Recommendation information pushing method and device, computer device and storage medium

Abstract

The application relates to a cost control method and device of recommendation information, a computer device and a storage medium. The method comprises the following steps: obtaining a target conversion cost value of recommendation information in a push cycle; the push cycle comprises a historical push sub-cycle in which information push has been performed and a future push sub-cycle in which information push has not been performed; reading historical push data of each push crowd in the historical push sub-cycle; determining a value interval of a pricing coefficient corresponding to each push crowd; determining a real-time pricing coefficient corresponding to each push crowd based on the target conversion cost value, the historical push data and the value interval; and determining a conversion cost value when each push crowd is pushed in the future push sub-cycle based on the real-time pricing coefficient and the target conversion cost value. The method can improve the accuracy of push cost control of recommendation information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a recommended information pushing method and device, computer equipment and storage medium. BACKGROUND

[0002] Recommended information is information delivered to the public through a certain form of media, including video, news, articles, advertisements and other media information. Previously, recommended information usually delivered its business-related information to the public through traditional media (such as newspapers, magazines, radio, television, etc.). In recent years, with the rapid development of the Internet, the Internet recommended information has become an important part of modern marketing strategy because of its fast speed and good effect.

[0003] The existing cost control method usually adjusts the recommended information bid at each push time as time goes on within any one recommended information pushing period, so that the actual conversion cost value of the pushed recommended information can reach the preset target conversion cost value at the end of the entire pushing period, thereby realizing the control of the recommended information pushing cost.

[0004] However, the same recommended information can bring different benefits to different people, and the existing cost control method only adjusts the price based on the recommended information granularity, thereby resulting in low accuracy of recommended information pushing cost control. SUMMARY

[0005] Therefore, it is necessary to provide a recommended information pushing method and device, computer equipment and storage medium capable of improving the information pushing effect to solve the above technical problems.

[0006] A recommended information cost control method, the method comprising:

[0007] obtaining a target conversion cost value of recommended information within a pushing period; the pushing period includes a historical pushing sub-period in which information pushing has been performed and a future pushing sub-period in which information pushing has not been performed;

[0008] reading historical pushing data of each pushing population within the historical pushing sub-period; the pushing population is obtained by grouping each pushing object according to a population division index, and the population division index includes a pushing index, and the pushing index includes a conversion rate and a return on investment;

[0009] determining a value range of a price adjustment coefficient corresponding to each pushing population;

[0010] determining a real-time price adjustment coefficient corresponding to each pushing population based on the target conversion cost value, the historical pushing data and the value range;

[0011] determining a conversion cost value when pushing to each of the push crowds in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value;

[0012] receiving an information recommendation request, the information recommendation request carrying user information of a requester;

[0013] determining a push crowd to which the requester belongs based on the user information;

[0014] obtaining a conversion cost value of the push crowd to which the requester belongs with respect to the recommended information;

[0015] obtaining conversion cost values of the push crowd to which the requester belongs with respect to other recommended information;

[0016] calculating a matching degree of the requester with the recommended information and a matching degree of the requester with the other recommended information;

[0017] calculating a push score value of the recommended information according to the conversion cost value of the recommended information and the matching degree of the requester with the recommended information, and calculating a push score value of the other recommended information according to the conversion cost value of the other recommended information and the matching degree of the requester with the other recommended information;

[0018] selecting to-be-recommended information from the recommended information and the other recommended information according to the push score value of the recommended information and the push score value of the other recommended information;

[0019] if the to-be-recommended information selected is the recommended information, pushing the recommended information to the requester according to the conversion cost value of the recommended information.

[0020] A cost control device of recommended information, the device comprising:

[0021] a target data obtaining module configured to obtain a target conversion cost value of recommended information in a push period; the push period comprising a historical push sub-period in which information has been pushed and a future push sub-period in which information has not been pushed;

[0022] a historical data obtaining module configured to read historical push data of each push crowd in the historical push sub-period;

[0023] a value interval determining module configured to determine a value interval of a pricing coefficient corresponding to each push crowd; the push crowd being obtained by group division of each push object according to a crowd division index; the crowd division index comprising a push index; the push index comprising a conversion rate and a return on investment;

[0024] A price adjustment coefficient determination module is configured to determine a real-time price adjustment coefficient corresponding to each of the push crowds based on the target conversion cost value, the historical push data, and the value range;

[0025] A conversion cost value determination module is configured to determine a conversion cost value for each of the push crowds in a future push sub-period based on the real-time price adjustment coefficient and the target conversion cost value;

[0026] A request receiving module is configured to receive an information recommendation request, which carries user information of a requester;

[0027] A crowd determination module is configured to determine a push crowd to which the requester belongs based on the user information;

[0028] A conversion cost value acquisition module is configured to acquire a conversion cost value of the recommendation information for the push crowd to which the requester belongs, and a conversion cost value of other recommendation information for the push crowd to which the requester belongs;

[0029] An information push module is configured to calculate a matching degree between the requester and the recommendation information, calculate a matching degree between the requester and other recommendation information, calculate a push score value of the recommendation information based on the conversion cost value of the recommendation information and the matching degree between the requester and the recommendation information, calculate a push score value of the other recommendation information based on the conversion cost value of the other recommendation information and the matching degree between the requester and the other recommendation information, select to-be-recommended information from the recommendation information and the other recommendation information based on the push score value of the recommendation information and the push score value of the other recommendation information, and push the recommendation information to the requester according to the conversion cost value of the recommendation information if the to-be-recommended information is the recommendation information.

[0030] A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0031] A target conversion cost value of recommendation information in a push period is acquired; the push period includes a historical push sub-period in which information push has been performed and a future push sub-period in which information push has not been performed;

[0032] Historical push data of each push crowd in the historical push sub-period is read; the push crowds are obtained by performing group division on each push object according to a crowd division index, and the crowd division index includes a push index, and the push index includes a conversion rate and a return on investment;

[0033] A value range of a price adjustment coefficient corresponding to each of the push crowds is determined;

[0034] Determine a real-time pricing coefficient corresponding to each of the push crowds based on the target conversion cost value, the historical push data and the value interval;

[0035] Determine a conversion cost value when pushing to each of the push crowds in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value;

[0036] Receive an information recommendation request, the information recommendation request carrying user information of a requester;

[0037] Determine a push crowd to which the requester belongs based on the user information;

[0038] Obtain a conversion cost value of the push crowd to which the requester belongs with respect to the recommended information;

[0039] Obtain conversion cost values of the push crowd to which the requester belongs with respect to other recommended information;

[0040] Calculate a matching degree of the requester with the recommended information and a matching degree of the requester with other recommended information;

[0041] Calculate a push score value of the recommended information based on the conversion cost value of the recommended information and the matching degree of the requester with the recommended information, and calculate push score values of the other recommended information based on the conversion cost values of the other recommended information and the matching degrees of the requester with the other recommended information;

[0042] Select to-be-recommended information from the recommended information and the other recommended information based on the push score value of the recommended information and the push score values of the other recommended information;

[0043] If the selected to-be-recommended information is the recommended information, push the recommended information to the requester according to the conversion cost value of the recommended information.

[0044] A computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the following steps:

[0045] Obtain a target conversion cost value of recommended information in a push period; the push period includes a historical push sub-period in which information has been pushed and a future push sub-period in which information has not been pushed;

[0046] Read historical push data of each push crowd in the historical push sub-period; the push crowd is obtained by grouping each push object according to a crowd grouping index, and the crowd grouping index includes a push index, the push index including a conversion rate and a return on investment;

[0047] determining a value range of a pricing coefficient corresponding to each of the push crowds;

[0048] determining a real-time pricing coefficient corresponding to each of the push crowds based on the target conversion cost value, the historical push data and the value range;

[0049] determining a conversion cost value when pushing to each of the push crowds in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value;

[0050] receiving an information recommendation request, the information recommendation request carrying user information of a requester;

[0051] determining a push crowd to which the requester belongs based on the user information;

[0052] obtaining a conversion cost value of the push crowd to which the requester belongs with respect to the recommended information;

[0053] obtaining conversion cost values of the push crowd to which the requester belongs with respect to other recommended information;

[0054] calculating a matching degree of the requester with the recommended information and a matching degree of the requester with other recommended information;

[0055] calculating a push score value of the recommended information according to the conversion cost value of the recommended information and the matching degree of the requester with the recommended information, and calculating a push score value of the other recommended information according to the conversion cost values of the other recommended information and the matching degree of the requester with the other recommended information;

[0056] selecting to-be-recommended information from the recommended information and the other recommended information according to the push score value of the recommended information and the push score value of the other recommended information;

[0057] if the selected to-be-recommended information is the recommended information, pushing the recommended information to the requester according to the conversion cost value of the recommended information.

[0058] A computer program, comprising computer instructions stored in a computer readable storage medium, a processor of a computer device reading the computer instructions from the computer readable storage medium, and the processor executing the computer instructions to enable the computer device to perform the following steps:

[0059] obtaining a target conversion cost value of recommended information in a push period; the push period comprising a historical push sub-period in which information has been pushed and a future push sub-period in which information has not been pushed;

[0060] read historical push data of each push group in the historical push sub-period; the push group is obtained by grouping each push object according to a group division index, and the group division index includes a push index, and the push index includes a conversion rate and a return on investment;

[0061] determine a value range of a pricing coefficient corresponding to each push group;

[0062] determine a real-time pricing coefficient corresponding to each push group based on the target conversion cost value, the historical push data and the value range;

[0063] determine a conversion cost value when pushing to each push group in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value;

[0064] receive an information recommendation request, and the information recommendation request carries user information of a requester;

[0065] determine a push group to which the requester belongs based on the user information;

[0066] obtain a conversion cost value of the push group about the recommended information;

[0067] obtain a conversion cost value of the push group about other recommended information;

[0068] calculate a matching degree of the requester and the recommended information, and calculate a matching degree of the requester and other recommended information;

[0069] calculate a push score value of the recommended information according to the conversion cost value of the recommended information and the matching degree of the requester and the recommended information, and calculate a push score value of the other recommended information according to the conversion cost value of the other recommended information and the matching degree of the requester and the other recommended information;

[0070] select a to-be-recommended information from the recommended information and the other recommended information according to the push score value of the recommended information and the push score value of the other recommended information;

[0071] if the selected to-be-recommended information is the recommended information, push the recommended information to the requester according to the conversion cost value of the recommended information.

[0072] The cost control method, device, computer device and storage medium of recommendation information provided in the above embodiments can obtain the target conversion cost value of the recommendation information in the push cycle, read the historical push data of each push crowd in the historical push sub-cycle, and determine the value range of the pricing coefficient corresponding to each push crowd, so as to determine the real-time pricing coefficient corresponding to each push crowd based on the target conversion cost value, the historical push data and the value range, and then determine the conversion cost value when each push crowd is pushed in the future push sub-cycle based on the real-time pricing coefficient and the target conversion cost value. When receiving an information recommendation request, the conversion cost value of each recommendation information of the push crowd to which the requester belongs can be obtained, and the push score value of each recommendation information can be determined based on the matching degree between the requester and each recommendation information and the conversion cost value of each recommendation information of the push crowd to which the requester belongs. The suitable recommendation information can be selected based on the push score value, and the selected recommendation information is recommended to the user in the user group, which can improve the push effect, reduce the number of invalid pushes, and greatly save network resources and computer resources. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 An application environment diagram of the cost control method of recommendation information in an embodiment;

[0074] Figure 2 A flowchart of the cost control method of recommendation information in an embodiment;

[0075] Figure 3 A flowchart of the push crowd division step in an embodiment;

[0076] Figure 4 A return rate statistical distribution diagram of a recommendation object in an embodiment;

[0077] Figure 5 A crowd division diagram in an embodiment;

[0078] Figure 6 A flowchart of the real-time pricing coefficient determination step in an embodiment;

[0079] Figure 7 A flowchart of the cost control method of recommendation information in another embodiment;

[0080] Figure 8 A structural block diagram of the cost control device of recommendation information in an embodiment;

[0081] Figure 9 A structural block diagram of the cost control device of recommendation information in another embodiment;

[0082] Figure 10 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0083] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0084] The cost control method of the recommended information provided by the present application can be realized based on cloud technology. The cloud technology refers to a series of resources such as hardware, software, network, etc. being unified in a wide area network or a local area network to realize the calculation, storage, processing and sharing of data. The cloud technology is a general term of network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model application, which can form a resource pool, and can be used on demand, and is flexible and convenient. Cloud computing technology will become an important support. The background service of the technical network system needs a large amount of computing and storage resources, such as video websites, picture websites and more portal websites. With the high development and application of the Internet industry, in the future, every item may have its own identification mark, and needs to be transmitted to the background system for logical processing. Different levels of data will be processed separately, and various industry data need strong system support, which can only be realized by cloud computing.

[0085] Further, the recommended information screening method provided by the present application can be realized based on big data processing in cloud computing. Big data refers to a collection of data that cannot be captured, managed and processed within a certain time range by conventional software tools, and is a massive, high-growth and diversified information asset that needs new processing mode to have stronger decision-making, insight discovery and process optimization capabilities. With the advent of the cloud era, big data has attracted more and more attention. Big data needs special technology to effectively process large amounts of data over time. The technologies suitable for big data include large-scale parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, the Internet and scalable storage systems.

[0086] The cost control method of the recommended information provided by the present application can be applied to, for example, Figure 1The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The server 104 obtains the target conversion cost value of the recommended information in the push cycle; the push cycle includes the historical push sub-period in which the information push has been carried out and the future push sub-period in which the information push has not been carried out; read the historical push data of each push crowd in the historical push sub-period; determine the value interval of the corresponding pricing coefficient of each push crowd; based on the target conversion cost value, the historical push data and the value interval, determine the real-time pricing coefficient corresponding to each push crowd; based on the real-time pricing coefficient and the target conversion cost value, determine the conversion cost value when pushing to each push crowd in the future push sub-period, push the recommended information to the terminal 102 according to the conversion cost value of the user belonging to the crowd corresponding to the terminal 102.

[0087] Among them, the terminal 102 can be but not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, the server 104 can be an independent physical server, also can be the server cluster composed of multiple service nodes in the block chain system, the service nodes form the peer-to-peer (P2P, Peer To Peer) network between each other, the P2P protocol is an application layer protocol running on the transmission control protocol (TCP, Transmission Control Protocol) protocol.

[0088] In addition, the server 104 can also be a server cluster composed of multiple physical servers, can be a cloud server providing cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, content delivery network (Content Delivery Network, CDN), and big data and artificial intelligence platform Basic cloud computing services such as platform.

[0089] In one embodiment, as Figure 2 shown, a cost control method of recommended information is provided, and the method is applied to the server in Figure 1 for example, including the following steps:

[0090] S202, obtaining the target conversion cost value of the recommended information in the push cycle; the push cycle includes the historical push sub-period in which the information push has been carried out and the future push sub-period in which the information push has not been carried out.

[0091] The recommendation information is information for pushing to a user, and the recommendation information includes media information such as a video, news, an article, and an advertisement. The user can also be referred to as a recommendation object. The push object is an information receiver of the recommendation information. The push cycle refers to a statistical cycle of pushing the recommendation information. In this embodiment of the application, the push cycle is a cycle in which the recommendation information is currently being pushed, and can also be referred to as a current push cycle. Specifically, the push cycle can be one hour, one day, one week, one month, or one year. The push cycle includes at least two sub-cycles. For example, the push cycle is one day, that is, the push cycle is 24 hours, and the push sub-cycle can be one hour.

[0092] The sub-cycle in which the recommendation information has been pushed in the push cycle is referred to as a historical push sub-cycle, and the sub-cycle in which the recommendation information has not been pushed in the push cycle is referred to as a future push sub-cycle. For example, the push cycle is 24 hours, and the start and end times of the push cycle are the first hour of the day to the last hour of the day. Assuming that the current time is the end of the tenth hour, the first 10 hours are the historical push sub-cycle, and the last 14 hours are the future push sub-cycle.

[0093] The target conversion cost value is a target value of a conversion cost that needs to be reached after the pushing of the recommendation information in the push cycle is completed. The conversion cost (CPA) refers to a fee that needs to be paid by an information pusher for one conversion of the recommendation information. The conversion refers to a task that a user performs based on the recommendation information after the information pusher pushes the recommendation information to the user. In other words, the user completes a conversion target. The conversion target can be a behavior such as browsing, registration, order placement, payment, and forwarding. For example, the user stays on a website for a certain period of time, registers or places an order on the website, the user leaves a message through the website or uses an online instant messaging tool of the website to consult, or the user actually makes a payment.

[0094] It can be understood that the information pusher needs to set the target conversion cost value in advance in order to obtain a higher return on investment. In order to achieve the target conversion cost value as much as possible, the bid of the recommendation information needs to be adjusted before each time the recommendation information is pushed in the push cycle. In other words, the conversion cost value of this time of pushing needs to be determined in advance before each time the recommendation information is pushed, and then the recommendation information is pushed according to the determined conversion cost value, so that the actual conversion cost value obtained after the pushing of the recommendation information in the entire push cycle is completed is close to the target conversion cost value.

[0095] In one embodiment, the bid for pushing the recommended information in each sub-period of the pushing period is adjusted before the pushing task of the sub-period starts, that is, the conversion cost value of the recommended information in the corresponding sub-period is adjusted before the start of each sub-period of the pushing period, and then the recommended information is pushed to the corresponding user according to the determined conversion cost value when the recommended information is pushed in the corresponding sub-period.

[0096] For example, the pushing period is 24 hours and the sub-period is 1 hour. Before the start of the first hour, the server can obtain the pre-set initial conversion cost value, determine the initial conversion cost value as the conversion cost value of the first hour, and push the recommended information according to the conversion cost value of the first hour in the first hour. Before the start of the second hour, the conversion cost value of the second hour is determined, and the recommended information is pushed according to the conversion cost value of the second hour in the second hour. Before the start of the third hour, the conversion cost value of the third hour is determined, and the recommended information is pushed according to the conversion cost value of the third hour in the third hour. Similarly, the conversion cost value of the 24th hour is determined, and the recommended information is pushed according to the conversion cost value of the 24th hour in the 24th hour. The actual conversion cost value of the recommended information in the entire pushing period is close to the target conversion cost value.

[0097] In determining the bid, the conversion cost value of each pushing can be determined based on the target conversion cost value and the adjustment coefficient. Referring to formula (1), the product of the target conversion cost value and the adjustment coefficient is taken as the conversion cost value of the pushing:

[0098] (1)

[0099] Wherein, CPA is the determined conversion cost value of the pushing, is the adjustment coefficient, is the target conversion cost value. It can be understood that the adjustment coefficient is variable, and by adjusting the adjustment coefficient , the adjustment of the conversion cost value of the pushing can be realized.

[0100] In one embodiment, the length of the sub-period in the pushing period can be determined according to the number of times of adjustment of the recommended information in the pushing period and the length of the pushing period. Specifically, the ratio of the length of the pushing period to the number of times of adjustment can be determined as the length of the sub-period of the pushing period. For example, the length of the pushing period is 24 hours and the number of times of adjustment is 12, then the pushing period has 12 sub-periods, and the length of each sub-period is 2 hours.

[0101] In an embodiment, the server obtains a historical conversion of the recommendation information in each historical push period in a preset historical period, and determines a target conversion cost value of the recommendation information in a current push period based on a target consumption of the recommendation information in the current push period and the historical conversion.

[0102] The preset historical period includes at least one historical push period, and the historical push period is another push period before the current push period. The target consumption refers to a target value of the total consumption that is reached after the push of the recommendation information in the push period is completed. The consumption refers to the resource consumed by the push of the recommendation information. The consumption can also be referred to as a push cost value. For example, if the total cost of the push of the recommendation information is 10,000 yuan, the consumption is 10,000 yuan, and the push cost value is also 10,000 yuan.

[0103] Specifically, the server can determine a target conversion of the current push period based on the historical conversion of each historical push period in the preset historical period, and determine a target conversion cost value of the recommendation information in the current push period based on the target consumption of the recommendation information in the current push period and the determined target conversion.

[0104] In an embodiment, after the server obtains the historical conversion of each historical push period in the preset historical period, the server calculates an average historical conversion in the preset historical period, determines the average historical conversion obtained by the calculation as a target conversion, and determines a target conversion cost value of the recommendation information in the current push period based on the determined target conversion and the target consumption of the recommendation information in the current push period.

[0105] In an embodiment, after the server obtains the historical conversion of each historical push period in the preset historical period, the server inputs the historical conversion of each historical push period in the preset historical period into a conversion prediction model, outputs a predicted target conversion through the conversion prediction model, and determines a target conversion cost value of the recommendation information in the current push period based on the predicted target conversion and the target consumption of the recommendation information in the current push period. The conversion prediction model can be a time series model. The time series model is to arrange the prediction objects in time sequence, and to infer the possibility and trend of future changes and change rules based on the change rules of the past time series.

[0106] As an example, the above embodiment is described. It is assumed that the recommended information is an advertisement A, the push cycle is one day, and the preset historical period is the previous month. The server obtains the historical conversion of the advertisement A in each day of the previous month, and inputs the historical conversion of each day of the previous month into the conversion prediction model in time sequence, predicts the target conversion of today through the conversion prediction model, and calculates the target conversion cost value of today based on the predicted target conversion of today and the target consumption of the advertisement A in the push of today. For example, the predicted target conversion of the advertisement A today is 5000 times, and the target consumption of the advertisement A in the push of today is 10000 yuan, so the target conversion cost of the advertisement A today is 2 yuan / time.

[0107] In one embodiment, the server obtains the conversion cost value of each historical push cycle of the recommended information in the preset historical period, and determines the target conversion cost value of the recommended information in the current push cycle based on the conversion cost value of each historical push cycle.

[0108] In one embodiment, after obtaining the conversion cost value of each historical push cycle of the recommended information in the preset historical period, the server calculates the average value of the conversion cost value in the preset historical period, and determines the calculated average value of the conversion cost value as the target conversion cost value of the recommended information in the current push cycle.

[0109] In one embodiment, after obtaining the conversion cost value of each historical push cycle of the recommended information in the preset historical period, the server inputs the obtained conversion cost value of each historical push cycle into a conversion cost prediction model, and outputs the predicted target conversion cost value through the conversion cost prediction model. The conversion cost prediction model can be a time series model, which is to arrange the prediction objects in time sequence, and to infer the possibility and trend of future changes and change rules from the change rules of the past time series.

[0110] As an example, the above embodiment is described. It is assumed that the recommended information is an advertisement B, the push cycle is one day, and the preset historical period is the previous month. The server obtains the conversion cost value of the advertisement B in each day of the previous month, and inputs the conversion cost value of each day of the previous month into the conversion cost prediction model in time sequence, and predicts the target conversion cost value of today through the conversion cost prediction model.

[0111] S204, read the historical push data of each push crowd in the historical push sub-cycle.

[0112] The push crowd refers to a group to which the push object belongs, the push object is an information receiver of the recommendation information, and each push object can be divided into different groups according to a crowd division index, thereby forming different push crowds. The crowd division index can be an attribute index or a push index. The attribute index is an index for crowd division based on attribute information of the push object, such as gender, age, and label. The label is an index for identifying the preference of the push object. The push index is an index determined based on the conversion of other recommendation information received by the push object, such as conversion rate and return on investment.

[0113] For example, when the crowd division index is gender, each push object can be divided into a male push crowd and a female push crowd. When the crowd division index is age, each push object can be divided into multiple push crowds of different age groups. When the crowd division index is conversion rate, each push object can be divided into multiple push crowds of different conversion rate intervals. When the crowd division index is return on investment, each push object can be divided into multiple push crowds of different return on investment intervals.

[0114] It can be understood that in the push period, the recommendation information can be pushed to different push crowds respectively. The determined conversion cost value can be different for different push crowds each time of pushing.

[0115] For example, the push period is 24 hours, the sub-period is 1 hour, the push crowds include a push crowd 1 and a push crowd 2, before the start of the first hour, an initial conversion cost value of the crowd 1 and an initial conversion cost value of the crowd 2 are obtained respectively, the initial conversion cost value of the crowd 1 is determined as the first hour conversion cost value of the crowd 1, and the initial conversion cost value of the crowd 2 is determined as the first hour conversion cost value of the crowd 2. In the first hour, the recommendation information is pushed to the push objects belonging to the crowd 1 according to the first hour conversion cost value of the crowd 1, and the recommendation information is pushed to the push objects belonging to the crowd 2 according to the first hour conversion cost value of the crowd 2. Before the start of the second hour, the second hour conversion cost value of the crowd 1 and the second hour conversion cost value of the crowd 2 are determined, in the second hour, the recommendation information is pushed to the push objects belonging to the crowd 1 according to the second hour conversion cost value of the crowd 1, and the recommendation information is pushed to the push objects belonging to the crowd 2 according to the second hour conversion cost value of the crowd 2. Similarly, before the start of the 24th hour, the 24th hour conversion cost value of the crowd 1 and the 24th hour conversion cost value of the crowd 2 are determined, in the 24th hour, the recommendation information is pushed to the push objects belonging to the crowd 1 according to the 24th hour conversion cost value of the crowd 1, and the recommendation information is pushed to the push objects belonging to the crowd 2 according to the 24th hour conversion cost value of the crowd 2.

[0116] The historical push data is historical data of different push crowds, for example, the push crowds include push crowd 1 and push crowd 2, and the historical push data includes the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the number of historical push sub-periods, and the number of sub-periods contained in the push period corresponding to the push crowd 1, and the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the number of historical push sub-periods, and the number of sub-periods contained in the push period corresponding to the push crowd 2. It can be understood that the number of historical push sub-periods and the number of sub-periods contained in the push period of the push crowd 1 are the same as the number of historical push sub-periods and the number of sub-periods contained in the push period of the push crowd 2.

[0117] The historical push data is historical data of different push crowds, for example, the push crowds include push crowd 1 and push crowd 2, and the historical push data includes the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the number of historical push sub-periods, and the number of sub-periods contained in the push period corresponding to the push crowd 1, and the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the number of historical push sub-periods, and the number of sub-periods contained in the push period corresponding to the push crowd 2. It can be understood that the number of historical push sub-periods and the number of sub-periods contained in the push period of the push crowd 1 are the same as the number of historical push sub-periods and the number of sub-periods contained in the push period of the push crowd 2.

[0118] The historical conversion cost value is the ratio of the consumption generated by pushing the recommended information to the conversion generated in the historical push sub-period, for example, the total cost of pushing the advertisement A to the push crowd 1 in the historical push sub-period is 1000 yuan, and the conversion of the push crowd 1 to the advertisement A in the historical push sub-period is 100 times, then the historical conversion cost value corresponding to the push crowd 1 is 10 yuan / time.

[0119] The historical price adjustment coefficient is used to determine the conversion cost value when the recommended information is pushed in the historical push sub-period.

[0120] The historical push cost value is the consumption generated by pushing the recommended information in the historical push sub-period, for example, the total cost of pushing the advertisement A to the push crowd 1 in the historical push sub-period is 1000 yuan, and the total cost of 1000 yuan is the historical push cost corresponding to the push crowd 1.

[0121] The number of historical push sub-periods refers to the number of sub-periods in which the recommended information has been pushed in the current push period. For example, the push period is 24 hours, the start and end time of the push period is the first hour of the day to the last hour of the day, the sub-period is 1 hour, the number of sub-periods contained in the push period is 24, and the current time is the end of the 10th hour. The number of historical push sub-periods is 10, and the number of future push sub-periods is 14.

[0122] S206, determine the value range of the price adjustment coefficient corresponding to each push crowd.

[0123] The pricing coefficient is a coefficient used to determine the conversion cost value each time the recommendation information is pushed. When determining the conversion cost value, the conversion cost value each time the recommendation information is pushed can be determined based on the target conversion cost value and the pricing coefficient. Referring to formula (1), the product of the target conversion cost value and the pricing coefficient is taken as the conversion cost value of the push. The pricing coefficient of each population has different values, and the pricing coefficient of the same population each time the push is performed can also be different, but the value interval of the pricing coefficient of the same population each time the push is performed is determined.

[0124] For example, there are 8 push populations, namely push population 1, push population 2,..., push population 7, and push population 8. The value interval of the pricing coefficient of each push population is shown in Table 1 as follows:

[0125]

[0126] It should be noted that the value interval of the corresponding pricing coefficient of the same push population can have multiple values. For example, for the advertisement A, the push cycle contains 24 sub-periods, the value interval of the pricing coefficient of the push population in the first 12 sub-periods is the first value interval, and the value interval of the pricing coefficient in the last 12 sub-periods is the second value interval. Alternatively, for the advertisement A, the push cycle contains 24 sub-periods, and each sub-period corresponds to a corresponding value interval of the pricing coefficient.

[0127] In S208, the real-time pricing coefficient corresponding to each push population is determined based on the target conversion cost value, the historical push data, and the value interval.

[0128] The real-time pricing coefficient is a coefficient used to determine the conversion cost value in real time. It can be understood that after the conversion cost value is determined in real time, the recommendation information will be pushed according to the determined conversion cost value when the recommendation information is pushed.

[0129] For example, the push cycle of the advertisement A is 24 hours, the start and end time of the push cycle is the first hour to the last hour of a day, the number of pricing times is 24, and the length of the sub-cycle is 1 hour. Assuming that the current time is the end of the 10th hour, the real-time pricing coefficient of the 11th hour is determined based on the target conversion cost value, the historical push data, and the value interval. The determined real-time pricing coefficient is used to determine the conversion cost value of the 11th hour. If the push population corresponding to the advertisement A includes push population 1 and push population 2, the pricing coefficient corresponding to push population 1 in the 11th hour is determined based on the target conversion cost value, the historical push data, and the value interval of push population 1, and the target conversion cost value, the historical push data of push population 2, and the value interval of push population 1, and the pricing coefficient corresponding to push population 2 in the 11th hour is determined based on the target conversion cost value, the historical push data, and the value interval of push population 2, and the target conversion cost value, the historical push data of push population 1, so as to obtain the real-time pricing coefficient of each push population.

[0130] In one embodiment, after obtaining the target conversion cost value of the recommendation information in the current push cycle, the historical push data of each push population, and the value interval of the pricing coefficient of each push population, the server determines the cost prediction value of the historical push sub-cycle and the cost prediction value of the future push sub-cycle based on the target conversion cost value and the historical push data of each push population, and then determines the real-time pricing coefficient of each push population based on the target conversion cost value, the cost prediction value of the historical push sub-cycle, the cost prediction value of the future push sub-cycle, the historical push data of each push population, and the value interval of the pricing coefficient of each push population.

[0131] The cost prediction value of the historical push sub-cycle is the predicted consumption of the pushed recommendation information in the historical push sub-cycle, and the cost prediction value of the future push sub-cycle is the predicted consumption of the pushed recommendation information in the future push sub-cycle.

[0132] S210, determining the conversion cost value when pushing to each push population in the future push sub-cycle based on the real-time pricing coefficient and the target conversion cost value.

[0133] Specifically, after determining the real-time pricing coefficient of each push population, the server can determine the real-time conversion cost value of each push population based on the real-time pricing coefficient and the target conversion cost value. The determined real-time conversion cost value is the conversion cost value when pushing the recommendation information to each push population in the future push sub-cycle, that is, the bid when pushing the recommendation information to each push population in the future push sub-cycle.

[0134] For example, the push cycle of the advertisement A is 24 hours, the start and end time of the push cycle is the first hour to the last hour of a day, the number of price adjustment is 24 times, and the length of the sub-cycle is 1 hour. Assuming that the current time is the end of the 10th hour, the real-time price adjustment coefficient of the 11th hour is determined based on the target conversion cost value, the historical push data and the value interval, and the conversion cost value of the 11th hour is determined based on the real-time price adjustment coefficient and the target conversion cost value. The advertisement A is pushed according to the conversion cost value of the 11th hour in the 11th hour. If the push crowd corresponding to the advertisement A includes the push crowd 1 and the push crowd 2, the conversion cost value of the 11th hour of the push crowd 1 is determined based on the real-time price adjustment coefficient and the target conversion cost value of the push crowd 1, and the advertisement A is pushed to the push object belonging to the push crowd 1 according to the conversion cost value of the 11th hour of the push crowd 1 in the 11th hour. The conversion cost value of the 11th hour of the push crowd 2 is determined based on the real-time price adjustment coefficient and the target conversion cost value of the push crowd 2, and the advertisement A is pushed to the push object belonging to the push crowd 2 according to the conversion cost value of the 11th hour of the push crowd 2 in the 11th hour.

[0135] In the cost control method of the recommendation information, the target conversion cost value of the recommendation information in the push cycle is obtained, the historical push data of each push crowd in the historical push sub-cycle is read, and the value interval of the price adjustment coefficient corresponding to each push crowd is determined. Therefore, the real-time price adjustment coefficient corresponding to each push crowd can be determined based on the target conversion cost value, the historical push data and the value interval, and the conversion cost value in the future push sub-cycle for pushing to each push crowd can be determined based on the real-time price adjustment coefficient and the target conversion cost value. For the recommendation information, the price of the recommendation information is adjusted based on the crowd granularity by fully considering the benefits brought by different crowds, so that the accuracy of the push cost control of the recommendation information is improved, and the recommendation information can be exposed more in the push crowd with a lower conversion cost, thereby improving the conversion rate while reducing the cost of information push.

[0136] In one embodiment, the cost control method of the recommendation information further includes a process of dividing the push objects into crowds, which specifically includes the following steps: obtaining the index value of each push object crowd division index, dividing each push object into a crowd based on the index value of each push object, and obtaining each push crowd.

[0137] The crowd division index can be an attribute index or a push index. The attribute index is an index for dividing crowds based on the attribute information of the push object, such as gender, age, label, etc. The label is an index for identifying the preference of the push object. The push index is an index determined based on the conversion of other recommendation information received by the push object, such as conversion rate, return on investment, etc.

[0138] Specifically, the server determines a statistical distribution of the push objects based on the index values of the push objects after obtaining the index values of the push objects, performs crowd division on the push objects based on the statistical distribution, and obtains each push crowd.

[0139] In one embodiment, the crowd division index is the return on investment (ROI). Figure 3 As shown in the cost control method of the recommendation information, the method further includes a process of crowd division on the push objects, and specifically includes the following steps:

[0140] S302, obtaining a return on investment (ROI) obtained after the recommendation information or other recommendation information is pushed to each push object in a historical push period.

[0141] The historical push period refers to other push periods before the current push period. The other recommendation information refers to other recommendation information of the same category as the recommendation information pushed in the current push period. For example, if the recommendation information in the current push period is advertisement A, the other recommendation information can be advertisement B, advertisement C, etc. If advertisement A belongs to a game advertisement, advertisements B and C also belong to a game advertisement. If advertisement A belongs to an investment and financial management advertisement, advertisements B and C also belong to an investment and financial management advertisement.

[0142] The return on investment (ROI) refers to the ratio of the income obtained by pushing the recommendation information or other recommendation information to the consumption spent on pushing the recommendation information or other recommendation information.

[0143] S304, determining a statistical distribution of the push objects in each ROI interval based on the ROI.

[0144] The statistical distribution is the number of push objects corresponding to each value of the statistical ROI, and the ROI distribution of the push objects is obtained with the ROI as the horizontal coordinate and the number as the vertical coordinate.

[0145] Figure 4 An ROI statistical distribution graph of the push objects in one embodiment is shown. Based on the ROI statistical distribution graph shown in Figure 4 As shown in the ROI statistical distribution graph, the cumulative number of push objects corresponding to each ROI interval and the total ROI interval in which the ROI is located can be obtained, such as Figure 4 In the ROI statistical distribution graph, the total ROI interval in which the ROI is located is (0, 0.2).

[0146] Specifically, the server obtains the ROI of each push object, counts the number of push objects corresponding to each ROI value, and obtains an ROI statistical distribution graph with the ROI as the horizontal coordinate and the number as the vertical coordinate. The statistical distribution of the push objects in each ROI interval is determined based on the ROI statistical distribution graph.

[0147] S306, performing crowd division on each push object based on the statistical distribution, to obtain each push crowd.

[0148] Specifically, the server can determine a total return rate interval of the return rate based on the statistical distribution, and perform crowd division on each push object according to the total return rate interval, to obtain each push crowd.

[0149] In one embodiment, S306 specifically includes the following steps: determining a total return rate interval based on the statistical distribution; dividing the total return rate interval into at least two return rate intervals; the return rate interval is a sub-interval of the total return rate interval; and dividing push objects with the same return rate interval into the same push crowd.

[0150] Specifically, after obtaining the total return rate interval, the server obtains a preset interval division manner, divides the total return rate interval into at least two return rate intervals based on the preset interval division manner, and divides push objects with the same return rate interval into the same push crowd.

[0151] The preset interval division manner includes a first interval division manner and a second interval division manner, the interval length of each return rate interval obtained by using the first interval division manner is the same, and the number of recommended objects included in different push crowds is significantly different; the interval length obtained by using the second interval division manner is different, and the number of recommended objects included in different push crowds is close.

[0152] S308, pushing the recommended information to each push crowd.

[0153] Specifically, within a push cycle, when the server receives an information recommendation request from a push object of a different push crowd, the server pushes recommended information to the corresponding push object in response to the received information recommendation request.

[0154] In the above embodiments, by obtaining the return rate of each push object and dividing according to the return rate, different push crowds for the return of the information recommendation party can be determined, based on the return of the information recommendation party by different push crowds, the conversion cost of each crowd can be adjusted, and information pushing is performed according to the adjusted conversion cost, so that the recommended information can get more exposure in the push crowd with lower conversion cost, which not only improves the accuracy of cost control, but also reduces the cost of information pushing while improving the conversion rate.

[0155] In one embodiment, the process of obtaining the return rate of the server after pushing the recommendation information or other recommendation information to each push object in the historical push cycle includes the following steps: obtaining the cumulative push cost value corresponding to each push object in the historical push cycle; determining the push benefit obtained after pushing the related information to each push object; the related information includes the recommendation information or other recommendation information; and determining the return rate based on the cumulative push cost value and the push benefit.

[0156] The number of historical push cycles is at least one, the cumulative push cost value is the total consumption of pushing the recommendation information or other recommendation information to a push object in the information recommendation direction, and the push benefit is the benefit generated by the push object based on the push recommendation information or other recommendation information after receiving the push recommendation information or other recommendation information.

[0157] Specifically, the server obtains the consumption generated after pushing the recommendation information or other recommendation information to each push object in the historical push cycle, and the benefit generated by each push object based on the received recommendation information or other recommendation information. For each push object, the corresponding investment return rate is calculated based on the corresponding benefit and consumption.

[0158] For example, the push cycle is one day, the recommendation information is advertisement A, advertisement A belongs to the game category, and in the previous day, advertisement B, advertisement C and advertisement D were pushed to push object 1, among which advertisement B and advertisement C belong to game advertisements, and advertisement D belongs to investment and financial advertisements. The consumption m spent on pushing advertisement B and advertisement C to push object 1 in the previous day is obtained, and the benefit n brought to the information pusher by the user based on the conversion of advertisement B and advertisement C is obtained. The ratio of the benefit n to the consumption m is determined as the investment return rate of push object 1 in the previous day.

[0159] In the above embodiment, the server obtains the cumulative push cost value corresponding to each push object in the historical push cycle, and determines the push benefit obtained after pushing the recommendation information or other recommendation information to each push object, so as to determine the return rate based on the cumulative push cost value and the push benefit. The push objects are divided into different push crowds based on the return rate, so as to obtain push crowds with different return rates for the information recommendation party, so that the push party adjusts the conversion cost of each crowd based on the different push crowd return rates, thereby improving the accuracy of cost control on the one hand, and reducing the cost of information push while improving the conversion rate on the other hand.

[0160] In one embodiment, after obtaining the total interval of the return rate, the server obtains a first interval division mode, divides the total return rate interval into at least two interval lengths of the same return rate interval according to the first interval division mode, and divides the push objects in the same return rate interval into the same push crowd.

[0161] The table below shows the population segmentation in one embodiment. The total return on investment range corresponding to the table is [0, 0.16]. The total return range is divided into eight return ranges: [0, 0.02), [0.02, 0.04), [0.04, 0.06), [0.06, 0.08), [0.08, 0.10), [0.10, 0.12), [0.12, 0.14], and [0.14, 0.16]. Recommended individuals within the same return range are identified as the same population.

[0162]

[0163] In one embodiment, after obtaining the total rate of return range, the server obtains a first interval division method and divides the total rate of return range into at least two rate of return ranges with the same interval length according to a second interval division method. The specific division process is as follows: determine the quantiles of the rate of return based on the statistical distribution; divide the total rate of return range using the quantiles as the interval endpoints to obtain at least two rate of return ranges.

[0164] Among them, quantiles are the values ​​corresponding to the division points in a statistical distribution that divides the target audience into several equal parts.

[0165] Figure 5 It is Figure 4 The target audience shown is divided into four equal parts. In the figure, Q1 is the first quartile, Q2 is the second quartile, and Q3 is the third quartile. The first quartile, also known as the "smaller quartile," is equal to the 25th percentile of the user interaction data arranged from smallest to largest. The second quartile, also known as the "median," is equal to the 50th percentile of the user interaction data arranged from smallest to largest. The third quartile, also known as the "larger quartile," is equal to the 75th percentile of the user interaction data arranged from smallest to largest. The resulting return on investment (ROI) range is [0, Q1), [Q1, Q2), [Q2, Q3), [Q3, 0.2]. Recommended users within the same ROI range are identified as the same user group.

[0166] As shown in the table below, this is the population segmentation in another embodiment. The total return on investment range corresponding to the table is [0, 0.16]. The total return range is divided into four return ranges: [0, 0.06), [0.06, 0.08), [0.08, 0.10], and [0.10, 0.16]. Recommended individuals within the same return range are identified as the same population.

[0167]

[0168] In the above embodiments, the server determines quantiles of the return rate based on the statistical distribution, divides the total return rate interval with the quantiles as interval endpoints to obtain at least two return rate intervals, so that the number of recommended objects included in each different push crowd is close, and the division of the crowd is more fine, so that when the push party adjusts the conversion cost of each crowd based on the return of different push crowds, the cost control can be more fine, and the accuracy of cost control is further improved.

[0169] In one embodiment, the historical push data includes: a historical conversion cost value, a historical price adjustment coefficient, a historical push cost value, a number of historical push sub-periods, and a number of sub-periods contained in the push period, as shown in Figure 6 S208 includes the following steps:

[0170] S602, input the number of historical push sub-periods and the number of sub-periods contained in the push period, and the target conversion cost value into the cost prediction model, and determine the cost prediction value of the historical push sub-period and the cost prediction value of the future push sub-period through the cost prediction model.

[0171] The cost prediction value of the historical push sub-period is the predicted consumption of the push recommendation information in the historical push sub-period, and the cost prediction value of the future push sub-period is the predicted consumption of the push recommendation information in the future push sub-period.

[0172] Specifically, for any push crowd, the server obtains a first target push sub-period corresponding to the historical push sub-period of the push period in the historical push period, and a second target push sub-period corresponding to the future push sub-period of the push period in the historical push period based on the number of historical push sub-periods and the number of sub-periods contained in the push period, and obtains the first historical conversion in the first target push sub-period and the second historical conversion in the second target push sub-period, and predicts the conversion prediction value of the historical push sub-period in the push period based on the first historical conversion through the cost prediction model, and predicts the conversion prediction value of the future push sub-period in the push period based on the second historical conversion, and then predicts the cost prediction value of the historical push sub-period in the push period based on the conversion prediction value of the historical push sub-period and the target conversion cost value, and predicts the cost prediction value of the future push sub-period in the push period based on the conversion prediction value of the future push sub-period and the target conversion cost value.

[0173] S604, determine the time influence factor value based on the cost prediction value of the historical push sub-period and the cost prediction value of the future push sub-period.

[0174] Specifically, the server determines the ratio of the cost prediction value of the future push sub-period to the cost prediction value of the historical push sub-period as the time influence factor.

[0175] S606, determine the real-time pricing coefficients corresponding to each push crowd based on the target conversion cost value, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, and the value interval.

[0176] Specifically, for each push crowd, the server calculates the real-time pricing coefficient corresponding to each push crowd respectively. For example, the push crowds include push crowd 1, push crowd 2, …, and push crowd 8. The server can first calculate the real-time pricing coefficient corresponding to push crowd 1, then calculate the real-time pricing coefficient of push crowd 2, and so on, until the real-time pricing coefficient of push crowd 8 is calculated.

[0177] In the above embodiment, the server inputs the target conversion cost value into the cost prediction model through the number of historical push sub-periods and the number of sub-periods contained in the push period, determines the cost prediction value of the historical push sub-period and the cost prediction value of the future push sub-period through the cost prediction model, and determines the time influence factor value based on the cost prediction value of the historical push sub-period and the cost prediction value of the future push sub-period. Thus, the real-time pricing coefficient corresponding to each push crowd can be determined in real time based on the target conversion cost value, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, and the value interval. Furthermore, the conversion cost value when pushing to each push crowd in the future push sub-period can be determined based on the real-time pricing coefficient, the recommended information is priced based on the crowd granularity by fully considering the revenue brought by different crowds, thereby improving the accuracy of the push cost control of the recommended information.

[0178] In one embodiment, S606 specifically includes the following steps: for each push crowd, selecting a candidate pricing coefficient from the value interval of the current push crowd; determining the candidate conversion cost value of each push crowd in the push period based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, and the time influence factor value; determining the candidate pricing coefficient corresponding to the candidate conversion cost value closest to the target conversion cost value as the real-time pricing coefficient of the current push crowd, until the real-time pricing coefficient corresponding to each push crowd is obtained.

[0179] Among them, the current push crowd is the crowd in each push crowd for which the real-time pricing coefficient is currently to be determined. For example, the push crowds include push crowd 1, push crowd 2, …, and push crowd 8. When the real-time pricing coefficient of push crowd 1 is to be determined, push crowd 1 is the current push crowd. When the real-time pricing coefficient of push crowd 2 is to be determined, push crowd 2 is the current push crowd.

[0180] The candidate price adjustment coefficient is any price adjustment coefficient value selected from the value range of the current push audience. For example, if the current push audience is push audience 1 and the value range of push audience 1 is (0,1), then any value between 0 and 1 can be selected as the candidate price adjustment coefficient.

[0181] The candidate conversion cost value refers to the conversion cost that will be achieved after pushing recommendation information to each target audience during the push period if the current push audience is adjusted using the candidate price adjustment coefficient.

[0182] Specifically, after obtaining the candidate price adjustment coefficients for the current push audience, as well as the historical conversion cost value, historical price adjustment coefficient, historical push cost value, and time influence factor value for each push audience, the server can input the obtained values ​​into the conversion cost value calculation model. The conversion cost value calculation model then determines the candidate conversion cost value for the current push audience during the push period. The conversion cost value calculation model calculates the candidate conversion cost value using the following formula:

[0183] (2)

[0184] (3)

[0185] (4)

[0186] (5)

[0187] (6)

[0188] (7)

[0189] (8)

[0190] In formula (2) This represents the candidate conversion cost value during the push period. This refers to the conversion cost value of the historical sub-cycles of the push notification period, specifically based on the historical conversion cost value of each push audience. To confirm, This represents the conversion cost value of future push sub-cycles within the push cycle, and specifically, the future push cost value for each push audience can be determined. The weighting factor for the historical push sub-cycles of the push cycle. The weight of future push sub-cycles within the push cycle.

[0191] In formula (3), there are a total of n target audiences. Let p be the weight of the p-th push audience. The price adjustment coefficient of the pth push crowd in the future push sub-period, The historical price adjustment coefficient of the pth push crowd in the historical push sub-period, when the historical push sub-period is multiple, The average of the historical price adjustment coefficients corresponding to the multiple historical push sub-periods, The historical conversion cost value of the pth push crowd in the historical push sub-period, when the historical push sub-period is multiple, The historical conversion cost value corresponding to the multiple historical push sub-periods.

[0192] In formula (4), The weight of the pth push crowd, The historical push cost value of the pth push crowd in the historical push sub-period, when the historical push sub-period is multiple, The average of the historical push cost values corresponding to the multiple historical push sub-periods.

[0193] In formula (5), The conversion of the historical push sub-period of the push period, which is specifically based on the historical conversion of each push crowd is determined, The conversion of the future push sub-period of the push period.

[0194] In formula (6), The push cost value of the pth push crowd in the future push sub-period, The historical push cost value of the historical push sub-period, which can be specifically determined based on the historical push cost value of each push crowd and the historical price adjustment coefficient .

[0195] In formula (7), The cost prediction value of the pth push crowd in the future push sub-period, that is, let to obtain the cost prediction value of the pth push crowd in the future push sub-period, The cost prediction value of the historical push sub-period in the push period, that is, let to obtain the cost prediction value in the historical push sub-period, which can be determined based on the cost prediction value of each push crowd in the historical push sub-period is a function expression about , is a function expression about , is a time influence factor value, which can be specifically represented by . Wherein, The expression form of is formula (8), wherein As a constant obtained from experiments, in one embodiment, The value is 2.5.

[0196] In formulas (2) to (7) above, and These are all intermediate process quantities, used to obtain... about The expression is obtained, and then based on formula (2) and formula (3), we get about The expression, if the current audience is the Pth push audience, then let Equal to the candidate price adjustment coefficient, thus based on about The expression calculates the value of the p-th push audience. When the candidate price adjustment coefficient is equal to the candidate conversion cost value during the push period, it represents the candidate conversion cost value.

[0197] In the above embodiments, the server selects candidate price adjustment coefficients from the value range of the current push audience for each push audience; based on the candidate price adjustment coefficients, historical conversion cost values, historical price adjustment coefficients, historical push cost values, and time influence factor values, it determines the candidate conversion cost value for the push cycle; the candidate price adjustment coefficient corresponding to the candidate conversion cost value that is closest to the target conversion cost value is determined as the real-time price adjustment coefficient for the current push audience, until the real-time price adjustment coefficients corresponding to each push audience are obtained. Thus, when determining the real-time price adjustment coefficient for a push audience, the impact of the audience and the push progress on the price adjustment is fully considered, wherein the push progress is represented by the time influence factor, thereby improving the accuracy of push cost control for recommendation information.

[0198] In one embodiment, the step of determining the candidate conversion cost value for each push audience during the push period based on the candidate price adjustment coefficient, historical conversion cost value, historical price adjustment coefficient, historical conversion cost value, and time influence factor value includes: determining the price adjustment coefficient for other push audiences in future push sub-periods based on the historical price adjustment coefficients of other push audiences outside the current push audience; inputting the candidate price adjustment coefficient, historical conversion cost value, historical price adjustment coefficient, historical conversion cost value, time influence factor value, and price adjustment coefficient for future push sub-periods into the conversion cost calculation model; and determining the candidate conversion cost value for the push period through the conversion cost calculation model.

[0199] Specifically, referring to formulas (2) to (8), the server obtains about After the expression, if the current audience is the Pth push audience, then let This is equal to the candidate price adjustment coefficient. Furthermore, for any push audience i other than the p-th push audience, let the price adjustment coefficient for that push audience i in future push sub-cycles be... wherein, is the historical pricing coefficient of the i-th push group in the historical push sub-period, and the candidate pricing coefficient of the P-th push group, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value, and the pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value of other push groups in the future push sub-period are input into the conversion cost calculation model, and the candidate conversion cost value of the push period is determined through the conversion cost calculation model.

[0200] In the above embodiment, the server determines the pricing coefficient of other push groups in the future push sub-period based on the historical pricing coefficient of other push groups other than the current push group; inputs the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value and the pricing coefficient of the future push sub-period into the conversion cost calculation model; and determines the candidate conversion cost value of the push period through the conversion cost calculation model, so that when determining the real-time pricing coefficient of a certain push group, the influence of the pricing of other push groups in the historical push sub-period on the push group is fully considered, thereby improving the accuracy of the push cost control of the recommended information.

[0201] In one embodiment, the server determines the candidate conversion cost value of each push group in the push period based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value and the time influence factor value, which includes: among other push groups other than the current push group, distinguishing a first other push group that has obtained a real-time pricing coefficient and a second other push group that has not obtained a real-time pricing coefficient; determining the pricing coefficient of the second other push group in the future push sub-period based on the historical pricing coefficient of the second other push group; inputting the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value, the real-time pricing coefficient of the first other push group and the pricing coefficient of the future push sub-period into the conversion cost calculation model; and determining the candidate conversion cost value of the push period through the conversion cost calculation model.

[0202] Specifically, referring to formulas (2) to (8), the server obtains After obtaining the expression of If the current group is the P-th push group, the first other push group that has obtained a real-time pricing coefficient and the second other push group that has not obtained a real-time pricing coefficient are determined among other push groups other than the current group, assuming that the first other push group is the 1st to the p-1th push group before the p-th push group, and the second push group is the p+1th to the nth push group after the p-th push group, for any one push group j in the first other push group, the real-time pricing coefficient has been obtained , then for any one of the second other push crowds i, the push crowd i is set to have the pricing coefficient , wherein, is the historical pricing coefficient of the i th push crowd in the historical push sub-period, and the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value of the P th push crowd, the real-time pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value of the first other push crowd in the future push sub-period, and the pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical conversion cost value, the time influence factor value of the second other push crowd in the future push sub-period are input into the conversion cost calculation model, and the candidate conversion cost value of the push period is determined by the conversion cost calculation model.

[0203] In the above embodiment, the server distinguishes the first other push crowd that has obtained the real-time pricing coefficient and the second other push crowd that has not obtained the real-time pricing coefficient among the other push crowds except the current push crowd; determines the pricing coefficient of the second other push crowd in the future push sub-period based on the historical pricing coefficient of the second other push crowd; inputs the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, the real-time pricing coefficient of the first other push crowd, and the pricing coefficient in the future push sub-period into the conversion cost calculation model; and determines the candidate conversion cost value of the push period by the conversion cost calculation model, so that when determining the real-time pricing coefficient of a push crowd, the influence of the pricing situation of the other push crowds in the historical push sub-period and the future push sub-period on the push crowd is fully considered, thereby improving the accuracy of the push cost control of the recommended information.

[0204] In one embodiment, the server determines the value range of the pricing coefficient corresponding to each push crowd, including: determining the progress of information push based on the push period and the historical push sub-period; and determining the value range of the pricing coefficient corresponding to each push crowd according to the progress.

[0205] It should be noted that the value range of the pricing coefficient corresponding to the same push crowd can have multiple value ranges, and for a certain push crowd, the push period can be divided into multiple progress intervals, each progress interval corresponding to a corresponding value range of the pricing coefficient.

[0206] Specifically, the server obtains the number of sub-periods contained in the push period and the number of historical push sub-periods, and determines the progress of pushing the recommended information according to the number of historical push sub-periods and the number of sub-periods of the push period, and obtains the value of the pricing coefficient matched with the progress.

[0207] For example, the push crowd has four value intervals of the pricing coefficient, respectively denoted as value interval 1, value interval 2, value interval 3 and value interval 4, wherein 0-20% of the progress of the push period corresponds to the value interval 1, 20%-50% of the progress of the push period corresponds to the value interval 2, 50%-80% of the progress of the push period corresponds to the value interval 3, and 80%-100% of the progress of the push period corresponds to the value interval 4. After the server obtains that the number of sub-periods of the push period is 10 and the number of historical push sub-periods is 3, the push progress is 30%, the value interval 2 corresponding to 30% is obtained, and the value interval 2 is the value interval of the pricing coefficient corresponding to the push crowd 1.

[0208] In the above embodiment, the server determines the progress of information push based on the push period and the historical push sub-period; and determines the value interval of the pricing coefficient corresponding to each push crowd according to the progress, so that when determining the real-time pricing coefficient of a certain push crowd, the influence of the push progress on the pricing is further considered, thereby improving the accuracy of the control of the push cost of the recommended information.

[0209] In one embodiment, after the server determines the conversion cost value of each push crowd in the future push sub-period, the server can further receive an information recommendation request in the future push sub-period, the information recommendation request carrying user information of a requester; determine the belonging push crowd of the requester based on the user information; obtain the conversion cost value of the recommended information of the belonging push crowd; and push the recommended information to the requester according to the conversion cost value of the recommended information.

[0210] The user information can be attribute information or user historical behavior data of the requester, wherein the attribute information can be gender, age, label, etc., the label is an index for identifying user preferences, and the user historical behavior data includes other recommended information received by the requester in a historical period and conversion based on the received other recommended information. The historical behavior data is used to determine the conversion rate, return on investment, etc. of the user.

[0211] Specifically, the server obtains a preset crowd division index, determines the index value of the crowd division index of the requester based on the user information, and determines the belonging recommended crowd of the requester according to the index value, then selects the conversion cost value corresponding to the belonging recommended crowd of the requester from the pre-determined conversion cost values of each push crowd, and pushes the recommended information to the requester according to the selected conversion cost value.

[0212] The population division index can be an attribute index or a push index. The attribute index is an index for dividing the population based on attribute information of the push object, such as gender, age, label, and the like. The label is an index for identifying the preference of the push object. The push index is an index determined based on the conversion of other recommendation information received by the push object, such as conversion rate, return on investment, and the like.

[0213] In one embodiment, the preset population division index is return on investment. The server determines the value of the return on investment of the requestor based on the user information of the requestor, determines the return on investment interval to which the value of the return on investment of the requestor belongs, determines the push population corresponding to the return on investment interval as the push population to which the requestor belongs, and selects the conversion cost value corresponding to the push population to which the requestor belongs from the conversion cost values of the push populations determined in advance, and pushes the recommendation information to the requestor according to the selected conversion cost value.

[0214] In the above embodiment, the server receives an information recommendation request, the information recommendation request carrying user information of the requestor, determines the push population to which the requestor belongs based on the user information, acquires the conversion cost value of the recommendation information for the push population, and pushes the recommendation information to the requestor according to the conversion cost value of the recommendation information, so that the push cost can be accurately controlled, and the actual conversion cost value after the end of the push period is close to the preset target conversion cost value.

[0215] In one embodiment, the server receives an information recommendation request, the information recommendation request carrying user information of the requestor, determines the push population to which the requestor belongs based on the user information in response to the information recommendation request, acquires a target conversion cost value of the recommendation information within a push period, the push period including a historical push sub-period in which information has been pushed and a future push sub-period in which information has not been pushed, reads historical push data of each push population within the historical push sub-period, determines a value interval of the pricing coefficient corresponding to the push population to which the requestor belongs, determines a real-time pricing coefficient corresponding to the push population to which the requestor belongs based on the target conversion cost value, the historical push data, and the value interval, determines a conversion cost value when the push population to which the requestor belongs is pushed in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value, and pushes the recommendation information to the requestor according to the conversion cost value of the recommendation information.

[0216] In one embodiment, the server can also acquire the conversion cost value of the other recommendation information for the push population, select the to-be-recommended information from the recommendation information and the other recommendation information based on the conversion cost value of the recommendation information and the conversion cost value of the other recommendation information, and if the selected to-be-recommended information is the recommendation information, execute the step of pushing the recommendation information to the requestor according to the conversion cost value of the recommendation information.

[0217] The recommendation information and the other recommendation information are information filtered from the recommendation information resource pool and pushed to the requester in response to the information recommendation request.

[0218] In an embodiment, the user information further includes search information of the requester, terminal information of the requester, and the like. The search information of the user can be keyword information currently searched by the user, and the terminal information of the requester can be operating system information of the terminal of the requester, a current network connection mode, and the like. After receiving the information push request, the server filters information matching at least one of the search information of the requester and the terminal information of the requester from the recommendation information resource pool in response to the information push request, and determines the filtered matching information as information that can be pushed to the requester. If the information that can be pushed to the requester includes the recommendation information and the other recommendation information, the step of obtaining the conversion cost value of the other recommendation information by the push target group is performed.

[0219] In an embodiment, the server further obtains the conversion cost value of the other recommendation information by the push target group, selects the to-be-recommended information from the recommendation information and the other recommendation information based on the size relationship between the conversion cost value of the recommendation information and the conversion cost value of the other recommendation information, and performs the step of pushing the recommendation information to the requester according to the conversion cost value of the recommendation information if the selected to-be-recommended information is the recommendation information.

[0220] In an embodiment, the server further obtains the conversion cost value of the other recommendation information by the push target group, calculates the matching degree of the requester and the recommendation information and the matching degree of the requester and the other recommendation information, calculates the push score value of the recommendation information according to the conversion cost value of the recommendation information and the matching degree of the requester and the recommendation information, calculates the push score value of the other recommendation information according to the conversion cost value of the other recommendation information and the matching degree of the requester and the other recommendation information, and selects the to-be-recommended information from the recommendation information and the other recommendation information according to the score value of the recommendation information and the score value of the other push information. If the selected to-be-recommended information is the recommendation information, the step of pushing the recommendation information to the requester according to the conversion cost value of the recommendation information is performed.

[0221] In the above embodiment, the server selects the to-be-recommended information from the recommendation information and the other recommendation information based on the conversion cost value of the recommendation information and the conversion cost value of the other recommendation information, so that the recommendation information can obtain more exposure in the push target group with lower conversion cost, thereby improving the conversion rate and reducing the cost of information push.

[0222] In an embodiment, as shown in Figure 7 a cost control method of recommendation information is further provided, and the method is applied to the server. Figure 1The server in the recommendation information push system is taken as an example for illustration, including the following steps:

[0223] S702, obtaining a return rate obtained after the recommended information or other recommended information is pushed to each push object in a historical push period.

[0224] S704, determining a statistical distribution of the recommended object on each return rate interval according to the return rate.

[0225] S706, dividing each push object into a push population based on the statistical distribution.

[0226] S708, pushing the recommended information to each push population.

[0227] S710, obtaining a target conversion cost value of the recommended information in a push period. The push period includes a historical push sub-period in which information has been pushed and a future push sub-period in which information has not been pushed.

[0228] S712, reading historical push data of each push population in the historical push sub-period.

[0229] S714, determining a value interval of the pricing coefficient corresponding to each push population.

[0230] S716, determining a real-time pricing coefficient corresponding to each push population based on the target conversion cost value, the historical push data and the value interval.

[0231] S718, determining a conversion cost value when pushing to each push population in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value.

[0232] S720, receiving an information recommendation request in the future push sub-period, the information recommendation request carrying user information of a requester.

[0233] S722, determining a push population to which the requester belongs based on the user information.

[0234] S724, obtaining a conversion cost value of the recommended information of the push population to which the requester belongs.

[0235] S726, pushing the recommended information to the requester according to the conversion cost value of the recommended information.

[0236] The application also provides an application scenario applying the recommended information cost control method.

[0237] The recommended information is an advertisement A of a certain game category, the recommended object is a user group corresponding to the game industry, the server obtains a return on investment (ROI) of each recommended object, and the recommended objects are divided into 8 push crowds based on the return on investment, and 8 push crowds of 8 ROI intervals are obtained, and a corresponding price adjustment coefficient value interval is determined for each push crowd. For reference, the following table, where a and b are the interval endpoints of the corresponding total interval.

[0238]

[0239] If the recommendation period is 24 hours, the current time is the end of the 10th hour, and if an information recommendation request is received at the current time, the requestor belongs to the recommended crowd according to the information recommendation request, if the recommended crowd belongs to the recommended crowd 5, the target conversion cost value of the advertisement A in the recommended period is obtained, and the historical push data of each push crowd about the advertisement A in the first 10 hours is obtained, and the value interval of the price adjustment coefficient of the recommended crowd 5 is determined. Then, the historical push data of the push crowd about the advertisement A is input into formulas (2) to (8) to obtain The expression of The value of the target The value of the target The value interval of the price adjustment coefficient of the recommended crowd 5, and the value interval of the target Make The value closest to the target conversion cost value Then the target Is determined as the real-time price adjustment coefficient of the recommended crowd 5, and the target Input formula (1) to obtain the conversion cost value of the recommended crowd 5 corresponding to the advertisement A, and push the advertisement A to the requestor according to the obtained conversion cost value of the advertisement A.

[0240] It should be understood that although each step in the flowcharts of Figure 2 , 3 , 6 and 7 is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, Figure 2 , 3 At least some of the steps in 6 and 7 can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0241] In one embodiment, as Figure 8As shown, a cost control device for recommendation information is provided. This device can be a software module, a hardware module, or a combination of both integrated into a computer device. Specifically, the device includes: a target data acquisition module 802, a historical data acquisition module 804, a value range determination module 806, a price adjustment coefficient determination module 808, and a conversion cost value determination module 810, wherein:

[0242] The target data acquisition module 802 is used to acquire the target conversion cost value of the recommendation information within the push period. The push period includes historical push sub-periods where information has been pushed and future push sub-periods where information has not been pushed.

[0243] The historical data acquisition module 804 is used to read the historical push data of each push group within the historical push sub-cycle.

[0244] The value range determination module 806 is used to determine the value range of the price adjustment coefficient corresponding to each push audience.

[0245] The price adjustment coefficient determination module 808 is used to determine the real-time price adjustment coefficient for each push audience based on the target conversion cost value, historical push data, and value range.

[0246] The conversion cost value determination module 810 is used to determine the conversion cost value when pushing to each push audience in future push sub-cycles based on the real-time price adjustment coefficient and the target conversion cost value.

[0247] The aforementioned cost control device for recommendation information obtains the target conversion cost value of the recommendation information within the push period, reads historical push data for each push audience within historical push sub-cycles, and determines the value range of the price adjustment coefficient corresponding to each push audience. Based on the target conversion cost value, historical push data, and value range, it can determine the real-time price adjustment coefficient for each push audience. Furthermore, based on the real-time price adjustment coefficient and the target conversion cost value, it determines the conversion cost value when pushing the recommendation information to each push audience in future push sub-cycles. By fully considering the benefits that different audiences can bring, and adjusting the price of recommendation information based on audience granularity, the accuracy of push cost control for recommendation information is improved.

[0248] In one embodiment, such as Figure 9 As shown, the device further includes: a rate of return acquisition module 812, used to acquire the rate of return obtained after pushing recommendation information or other recommendation information to each push object in a historical push period; a statistical distribution determination module 814, used to determine the statistical distribution of the recommendation object in each rate of return interval based on the rate of return; a user segmentation module 816, used to segment each push object into users based on the statistical distribution to obtain each push user group; and an information push module 818, used to push recommendation information to each push user group.

[0249] In an embodiment, the return rate obtaining module 812 is further configured to: obtain a cumulative push cost value corresponding to each push object in a historical push period; determine a push benefit obtained after pushing the recommendation information or other recommendation information to each push object; and determine a return rate based on the cumulative push cost value and the push benefit.

[0250] In an embodiment, the crowd dividing module 816 is further configured to: determine a total return rate interval based on the statistical distribution; divide the total return rate interval into at least two return rate intervals; the return rate interval is a subinterval of the total return rate interval; and divide push objects with the same return rate interval into the same push crowd.

[0251] In an embodiment, the crowd dividing module 816 is further configured to: determine quantiles of the return rate based on the statistical distribution; divide the total return rate interval into at least two return rate intervals by taking the quantiles as interval endpoints.

[0252] In an embodiment, the historical push data includes: a historical conversion cost value, a historical pricing coefficient, a historical push cost value, a number of historical push subperiods, and a number of subperiods included in the push period; the pricing coefficient determining module 808 is further configured to: input the number of historical push subperiods, the number of subperiods included in the push period, and the target conversion cost value into a cost prediction model, determine a cost prediction value of the historical push subperiod and a cost prediction value of the future push subperiod through the cost prediction model, and determine a time influence factor value based on the cost prediction value of the historical push subperiod and the cost prediction value of the future push subperiod; and determine the real-time pricing coefficient corresponding to each push crowd based on the target conversion cost value, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, and a value interval.

[0253] In an embodiment, the pricing coefficient determining module 808 is further configured to: for each push crowd, select a candidate pricing coefficient from the value interval of the current push crowd; determine a candidate conversion cost value of the push period based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, and the time influence factor value; and determine the real-time pricing coefficient of the current push crowd as the candidate pricing coefficient corresponding to the candidate conversion cost value closest to the target conversion cost value, until the real-time pricing coefficient corresponding to each push crowd is obtained.

[0254] In an embodiment, the price adjustment coefficient determination module 808 is further configured to: determine, based on historical price adjustment coefficients of other push crowds other than the current push crowd, price adjustment coefficients of the other push crowds in a future push sub-period; input the candidate price adjustment coefficient, the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the time influence factor value, and the price adjustment coefficients of the future push sub-period into the conversion cost calculation model; and determine, by the conversion cost calculation model, the candidate conversion cost value of the push period.

[0255] In an embodiment, the price adjustment coefficient determination module 808 is further configured to: distinguish, among the other push crowds other than the current push crowd, a first other push crowd that has obtained a real-time price adjustment coefficient and a second other push crowd that has not obtained a real-time price adjustment coefficient; determine, based on historical price adjustment coefficients of the second other push crowd, price adjustment coefficients of the second other push crowd in a future push sub-period; input the candidate price adjustment coefficient, the historical conversion cost value, the historical price adjustment coefficient, the historical push cost value, the time influence factor value, the real-time price adjustment coefficient of the first other push crowd, and the price adjustment coefficients of the future push sub-period into the conversion cost calculation model; and determine, by the conversion cost calculation model, the candidate conversion cost value of the push period.

[0256] In an embodiment, the value range determination module 806 is further configured to: determine a progress of information pushing based on the push period and the historical push sub-period determination information; and determine, according to the progress, a value range of the price adjustment coefficient corresponding to each push crowd.

[0257] In an embodiment, as Figure 9 The apparatus further includes: a request receiving module 820 configured to receive an information recommendation request, the information recommendation request carrying user information of a requester; a crowd determining module 822 configured to determine, based on the user information, a push crowd to which the requester belongs; a conversion cost value obtaining module 824 configured to obtain a conversion cost value of the push crowd with respect to the recommended information; and an information pushing module 818 configured to push the recommended information to the requester according to the conversion cost value of the recommended information.

[0258] Specific limitations of the cost control apparatus for the recommended information can be referred to the limitations of the cost control method for the recommended information in the foregoing, which will not be repeated here. Each module in the cost control apparatus for the recommended information can be realized by software, hardware, or a combination thereof, in whole or in part. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in the computer device in software form, so as to be called and executed by the processor to perform operations corresponding to each module.

[0259] In an embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as Figure 10As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store information push data. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a recommended information cost control method.

[0260] Those skilled in the art can understand that, Figure 10 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0261] In one embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0262] In one embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program is executed by the processor to implement the steps in the above method embodiments.

[0263] In one embodiment, a computer program product or computer program is provided, which includes computer instructions 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 executes the computer instructions, so that the computer device executes the steps in the above method embodiments.

[0264] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0265] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of each technical feature in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0266] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method of pushing a recommendation information, characterized by, The method is executed by a server, and comprises: obtaining a target conversion cost value of recommendation information in a push cycle; the push cycle comprises a historical push sub-cycle in which information has been pushed and a future push sub-cycle in which information has not been pushed; reading historical push data of each push crowd in the historical push sub-cycle; the push crowd is obtained by grouping each push object according to a crowd division index; the crowd division index comprises a push index; the push index comprises a conversion rate and a return on investment; the historical push data comprises a historical conversion cost value, a historical pricing coefficient, a historical push cost value, a number of the historical push sub-cycle and a number of sub-cycles included in the push cycle; determining a value interval of a corresponding pricing coefficient of each push crowd; inputting the number of the historical push sub-cycle, the number of sub-cycles included in the push cycle and the target conversion cost value into a cost prediction model, and determining a cost prediction value of the historical push sub-cycle and a cost prediction value of the future push sub-cycle through the cost prediction model; determining a time influence factor value based on the cost prediction value of the historical push sub-cycle and the cost prediction value of the future push sub-cycle; for each push crowd, selecting a candidate pricing coefficient from the value interval of the current push crowd; determining a candidate conversion cost value of the push cycle based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value and the time influence factor value; determining a real-time pricing coefficient of the current push crowd as a candidate pricing coefficient corresponding to a candidate conversion cost value closest to the target conversion cost value, until real-time pricing coefficients corresponding to each push crowd are obtained; determining a conversion cost value when each push crowd is pushed in the future push sub-cycle based on the real-time pricing coefficient and the target conversion cost value; receiving an information recommendation request; the information recommendation request carries user information of a requester; determining a belonging push crowd of the requester based on the user information; obtaining a conversion cost value of the recommendation information about the belonging push crowd; obtaining conversion cost values of other recommendation information about the belonging push crowd; calculating a matching degree of the requester and the recommendation information; and calculating matching degrees of the requester and other recommendation information; calculating a push score value of the recommendation information according to the conversion cost value of the recommendation information and the matching degree of the requester and the recommendation information; and calculating push score values of the other recommendation information according to the conversion cost values of the other recommendation information and the matching degrees of the requester and the other recommendation information; selecting to-be-recommended information from the recommendation information and the other recommendation information according to the push score value of the recommendation information and the push score values of the other recommendation information; if the to-be-recommended information is the recommendation information, pushing the recommendation information to the requester according to the conversion cost value of the recommendation information.

2. The method of claim 1, wherein, The method further comprises: Obtain a return rate obtained after the recommended information or other recommended information is pushed to each push object in a historical push cycle; Determine a statistical distribution of each push object in each return rate interval according to the return rate; Divide each push object into a push crowd based on the statistical distribution to obtain each push crowd; Push recommended information to each push crowd.

3. The method of claim 2, wherein, The obtaining of the return rate obtained after the recommended information or other recommended information is pushed to each push object in a historical push cycle includes: Obtain a cumulative push cost value corresponding to each push object in the historical push cycle; Determine a push benefit obtained after the recommended information or the other recommended information is pushed to each push object; Determine the return rate based on the cumulative push cost value and the push benefit.

4. The method of claim 2, wherein, The division of each push object into a push crowd based on the statistical distribution includes: Determine a total return rate interval based on the statistical distribution; Divide the total return rate interval into at least two return rate intervals; the return rate interval is a subinterval of the total return rate interval; Divide push objects with the same return rate interval into the same push crowd.

5. The method of claim 4, wherein, The division of the total return rate interval into at least two return rate intervals includes: Determine quantiles of the return rate based on the statistical distribution; Divide the total return rate interval into at least two return rate intervals by taking the quantiles as interval endpoints.

6. The method of claim 1, wherein, The determination of the candidate conversion cost value of the push cycle based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, and the time influence factor value includes: Determine a pricing coefficient of other push crowds except the current push crowd in the future push sub-cycle based on historical pricing coefficients of the other push crowds; Input the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, and the pricing coefficient of the future push sub-cycle into a conversion cost calculation model; Determine the candidate conversion cost value of the push cycle through the conversion cost calculation model.

7. The method of claim 1, wherein, The determination of the candidate conversion cost value of the push cycle based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, and the time influence factor value includes: Among other push crowds except the current push crowd, distinguish a first other push crowd that has obtained a real-time pricing coefficient and a second other push crowd that has not obtained a real-time pricing coefficient; Determine a pricing coefficient of the second other push crowd in the future push sub-cycle based on a historical pricing coefficient of the second other push crowd; Input the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, the real-time pricing coefficient of the first other push crowd, and the pricing coefficient of the future push sub-cycle into a conversion cost calculation model; Determine the candidate conversion cost value of the push cycle through the conversion cost calculation model.

8. The method of claim 1, wherein, The determining of the value range of the pricing coefficient corresponding to each of the push crowds comprises: determining a progress of information pushing based on the push period and the historical push sub-period determination information; determining the value range of the pricing coefficient corresponding to each of the push crowds according to the progress interval to which the progress belongs; wherein the push period of each of the push crowds comprises a plurality of progress intervals, and each progress interval corresponds to a corresponding value range of the pricing coefficient.

9. A device for pushing recommendation information, characterized in that, The device comprises: a target data acquisition module configured to acquire a target conversion cost value of recommendation information in a push period; the push period comprises a historical push sub-period in which information pushing has been performed and a future push sub-period in which information pushing has not been performed; a historical data acquisition module configured to read historical push data of each push crowd in the historical push sub-period; a value range determination module configured to determine a value range of a pricing coefficient corresponding to each of the push crowds; the push crowds are obtained by grouping each push object according to a crowd division index; the crowd division index comprises a push index, and the push index comprises a conversion rate and a return on investment; the historical push data comprises a historical conversion cost value, a historical pricing coefficient, a historical push cost value, a number of the historical push sub-period, and a number of sub-periods included in the push period; a pricing coefficient determination module configured to input the number of the historical push sub-period, the number of sub-periods included in the push period, and the target conversion cost value into a cost prediction model, determine a cost prediction value of the historical push sub-period and a cost prediction value of the future push sub-period through the cost prediction model, determine a time influence factor value based on the cost prediction value of the historical push sub-period and the cost prediction value of the future push sub-period, select a candidate pricing coefficient from the value range of the current push crowd for each of the push crowds, determine a candidate conversion cost value of the push period based on the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, and the time influence factor value, and determine a real-time pricing coefficient of the current push crowd corresponding to the candidate pricing coefficient closest to the target conversion cost value, until real-time pricing coefficients corresponding to each of the push crowds are obtained; a conversion cost value determination module configured to determine a conversion cost value when each of the push crowds is pushed in the future push sub-period based on the real-time pricing coefficient and the target conversion cost value; a request receiving module configured to receive an information recommendation request; the information recommendation request carries user information of a requester; a crowd determination module configured to determine a belonging push crowd of the requester based on the user information; a conversion cost value acquisition module configured to acquire a conversion cost value of the recommendation information about the belonging push crowd and a conversion cost value of other recommendation information about the belonging push crowd. The information pushing module is configured to: calculate a matching degree between the requestor and the recommended information; calculate a matching degree between the requestor and other recommended information; calculate a pushing score value of the recommended information according to the conversion cost value of the recommended information and the matching degree between the requestor and the recommended information; calculate a pushing score value of the other recommended information according to the conversion cost value of the other recommended information and the matching degree between the requestor and the other recommended information; select a to-be-recommended information from the recommended information and the other recommended information according to the pushing score value of the recommended information and the pushing score value of the other recommended information; and push the recommended information to the requestor according to the conversion cost value of the recommended information if the to-be-recommended information is the recommended information.

10. The apparatus of claim 9, wherein, The device further comprises: The return rate obtaining module is configured to obtain a return rate obtained after the recommended information or other recommended information is pushed to each pushing object in a historical pushing period. The statistical distribution determining module is configured to determine a statistical distribution of each pushing object in each return rate interval according to the return rate. The crowd dividing module is configured to divide each pushing object into a pushing crowd based on the statistical distribution. The information pushing module is configured to push recommended information to each pushing crowd.

11. The apparatus of claim 10, wherein, The return rate obtaining module is further configured to: obtain a cumulative pushing cost value corresponding to each pushing object in the historical pushing period; determine a pushing income obtained after the recommended information or the other recommended information is pushed to each pushing object; and determine the return rate based on the cumulative pushing cost value and the pushing income.

12. The apparatus of claim 10, wherein, The crowd dividing module is further configured to: determine a total return rate interval based on the statistical distribution; divide the total return rate interval into at least two return rate intervals; the return rate interval is a subinterval of the total return rate interval; divide pushing objects with the same return rate interval into the same pushing crowd.

13. The apparatus of claim 12, wherein, The crowd dividing module is further configured to: determine quantiles of the return rate based on the statistical distribution; divide the total return rate interval into at least two return rate intervals by taking the quantiles as interval endpoints.

14. The apparatus of claim 9, wherein, The pricing coefficient determining module is further configured to: determine a pricing coefficient of other pushing crowds except the current pushing crowd in the future pushing sub-period based on historical pricing coefficients of the other pushing crowds; input the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical pushing cost value, the time influence factor value and the pricing coefficient of the future pushing sub-period into a conversion cost calculation model; and determine a candidate conversion cost value of the pushing period by the conversion cost calculation model.

15. The apparatus of claim 9, wherein, The pricing coefficient determining module is further configured to: distinguish a first other pushing crowd that has obtained a real-time pricing coefficient and a second other pushing crowd that has not obtained a real-time pricing coefficient from other pushing crowds except the current pushing crowd; determine a pricing coefficient of the second other pushing crowd in the future pushing sub-period based on historical pricing coefficients of the second other pushing crowd. inputting the candidate pricing coefficient, the historical conversion cost value, the historical pricing coefficient, the historical push cost value, the time influence factor value, a real-time pricing coefficient of the first other pusher group, and a pricing coefficient of a future push sub-period into a conversion cost calculation model; determining, by the conversion cost calculation model, a candidate conversion cost value of the push period.

16. The apparatus of claim 9, wherein, The value interval determination module is further configured to: determine a progress of information push based on the push period and the historical push sub-period determination information; determine a value interval of the pricing coefficient corresponding to each pusher group according to a progress interval to which the progress belongs; wherein the push period of each pusher group includes a plurality of progress intervals, and each progress interval corresponds to a corresponding value interval of the pricing coefficient. 17.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-16. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 8.

18. A computer readable storage medium storing a computer program, wherein the computer program comprises instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 17. The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 8.

Images