can be converted into vector features to obtain the object features corresponding to each registered object; and then the business attribute information corresponding to each business to be recommended can be converted into vector features to obtain the business features corresponding to each business to be recommended.
[0088] Among them, object features can be used to characterize the attributes and characteristics related to the registered objects, and can be used to distinguish the differences between different registered objects; similarly, business features can be used to characterize the attributes and characteristics of the services to be recommended, and can be used to distinguish the differences between different services to be recommended. After obtaining the object features corresponding to each registered object and the business features corresponding to each service to be recommended, the business features corresponding to each service to be recommended can be calculated in turn for similarity with the object features corresponding to each registered object. For example, the business features corresponding to service 1 to be recommended can be calculated for similarity with the object features corresponding to each registered object, and the business features corresponding to service 2 to be recommended can be calculated for similarity with the object features corresponding to each registered object, until the traversal of the business features corresponding to N services to be recommended is completed; in an embodiment of the present application, one or more methods such as Euclidean distance, Manhattan distance, Minkowski distance, and cosine similarity can be used to calculate the similarity between business features and object features.
[0089] Taking any one of the N businesses to be recommended (for example, business i to be recommended) as an example, the method for determining the delivery area range and delivery duration corresponding to the business to be recommended i is described. The registered objects corresponding to the object features with a similarity greater than the similarity threshold can be combined into the candidate delivery range corresponding to the business to be recommended i, and the delivery area range of the business to be recommended i can be determined within the candidate delivery range based on the recommended traffic corresponding to the business to be recommended i; wherein the similarity threshold is a pre-set parameter, and the specific value can be determined according to the actual situation. The candidate delivery range can be understood as the range composed of the target audience associated with the business to be recommended on the business platform; the delivery area range can be understood as the range composed of the delivery objects corresponding to the business to be recommended when it is finally pushed on the business platform. Generally speaking, the candidate delivery range is greater than or equal to the delivery area range. Assuming that the candidate delivery range includes 1,000 registered objects, and the recommended traffic corresponding to the recommended business i is 500 pushes, then the 500 registered objects ranked with the highest similarity can be selected from the 1,000 registered objects and determined as the delivery area range for the recommended business i; the total delivery duration corresponding to the N recommended businesses can be obtained, and the delivery duration corresponding to the recommended business i can be determined based on the recommended traffic corresponding to the recommended business i. For example, if the total delivery duration is 100 minutes, the recommended traffic corresponding to the recommended business i is a 50-minute push duration, and the second traffic allocation ratio corresponding to the recommended business i is 50%, then the delivery duration corresponding to the recommended business i is set to 50 minutes. Then, according to the delivery area range and delivery duration corresponding to the recommended business i, the recommended business i can be pushed to the registered objects corresponding to the business platform.
[0090] Please refer to Figure 5, which is a schematic diagram of an interface for a push service provided in an embodiment of the present application. Taking the example of a service platform being a video client and the services being pushed including To-Be-Recommended Service 1, To-Be-Recommended Service 2, and To-Be-Recommended Service 3, the service platform's service push display interface will be described. As shown in Figure 5, the user corresponding to terminal device 10a is Registered User 1, and the user corresponding to terminal device 10b is Registered User 2. When Registered User 1 or Registered User 2 requests video playback, the terminal device triggers video client 20. In response to the triggering operation on video client 20, the terminal device displays the video display interface corresponding to video client 20. The video display interface may include a menu bar 21. Registered users can select any live broadcast room to watch a live video by triggering the live broadcast controls in menu bar 21. For example, a registered user can trigger live broadcast room 22 to watch a live broadcast of host A. Simultaneously, video client 20 may also push recommended services to registered users in service push area 23. It will be understood that the greater the second traffic allocation ratio corresponding to a recommended service, the greater the recommended traffic volume received, and therefore, the wider the delivery area for To-Be-Recommended Service 1 on video client 20. Assuming that, according to the second traffic allocation ratio corresponding to each service to be recommended, it is determined that the delivery area range corresponding to the service to be recommended 1 includes the registered object 1 and the registered object 2, the delivery area range corresponding to the service to be recommended 2 includes the registered object 1 and the registered object 2, and the delivery area range corresponding to the service to be recommended 3 includes the registered object 1, as shown in FIG5 , the service to be recommended 1 and the service to be pushed 2 can be pushed synchronously to the registered object 1 and the registered object 2 in the service push area 23 of the video client 20; the service to be recommended 3 is pushed to the registered object 1 in the service push area 23 of the video client 20, and the service to be recommended 3 is not pushed to the registered object 2.
[0091] Please also refer to Figure 6, which is a schematic diagram of the interface of another push service provided in an embodiment of the present application. As shown in Figure 6, the user corresponding to terminal device 10a is registered object 1. It can be understood that the larger the second traffic allocation ratio corresponding to the recommended service, the greater the recommended traffic obtained, and therefore the longer the recommended service 1 is delivered to video client 20. Assume that the video client 20 will push services 1, 2, and 3 to be recommended to the registered subject 1 during the time period of 18:00-18:30. Based on the second traffic allocation ratio corresponding to each service to be recommended, the corresponding delivery duration of service 1 is determined to be 30 minutes, the corresponding delivery duration of service 2 is 20 minutes, and the corresponding delivery duration of service 1 is 10 minutes. As shown in FIG6 , when the system time is "18:00", services 1, 2, and 3 can be simultaneously pushed to the registered subject 1 in the service push area 23 of the video client 20. When the system time is "18:10", the push of service 3 can be canceled in the service push area 23, and services 1 and 2 can be pushed to the registered subject 1. When the system time is "18:20", the push of service 2 can be disabled in the service push area 23, and service 1 can be pushed to the registered subject 1.
[0092] In an embodiment of the present application, after obtaining N services to be recommended and M indicator sampling values corresponding to each service to be recommended, the M indicator sampling values associated with the N services to be recommended can be compared with the recommendation constraints. If one or more of the M indicator sampling values corresponding to the N services to be recommended satisfy the recommendation constraints, the recommendation probability of each service to be recommended under the premise of satisfying the recommendation constraints can be obtained, and then the first traffic allocation ratio of each service to be recommended under the premise of satisfying the recommendation constraints can be determined based on the recommendation probability. The final traffic allocation ratio of each service to be recommended is obtained by weighted summing the first traffic allocation ratio and the traffic reference allocation ratio (the traffic allocation ratio of each service to be recommended under the premise of not satisfying the recommendation constraints). In the process of traffic allocation for N services to be recommended, by setting multiple indicators for the N services to be recommended, each indicator can be configured with its corresponding recommendation constraints, and using multiple indicators and the recommendation constraints corresponding to each indicator as comprehensive considerations in traffic allocation can improve the rationality of the final traffic allocation ratio of each service to be recommended, thereby improving the traffic utilization rate for the N services to be recommended.
[0093] Please refer to Figure 7, which is a flow chart of another business data processing method provided in an embodiment of the present application. It can be understood that the data processing method is executed by a computer device, which can be a terminal device (for example, terminal device 10a, terminal device 10b, and terminal device 10c in the embodiment corresponding to Figure 1), or a server (for example, server 10d in the embodiment corresponding to Figure 1). As shown in Figure 7, the business data processing method can include the following steps S201-S208:
[0094] Step S201: Obtain N services to be recommended, and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1.
[0095] The specific implementation of step S201 may refer to the specific implementation of step S101 in FIG3 , and will not be described in detail here.
[0096] Step S202: If all M indicator sampling values corresponding to the service to be recommended satisfy the recommendation constraint conditions, the constraint pass times are updated; the constraint pass times are used to represent the number of times in the total number of sampling times that all M indicator sampling values corresponding to the service to be recommended satisfy the recommendation constraint conditions.
[0097] The total number of sampling times can be understood as the total number of times the indicators of N recommended businesses are sampled. The total number of sampling times can be recorded as Time (s) After each sampling is completed, the total number of sampling times can be increased by 1 (Time (s) +=1), thereby updating the total number of sampling times Time (s) The total number of samplings can be set according to the actual situation. For example, the total number of samplings can be set to 1000 times, or it can be set to 10000 times. This embodiment of the present application does not limit this. In these samples, there may be some cases where the index sampling values of the services to be recommended meet the recommendation constraints, or there may be cases where each service to be recommended has an index sampling value that does not meet the recommendation constraints. The number of constraint passes records how many times the index sampling values meet the recommendation constraints in these samples. The number of constraint passes can be recorded as Time (t) After each sampling is completed, the M indicator sampling values associated with the N recommended services can be compared with the recommendation constraints. If there is at least one indicator sampling value corresponding to the recommended service that meets the recommendation constraints, it means that the sampling passes the constraint, and the number of constraint passes can be increased by 1 (Time (t) +=1), thereby updating the constraint passing times Time (t) .
[0098] Step S203: The recommended services that meet the recommendation constraints are grouped into a recommended service set. If the M indicator sampling values corresponding to the recommended service i in the recommended service set are greater than or equal to the M indicator sampling values corresponding to the recommended services other than the recommended service i in the recommended service set, the recommended service i is determined as the optimal recommended service.
[0099] The optimal recommended business can be understood as the business to be recommended that satisfies the recommendation constraints and has the highest indicator sampling value in the set of recommended businesses. In some possible implementations, the optimal recommended business can be considered the "optimal arm" in a multi-armed bandit (MAB) problem. In other words, the optimal recommended business can be considered the arm with the highest average return in the MAB. The MAB problem is a classic reinforcement learning problem, which seeks to maximize returns (EE trade-off) between exploring potential high returns and leveraging known high returns.
[0100] In one sampling, the recommended services that meet the recommendation constraints can be added to the recommended service set, so that all the recommended services included in the recommended service set are recommended services that meet the recommendation constraints. In the embodiment of the present application, the Pareto optimal method can be used to determine the optimal recommended service from the recommended service set. For example, the recommended service set includes services i, j, and k to be recommended. If the M indicator sampling values corresponding to the recommended service i are are greater than or equal to the corresponding indicator sampling value of the recommended business j And the indicator sampling value corresponding to the recommended business i are also greater than or equal to the corresponding index sampling value of the recommended business k Then, the recommended service i can be determined as the optimal recommended service for this sampling. It is understood that when comparing the index sampling values corresponding to each recommended service in the recommended service set that meets the recommendation constraint conditions, the index sampling values corresponding to the same index are compared, while the index sampling values corresponding to different indexes are not compared.
[0101] If one or more of the M indicator sample values corresponding to the recommended service i is smaller than the indicator sample value corresponding to the recommended service j (for example, Among the M index sample values corresponding to the recommended service j, there is one or more index sample values that are smaller than the index sample values corresponding to the recommended service k (for example, ), there are one or more of the M indicator sampling values corresponding to the recommended business k that are smaller than the indicator sampling value corresponding to the recommended business i (for example, ), it can be determined that there is no M indicator sampling value corresponding to any of the to-be-recommended services in the recommended service set, which is greater than or equal to the M indicator sampling values corresponding to the remaining to-be-recommended services in the recommended service set. In this case, when the Pareto optimal method is used to determine the optimal recommended service from the recommended service set, it is impossible to uniquely find the optimal recommended service. At this time, the to-be-recommended services (including to-be-recommended service i, to-be-recommended service j and to-be-recommended service k) contained in the recommended service set can all be determined as the optimal recommended services, and the optimal number of recommendations corresponding to the to-be-recommended services contained in the recommended service set can be updated at the same time. The method for updating the optimal number of recommendations corresponding to the to-be-recommended services contained in the recommended service set can refer to the method for updating the optimal number of recommendations corresponding to the to-be-recommended service i in step S204, and will not be repeated here.
[0102] In an embodiment of the present application, when it is impossible to determine the only optimal recommended service, the services to be recommended included in the recommended service set are regarded as the optimal recommended services. This can improve the efficiency and rationality of searching for the optimal recommended services, and thus help improve the calculation efficiency of the final traffic allocation ratio of each service to be recommended without affecting the rationality of the final traffic allocation ratio of each service to be recommended.
[0103] Step S204: Update the optimal recommendation times corresponding to the recommended business i, and determine the ratio between the optimal recommendation times and the constraint passing times as the recommendation probability corresponding to the recommended business i; the optimal recommendation times is used to represent the number of times the recommended business i is the optimal recommended business in the total number of sampling times.
[0104] In the embodiment of the present application, the optimal number of recommendations corresponding to the recommended service i can be recorded as After a sampling is completed, if the recommended business i is determined to be the best recommended business in this sampling, the best recommendation times corresponding to the recommended business i can be increased by 1. After reaching the total number of sampling times, you can pass the number of constraints according to Time (t) The optimal number of recommendations corresponding to the recommended business i Calculate the recommendation probability P(Arm i(Best) ), where P(Arm i(Best) )= In other words, the optimal number of recommendations corresponding to the recommended business i can be and the number of constraint passes Time (t) The ratio between them is determined as the recommendation probability P(Arm i(Best) It is understandable that the calculation method of the recommendation probabilities of other services among the N services to be recommended is similar to the calculation method of the recommendation probability of the service to be recommended i, and will not be described in detail here.
[0105] In an embodiment of the present application, when calculating the recommendation probabilities of N services to be recommended, it is only necessary to calculate and screen the probabilities corresponding to multiple services to be recommended that meet the recommendation constraints corresponding to each indicator. Therefore, the calculation efficiency of the recommendation probability can be improved, which in turn helps to improve the calculation efficiency of the final traffic distribution ratio of each service to be recommended.
[0106] Step S205: determining a first traffic allocation ratio corresponding to the N services to be recommended according to the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended.
[0107] The specific implementation of step S205 may refer to the specific implementation of step S102 in FIG3 , and will not be described in detail here.
[0108] Step S206: Determine the ratio between the number of constraint passing times and the total number of sampling times as the first allocation weight corresponding to the first traffic allocation ratio.
[0109] Step S207: Determine the ratio between the number of times the constraint fails and the total number of sampling times as the second allocation weight corresponding to the reference flow allocation ratio.
[0110] Optionally, if each business to be recommended has an indicator sampling value that does not meet the recommendation constraint conditions, the constraint failure times are updated. As mentioned above, in the process of multiple sampling, there may be some indicator sampling values of the business to be recommended that meet the recommendation constraint conditions, or there may be indicator sampling values of each business to be recommended that do not meet the recommendation constraint conditions. The constraint failure times records the number of times that each business to be recommended has an indicator sampling value that does not meet the recommendation constraint conditions in these samples. The constraint failure times can be recorded as Time (w) In addition, the number of times the constraint fails is Time (w) It can also be understood as the total number of sampling times Time (s) and the number of constraint passes Time (t) In other words, the number of times the constraint fails can also be expressed as Time (w) =Time (s) -Time (t) After each sampling is completed, the M indicator sampling values associated with the N recommended services can be compared with the recommendation constraints. If each recommended service has an indicator sampling value that does not meet the recommendation constraints, it means that the sampling has failed the constraint. At this time, the number of constraint failures can be increased by 1 (Time (s) +=1), thereby updating the constraint failure count Time (s) .
[0111] Furthermore, M indicator constraint weights corresponding to the recommended business i among the N recommended businesses can be obtained; wherein, each of the M indicators has an indicator constraint weight, and the indicator constraint weights corresponding to different indicators can be the same or different. The indicator constraint weights can be set according to the business expert rules. For example, the indicator constraint weight corresponding to each indicator can be determined based on the expert's evaluation opinion on each indicator. For example, based on prior knowledge, it is believed that the importance of indicator 1 is twice that of indicator 2, then the indicator constraint weight of indicator 1 can be set to 2, and the indicator constraint weight of indicator 2 can be set to 1. Alternatively, a statistical model (such as a regression model) can be used to determine the contribution of each indicator, and then the corresponding indicator constraint weight can be derived.
[0112] The weighted sum of the M indicator sampling values and the M indicator constraint weights corresponding to each recommended business is performed to obtain the recommended evaluation value corresponding to each recommended business; wherein the recommended evaluation value can be understood as a weighted average comprehensive indicator. For example, the recommended evaluation value corresponding to the recommended business i can be calculated in, W (i1) It can be understood as the indicator constraint weight corresponding to indicator 1 of the recommended business i. It can be understood as the index sampling value of the recommended business i for index 1; W (i2) It can be understood as the indicator constraint weight corresponding to indicator 2 of the recommended business i. It can be understood as the index sampling value of the recommended business i for index 2; similarly, the recommendation evaluation value corresponding to the recommended business j can be calculated And the recommended evaluation value corresponding to the recommended business k If the recommended evaluation value of the recommended business i Greater than or equal to the recommendation evaluation value corresponding to the recommended business j And greater than or equal to the recommendation evaluation value corresponding to the recommended business k The service i to be recommended can be determined as a candidate service for recommendation. The candidate service for recommendation can be understood as the service to be recommended corresponding to the largest recommendation evaluation value in a sampling when all N services to be recommended have indicator sampling values that do not meet the recommendation constraint conditions.
[0113] Then, the candidate recommendation times corresponding to the service to be recommended i can be updated, and the ratio between the candidate recommendation times and the constraint failure times is determined as the reference allocation probability corresponding to the service to be recommended i; wherein, the candidate recommendation times is used to represent the number of times the service to be recommended i is a candidate recommendation service in the total number of sampling times. In the embodiment of the present application, the candidate recommendation times corresponding to the service to be recommended i can be recorded as After a sampling is completed, if the service to be recommended i is determined as a candidate service to be recommended in this sampling, the number of candidate recommendations corresponding to the service to be recommended i can be increased. After reaching the total number of sampling times, you can (w) The number of candidate recommendations corresponding to the recommended business i Calculate the reference allocation probability P(Arm i(cand) ),in, In other words, the candidate recommendation times corresponding to the recommended business i can be Time and the number of times the constraint fails (w) The ratio between them is determined as the reference allocation probability P(Arm i(cand) It is understandable that the calculation method of the reference allocation probabilities corresponding to other services to be recommended among the N services to be recommended is similar to the calculation method corresponding to the service to be recommended i, and will not be repeated here.
[0114] Furthermore, the reference allocation ratio of traffic corresponding to the N services to be recommended can be determined based on the proportion of the reference allocation probability corresponding to each service to be recommended in the cumulative value of the reference allocation probability corresponding to the N services to be recommended. The cumulative value of the reference allocation probability can be understood as the sum of the probabilities of each service to be recommended being selected when the recommendation constraint conditions are not met; in other words, after calculating the reference allocation probability corresponding to each service to be recommended, the reference allocation probability corresponding to each service to be recommended can be added to obtain the cumulative value of the reference allocation probability. Taking any one of the N services to be recommended (for example, service i to be recommended) as an example, the calculation method of the reference allocation ratio of traffic corresponding to service i to be recommended is described. After obtaining the reference allocation probability P (Arm i(cand) The cumulative value of the reference allocation probability corresponding to N recommended services After that, the reference traffic allocation ratio U corresponding to the recommended service i can be calculated i(c) ,in, In other words, the reference allocation probability P(Arm i(cand) The cumulative value of recommendation probability corresponding to N recommended businesses The ratio between them is determined as the traffic reference allocation ratio U corresponding to the recommended service i i(c) .
[0115] In the embodiment of the present application, the first allocation weight corresponding to the first traffic allocation ratio can be recorded as W (t) , where W (t) =Time (t) / Time (s)In other words, the constraint can be passed through the number of times Time (t) and the total number of sampling times Time (s) The ratio between them is determined as the first allocation weight W (t) Similarly, the second allocation weight corresponding to the traffic reference allocation ratio can be recorded as W (c) , where W (c) =Time (w) / Time (s) In other words, the number of times the constraint fails can be expressed as Time (w) and the total number of sampling times Time (s) The ratio between them is determined as the second allocation weight W (c) Optionally, the second allocation weight W (c) It can also be expressed as W (c) =1-W (t) .
[0116] Step S208: The product of the first traffic allocation ratio and the first allocation weight, and the sum of the product of the traffic reference allocation ratio and the second allocation weight are determined as the second traffic allocation ratio corresponding to the N recommended services, and the N recommended services are pushed in the service platform according to the second traffic allocation ratio.
[0117] In the embodiment of the present application, the first flow distribution ratio can be recorded as U i(t) , the traffic reference allocation ratio is recorded as U i(c) , the second flow distribution ratio is recorded as U i Among them, U i =W (t) *U i(t) +W (c) *U i(c) In other words, the first flow distribution ratio U i(t) and the first allocation weight W (t) The product of the two, and the flow reference distribution ratio U i(c) and the second allocation weight W (c) The sum of the products between them is determined as the second traffic allocation ratio U corresponding to the N recommended services. i The specific method of pushing the N recommended services on the service platform can be referred to the description of step S103 and will not be repeated here.
[0118] Please refer to Figure 8, which is a schematic diagram of a method for calculating the traffic distribution ratio provided by an embodiment of the present application. As shown in Figure 8, the embodiment of the present application mainly relates to the calculation of the traffic distribution ratio corresponding to each arm in the Thompson sampling multi-armed bandit (CMO-TS-MAB) for multi-objective (M>1) optimization. It can be understood that the embodiment of the present application is also applicable to the traffic distribution scenario of single-objective (M=1) optimization. Among them, the multi-armed bandit can include N+1 groups of arms, specifically 1 control group arm, and N experimental group arms. The control group arm here can be understood as the control business involved in the above text, and the N experimental group arms can be understood as the N recommended businesses involved in the above text. In addition, unless otherwise specified, when the embodiment of the present application mentions "arm" alone, it is assumed to be the experimental group arm. Among them, each arm corresponds to M indicators, in other words, each arm corresponds to M indicator sampling values.
[0119] As shown in Figure 8, the prior distribution of each indicator of each arm in the M indicators can be initialized first, and then the prior distribution of each arm in each indicator can be corrected according to the online experimental data to obtain the posterior distribution of each arm in each indicator. The prior distribution here can be understood as the first probability distribution mentioned above, the online experimental data can be understood as the platform collection data mentioned above, and the posterior distribution can be understood as the second probability distribution. The specific correction method can refer to the description in the aforementioned step S101, which will not be repeated here. In an embodiment of the present application, the total number of sampling times can be set to 10,000 times, and the Monte Carlo sampling strategy can be used to obtain the M indicator sampling values corresponding to each arm. The design method of the recommended constraint conditions can refer to the description in the aforementioned step S102, which will not be repeated here.
[0120] In the process of judging whether there is at least one index sampling value corresponding to arm i to arm N in N arms that satisfies the recommended constraint C corresponding to the jth index (any index among M indexes) result When the index sampling value I of each arm for the jth index is i Converted into sampling index vector in, is an N×1 vector, representing the index sampling value from arm i to arm N based on the jth index distribution among N arms; C result It is also an N×1 vector, indicating whether arm i to arm N in the N arms meets the recommendation constraint. Then calculate the weighted sample vector corresponding to the jth indicator for each arm Among them, ⊙ represents the Hadamard product operation, which is also called element-by-element product or matrix dot multiplication. It refers to the new matrix or vector obtained by multiplying two matrices or vectors with the same dimension one by one according to the element position. Then, we can use the weighted sample vector The value of arm i to arm N is used to determine whether there is at least one index sampling value corresponding to an arm that satisfies the recommended constraint condition C corresponding to the jth index. result ; Specifically, when the weighted sample vector When the value of is 1, there is at least one index sampling value corresponding to arm i to arm N in the N arms that satisfies the recommended constraint condition C corresponding to the jth index. result , weighted sample vector When the value of is 0, there are arms from arm i to arm N in the N arms that do not meet the recommended constraint condition C result The indicator sampling value.
[0121] As shown in FIG8 , after 10,000 samplings are completed, the number of times the recommended constraint conditions are met (also called the number of constraint passes) and the number of times the recommended constraint conditions are not met (also called the number of constraint failures) are counted. For example, the number of constraint passes can be Time (t) The number of times the constraint fails can be (10000-Time (t) ) times. If a sampling passes the constraint, the optimal arm can be determined from the arms that meet the recommendation constraint conditions. After 10,000 samplings are completed, the optimal arm probability of each arm is calculated, and the diversion ratio is calculated based on the optimal arm probability of each arm. The optimal arm here can be understood as the optimal recommended business mentioned above, the optimal arm probability can be understood as the recommendation probability mentioned above, and the diversion ratio can be understood as the first diversion ratio U designed above. i(t) The method for obtaining the optimal arm, optimal arm probability, and diversion ratio of the embodiment of the present application can be referred to the description in the aforementioned step S205, which will not be repeated here. After 10,000 samplings are completed, the original diversion ratio can be calculated based on the number of constraint failures. The original diversion ratio here can be understood as the traffic reference allocation ratio U mentioned above. i(c) Therefore, the calculation method of the original diversion ratio in the embodiment of the present application can refer to the description in the aforementioned step S207, which will not be repeated here. Then the weight corresponding to the diversion ratio is calculated (weight W (t) =Time (t) / 10000), and the weight corresponding to the original diversion ratio (weight W (c) =1-W (t) ), and finally adopt a weighted summation processing method to determine the final diversion ratio of each arm, and redistribute the arms according to the final diversion ratio; the final diversion ratio here can be understood as the second flow distribution ratio Ui mentioned above. Therefore, the calculation method of the final diversion ratio of the embodiment of the present application can refer to the description in the aforementioned step S208, which will not be repeated here.
[0122] The method for determining the traffic allocation ratio of multiple services to be recommended involved in the embodiment of the present application can be applied to various recommendation scenarios. In order to verify the effect of the traffic allocation scheme corresponding to the multiple services to be recommended provided in the embodiment of the present application (for the sake of ease of description, it can be referred to as this scheme), this scheme and the single-target Thompson multi-armed bandit algorithm (single-target-TS-MAB) can be applied to the delivery of traffic products of XX Communications Company, and the delivery time on the business platform is 20XX.XX.03-20XX.XX.08. This scheme uses transaction conversion rate and average cost as indicators, and single-target-TS-MAB uses average cost as an indicator; the control groups of this scheme and single-target-TS-MAB select the traffic product with the largest transaction volume in the same period last month or the traffic product recommended by the business expert rules. The performance effects of this scheme and single-target-TS-MAB are shown in Table 3 below:
[0123] Table 3
[0124] As can be seen from Table 3, compared to the single-objective TS-MAB approach, this solution achieves positive improvements in both transaction conversion rate and average cost, demonstrating that it can improve the traffic utilization of each traffic product, thereby enhancing the effectiveness of each traffic product. Furthermore, transaction conversion rate and average cost are negatively correlated metrics, demonstrating that this solution is also effective in recommendation scenarios where optimization objectives are negatively correlated.
[0125] In an embodiment of the present application, in the process of allocating traffic for N services to be recommended, by setting multiple indicators for the N services to be recommended, each indicator can be configured with a corresponding recommendation constraint condition, and multiple indicators and the recommendation constraint conditions corresponding to each indicator are used as comprehensive considerations in traffic allocation. The rationality of the final traffic allocation ratio of each service to be recommended can be improved, and the traffic utilization rate for the N services to be recommended can be improved.
[0126] It is understandable that in the specific implementation of this application, relevant information of the user in the business platform may be involved (for example, the user's login information, basic information, etc. in the business platform). When the above embodiments of this application are applied to specific products or technologies, the user's permission or consent must be obtained, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0127] Please refer to Figure 9, which is a schematic diagram of the structure of a business data processing device provided in an embodiment of the present application. It is understood that the business data processing device can be applied to the server 10d or terminal device (for example, terminal device 10a) shown in Figure 1. As shown in Figure 9, the business data processing device 1 may include: an indicator sampling value acquisition module 11, a first traffic distribution ratio acquisition module 12, and a second traffic distribution ratio acquisition module 13, wherein:
[0128] The indicator sampling value acquisition module 11 is used to obtain N services to be recommended and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1;
[0129] The first traffic allocation ratio acquisition module 12 is configured to obtain recommendation probabilities corresponding to N services to be recommended if all M indicator sampling values corresponding to the services to be recommended satisfy the recommendation constraint conditions, and determine the first traffic allocation ratios corresponding to the N services to be recommended based on the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended;
[0130] The second traffic allocation ratio acquisition module 13 is configured to perform a weighted summation of the first traffic allocation ratio and the reference traffic allocation ratio to obtain second traffic allocation ratios corresponding to N services to be recommended, and push the N services to be recommended on the service platform according to the second traffic allocation ratios;
[0131] The traffic reference allocation ratio refers to the traffic allocation ratio adopted by N services to be recommended when each service to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
[0132] Among them, the specific functional implementation methods of the indicator sampling value acquisition module 11, the first flow distribution ratio acquisition module 12 and the second flow distribution ratio acquisition module 13 can be referred to steps S101 to S103 in the embodiment corresponding to Figure 3 above, and will not be repeated here.
[0133] In some possible implementations, the indicator sampling value acquisition module 11 may include: a probability distribution construction unit 111, a probability distribution correction unit 112, and a sampling processing unit 113, wherein:
[0134] The probability distribution construction unit 111 is configured to obtain M indicators associated with the N services to be recommended, and to construct a first probability distribution corresponding to the jth indicator among the M indicators; j is a positive integer less than or equal to M;
[0135] The probability distribution correction unit 112 is used to obtain the platform collected data corresponding to the j-th indicator, and correct the first probability distribution according to the platform collected data to obtain a second probability distribution;
[0136] The sampling processing unit 113 is configured to perform sampling processing on each of the N services to be recommended according to the second probability distribution, and obtain an indicator sampling value of each service to be recommended for the j-th indicator.
[0137] The specific functional implementation of the probability distribution construction unit 111, the probability distribution correction unit 112 and the sampling processing unit 113 can refer to step S101 in the embodiment corresponding to FIG3 above, which will not be repeated here.
[0138] In some possible implementations, the business data processing device 1 may further include: an indicator conversion module 14, a first comparison module 15, and a second comparison module 16, wherein:
[0139] The indicator conversion module 14 is used to convert each of the M indicators associated with the N services to be recommended into a sub-constraint condition; a is a positive integer;
[0140] The first comparison module 15 is configured to determine that the j-th indicator sample value of the to-be-recommended service i satisfies the recommendation constraint condition corresponding to the j-th indicator if the j-th indicator sample value of the to-be-recommended service i satisfies a sub-constraint conditions corresponding to the j-th indicator; the to-be-recommended service i belongs to N to-be-recommended services, the j-th indicator sample value refers to the indicator sample value of the to-be-recommended service i for the j-th indicator, different indicators correspond to different recommendation constraints, and j is a positive integer less than or equal to M;
[0141] The second comparison module 16 is configured to determine that the recommended service i does not satisfy the recommendation constraint corresponding to the jth indicator if the jth indicator sampling value of the recommended service i does not satisfy one or more sub-constraints among the a sub-constraints corresponding to the jth indicator.
[0142] The specific functional implementation of the indicator conversion module 14 , the first comparison module 15 and the second comparison module 16 can be found in step S102 of the embodiment corresponding to FIG3 , which will not be described in detail here.
[0143] In some possible implementations, the indicator conversion module 14 may further include: an indicator type acquisition unit 141, a first conversion unit 142, and a second conversion unit 143, wherein:
[0144] The indicator type acquisition unit 141 is used to acquire the indicator type corresponding to the j-th indicator among the M indicators associated with the N services to be recommended;
[0145] A first conversion unit 142 is configured to obtain a first constraint design strategy corresponding to the ratio type if the indicator type is a ratio type, and convert the jth indicator into a sub-constraint conditions according to the first constraint design strategy;
[0146] The second conversion unit 143 is configured to obtain a second constraint design strategy corresponding to the mean type if the indicator type is the mean type, and convert the j-th indicator into a sub-constraint conditions according to the second constraint design strategy.
[0147] The specific functional implementation of the indicator type acquisition unit 141 , the first conversion unit 142 and the second conversion unit 143 can be found in step S102 in the embodiment corresponding to FIG. 3 , and will not be described in detail here.
[0148] In some possible implementations, the indicator conversion module 14 is specifically configured to:
[0149] Determine the target value or target range corresponding to the jth indicator among the M indicators associated with the N services to be recommended;
[0150] Convert the target value or target range into corresponding sub-constraints to obtain a sub-constraints corresponding to the j-th indicator.
[0151] In some possible implementations, the first traffic distribution ratio acquisition module 12 may include: a pass count updating unit 121, a first determining unit 122, and a recommendation probability determining unit 123, wherein:
[0152] The pass count updating unit 121 is configured to update the constraint pass count if all M indicator sample values corresponding to the service to be recommended satisfy the recommendation constraint condition. The constraint pass count indicates the number of times, within the total number of sampling times, that all M indicator sample values corresponding to the service to be recommended satisfy the recommendation constraint condition.
[0153] The first determining unit 122 is configured to group the services to be recommended that meet the recommendation constraint conditions into a recommended service set, and if the M indicator sample values corresponding to the service to be recommended i in the recommended service set are greater than or equal to the M indicator sample values corresponding to the services to be recommended other than the service to be recommended i in the recommended service set, then determine the service to be recommended i as the optimal recommended service;
[0154] The recommendation probability determination unit 123 is used to update the optimal recommendation number corresponding to the service to be recommended i, and determine the ratio between the optimal recommendation number and the constraint passing number as the recommendation probability corresponding to the service to be recommended i; the optimal recommendation number is used to represent the number of times that the service to be recommended i is the optimal recommended service in the total number of sampling times.
[0155] The specific functional implementation of the pass count updating unit 121 , the first determining unit 122 and the recommendation probability determining unit 123 may refer to steps S202 to S204 in the embodiment corresponding to FIG. 7 , which will not be described in detail here.
[0156] In some possible implementations, the first traffic distribution ratio acquisition module 12 may further include: a second determination unit 124, wherein:
[0157] The second determination unit 124 is used to determine all the to-be-recommended businesses contained in the recommended business set as optimal recommended businesses, and update the optimal number of recommendations corresponding to the to-be-recommended businesses contained in the recommended business set, if the M indicator sampling values corresponding to the to-be-recommended businesses do not exist in the recommended business set and are greater than or equal to the M indicator sampling values corresponding to the remaining to-be-recommended businesses in the recommended business set.
[0158] The specific functional implementation of the second determining unit 124 can refer to step S204 in the embodiment corresponding to FIG. 7 , which will not be described in detail here.
[0159] In some possible implementations, the service data processing device 1 may further include: a failure count updating module 17, a weighted summing module 18, a candidate recommended service determining module 19, a reference allocation probability acquiring module 20, and a reference allocation ratio acquiring module 21, wherein:
[0160] The failure count update module 17 is configured to update the constraint failure count if each service to be recommended has an indicator sampling value that does not satisfy the recommendation constraint condition; the constraint failure count refers to the difference between the total number of sampling times and the constraint pass count;
[0161] The weighted summation module 18 is used to obtain M indicator constraint weights corresponding to the recommended business i among the N recommended businesses, and perform weighted summation on the M indicator sampling values and the M indicator constraint weights corresponding to the recommended business i to obtain a recommendation evaluation value corresponding to the recommended business i;
[0162] The candidate recommended service determination module 19 is configured to determine the service i as a candidate recommended service if the recommendation evaluation value corresponding to the service i is greater than or equal to the recommendation evaluation value corresponding to the N services other than the service i.
[0163] The reference allocation probability acquisition module 20 is used to update the candidate recommendation count corresponding to the service to be recommended i, and determine the reference allocation probability corresponding to the service to be recommended i as the ratio between the candidate recommendation count and the constraint failure count; the candidate recommendation count is used to represent the number of times the service to be recommended i is a candidate recommendation service in the total number of sampling times;
[0164] The reference allocation ratio acquisition module 21 is used to determine the traffic reference allocation ratio corresponding to the N services to be recommended according to the proportion of the reference allocation probability corresponding to each service to be recommended in the cumulative value of the reference allocation probabilities corresponding to the N services to be recommended.
[0165] Among them, the specific functional implementation methods of the failure number update module 17, the weighted sum module 18, the candidate recommended business determination module 19, the reference allocation probability acquisition module 20 and the reference allocation ratio acquisition module 21 can be found in step S205 of the embodiment corresponding to Figure 7 above, and will not be repeated here.
[0166] In some possible implementations, the second traffic distribution ratio acquisition module 13 may include: a first distribution weight determination unit 131, a second distribution weight determination unit 132, and a traffic distribution ratio determination unit 133, wherein:
[0167] A first allocation weight determining unit 131 is configured to determine a ratio between the number of constraint passes and the total number of sampling times as a first allocation weight corresponding to the first traffic allocation ratio;
[0168] A second allocation weight determination unit 132 is configured to determine a ratio between the number of constraint failures and the total number of samplings as a second allocation weight corresponding to the reference flow allocation ratio;
[0169] The traffic allocation ratio determining unit 133 is configured to determine the sum of the product of the first traffic allocation ratio and the first allocation weight and the product of the reference traffic allocation ratio and the second allocation weight as the second traffic allocation ratio corresponding to the N services to be recommended.
[0170] Among them, the specific functional implementation of the first allocation weight determination unit 131, the second allocation weight determination unit 132 and the flow distribution ratio determination unit 133 can refer to steps S206 to S208 in the embodiment corresponding to Figure 7 above, and will not be repeated here.
[0171] In some possible implementations, the second traffic allocation ratio acquisition module 13 may include: a recommended traffic determination unit 134 and a service push unit 135, wherein:
[0172] The recommended traffic flow determination unit 134 is configured to determine the recommended traffic flow corresponding to each of the N services to be recommended according to the second traffic flow allocation ratio, and determine the delivery area and delivery duration corresponding to each service to be recommended according to the recommended traffic flow;
[0173] The service push unit 135 is configured to push N services to be recommended in the service platform according to the delivery area and delivery duration.
[0174] The specific functional implementation of the recommended traffic determination unit 134 and the service push unit 135 can refer to step S103 in the embodiment corresponding to FIG. 3 , which will not be described in detail here.
[0175] In some possible implementations, the recommended traffic determination unit 134 may include: an object feature acquisition subunit 1341, a similarity determination subunit 1342, a delivery area determination subunit 1343, and a delivery duration determination subunit 1344, wherein:
[0176] The object feature acquisition subunit 1341 is used to acquire a set of registered objects in the business platform, and acquire object features corresponding to each registered object in the set of registered objects;
[0177] The similarity determination subunit 1342 is configured to obtain a service feature corresponding to a service to be recommended i among the N services to be recommended, and determine the similarity between the service feature corresponding to the service to be recommended i and the object feature corresponding to each registered object;
[0178] The delivery area determination subunit 1343 is configured to group the registered objects corresponding to the object features having a similarity greater than a similarity threshold into candidate delivery areas corresponding to the service to be recommended i, and determine the delivery area of the service to be recommended i within the candidate delivery areas based on the recommended traffic corresponding to the service to be recommended i;
[0179] The delivery duration determining subunit 1344 is configured to obtain the total delivery durations corresponding to N services to be recommended, and determine the delivery duration corresponding to the service to be recommended i according to the recommended traffic corresponding to the service to be recommended i.
[0180] Among them, the specific functional implementation methods of the object feature acquisition subunit 1341, the similarity determination subunit 1342, the delivery area range determination subunit 1343 and the delivery duration determination subunit 1344 can be referred to step S103 in the embodiment corresponding to Figure 3 above, and will not be repeated here.
[0181] According to one embodiment of the present application, the various modules in the business data processing device 1 shown in Figure 9 can be separately or all combined into one or several units to constitute, or one (some) of the units can be further divided into at least two functionally smaller sub-units, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above modules are divided based on logical functions. In actual applications, the functions of one module can also be implemented by at least two units, or the functions of at least two modules can be implemented by one unit. In other embodiments of the present application, the data processing device 1 can also include other units. In actual applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of at least two units.
[0182] In an embodiment of the present application, after obtaining N services to be recommended and M indicator sampling values corresponding to each service to be recommended, the M indicator sampling values associated with the N services to be recommended can be compared with the recommendation constraints. If one or more of the M indicator sampling values corresponding to the N services to be recommended satisfy the recommendation constraints, the recommendation probability of each service to be recommended under the premise of satisfying the recommendation constraints can be obtained, and then the first traffic allocation ratio of each service to be recommended under the premise of satisfying the recommendation constraints can be determined based on the recommendation probability. The final traffic allocation ratio of each service to be recommended is obtained by weighted summing the first traffic allocation ratio and the traffic reference allocation ratio (the traffic allocation ratio of each service to be recommended under the premise of not satisfying the recommendation constraints). In the process of traffic allocation for N services to be recommended, by setting multiple indicators for the N services to be recommended, each indicator can be configured with its corresponding recommendation constraints, and using multiple indicators and the recommendation constraints corresponding to each indicator as comprehensive considerations in traffic allocation can improve the rationality of the final traffic allocation ratio of each service to be recommended, and thus improve the traffic utilization rate for the N services to be recommended.
[0183] Please refer to Figure 10, which is a structural diagram of a computer device provided in an embodiment of the present application. As shown in Figure 10, the computer device 1000 can be the server 10d in the embodiment corresponding to Figure 1, or it can be the terminal device 10a in the embodiment corresponding to Figure 1, and it will not be limited here. For ease of understanding, the embodiment of the present application takes the computer device as an example of a terminal device. The computer device 1000 may include: a processor 1001, a network interface 1004 and a memory 1005. In addition, the computer device 1000 may also include: a user interface 1003, and at least one communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. Among them, the user interface 1003 may include a display screen (Display), a keyboard (Keyboard), and optionally, the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 1005 may optionally be at least one storage device located away from the processor 1001. As shown in FIG10 , the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a device control application.
[0184] In the computer device 1000 shown in FIG10 , the network interface 1004 can provide network communication functions; the user interface 1003 is mainly used to provide an interface for user input; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:
[0185] Obtain N services to be recommended, and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1;
[0186] If all the sample values of the indicators corresponding to the services to be recommended satisfy the recommendation constraints, obtain the recommendation probabilities corresponding to the N services to be recommended, and determine the first traffic allocation ratios corresponding to the N services to be recommended based on the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended;
[0187] Performing a weighted summation of the first traffic allocation ratio and the reference traffic allocation ratio to obtain a second traffic allocation ratio corresponding to N services to be recommended, and pushing the N services to be recommended on the service platform according to the second traffic allocation ratio;
[0188] The traffic reference allocation ratio refers to the traffic allocation ratio adopted by N services to be recommended when each service to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
[0189] It should be understood that the computer device 1000 described in the embodiments of the present application can execute the description of the business data processing method in the embodiments corresponding to Figures 3 and 7 above, and can also execute the description of the business data processing device 1 in the embodiment corresponding to Figure 9 above, and will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated here.
[0190] In addition, it should be noted that the present application embodiment also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program executed by the business data processing device 1 mentioned above, and the computer program includes program instructions. When the processor executes the program instructions, it can execute the description of the business data processing method in any of the embodiments corresponding to Figures 3 and 7 above. Therefore, it will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application. As an example, the program instructions can be deployed on a computing device for execution, or executed on multiple computing devices located at one location, or executed on multiple computing devices distributed at multiple locations and interconnected by a communication network. Multiple computing devices distributed at multiple locations and interconnected by a communication network can constitute a blockchain system.
[0191] In addition, it should be noted that: the embodiment of the present application also provides a computer program product or computer program, which may include computer instructions, and the computer instructions may be stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor can execute the computer instructions, so that the computer device executes the description of the business data processing method in the embodiment corresponding to any one of Figures 3 and 7 above, and therefore, will not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated. For technical details not disclosed in the computer program product or computer program embodiment involved in this application, please refer to the description of the method embodiment of this application.
[0192] The terms "first", "second", etc. in the description, claims, and drawings of the embodiments of the present application are used to distinguish different media contents, rather than to describe a specific order. In addition, the terms "including" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules that are not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.
[0193] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0194] The method and related apparatus provided in the embodiment of the present application are described with reference to the method flow chart and / or structural diagram provided in the embodiment of the present application, and specifically can be implemented by computer program instructions for each process and / or box of the method flow chart and / or structural diagram, and the combination of the process and / or box in the flow chart and / or block diagram. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the function specified in one process or multiple processes of the flow chart and / or one box or multiple boxes of the structural diagram. These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, and the instruction device implements the function specified in one process or multiple processes of the flow chart and / or one box or multiple boxes of the structural diagram. These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the structural diagram.
[0195] The above disclosure is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.
Claims
1. A business data processing method, It is characterized in that The method is performed by a computer device, and the method comprises: Obtain N services to be recommended, and obtain M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1; If there are M indicator sampling values corresponding to the services to be recommended that all satisfy the recommendation constraint conditions, then obtaining the recommendation probabilities corresponding to the N services to be recommended, and determining the first traffic allocation ratios corresponding to the N services to be recommended according to the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended; Performing a weighted summation of the first traffic allocation ratio and the traffic reference allocation ratio to obtain a second traffic allocation ratio corresponding to the N services to be recommended, and pushing the N services to be recommended in the service platform according to the second traffic allocation ratio; The traffic reference allocation ratio refers to the traffic allocation ratio adopted by the N services to be recommended when each of the services to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
2. The method according to claim 1, It is characterized in that The obtaining of M indicator sampling values corresponding to each of the N services to be recommended includes: Acquire M indicators associated with the N services to be recommended, and construct a first probability distribution corresponding to the jth indicator among the M indicators; j is a positive integer less than or equal to M; Acquire the platform collected data corresponding to the j-th indicator, and modify the first probability distribution according to the platform collected data to obtain a second probability distribution; A sampling process is performed on each of the N services to be recommended according to the second probability distribution to obtain an indicator sampling value of each of the services to be recommended for the j-th indicator.
3. The method according to claim 1, It is characterized in that The method further comprises: Each of the M indicators associated with the N services to be recommended is converted into a sub-constraint condition; a is a positive integer; If the j-th indicator sampling value of the service i to be recommended satisfies a sub-constraints corresponding to the j-th indicator, it is determined that the j-th indicator sampling value of the service i to be recommended satisfies the recommendation constraint corresponding to the j-th indicator; the service i to be recommended belongs to the N services to be recommended, the j-th indicator sampling value refers to the indicator sampling value of the service i to be recommended for the j-th indicator, different indicators correspond to different recommendation constraints, and j is a positive integer less than or equal to M; If the j-th indicator sampling value of the service to be recommended i does not satisfy one or more sub-constraints among the a sub-constraints corresponding to the j-th indicator, it is determined that the service to be recommended i does not satisfy the recommendation constraint corresponding to the j-th indicator.
4. The method according to claim 3, It is characterized in that The step of converting each of the M indicators associated with the N services to be recommended into a constraint condition includes: Obtaining the indicator type corresponding to the jth indicator among the M indicators associated with the N services to be recommended; If the indicator type is a ratio type, obtaining a first constraint design strategy corresponding to the ratio type, and converting the j-th indicator into a sub-constraint conditions according to the first constraint design strategy; If the indicator type is a mean type, a second constraint design strategy corresponding to the mean type is obtained, and the j-th indicator is converted into a sub-constraint conditions according to the second constraint design strategy.
5. The method according to claim 3, It is characterized in that The step of converting each of the M indicators associated with the N services to be recommended into a constraint condition includes: Determine a target value or a target range corresponding to the jth indicator among the M indicators associated with the N services to be recommended; The target value or target range is converted into corresponding sub-constraint conditions to obtain a sub-constraint conditions corresponding to the j-th indicator.
6. The method according to claim 1, It is characterized in that If all the M indicator sampling values corresponding to the services to be recommended satisfy the recommendation constraint conditions, obtaining the recommendation probabilities corresponding to the N services to be recommended includes: If there are M indicator sampling values corresponding to the service to be recommended that all satisfy the recommendation constraint conditions, then the constraint passing number is updated; the constraint passing number is used to represent the number of times in the total number of samplings that the M indicator sampling values corresponding to the service to be recommended all satisfy the recommendation constraint conditions; The services to be recommended that meet the recommendation constraint conditions are formed into a recommended service set. If the M index sampling values corresponding to the service to be recommended i in the recommended service set are greater than or equal to the M index sampling values corresponding to the services to be recommended other than the service to be recommended i in the recommended service set, the service to be recommended i is determined as the optimal recommended service; Update the optimal recommendation number corresponding to the service to be recommended i, and determine the ratio between the optimal recommendation number and the constraint passing number as the recommendation probability corresponding to the service to be recommended i; the optimal recommendation number is used to characterize the number of times that the service to be recommended i is the optimal recommended service in the total sampling number.
7. The method according to claim 6, It is characterized in that The method further comprises: If there are no M indicator sampling values corresponding to the to-be-recommended services in the recommended service set that are greater than or equal to the remaining M indicator sampling values corresponding to the to-be-recommended services in the recommended service set, then all the to-be-recommended services contained in the recommended service set are determined as the optimal recommended services, and the optimal number of recommendations corresponding to the to-be-recommended services contained in the recommended service set is updated.
8. The method according to claim 6, It is characterized in that The method further comprises: If each of the services to be recommended has an indicator sampling value that does not satisfy the recommendation constraint condition, then the constraint failure times are updated; the constraint failure times refer to the difference between the total sampling times and the constraint passing times; Obtaining M indicator constraint weights corresponding to a service to be recommended i among the N services to be recommended, performing weighted summation on the M indicator sampling values corresponding to the service to be recommended i and the M indicator constraint weights, to obtain a recommendation evaluation value corresponding to the service to be recommended i; If the recommendation evaluation value corresponding to the to-be-recommended service i is greater than or equal to the recommendation evaluation value corresponding to the to-be-recommended service other than the to-be-recommended service i among the N to-be-recommended services, then the to-be-recommended service i is determined as a candidate recommended service; Updating the candidate recommendation times corresponding to the service to be recommended i, and determining the ratio between the candidate recommendation times and the constraint failure times as the reference allocation probability corresponding to the service to be recommended i; the candidate recommendation times are used to represent the number of times the service to be recommended i is a candidate recommendation service in the total number of sampling times; The reference allocation ratio of traffic corresponding to the N services to be recommended is determined according to the proportion of the reference allocation probability corresponding to each service to be recommended in the cumulative value of the reference allocation probabilities corresponding to the N services to be recommended.
9. The method according to claim 1, It is characterized in that The step of performing a weighted summation on the first traffic allocation ratio and the traffic reference allocation ratio to obtain a second traffic allocation ratio corresponding to the N services to be recommended includes: Determine the ratio between the number of constraint passes and the total number of samplings as a first allocation weight corresponding to the first flow allocation ratio; Determine the ratio between the number of times the constraint is not passed and the total number of sampling times as the second allocation weight corresponding to the reference flow allocation ratio; The sum of the product of the first traffic allocation ratio and the first allocation weight and the product of the reference traffic allocation ratio and the second allocation weight is determined as the second traffic allocation ratio corresponding to the N services to be recommended.
10. The method according to claim 1, It is characterized in that The pushing the N to-be-recommended services in the service platform according to the second traffic allocation ratio includes: Determine the recommended traffic corresponding to each of the N services to be recommended according to the second traffic allocation ratio, and determine the delivery area range and delivery duration corresponding to each of the services to be recommended according to the recommended traffic; According to the delivery area range and the delivery duration, the N services to be recommended are pushed in the service platform.
11. The method according to claim 10, It is characterized in that The determining, according to the recommended traffic, a delivery area range and delivery duration corresponding to each of the services to be recommended, includes: Acquire a set of registered objects in the business platform, and acquire object characteristics corresponding to each registered object in the set of registered objects; Obtaining a service feature corresponding to a service to be recommended i among the N services to be recommended, and determining a similarity between the service feature corresponding to the service to be recommended i and an object feature corresponding to each registered object; The registered objects corresponding to the object features whose similarity is greater than the similarity threshold are combined into the candidate delivery range corresponding to the to-be-recommended service i, and the delivery area range of the to-be-recommended service i is determined within the candidate delivery range according to the recommended traffic corresponding to the to-be-recommended service i; The total delivery durations corresponding to the N services to be recommended are obtained, and the delivery duration corresponding to the service to be recommended i is determined according to the recommended traffic corresponding to the service to be recommended i.
12. A business data processing device, It is characterized in that include: An indicator sampling value acquisition module, used to acquire N services to be recommended, and acquire M indicator sampling values corresponding to each of the N services to be recommended; M and N are both integers greater than 1; A first traffic allocation ratio acquisition module is used to obtain the recommendation probabilities corresponding to the N services to be recommended if all the M indicator sampling values corresponding to the services to be recommended satisfy the recommendation constraint conditions, and determine the first traffic allocation ratios corresponding to the N services to be recommended according to the proportion of the recommendation probability corresponding to each service to be recommended in the cumulative value of the recommendation probabilities corresponding to the N services to be recommended; A second traffic allocation ratio acquisition module is used to perform a weighted summation of the first traffic allocation ratio and the traffic reference allocation ratio to obtain a second traffic allocation ratio corresponding to the N services to be recommended, and push the N services to be recommended in the service platform according to the second traffic allocation ratio; The traffic reference allocation ratio refers to the traffic allocation ratio adopted by the N services to be recommended when each of the services to be recommended has an indicator sampling value that does not meet the recommendation constraint condition.
13. A computer device, It is characterized in that The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 11.
14. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 11 are executed.
15. A computer program product, It is characterized in that The method comprises a computer program / instruction, which implements the method according to any one of claims 1 to 11 when executed by a processor.
Citation Information
Patent Citations
Flow distribution method and apparatus and service distribution method and apparatus for information pushing orders
CN108428144A
Content recommendation method, apparatus and system, and electronic device
CN108563681A
Information recommendation method, related device, equipment and storage medium
CN114461893A
Dynamic Content Recommendations
US20150370818A1
