CN109657132B

CN109657132B - Recommended information cost control method, recommended information cost control device, computer equipment and storage medium

Info

Publication number: CN109657132B
Application number: CN201710942481.1A
Authority: CN
Inventors: 符永顺; 宋元峰; 胡健; 陈功; 黄识; 张必锋
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2017-10-11
Filing date: 2017-10-11
Publication date: 2023-04-25
Anticipated expiration: 2037-10-11
Also published as: CN109657132A

Abstract

The application relates to a recommended information cost control method, which comprises the following steps: the method comprises the steps of obtaining accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change; acquiring conversion numbers corresponding to the recommended information in a preset time period; calculating according to the accumulated exposure cost and the conversion number to obtain average conversion cost; calculating according to the average conversion cost and the standard conversion unit price to obtain a current error value; according to the method, the conversion unit price can be effectively adjusted according to the error value, and the adjustment accuracy is improved. In addition, a recommended information cost control device, a computer device and a storage medium are also provided.

Description

Recommended information cost control method, recommended information cost control device, computer equipment and storage medium

Technical Field

The present invention relates to the field of computer processing technologies, and in particular, to a method and apparatus for controlling recommended information cost, a computer device, and a storage medium.

Background

With the development of internet technology, people's lives are more and more closely linked with the internet. The internet is one of the important ways for people to obtain information, and in order to recommend information to interested users, many fields will recommend information by means of a network platform. The information recommending party recommends the information to the terminal user by means of a third party network platform. The third party network platform performs data statistics according to the predicted cost in the recommending process, and is likely to be inconsistent with or even far from the actual cost data obtained by the information recommending party.

The conventional solution to the above problem is to simply adjust according to the ratio of the predicted cost data to the actual cost data, and the accuracy of the adjustment is low.

Disclosure of Invention

Based on this, it is necessary to provide a recommended information cost control method, apparatus, computer device and storage medium with high adjustment accuracy, in order to solve the problem of low adjustment accuracy.

A recommended information cost control method, comprising:

Acquiring accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change;

acquiring the conversion number corresponding to the recommended information in the preset time period;

calculating to obtain average conversion cost according to the accumulated exposure cost and the conversion number;

calculating to obtain a current error value according to the average conversion cost and the standard conversion unit price;

and calculating a current conversion unit price adjustment value according to the current error value, and calculating a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price.

A recommended information cost control apparatus, the apparatus comprising:

the system comprises an exposure cost acquisition module, a calculation module and a calculation module, wherein the exposure cost acquisition module is used for acquiring accumulated exposure cost corresponding to recommended information in a preset time period, the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change;

the conversion number acquisition module is used for acquiring the conversion number corresponding to the recommended information in the preset time period;

The average conversion cost calculation module is used for calculating and obtaining average conversion cost according to the accumulated exposure cost and the conversion number;

the current error value calculation module is used for calculating and obtaining a current error value according to the average conversion cost and the standard conversion unit price;

the target conversion unit price calculation module is used for calculating a current conversion unit price adjustment value according to the current error value, and calculating a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price.

A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to:

A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:

According to the recommended information cost control method, the recommended information cost control device, the computer equipment and the storage medium, average conversion cost is obtained through calculation by obtaining accumulated exposure cost and conversion number corresponding to recommended information in a preset time period, then a current error value is obtained through calculation according to the average conversion cost and standard conversion unit price, a current conversion unit price adjustment value is obtained through calculation according to the current error value, and then target conversion unit price corresponding to the next period is obtained through calculation according to the current conversion unit price adjustment value and the standard conversion unit price. The conversion unit price is continuously adjusted according to the calculated error value, so that the error value between the conversion unit price and the standard conversion unit price is gradually smaller, and finally the average conversion cost corresponding to the recommended information is close to or consistent with the standard conversion unit price.

Drawings

FIG. 1 is an application environment diagram of a recommendation information cost control method in one embodiment;

FIG. 2 is a flow chart of a method of recommendation information cost control in one embodiment;

FIG. 3 is a schematic diagram of an interface for information recommendation in one embodiment;

FIG. 4 is a flowchart of a method for acquiring cumulative exposure costs corresponding to recommended information in a predetermined period of time according to one embodiment;

FIG. 5 is a flowchart of a method for determining an exposure cost corresponding to each exposure in a predetermined period of time according to an embodiment;

FIG. 6 is a flowchart of a method for calculating a current conversion unit price adjustment value based on a current error value in one embodiment;

FIG. 7 is a flowchart of a method for calculating a current conversion unit price adjustment value according to a current error value in another embodiment;

FIG. 8 is a schematic diagram of a process regulated by PID control in one embodiment;

FIG. 9 is a flow chart of a method for calculating a current conversion unit price adjustment value based on a current proportional control value, a current integral control value, and a current derivative control value, in one embodiment;

FIG. 10 is a flow chart of a method of cost control of recommended information in another embodiment;

FIG. 11 is a flow diagram of controlling ad conversion costs in one embodiment;

FIG. 12 is a flow chart of a method of cost control of recommended information in yet another embodiment;

FIG. 13 is a block diagram showing a configuration of a recommended information cost control device in one embodiment;

FIG. 14 is a block diagram of an exposure cost acquisition module in one embodiment;

FIG. 15 is a block diagram of the target conversion unit price calculation module in one embodiment;

FIG. 16 is a block diagram of a target conversion unit price calculation module in another embodiment;

FIG. 17 is a block diagram showing another embodiment of a recommended information cost control apparatus;

fig. 18 is an internal structural view of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

FIG. 1 is a diagram of an application environment for a method of cost control of recommended information in one embodiment. Referring to fig. 1, the recommended information cost control method is applied to a recommended information cost control system. The recommended information cost control system includes a plurality of terminals 110 and a server 120. The terminal 110 and the server 120 are connected through a network. The terminal 110 may be a desktop terminal or a mobile terminal, and the mobile terminal may be at least one of a mobile phone, a tablet computer, a notebook computer, and the like. The server 120 may be implemented as a stand-alone server or as a server cluster composed of a plurality of servers. The server 120 transmits recommended information to the terminal 110, and each time the recommended information is transmitted to one terminal, the corresponding recommended information corresponds to one exposure, and each exposure generates a corresponding exposure cost. Specifically, the server 102 acquires the cumulative exposure cost corresponding to the recommended information in the preset period, the cumulative exposure cost being superimposed by each exposure cost corresponding to the recommended information, and each exposure cost being calculated according to the conversion unit price corresponding to the period of the dynamic adjustment change. The server 102 also needs to obtain the conversion number corresponding to the recommended information in the preset time period, calculate to obtain an average conversion cost according to the accumulated exposure cost and the conversion number, calculate to obtain a current error value according to the average conversion cost and the standard conversion unit price, calculate to obtain a current conversion unit price adjustment value according to the current error value, and calculate to obtain the target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price. And calculating the cost of each exposure of the next period according to the adjusted target conversion unit price so as to enable the overall average conversion cost corresponding to the recommended information to be close to or consistent with the standard conversion unit price.

As shown in FIG. 2, in one embodiment, a method of recommendation information cost control is provided. The present embodiment is mainly exemplified by the application of the method to the server 120 in fig. 1. Referring to fig. 2, the method for controlling the recommended information cost specifically includes the following steps:

s202, acquiring accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change.

The recommendation information is information which is actively recommended to the terminal user. The recommendation information may be information such as articles, advertisements, music, etc. The exposure cost refers to exposure fees charged after the third party platform pushes the recommended information to the end user. For example, advertisers may place advertisements via a third party platform that, after placing advertisements to users, may be charged a corresponding placement fee (i.e., exposure fee). The accumulated exposure cost is obtained by superposing the exposure cost of each time corresponding to the recommended information, the recommended information is pushed to one end user, which is equivalent to the exposure of the recommended information once, and similarly, the recommended information is correspondingly exposed for N times by pushing the recommended information to N end users. The cost per exposure refers to the exposure cost charged by pushing recommended information to one end user. The cost per exposure is positively correlated to the conversion unit price, which is dynamically adjusted with the period.

The conversion unit price refers to the conversion cost corresponding to one conversion. The conversion refers to specific actions of the user on the recommended information after the recommended information is pushed. The specific behavior can be set according to the actual situation. For example, if the recommended information is an article, the user can click to read the article as a corresponding specific behavior, and the user can click to read the article is conversion; if the recommendation information is an advertisement of the mobile application APP, the specific behavior can be set as the behavior of downloading the APP by the user, and the APP is downloaded once by the user and is recorded as one conversion.

Specifically, a sampling period is preset, and the conversion unit price is adjusted periodically according to the sampling period. For example, setting the sampling period to 5 minutes, the conversion unit price is adjusted every 5 minutes. And each adjustment needs to first acquire the accumulated exposure cost corresponding to the recommended information in the preset time period. The preset time period may be set according to the actual situation, for example, a time period from the start time to the current time may be set as the preset time period, or a time period nearest to the current time may be set as the preset time period.

S204, obtaining the conversion number corresponding to the recommended information in the preset time period.

The conversion number refers to the conversion number of the recommended information by the end user after the recommended information is recommended to the end user. Conversion refers to the specific behavior of a user on recommended information after the recommended information is pushed to the user. The specific behavior can be set according to the actual situation. For example, if the recommendation information is an advertisement of a certain product, the behavior that the user purchased the product may be set as a specific behavior, and if the recommendation information is an advertisement of a certain APP, the specific behavior may be set as a behavior that the user downloads the APP.

The conversion number corresponding to the recommended information in the preset time period is obtained, namely the total conversion times in the preset time period are obtained, and each conversion is recorded as 1. Specifically, the conversion number may be obtained by the conversion number fed back by the information recommender (e.g., advertiser), or may be obtained by the server monitoring the conversion behavior of the user on the recommended information, and the server records the conversion behavior of the user on the recommended information once every time the server monitors the conversion behavior. For example, the third party platform performs information promotion for the information recommender, and supposedly pushes the recommended information to 1000 end users, and then the number of users who convert the recommended information in the 1000 end users, that is, the corresponding conversion times, needs to be counted. For example, assuming that the recommendation information is an advertisement for a product, if the end user purchases the product to convert the recommendation information, if 10 end users out of 1000 end users pushed purchase the product, the corresponding conversion number is 10. FIG. 3 is a schematic diagram of an interface for information recommendation via a third party platform in one embodiment.

S206, calculating to obtain the average conversion cost according to the accumulated exposure cost and the conversion number.

The average conversion cost refers to the actual average conversion cost of each conversion of the recommended information in the preset time period. The average conversion cost is calculated from the ratio of the cumulative exposure cost to the conversion number. Specifically, after the cumulative exposure cost and the conversion number in the preset time period are obtained, the average conversion cost is calculated according to the ratio of the cumulative exposure cost to the conversion number. The cumulative exposure cost refers to the total cost in a preset time period, the conversion number refers to the total conversion times in the preset time period, and the ratio of the total conversion times to the conversion times is the calculated average conversion cost.

And S208, calculating according to the average conversion cost and the standard conversion unit price to obtain the current error value.

The standard conversion unit price refers to preset conversion cost per conversion. And average conversion cost refers to the calculated cost per actual conversion over a period of time. Specifically, the current error value is calculated according to the difference between the standard conversion unit price and the average conversion cost. The current error value is used to measure the gap between the theoretical conversion cost (standard conversion unit price) and the actual conversion cost. Wherein the standard conversion unit price is a cost of one conversion set by an information recommender (e.g., advertiser).

S210, calculating to obtain a current conversion unit price adjustment value according to the current error value, and calculating to obtain a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price.

In order to make the total average conversion cost corresponding to the recommended information approximate to or keep the total average conversion cost consistent with the standard conversion unit price, namely to make the actual conversion cost corresponding to the recommended information finally approximate to the standard conversion unit price, the conversion unit price is periodically adjusted according to the set sampling period so as to adjust the error value towards the direction of reducing. In one embodiment, after the current error value is calculated, a corresponding proportional control coefficient is obtained, and the current conversion unit price adjustment value is calculated according to the product of the proportional control coefficient and the current error value. And then calculating to obtain a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price, so that the exposure cost can be conveniently adjusted according to the target conversion unit price in the next period. In one embodiment, the sum of the current conversion unit price adjustment value and the standard conversion unit price is taken as the target conversion unit price for the next cycle. The conversion unit price of each period is continuously adjusted so that the total average conversion cost of the recommended information is consistent or equivalent to the standard conversion unit price.

For example, let initial conversion unit price y0=standard conversion unit price, first, calculate the accumulated exposure costs in the first period according to y0, wherein each accumulated exposure costs in the first period is positively correlated with y 0. Assuming that the preset period is just one cycle, then the average conversion cost in the first cycle=the cumulative exposure cost in the first cycle/the conversion number in the first cycle. If the calculated average conversion cost is greater than the standard conversion unit price, the deduction of the exposure cost in the first period is higher, the corresponding conversion unit price needs to be regulated down later, and the conversion unit price and the cost of each exposure are in positive correlation, so that the cost of each subsequent exposure is correspondingly regulated down. If the calculated average conversion cost is smaller than the standard conversion cost, the deduction of the exposure cost in the first period is lower, the corresponding conversion unit price needs to be increased later, and the conversion unit price and the cost of each exposure are in positive correlation, so that the cost of each subsequent exposure is correspondingly increased. Specifically, the adjustment is performed according to the calculated error value, if the error value is larger, the corresponding adjustment amplitude is also relatively larger, and the adjustment is performed at intervals of preset time (for example, 5 min) according to a preset sampling period, so that the error value is adjusted towards a reduced direction, and the final average conversion cost of the recommended information is kept consistent with the standard conversion unit price.

According to the recommended information cost control method, the average conversion cost is obtained through calculation by acquiring the accumulated exposure cost and the conversion number corresponding to the recommended information in the preset time period, then the current error value is obtained through calculation according to the average conversion cost and the standard conversion unit price, the current conversion unit price adjustment value is obtained through calculation according to the current error value, and then the target conversion unit price corresponding to the next period is obtained through calculation according to the current conversion unit price adjustment value and the standard conversion unit price. The conversion unit price is continuously adjusted according to the calculated error value, so that the error value between the conversion unit price and the standard conversion unit price is gradually smaller, and finally the average conversion cost corresponding to the recommended information is close to or consistent with the standard conversion unit price.

As shown in fig. 4, in one embodiment, the step S202 of obtaining the cumulative exposure cost corresponding to the recommended information in the preset time period includes:

S202A, acquiring the current time, taking the current time as a termination endpoint, determining a corresponding starting endpoint according to a preset time interval, and obtaining a preset time period according to the termination endpoint and the starting endpoint.

In order to improve the adjustment sensitivity of the recommended information cost control, when calculating the average conversion cost, the accumulated cost of exposure in the last period of time is adopted, and compared with the traditional statistical mode of statistics from the recommended beginning to the current moment, the accumulated cost of exposure in the last period of time can reflect the change trend of the average conversion cost in the last period of time more sensitively, so that the average conversion cost can be adjusted more accurately. Specifically, the current time is obtained, the current time is taken as a termination endpoint, a corresponding starting endpoint is obtained through calculation according to a preset time interval, and then a time period between the starting endpoint and the termination endpoint is taken as a preset time period.

S202B, acquiring the exposure cost corresponding to each exposure in a preset time period, and superposing the exposure cost to obtain the accumulated exposure cost in the preset time period.

The cost of each exposure is not necessarily the same, so the cost of each exposure in the preset time period needs to be obtained, and then the cost of each exposure is superimposed to obtain the accumulated cost of exposure in the preset time period. In one embodiment, the cost per exposure is related to the conversion unit price, the predicted conversion rate, wherein the cost per exposure is directly related to the conversion unit price, and also directly related to the predicted conversion rate. The preset time period may include a plurality of periods, and the conversion unit price is dynamically adjusted and changed along with the periods, so that the conversion unit prices corresponding to the different periods are likely to be different. The predicted conversion rate refers to the probability of converting the recommended information by the predicted user, and since each user has its own personalized user characteristics, the predicted conversion rate is predicted according to the user characteristics, so that the predicted conversion rate corresponding to each user may be different. The cost of each exposure may be different and therefore the exposure cost for each exposure needs to be obtained. In one embodiment, the cost per exposure = conversion unit price predicts the conversion rate. The conversion unit price is dynamically adjusted along with the period, and the predicted conversion rate is the predicted probability of converting the recommended information to the user according to the characteristic behavior of the user. Specifically, a predictive conversion model is obtained in advance according to user characteristic training, the probability of user conversion recommendation information is predicted according to the predictive conversion model according to user characteristics, and the corresponding predictive conversion rates are different due to different user characteristics of each user.

As shown in fig. 5, in one embodiment, the step S202B of obtaining the exposure cost corresponding to each exposure in the preset period of time includes:

s502, obtaining a predicted click rate, a predicted conversion rate and a corresponding conversion unit price corresponding to the current exposure.

The exposure cost corresponding to each exposure is proportional to the corresponding predicted click rate and conversion rate, and is also proportional to the corresponding conversion unit price. The predicted click rate and the predicted conversion rate are obtained by predicting the probability of clicking recommended information and the probability of converting recommended information of a user according to the behavior characteristics of the user through the trained click prediction model and the trained conversion prediction model. The conversion unit price is dynamically adjusted and changed along with the period. The conversion unit price corresponding to the current exposure is determined according to the period in which the current exposure is located.

S504, calculating the current exposure cost according to the product of the predicted click rate, the predicted conversion rate and the corresponding conversion unit price.

Specifically, the cost per exposure=predicted click rate×predicted conversion rate×conversion unit price, and then the cost per exposure is calculated as the product of the predicted click rate, the predicted conversion rate, and the conversion unit price corresponding to the current period. Because the predicted click rate and the predicted conversion rate corresponding to each user are different, when the cost of each exposure is calculated, the predicted click rate and the predicted conversion rate corresponding to the current user and corresponding to the current exposure need to be obtained, and then the conversion unit price of the period in which the current exposure is located needs to be obtained. Then the product of the three can be used to calculate the current exposure cost.

As shown in fig. 6, in one embodiment, the step S210 of calculating the current conversion unit price adjustment value according to the current error value includes:

S210A, a proportion control coefficient is obtained, and a current proportion control value is obtained through calculation according to the proportion control coefficient and the current error value.

The objective of the recommended information cost control is to keep the average conversion cost of the recommended information consistent with the standard conversion unit price, and obtain a proportional control coefficient after calculating to obtain a current error value, and calculate to obtain the current proportional control value according to the proportional control coefficient and the current error value. The current proportional control value is positively correlated with the current proportional control coefficient and is positively correlated with the current error value. The current proportional control value is obtained by adjusting the current error value according to the proportional control coefficient, and the current error value is adjusted towards the direction of reduction, the error reduction speed is dependent on the proportional control coefficient Kp, and the larger the proportional control coefficient Kp is, the faster the error reduction is.

S210B, performing integral calculation according to the current error value to obtain a current integral control value.

The integral control value and the integral of the error value are in a proportional relation, and each time the integral of the error is calculated, a current error value is added on the basis of the original integral value. The current error value may be positive or negative, and the increment of integration is negative when the current error value is negative. The integral term can eliminate steady-state errors which occur when proportional control is used alone, so that the control precision is further improved.

And S210C, calculating to obtain a current conversion unit price adjustment value according to the current proportion control value and the current integral control value.

The current conversion unit price adjustment value refers to a calculated amplitude value which needs to be adjusted for the conversion unit price. The current conversion unit price adjustment is positively correlated with the current proportional control value and is positively correlated with the current integral control value. In one embodiment, the current conversion unit price adjustment value may be calculated by a weighted sum of the current proportional control value and the current integral control value. In another embodiment, in order to facilitate control of the current conversion unit price adjustment value to be too large or too small, a range of conversion unit price adjustment values is preset, a maximum endpoint value in the range is taken as a first threshold value, and a minimum endpoint value in the range is taken as a second threshold value. When the calculated current conversion unit price adjustment value is larger than a preset first threshold value, the first threshold value is used as the current conversion unit price adjustment value, and when the calculated current conversion unit price adjustment value is smaller than a preset second threshold value, the second threshold value is used as the current conversion unit price adjustment value. For example, if the conversion unit price adjustment value is set in advance to a range of [ -7,7], then if the current conversion unit price adjustment value is greater than 7, the maximum value 7 in the range is taken as the current conversion unit price adjustment value. And if the current conversion unit price adjustment value is smaller than-7, taking the minimum value-7 in the range as the current conversion unit price adjustment value. If the calculated current conversion unit price adjustment value is within the preset range, the obtained current conversion unit price adjustment value is directly used.

In one embodiment, the step S210B of performing an integral calculation according to the current error value to obtain a current integral control value includes: determining a current weight coefficient according to the current moment corresponding to the current error value, and determining a current target error value according to the current error value and the current weight coefficient; and acquiring a historical target error value corresponding to the historical period, and calculating to obtain a current integral control value according to the historical target error value and the current target error value.

Specifically, a weight coefficient associated with time is preset, a current weight coefficient is determined according to a current moment corresponding to the current error value, and a current target error value is obtained through calculation according to the current error value and the current weight coefficient. The current target error value is positively correlated with the current error value and the current weight coefficient. And simultaneously acquiring a historical target error value corresponding to the historical period, wherein the historical target error value is also determined by the corresponding weight coefficient and the historical error value. And calculating to obtain a current integral control value according to the historical target error value and the current target error value. In one embodiment, the current integral control value

Where w (t) is a weight coefficient associated with time. Because the accuracy of the prediction click model and the prediction conversion model is different in the different periods before and after information recommendation, the deviation of the different periods is also different. In general, the predicted click rate and the predicted conversion rate in the early information recommendation stage are relatively inaccurate, so that the calculated error is larger, and the predicted click rate and the predicted conversion rate are relatively accurate and the error is relatively smaller in the later information recommendation stage because of more click and conversion data. Therefore, when calculating the integral control value, the errors in different periods are added with different weight coefficients when accumulating, for example, the weight coefficients attenuated with time are added, so that the control accuracy is further improved.

As shown in fig. 7, in one embodiment, the step S210 of calculating the current conversion unit price adjustment value according to the current error value further includes:

S210D, a first error value corresponding to the previous period is obtained.

Each period corresponds to a corresponding error value, and an error value corresponding to the previous period is obtained, so that the error value is called a first error value for convenience of distinguishing. Since the conversion unit price is dynamically adjusted along with the period, and the adjustment of the conversion unit price requires calculation to obtain the error value of the average conversion cost and the standard conversion unit price in the preset time period, each period corresponds to a corresponding error value.

S210E, calculating to obtain a current differential control value according to the first error value and the current error value.

Wherein the differential control value is in direct proportion to the rate of change of the error. If the control is performed by using the proportional and integral terms, hysteresis is generated, so that the control is excessive, namely overshoot, and the differential control is correspondingly adjusted according to the change trend of the error, so that the overshoot is gradually reduced, and therefore, the differential term has the characteristics of advance and prediction. The cost of the recommended information can be controlled more precisely by differentiating the control value.

The step S210C of calculating the current conversion unit price adjustment value according to the current proportional control value and the current integral control value includes: and calculating according to the current proportional control value, the current integral control value and the current differential control value to obtain a current conversion unit price adjustment value.

Specifically, in order to calculate the current conversion unit price adjustment value more accurately, the current conversion unit price adjustment value is calculated simultaneously using the proportional control value, the integral control value, and the derivative control value together. I.e. the current conversion unit price adjustment value is calculated by means of PID control, wherein the P (report) table is the scale, I (Integration) the integral and D (Differentiation) the derivative. FIG. 8 is a schematic diagram of a process for regulating by PID control in one embodiment. Firstly, receiving accumulated exposure cost and corresponding conversion number in a preset time period, calculating to obtain corresponding average conversion cost, then calculating to obtain a current error value e (t) with standard conversion unit price, then respectively calculating corresponding current proportional control value, current integral control value and current differential control value according to the current error value e (t), and then obtaining corresponding current conversion unit price adjustment value u (t) according to the proportional control value, the integral control value and the differential control value. And then calculating the target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value u (t), and continuously adjusting the target conversion unit price of the later period in such a way that the average conversion cost of the finally obtained overall is close to or consistent with the standard conversion cost.

As shown in fig. 9, in one embodiment, the step S210C of calculating the current conversion unit price adjustment value according to the current proportional control value, the current integral control value, and the current derivative control value includes:

s902, acquiring a sampling period, an integral time coefficient and a differential time coefficient, and calculating to obtain the integral coefficient according to the sampling period, the integral time coefficient and the proportional control coefficient.

Specifically, after the current proportional control value, the current integral control value, and the current differential control value are calculated, a sampling period, an integral time coefficient, and a differential time coefficient need to be obtained. The sampling period refers to sampling every preset time, for example, 5 minutes. The integral time coefficient and the differential time coefficient are obtained by carrying out parameter setting through experiments, wherein the integral time coefficient is a coefficient related to an integral term, and the differential time coefficient is a coefficient related to a differential term.

Wherein the integral coefficient is positively correlated with the sampling period and the proportional control coefficient, and inversely correlated with the integral time coefficient. In one embodiment, where the integration coefficient is equal to the product of the sampling period and the proportional control coefficient and then divided by the corresponding integration time coefficient, a specific formula may be expressed as follows: k (K) _i ＝K _p *T ₀ /T _i Wherein ki denotes an integral coefficient, kp denotes a proportional control coefficient, T ₀ Representing the period of use, T _i Representing the integration time coefficient.

S904, calculating a differential coefficient according to the sampling period, the differential time coefficient and the proportional control coefficient.

Wherein the differential coefficient is positively correlated with the differential time coefficient and the proportional control coefficient, and inversely correlated with the sampling period. In one embodiment, the differential coefficient is obtained by multiplying the differential time coefficient by the proportional control coefficient and then dividing by the sampling period. The specific formula can be expressed as follows: k (K) _d ＝K _p *T _d /T ₀ Wherein Kd represents a differential coefficient, kp represents a proportional control coefficient, T ₀ Representing the period of use, T _d Representing the differential time coefficient.

S906, calculating to obtain a current conversion unit price adjustment value according to the integral coefficient, the differential coefficient, the current proportional control value, the current integral control value and the current differential control value.

After the integral coefficient and the differential coefficient are determined, a target integral control value can be obtained through calculation by the product of the integral coefficient and the current integral control value, a target differential control value can be obtained through calculation by the product of the differential coefficient and the current differential control value, and a current conversion unit price adjustment value can be obtained through calculation according to the current proportional control value, the target differential control value and the target integral control value. Specifically, the current conversion unit price adjustment value is equal to the sum of the current proportional control value, the target differential control value, and the target integral control value. In one embodiment, the current conversion unit price adjustment value may be calculated using the following formula.

Wherein Kp represents a proportional control coefficient, T ₀ Representing the sampling period, T _i Representing the integral time coefficient, T _d Representing the differential time coefficient. Wherein Kp, T ₀ 、T _i And T _d Are all values determined in advance by experiments, in one embodiment, kp takes the value of 6, T ₀ The value is 5min, T _i The value is 60min, T _d The value is 15min.

As shown in fig. 10, in one embodiment, the above recommended information cost control method further includes:

and S212, when a plurality of recommended information are provided, acquiring the target conversion unit price corresponding to the current period of each recommended information in real time.

When a plurality of recommendation information are provided, the recommendation value of each user for each recommendation information needs to be calculated respectively, and then pushing is carried out according to the calculated recommendation value, so that the recommendation accuracy is improved. The recommended value of each recommended information is related to the target conversion unit price corresponding to the current period, so that the target conversion unit price corresponding to the current period of each recommended information needs to be acquired first.

S214, obtaining the predicted click rate and the predicted conversion rate of each piece of recommended information of the current user.

The prediction click rate and the prediction conversion rate of different users on the same recommended information are different, and the prediction click rate and the prediction conversion rate of the same user on different recommended information are also different. In order to calculate the recommendation value of the current user for each recommendation information, the predicted click rate and the predicted conversion rate of the current user for each recommendation information also need to be obtained. The predicted click rate and the predicted conversion rate are obtained by predicting the probability of clicking the recommended information of the user and the probability of converting the recommended information according to the behavior characteristics of the user through the trained click prediction model and conversion prediction model. The conversion unit price is dynamically adjusted and changed along with the period. The conversion unit price corresponding to the current exposure is determined according to the period in which the current exposure is located.

And S216, calculating to obtain a recommended value of each recommended information corresponding to the current user according to the conversion unit price, the predicted click rate and the predicted conversion rate.

The recommendation value reflects the income which can be brought by recommending the corresponding recommendation information. The magnitude of the recommended value is positively correlated with the conversion unit price, and is positively correlated with the predicted click rate and the predicted conversion rate. The predicted click rate and the predicted conversion rate are obtained through corresponding click prediction models and conversion prediction models. The conversion unit price is dynamically adjusted and changed along with the period. In one embodiment, recommended value = conversion unit price × predicted click rate × predicted conversion rate. In another embodiment, the corresponding recommended value may also be determined by calculating the corresponding eCPM (effectivecost per mille). eCPM refers to the revenue that can be obtained per thousand impressions. And taking the eCPM value corresponding to each piece of calculated recommendation information as a corresponding recommendation value. I.e., ecpm=conversion unit price predicted click rate predicted conversion rate 1000.

S218, sorting the plurality of recommendation information according to the recommendation values, and recommending according to the sorting result.

Specifically, after the recommendation value corresponding to each piece of recommendation information is obtained through calculation, the corresponding pieces of recommendation information are ranked according to the sequence from the big recommendation value to the small recommendation value, and a corresponding ranking result is obtained. In one embodiment, the first recommendation information (i.e., the recommendation information with the largest recommendation value) in the ranking result is recommended to the current user. After recommending the recommended information to the corresponding terminal, the corresponding exposure cost of the exposure is correspondingly determined. And the accumulated exposure cost in a period of time is conveniently counted later, then the corresponding conversion number is obtained, the average conversion cost of the current period is calculated according to the accumulated exposure cost and the conversion number, and then the target conversion unit price corresponding to the next period is determined by comparing the average conversion cost with the standard conversion unit price. By such a cycle, the final corresponding average conversion cost can be more effectively controlled to be close to the standard conversion unit price.

In a particular embodiment, in the field of advertising, the oCPM (optimized CPM) bid pattern is a bid pattern for advertising, and the oCPM bid pattern is a bid per conversion, but per conversionExposure is performed in a bidding mode of deduction. Since advertisers bid on conversion and advertising platforms deduct fees on exposure, i.e., the bidding and deduction methods are decoupled, the resulting conversion costs of the advertisement may not be consistent with, or even far from, the advertiser's bid. Therefore, the advertising platform is required to control the conversion cost for the advertisement by controlling the cost of each exposure so that the final average conversion cost is consistent with the conversion bid of the advertiser, thereby improving the conversion efficiency. In the process of advertising, the conversion unit price is continuously adjusted to adjust the cost of each exposure. Where cost per exposure = conversion unit price × predicted click rate × predicted conversion rate. FIG. 11 is a flow diagram illustrating the control of ad conversion costs, in one embodiment. 1) And acquiring exposure deduction (namely accumulated exposure cost) and corresponding conversion number in a preset time period. 2) And calculating to obtain the corresponding average conversion cost. The specific formula is as follows:

Wherein cpa (t) represents the average conversion cost,/->

Refers to->

The cumulative exposure cost over the period of time, paid refers to the cost per exposure. />

Representation->

Number of conversions over a period of time. 3) The error value between the average conversion cost and the standard conversion unit price is calculated as follows: e (t) =bid-cpa (t), where bid represents the standard conversion bid and e (t) represents the error value of both. 4) The corresponding proportional control value is calculated from the error value e (t), the proportional control value being proportional to the proportional control coefficient, i.e. proportional control value=kp×e (t). 5) Calculating corresponding integral according to error valueControl values and differential control values. The integral control value is equal to the accumulation of the error value corresponding to each cycle of the history, and the corresponding error value of each cycle is multiplied by a weight coefficient which decays with time in order to calculate the corresponding integral term more accurately. The specific formula is as follows: integral control value->

Where w (t) is a weight coefficient associated with time. The differential control value is equal to the current error value and the last period error value, differential control value=e (t) -e (t-1). 6) And calculating the conversion unit price adjustment value according to the proportional control value, the integral control value and the differential control value. Specifically, the following formula is adopted to calculate the corresponding conversion unit price adjustment value u (t). / >

Wherein Kp represents a proportional control coefficient, T ₀ Representing the sampling period, T _i Representing the integral time coefficient, T _d Representing the differential time coefficient. Wherein Kp, T ₀ 、T _i And T _d Are all values determined experimentally in advance. The parameters can be determined experimentally in advance according to actual conditions. 7) The conversion unit price is updated as the conversion unit price of the next cycle according to the conversion unit price adjustment value u (t), and is specifically expressed as follows: calibrationBid (t) = bid + u (t), where calibrationBid (t) denotes the updated conversion unit price. 8) The eCPM value of the advertisement is calculated. Specifically, ecpm= calibrationBid (t) ×pctr×pcvr×1000, wherein pCTR represents a predicted click rate and pCVR represents a predicted conversion rate. 9) And then sorting according to the calculated eCPM value. 10 According to the sequencing result, the corresponding exposure cost of each recommendation can be determined after the recommendation is performed, namely the exposure cost of each time. And then, circularly entering the step 1), and continuously adjusting the conversion unit price corresponding to each period to finally enable the average cost of advertisement delivery to be close to or consistent with the standard conversion unit price. />

As shown in fig. 12, a recommended information cost control method is proposed, which includes the steps of:

S1201, acquiring the current time, taking the current time as a termination endpoint, determining a corresponding starting endpoint according to a preset time interval, and obtaining a preset time period according to the termination endpoint and the starting endpoint.

S1202, acquiring the exposure cost corresponding to each exposure in a preset time period, and superposing the exposure cost to obtain the accumulated exposure cost in the preset time period.

S1203, obtaining the conversion number corresponding to the recommended information in the preset time period.

And S1204, calculating to obtain the average conversion cost according to the accumulated exposure cost and the conversion number.

S1205, calculating to obtain the current error value according to the average conversion cost and the standard conversion unit price.

S1206, acquiring a proportion control coefficient, and calculating to obtain a current proportion control value according to the proportion control coefficient and the current error value;

s1207, determining a current weight coefficient according to the current moment corresponding to the current error value, and determining a current target error value according to the current error value and the current weight coefficient;

s1208, acquiring a historical target error value corresponding to the historical period, and calculating to obtain a current integral control value according to the historical target error value and the current target error value.

S1209, a first error value corresponding to the previous period is acquired.

S1210, calculating the current differential control value according to the first error value and the current error value.

S1211, acquiring a sampling period, an integral time coefficient and a differential time coefficient, and calculating to obtain the integral coefficient according to the sampling period, the integral time coefficient and the proportional control coefficient.

S1212, calculating a differential coefficient according to the sampling period, the differential time coefficient and the proportional control coefficient.

S1213, calculating to obtain a current conversion unit price adjustment value according to the integral coefficient, the differential coefficient, the current proportional control value, the current integral control value and the current differential control value.

It should be understood that, although the steps in the above-described flowcharts are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps may comprise a plurality of sub-steps or phases, which are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the sub-steps or phases are performed necessarily performed in sequence, but may be performed alternately or at the bottom with at least some of the other steps or phases of other steps.

As shown in fig. 13, there is proposed a recommended information cost control apparatus including:

an exposure cost obtaining module 1302, configured to obtain an accumulated exposure cost corresponding to the recommended information in a preset time period, where the accumulated exposure cost is obtained by superimposing exposure costs corresponding to the recommended information each time, and each exposure cost is obtained by calculating a conversion unit price corresponding to a period according to a dynamic adjustment change;

a conversion number obtaining module 1304, configured to obtain a conversion number corresponding to the recommended information in the preset time period;

an average conversion cost calculation module 1306, configured to calculate an average conversion cost according to the accumulated exposure cost and the conversion number;

a current error value calculation module 1308, configured to calculate a current error value according to the average conversion cost and the standard conversion unit price;

the target conversion unit price calculation module 1310 is configured to calculate a current conversion unit price adjustment value according to the current error value, and calculate a target conversion unit price corresponding to the next cycle according to the current conversion unit price adjustment value and the standard conversion unit price.

As shown in fig. 14, in one embodiment, the exposure cost acquisition module 1302 includes:

A preset time period determining module 1302A, configured to obtain a current time, determine a corresponding starting endpoint according to a preset time interval with the current time as a termination endpoint, and obtain a preset time period according to the termination endpoint and the starting endpoint;

the exposure cost superimposing module 1302B is configured to obtain an exposure cost corresponding to each exposure in the preset time period, and superimpose the exposure cost each time to obtain an accumulated exposure cost in the preset time period.

In one embodiment, the exposure cost superimposing module is further configured to obtain a predicted click rate, a predicted conversion rate, and a corresponding conversion unit price corresponding to the current exposure, and calculate the current exposure cost according to a product of the predicted click rate, the predicted conversion rate, and the corresponding conversion unit price.

As shown in fig. 15, in one embodiment, the target conversion unit price calculation module 1310 includes:

the proportion calculating module 1310A is configured to obtain a proportion control coefficient, and calculate a current proportion control value according to the proportion control coefficient and the current error value;

the integral calculation module 1310B is configured to perform integral calculation according to the current error value to obtain a current integral control value;

and the adjustment value calculating module 1310C is configured to calculate a current conversion unit price adjustment value according to the current proportional control value and the current integral control value.

In one embodiment, the integral calculation module 1310B is further configured to determine a current weight coefficient according to a current time corresponding to the current error value, determine a current target error value according to the current error value and the current weight coefficient, obtain a historical target error value corresponding to the historical period, and calculate a current integral control value according to the historical target error value and the current target error value.

As shown in fig. 16, in one embodiment, the target conversion unit price calculation module 1310 further includes:

the differential calculation module 1310D is configured to obtain a first error value corresponding to a previous period, and calculate a current differential control value according to the first error value and the current error value;

the adjustment value calculating module 1310C is further configured to calculate a current conversion unit price adjustment value according to the current proportional control value, the current integral control value, and the current derivative control value.

In one embodiment, the adjustment value calculating module 1310C is further configured to obtain a sampling period, an integration time coefficient, and a differential time coefficient, calculate an integration coefficient according to the sampling period, the integration time coefficient, and the proportional control coefficient, calculate a differential coefficient according to the sampling period, the differential time coefficient, and the proportional control coefficient, and calculate a current conversion unit price adjustment value according to the integration coefficient, the differential coefficient, the current proportional control value, the current integral control value, and the current differential control value.

As shown in fig. 17, in one embodiment, the recommended information cost control device further includes:

and the ranking recommendation module 1312 is configured to, when there are multiple recommendation information, obtain, in real time, a target conversion unit price corresponding to a current period of each recommendation information, obtain a predicted click rate and a predicted conversion rate of each recommendation information by a current user, calculate a recommendation value corresponding to the current user according to the conversion unit price, the predicted click rate and the predicted conversion rate, rank the multiple recommendation information according to the recommendation value, and recommend according to a ranking result.

FIG. 18 illustrates an internal block diagram of a computer device in one embodiment. The computer device may be specifically the server 120 of fig. 1. As shown in fig. 18, the computer device includes a processor, a memory, a network interface, an input device, and a display screen connected by a system bus. The memory includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program which, when executed by a processor, causes the processor to implement a recommended information cost control method. The internal memory may also have stored therein a computer program which, when executed by the processor, causes the processor to perform the recommended information cost control method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the structure shown in fig. 18 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the computer device to which the present application is applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, the present application provides that the recommended information cost control apparatus may be implemented in the form of a computer program that is executable on a computer device as shown in fig. 18. The memory of the computer device may store various program modules constituting the recommended information cost control apparatus, such as an exposure cost acquisition module 1302, a conversion number acquisition module 1304, an average conversion cost calculation module 1306, a current error value calculation module 1308, and a target conversion unit price calculation module 1310. The computer program constituted by the respective program modules causes the processor to execute the steps in the recommended information cost control method of the respective embodiments of the present application described in the present specification.

For example, the computer apparatus shown in fig. 18 may perform acquisition of an accumulated exposure cost corresponding to the recommended information for a preset period of time, which is superimposed by each exposure cost corresponding to the recommended information, which is calculated from conversion unit price corresponding to the period, which is dynamically adjusted to change, by the exposure cost acquisition module 1302 in the recommended information cost control device shown in fig. 13; acquiring the conversion number corresponding to the recommended information in the preset time period through a conversion number acquisition module 1304; performing calculation according to the accumulated exposure cost and the conversion number by an average conversion cost calculation module 1306 to obtain an average conversion cost; performing a calculation by a current error value calculation module 1308 based on the average conversion cost and the standard conversion unit price to obtain a current error value; and performing calculation according to the current error value by the target conversion unit price calculation module 1310 to obtain a current conversion unit price adjustment value, and calculating according to the current conversion unit price adjustment value and the standard conversion unit price to obtain the target conversion unit price corresponding to the next period.

In one embodiment, a computer device is presented comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of: acquiring accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change; acquiring the conversion number corresponding to the recommended information in the preset time period; calculating to obtain average conversion cost according to the accumulated exposure cost and the conversion number; calculating to obtain a current error value according to the average conversion cost and the standard conversion unit price; and calculating a current conversion unit price adjustment value according to the current error value, and calculating a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price.

In one embodiment, the step of obtaining the accumulated exposure costs corresponding to the recommended information in the preset time period includes: acquiring a current time, taking the current time as a termination endpoint, determining a corresponding starting endpoint according to a preset time interval, and obtaining a preset time period according to the termination endpoint and the starting endpoint; and acquiring the exposure cost corresponding to each exposure in the preset time period, and superposing the exposure cost to obtain the accumulated exposure cost in the preset time period.

In one embodiment, the step of obtaining the exposure cost corresponding to each exposure in the preset time period includes: obtaining a predicted click rate, a predicted conversion rate and a corresponding conversion unit price corresponding to the current exposure; and calculating the current exposure cost according to the product of the predicted click rate, the predicted conversion rate and the corresponding conversion unit price.

In one embodiment, the step of calculating the current conversion unit price adjustment value according to the current error value includes: acquiring a proportional control coefficient, and calculating to obtain a current proportional control value according to the proportional control coefficient and the current error value; performing integral calculation according to the current error value to obtain a current integral control value; and calculating to obtain a current conversion unit price adjustment value according to the current proportion control value and the current integral control value.

In one embodiment, the step of performing integral calculation according to the current error value to obtain a current integral control value includes: determining a current weight coefficient according to the current moment corresponding to the current error value, and determining a current target error value according to the current error value and the current weight coefficient; and acquiring a historical target error value corresponding to the historical period, and calculating to obtain a current integral control value according to the historical target error value and the current target error value.

In one embodiment, the computer program is executed by a processor, the processor being further configured to perform the steps of: acquiring a first error value corresponding to the previous period; calculating a current differential control value according to the first error value and the current error value; the step of calculating the current conversion unit price adjustment value according to the current proportion control value and the current integral control value comprises the following steps: and calculating to obtain a current conversion unit price adjustment value according to the current proportion control value, the current integral control value and the current differential control value.

In one embodiment, the step of calculating the current conversion unit price adjustment value according to the current proportional control value, the current integral control value and the current derivative control value includes: acquiring a sampling period, an integral time coefficient and a differential time coefficient, and calculating to obtain an integral coefficient according to the sampling period, the integral time coefficient and the proportional control coefficient; calculating to obtain a differential coefficient according to the sampling period, the differential time coefficient and the proportional control coefficient; and calculating according to the integral coefficient, the differential coefficient, the current proportional control value, the current integral control value and the current differential control value to obtain a current conversion unit price adjustment value.

In one embodiment, the computer program is executed by a processor, the processor being further configured to perform the steps of: when a plurality of recommended information are provided, acquiring target conversion unit price corresponding to the current period of each recommended information in real time; the method comprises the steps of obtaining the predicted click rate and the predicted conversion rate of each piece of recommended information of a current user; calculating to obtain a recommended value of each recommended information corresponding to the current user according to the conversion unit price, the predicted click rate and the predicted conversion rate; and sequencing the plurality of recommendation information according to the recommendation value, and recommending according to the sequencing result.

A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of: acquiring accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, and each exposure cost is obtained by calculation according to conversion unit price corresponding to a period of dynamic adjustment change; acquiring the conversion number corresponding to the recommended information in the preset time period; calculating to obtain average conversion cost according to the accumulated exposure cost and the conversion number; calculating to obtain a current error value according to the average conversion cost and the standard conversion unit price; and calculating a current conversion unit price adjustment value according to the current error value, and calculating a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price.

Those skilled in the art will appreciate that all or part of the processes in the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, where the program may be stored in a non-volatile computer readable storage medium, and where the program, when executed, may include processes in the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.

Claims

1. An information recommendation method, performed by a server, comprising:

the method comprises the steps of obtaining accumulated exposure cost corresponding to recommended information in a preset time period, wherein the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, each exposure cost is obtained by calculation according to a predicted conversion rate corresponding to each exposure and conversion unit price corresponding to a period, which is dynamically adjusted and changed, and the predicted conversion rate corresponding to each exposure is obtained by prediction according to user characteristics of a pushing user corresponding to each exposure; the conversion unit price refers to conversion cost corresponding to one time of conversion, and the conversion refers to specific behavior of a user on recommended information after the recommended information is pushed;

acquiring a proportional control coefficient, and calculating to obtain a current proportional control value according to the proportional control coefficient and the current error value;

determining a current weight coefficient according to the current moment corresponding to the current error value, and determining a current target error value according to the current error value and the current weight coefficient;

acquiring a historical target error value corresponding to a historical period, and calculating to obtain a current integral control value according to the historical target error value and the current target error value; the historical target error value is obtained according to the historical error value and a corresponding historical weight coefficient, the current weight coefficient and the historical weight coefficient are different weight coefficients, and the weight coefficients decay with time;

calculating a current conversion unit price adjustment value according to the current proportion control value and the current integral control value, wherein the method comprises the following steps: calculating according to the current proportion control value and the current integral control value to obtain an initial conversion unit price adjustment value; when the initial conversion unit price adjustment value is larger than a preset first threshold value, the preset first threshold value is used as a current conversion unit price adjustment value; when the initial conversion unit price adjustment value is smaller than a preset second threshold value, the preset second threshold value is used as a current conversion unit price adjustment value; the preset first threshold value is larger than the preset second threshold value; when the initial conversion unit price adjustment value is smaller than or equal to a preset first threshold value and larger than or equal to a preset second threshold value, the initial conversion unit price adjustment value is used as a current conversion unit price adjustment value;

Calculating according to the current conversion unit price adjustment value and the standard conversion unit price to obtain a target conversion unit price corresponding to the next period; the target conversion unit price is used for reducing the current error value obtained by calculation of the next period, the target conversion unit price is also used for calculating the recommended value of the recommended information corresponding to the current user, the recommended value is obtained by calculation according to the predicted click rate and the predicted conversion rate of the recommended information by the current user and the target conversion unit price, and the server is used for recommending the recommended information with the highest recommended value to the terminal corresponding to the current user.

2. The method according to claim 1, wherein the step of obtaining the cumulative exposure cost corresponding to the recommended information in the preset time period includes:

acquiring a current time, taking the current time as a termination endpoint, determining a corresponding starting endpoint according to a preset time interval, and obtaining a preset time period according to the termination endpoint and the starting endpoint;

and acquiring the exposure cost corresponding to each exposure in the preset time period, and superposing the exposure cost to obtain the accumulated exposure cost in the preset time period.

3. The method according to claim 2, wherein the step of obtaining the exposure cost corresponding to each exposure in the preset time period includes:

Obtaining a predicted click rate, a predicted conversion rate and a corresponding conversion unit price corresponding to the current exposure;

and calculating the current exposure cost according to the product of the predicted click rate, the predicted conversion rate and the corresponding conversion unit price.

4. The method of claim 3, wherein the predicted click rate is predicted from the user characteristics of the pushing user corresponding to the current exposure by a click prediction model, and the predicted conversion rate is predicted from the user characteristics of the pushing user corresponding to the current exposure by a conversion prediction model.

5. The method of claim 1, wherein the step of calculating a current conversion unit price adjustment value from the current proportional control value and the current integral control value further comprises:

acquiring a first error value corresponding to the previous period;

calculating a current differential control value according to the first error value and the current error value;

and calculating to obtain a current conversion unit price adjustment value according to the current proportion control value, the current integral control value and the current differential control value.

6. The method of claim 5, wherein the step of calculating a current conversion unit price adjustment value based on the current proportional control value, the current integral control value, and the current derivative control value comprises:

Acquiring a sampling period, an integral time coefficient and a differential time coefficient, and calculating to obtain an integral coefficient according to the sampling period, the integral time coefficient and the proportional control coefficient;

calculating to obtain a differential coefficient according to the sampling period, the differential time coefficient and the proportional control coefficient;

and calculating according to the integral coefficient, the differential coefficient, the current proportional control value, the current integral control value and the current differential control value to obtain a current conversion unit price adjustment value.

7. The method of claim 6, wherein the integration coefficient is positively correlated with the sampling period, the proportional control coefficient, and wherein the integration coefficient is negatively correlated with the integration time coefficient; the differential coefficient is positively correlated with the differential time coefficient and the proportional control coefficient, and the differential coefficient is negatively correlated with the sampling period.

8. The method according to claim 1, wherein the method further comprises:

when a plurality of recommended information are provided, acquiring target conversion unit price corresponding to the current period of each recommended information in real time;

the method comprises the steps of obtaining the predicted click rate and the predicted conversion rate of each piece of recommended information of a current user;

Calculating to obtain a recommended value of each recommended information corresponding to the current user according to the target conversion unit price, the predicted click rate and the predicted conversion rate;

and sequencing the plurality of recommendation information according to the recommendation value, and recommending according to the sequencing result.

9. An information recommendation apparatus, the apparatus comprising:

the system comprises an exposure cost acquisition module, a storage module and a storage module, wherein the exposure cost acquisition module is used for acquiring accumulated exposure cost corresponding to recommended information in a preset time period, the accumulated exposure cost is obtained by superposition of each exposure cost corresponding to the recommended information, each exposure cost is obtained by calculation according to a predicted conversion rate corresponding to each exposure and conversion unit price corresponding to a period of dynamic adjustment change, and the predicted conversion rate corresponding to each exposure is obtained by prediction according to user characteristics of a pushing user corresponding to each exposure; the conversion unit price refers to conversion cost corresponding to one time of conversion, and the conversion refers to specific behavior of a user on recommended information after the recommended information is pushed;

the target conversion unit price calculation module is used for acquiring a proportion control coefficient and calculating a current proportion control value according to the proportion control coefficient and the current error value; determining a current weight coefficient according to the current moment corresponding to the current error value, and determining a current target error value according to the current error value and the current weight coefficient; acquiring a historical target error value corresponding to a historical period, and calculating to obtain a current integral control value according to the historical target error value and the current target error value; the historical target error value is obtained according to the historical error value and a corresponding historical weight coefficient, the current weight coefficient and the historical weight coefficient are different weight coefficients, and the weight coefficients decay with time; calculating a current conversion unit price adjustment value according to the current proportion control value and the current integral control value, and calculating a target conversion unit price corresponding to the next period according to the current conversion unit price adjustment value and the standard conversion unit price; the target conversion unit price is used for reducing the current error value obtained by calculation of the next period, the target conversion unit price is also used for calculating a recommended value of the recommended information corresponding to the current user, the recommended value is obtained by calculation according to the predicted click rate and the predicted conversion rate of the recommended information by the current user and the target conversion unit price, and the server is used for recommending the recommended information with the highest recommended value to the terminal corresponding to the current user;

Calculating a current conversion unit price adjustment value according to the current proportion control value and the current integral control value, wherein the method comprises the following steps: calculating according to the current proportion control value and the current integral control value to obtain an initial conversion unit price adjustment value; when the initial conversion unit price adjustment value is larger than a preset first threshold value, the preset first threshold value is used as a current conversion unit price adjustment value; when the initial conversion unit price adjustment value is smaller than a preset second threshold value, the preset second threshold value is used as a current conversion unit price adjustment value; the preset first threshold value is larger than the preset second threshold value; and when the initial conversion unit price adjustment value is smaller than or equal to a preset first threshold value and larger than or equal to a preset second threshold value, the initial conversion unit price adjustment value is used as a current conversion unit price adjustment value.

10. The apparatus of claim 9, wherein the exposure cost acquisition module comprises:

the preset time period determining module is used for obtaining the current time, taking the current time as a termination endpoint, determining a corresponding starting endpoint according to a preset time interval, and obtaining a preset time period according to the termination endpoint and the starting endpoint;

the exposure cost superposition module is used for acquiring the exposure cost corresponding to each exposure in the preset time period, and superposing the exposure cost to obtain the accumulated exposure cost in the preset time period.

11. The apparatus according to claim 10, wherein the exposure cost superimposing module is further configured to obtain a predicted click rate, a predicted conversion rate, and a corresponding conversion unit price corresponding to the current exposure, and calculate the current exposure cost according to a product of the predicted click rate, the predicted conversion rate, and the corresponding conversion unit price.

12. The apparatus of claim 11, wherein the predicted click rate is predicted from a user characteristic of the push user corresponding to the current exposure by a click prediction model, and wherein the predicted conversion rate is predicted from a user characteristic of the push user corresponding to the current exposure by a conversion prediction model.

13. The apparatus of claim 9, wherein the target conversion unit price calculation module is further configured to obtain a first error value corresponding to a previous cycle; calculating a current differential control value according to the first error value and the current error value; and calculating to obtain a current conversion unit price adjustment value according to the current proportion control value, the current integral control value and the current differential control value.

14. The apparatus of claim 13, wherein the target conversion unit price calculation module is further configured to obtain a sampling period, an integration time coefficient, and a differential time coefficient, and calculate an integration coefficient according to the sampling period, the integration time coefficient, and the proportional control coefficient; calculating to obtain a differential coefficient according to the sampling period, the differential time coefficient and the proportional control coefficient; and calculating according to the integral coefficient, the differential coefficient, the current proportional control value, the current integral control value and the current differential control value to obtain a current conversion unit price adjustment value.

15. The apparatus of claim 14, wherein the integration coefficient is positively correlated with the sampling period, the proportional control coefficient, and wherein the integration coefficient is negatively correlated with the integration time coefficient; the differential coefficient is positively correlated with the differential time coefficient and the proportional control coefficient, and the differential coefficient is negatively correlated with the sampling period.

16. The apparatus of claim 9, wherein the apparatus further comprises:

the sequencing recommendation module is used for acquiring the target conversion unit price corresponding to the current period of each recommendation information in real time when a plurality of recommendation information are provided; the method comprises the steps of obtaining the predicted click rate and the predicted conversion rate of each piece of recommended information of a current user; calculating to obtain a recommended value of each recommended information corresponding to the current user according to the target conversion unit price, the predicted click rate and the predicted conversion rate; and sequencing the plurality of recommendation information according to the recommendation value, and recommending according to the sequencing result.

17. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1 to 8.

18. A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method of any one of claims 1 to 8.