Method and device for determining merchant advertiser conversion rate

By combining industrial and commercial registration information, main category information and same-to-month information, Bayesian smoothing processing and neural network model to calculate the advertiser conversion rate of merchants, the problem of low accuracy of advertisers in the existing technology is solved, and the realization of the realization of advertisers' potential tapping is achieved. More accurate advertising mining.

CN119090562APending Publication Date: 2024-12-06BEIJING JINGDONG QIANSHITECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310666242.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-06-06
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing advertiser's potential tapping method only considers the merchant's order quantity, ignores the merchant's attributes and potential rules, resulting in a low accuracy rate of advertisers' potential tapping.

Method used

By obtaining the merchant's industrial and commercial registration information, main category information, same-to-month information and the advertiser mining rate of target miners, Bayesian smoothing processing and neural network model are used to calculate the advertiser conversion rate of merchants.

Benefits of technology

It improves the accuracy of advertisers' potential tapping, enables explorers to accurately visit high-potential merchants, and improves the calculation accuracy of advertisers' conversion rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090562A_ABST
    Figure CN119090562A_ABST
Patent Text Reader

Abstract

The invention discloses a method and device for determining a merchant advertiser conversion rate, and relates to the technical field of big data intelligent processing. A specific implementation mode of the method comprises the steps of obtaining dimension data of a target merchant, wherein the dimension data comprises industrial and commercial registration information, camp category information, same-link ratio information and an advertiser mining rate of a target mining person corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated by adopting Bayesian smoothing processing; processing the dimension data of the target merchant into dimension features; inputting the dimension features of the target commercial tenant into an advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target commercial tenant; wherein the advertiser conversion rate model is obtained by training a neural network model. According to the embodiment, the technical problem of low potential tapping accuracy of advertisers can be solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of big data intelligent processing, and in particular to a method and device for determining a merchant advertiser conversion rate. Background Art

[0002] On the one hand, offline advertising can expand merchants' advertising channels, increase merchant product exposure, and further increase merchant revenue. On the other hand, for enterprises, facing a large number of existing cooperative merchants, if they develop existing advertisers, they can increase advertising revenue for enterprises, thus achieving a win-win situation.

[0003] Currently, the way to tap advertisers' potential is to directly tap based on the daily average order volume, monthly average order volume and revenue ranking, and tap the advertiser potential of merchants with relatively large order volume and revenue, and recommend them to use offline advertising products. However, this mining method only considers the merchant's own order volume, ignoring the merchant's attributes and the potential rules of order volume, resulting in a low accuracy rate in tapping advertisers' potential. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and device for determining a merchant advertiser conversion rate, so as to solve the technical problem of low accuracy in potential tapping of advertisers.

[0005] To achieve the above objective, according to one aspect of an embodiment of the present invention, a method for determining a merchant advertiser conversion rate is provided, comprising:

[0006] Acquire dimensional data of the target merchant, the dimensional data including business registration information, main category information, year-on-year and month-on-month information, and the advertiser mining rate of the target mining personnel corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing;

[0007] Processing the dimensional data of the target merchant into dimensional features;

[0008] The dimensional features of the target merchant are input into an advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is obtained by training with a neural network model.

[0009] Optionally, the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing, including:

[0010] Calculate the advertiser signing rate of each mining personnel in the past period of time, and then calculate the mean and variance of the advertiser signing rate;

[0011] Calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate;

[0012] The advertiser mining rate of the target mining personnel is recalculated according to the Bayesian smoothing coefficient.

[0013] Optionally, the advertiser signing rate of each mining personnel in the past period of time is calculated, so as to calculate the advertiser signing rate mean and the advertiser signing rate variance, including:

[0014] For each mining personnel, the number of advertiser contracts signed by the mining personnel in the past period of time is divided by the number of merchants visited, thereby obtaining the advertiser contract signing rate of the mining personnel in the past period of time;

[0015] Calculate the average of the advertiser signing rates according to the advertiser signing rates of each mining personnel in the past period of time;

[0016] The advertiser signing rate variance is calculated based on the advertiser signing rate of each mining personnel in the past period of time and the average of the advertiser signing rate.

[0017] Optionally, calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate includes:

[0018] The first Bayesian smoothing coefficient and the second Bayesian smoothing coefficient are calculated using the following formulas:

[0019] α=mean*((mean*(1-mean)) / (var*var)-1)

[0020] β=(1-mean)*((mean*(1-mean)) / (var*var)-1)

[0021] Among them, mean is the mean of advertiser signing rate, var is the variance of advertiser signing rate, α is the first Bayesian smoothing coefficient, and β is the second Bayesian smoothing coefficient.

[0022] Optionally, recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient includes:

[0023] The advertiser mining rate of the target mining personnel is calculated using the following formula:

[0024] M=(S+α) / (P+α+β)

[0025] Among them, S is the number of advertisers signed, and P is the number of merchants visited.

[0026] Optionally, processing the dimension data of the target merchant into dimension features includes:

[0027] Processing the dimensional data of the target merchant into dimensional features by using a multi-hot encoding method;

[0028] The pipeline mechanism is used to serialize the dimensional features of each dimension and aggregate them into a total dimensional feature.

[0029] Optionally, before obtaining the dimension data of the target merchant, the following steps are also included:

[0030] Extracting the dimension data of each positive sample merchant and the dimension data of each negative sample merchant according to a preset ratio; wherein the same period last year information includes daily period last year information, weekly period last year information, and monthly period last year information;

[0031] Processing the dimensional data of each of the positive sample merchants and the dimensional data of each of the negative sample merchants into dimensional features;

[0032] The dimensional features of each positive sample merchant and the dimensional features of each negative sample merchant are input into a neural network model for supervised training, so as to obtain an advertiser conversion rate model through training.

[0033] In addition, according to another aspect of an embodiment of the present invention, a device for determining a merchant advertiser conversion rate is provided, comprising:

[0034] An acquisition module is used to acquire dimension data of a target merchant, wherein the dimension data includes business registration information, main business category information, year-on-year and month-on-month information, and an advertiser mining rate of a target mining person corresponding to the target merchant; wherein the advertiser mining rate of the target mining person is calculated by Bayesian smoothing processing;

[0035] A processing module, used for processing the dimension data of the target merchant into dimension features;

[0036] A calculation module is used to input the dimensional features of the target merchant into an advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is obtained by training a neural network model.

[0037] Optionally, the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing, including:

[0038] Calculate the advertiser signing rate of each mining personnel in the past period of time, and then calculate the mean and variance of the advertiser signing rate;

[0039] Calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate;

[0040] The advertiser mining rate of the target mining personnel is recalculated according to the Bayesian smoothing coefficient.

[0041] Optionally, the advertiser signing rate of each mining personnel in the past period of time is calculated, so as to calculate the advertiser signing rate mean and the advertiser signing rate variance, including:

[0042] For each mining personnel, the number of advertiser contracts signed by the mining personnel in the past period of time is divided by the number of merchants visited, thereby obtaining the advertiser contract signing rate of the mining personnel in the past period of time;

[0043] Calculate the average of the advertiser signing rates according to the advertiser signing rates of each mining personnel in the past period of time;

[0044] The advertiser signing rate variance is calculated based on the advertiser signing rate of each mining personnel in the past period of time and the average of the advertiser signing rate.

[0045] Optionally, calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate includes:

[0046] The first Bayesian smoothing coefficient and the second Bayesian smoothing coefficient are calculated using the following formulas:

[0047] α=mean*((mean*(1-mean)) / (var*var)-1)

[0048] β=(1-mean)*((mean*(1-mean)) / (var*var)-1)

[0049] Among them, mean is the mean of advertiser signing rate, var is the variance of advertiser signing rate, α is the first Bayesian smoothing coefficient, and β is the second Bayesian smoothing coefficient.

[0050] Optionally, recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient includes:

[0051] The advertiser mining rate of the target mining personnel is calculated using the following formula:

[0052] M=(S+α) / (P+α+β)

[0053] Among them, S is the number of advertisers signed, and P is the number of merchants visited.

[0054] Optionally, the processing module is further used for:

[0055] Processing the dimensional data of the target merchant into dimensional features by using a multi-hot encoding method;

[0056] The pipeline mechanism is used to serialize the dimensional features of each dimension and aggregate them into a total dimensional feature.

[0057] Optionally, a training module is also included, for:

[0058] Extracting the dimension data of each positive sample merchant and the dimension data of each negative sample merchant according to a preset ratio; wherein the same period last year information includes daily period last year information, weekly period last year information, and monthly period last year information;

[0059] Processing the dimensional data of each of the positive sample merchants and the dimensional data of each of the negative sample merchants into dimensional features;

[0060] The dimensional features of each positive sample merchant and the dimensional features of each negative sample merchant are input into a neural network model for supervised training, so as to obtain an advertiser conversion rate model through training.

[0061] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, including:

[0062] one or more processors;

[0063] a storage device for storing one or more programs,

[0064] When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the above embodiments.

[0065] According to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0066] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0067] One embodiment of the above invention has the following advantages or beneficial effects: because the industrial and commercial registration information, main category information, year-on-year information and the advertiser mining rate of the target mining personnel corresponding to the target merchant are all input as feature data into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant, and the technical means of Bayesian smoothing of the advertiser mining rate are adopted, so the technical problem of low accuracy of advertiser potential mining in the prior art is overcome. The embodiment of the present invention uses the industrial and commercial registration information, main category information, year-on-year information and the advertiser mining rate of the target mining personnel corresponding to the target merchant as feature data and adopts Bayesian smoothing to process the advertiser mining rate, and accurately calculates the probability of each merchant becoming an advertiser through the advertiser conversion rate model, thereby providing a judgment basis for mining personnel to mine high-potential advertisers, so that mining personnel can visit accurately.

[0068] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0070] Figure 1 is a flow chart of a method for determining a merchant advertiser conversion rate according to an embodiment of the present invention;

[0071] Figure 2 is a flow chart of a method for determining a merchant advertiser conversion rate according to a reference embodiment of the present invention;

[0072] Figure 3 is a schematic diagram of an apparatus for determining a merchant advertiser conversion rate according to an embodiment of the present invention;

[0073] Figure 4 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0074] Figure 5 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0075] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0076] It should be noted that in the technical solution of this application, the acquisition, storage, use, and processing of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0077] Figure 1 is a flow chart of a method for determining a merchant advertiser conversion rate according to an embodiment of the present invention. As an embodiment of the present invention, Figure 1 As shown, the method for determining the merchant advertiser conversion rate may include:

[0078] Step 101, obtaining dimensional data of the target merchant, the dimensional data including industrial and commercial registration information, main category information, year-on-year and month-on-month information and the advertiser mining rate of the target mining personnel corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated using Bayesian smoothing processing.

[0079] First, obtain the dimensional data of the target merchant, which may include industrial and commercial registration information, main business category information, year-on-year information, and the advertiser mining rate of the target mining personnel corresponding to the target merchant; among which, the industrial and commercial registration information mainly includes the legal representative's appointment information, business status, establishment date, registered capital, paid-in capital, business term, taxpayer qualifications, enterprise type, industry, staff size, number of insured persons, business scope, etc. It should be pointed out that the merchant's industrial and commercial registration information can be obtained by calling the industrial and commercial registration information interface. The main category information may include primary categories, secondary categories, tertiary categories, etc.

[0080] After obtaining the business registration information and main category information of the merchant, it is necessary to perform Chinese word segmentation on the business registration information and main category information. Take the main category information as an example. For example, the main category information of a merchant is: household daily necessities / home decoration building materials / household cleaning / paper products / kitchenware. After Chinese word segmentation, it becomes [furniture daily necessities, home decoration building materials, household cleaning, paper products, kitchenware]. The features after Chinese word segmentation show the merchant's main category information in a more detailed and granular manner.

[0081] Optionally, the year-on-year information includes daily year-on-year information, weekly year-on-year information, and monthly year-on-year information. This part of the features includes logistics-side derived features, and the main derived features include single quantity year-on-year features and package year-on-year information features. For example, the single quantity information can be traced back to the recent daily year-on-year information, the recent weekly year-on-year information, and the recent monthly year-on-year information, which helps to fully understand the dimensions of recent single quantity information, thereby playing a positive role in mining advertiser information. The following is a supplement to the year-on-year information with single quantity and package quantity as features for derivation.

[0082] Among them, the order volume and package volume include the recent and long-term total volume characteristics, and the year-on-year information in different time dimensions. The time dimension of the year-on-year information is mainly based on the daily granularity, weekly granularity, and monthly granularity. The detailed calculation method is as follows:

[0083] Based on the daily granularity, the calculation method for the current day's daily order volume compared to yesterday is as follows

[0084]

[0085] Based on the weekly granularity, if the current day is the kth day of this week, the average daily order volume statistics for this week up to today are as follows:

[0086]

[0087] Among them, n 01 to n 0k They are the order volumes from the 1st day to the kth day of this week.

[0088] The average daily order volume statistics for the previous week as of the same day are as follows:

[0089]

[0090] Among them, n 11 to n 1k They are the order volumes from the 1st day to the kth day of last week.

[0091] The calculation method for the average daily order volume of this week ending today compared with the average daily order volume of last week ending the same day is:

[0092]

[0093] Based on the average daily order volume, a month-on-month comparison of the average daily orders from the 1st to the 4th week is established as of this week, based on the month-on-month comparison of the total order volume of this week compared with the total order volume of previous weeks, and the year-on-year comparison of the average daily order volume from this week to that day compared with the same week last year.

[0094] Based on the monthly granularity, the calculation method is similar to the weekly granularity. It includes the month-on-month characteristics calculated as of a certain day of a certain month compared with the same day of the previous month, the month-on-month comparison of the total volume of a certain month relative to the total volume of the previous month, the year-on-year comparison of the daily average volume as of a certain day of this month relative to the same day of the same month last year, and the year-on-year comparison of the daily order volume of the current day relative to the same day of the previous month. Among them, the year-on-year calculation method for the daily order volume of the current day relative to the same day of the previous month is

[0095]

[0096] Therefore, the year-on-year information can be selected from at least one of the following: the average year-on-year increase in the number of orders from this week to the same day relative to the same day last week, the average year-on-year increase in the number of packages from this week to the same day relative to the same day last week, the average year-on-year increase in the number of orders from this week to the same day relative to the same day two weeks ago, the average year-on-year increase in the number of packages from this week to the same day relative to the same day two weeks ago, the average year-on-year increase in the number of orders from this week to the same day relative to the same day three weeks ago, the average year-on-year increase in the number of packages from this week to the same day relative to the same day three weeks ago, The average daily order volume as of this week compared to the same day four weeks ago, the average daily package volume as of this week compared to the same day four weeks ago, the average weekly order volume last week compared to the weekly order volume three weeks ago, the average weekly order volume last week compared to the weekly package volume three weeks ago, the average weekly order volume last week compared to the weekly order volume four weeks ago, the average weekly order volume last week compared to the weekly package volume four weeks ago, the average weekly order volume last week compared to the weekly order volume five weeks ago, the average weekly order volume last week compared to the weekly package volume five weeks ago, the average weekly order volume three weeks ago compared to the weekly order volume four weeks ago Week-on-week comparison of order volume, week-on-week comparison of order volume three weeks ago relative to package volume four weeks ago, week-on-week comparison of order volume four weeks ago relative to order volume five weeks ago, week-on-week comparison of order volume four weeks ago relative to package volume five weeks ago, average daily month-on-month comparison of order volume as of today relative to that of last month, average daily month-on-month comparison of package volume as of today relative to that of last month, average daily month-on-month comparison of order volume as of today relative to that of two months ago, average daily month-on-month comparison of package volume as of today relative to that of two months ago, average daily month-on-month comparison of order volume as of today relative to that of two months ago, average daily month-on-month comparison of order volume as of today relative to that of two months ago The average daily month-on-month change in order volume from three months ago to the same day, the average daily month-on-month change in package volume from this month to the same day relative to the same day three months ago, the average daily month-on-month change in order volume from this month to the same day relative to the same month last year, the average daily month-on-month change in package volume from this month to the same day relative to the same month last year, the month-on-month change in order volume from last month to the month two months ago, the month-on-month change in order volume from last month to the month two months ago, the month-on-month change in order volume from two months ago to the month three months ago, and the month-on-month change in order volume from last month to the month two months ago.

[0097] Optionally, the dimensional data may also include long-term order quantity information (such as today's order quantity, today's package quantity, yesterday's order quantity, yesterday's package quantity, this week's order quantity, this week's package quantity, last week's order quantity, last week's package quantity, etc.) and recent order quantity information (such as last month's order quantity, last month's package quantity, order quantity two weeks ago, package quantity two weeks ago, order quantity three weeks ago, package quantity three weeks ago, order quantity two months ago, package quantity two months ago, etc.).

[0098] In the embodiment of the present invention, each merchant has its corresponding mining personnel, and the two correspond one to one. The mining personnel mine the merchant to make it an advertiser. The embodiment of the present invention also uses the mining personnel's advertiser mining rate as one of the merchant's characteristics. This feature indicates the mining personnel's advertiser mining ability, which helps to improve the calculation accuracy of the advertiser conversion rate.

[0099] The advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing. If the advertiser mining rate of the mining personnel is directly used as a feature, then if all the merchants visited by the mining personnel with a small number of merchants are converted, the confidence of the model will be reduced. Therefore, the embodiment of the present invention uses Bayesian smoothing to process the advertiser mining rate of the mining personnel, thereby improving the calculation accuracy of the advertiser conversion rate.

[0100] Optionally, the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing, including: calculating the advertiser signing rate of each mining personnel in the past period of time, thereby calculating the mean of the advertiser signing rate and the variance of the advertiser signing rate; calculating the Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate; recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient. First, the advertiser signing rate of each mining personnel in the past period of time is calculated by the following formula: advertiser signing rate = number of advertiser signings / number of visited merchants; then, the mean of the advertiser signing rate and the variance of the advertiser signing rate are calculated according to the advertiser signing rate of each mining personnel in the past period of time, and then the Bayesian smoothing coefficient is calculated according to the mean of the advertiser signing rate and the variance of the advertiser signing rate, and finally, the advertiser mining rate of the target mining personnel is recalculated according to the Bayesian smoothing coefficient, thereby realizing the Bayesian smoothing of the advertiser mining rate of the target mining personnel.

[0101] Optionally, the advertiser signing rate of each mining personnel in the past period of time is calculated, so as to calculate the mean advertiser signing rate and the variance of the advertiser signing rate, including: for each mining personnel, dividing the number of advertiser signings of the mining personnel in the past period of time by the number of merchants visited, so as to obtain the advertiser signing rate of the mining personnel in the past period of time; calculating the mean advertiser signing rate based on the advertiser signing rate of each mining personnel in the past period of time; calculating the variance of the advertiser signing rate based on the advertiser signing rate of each mining personnel in the past period of time and the mean advertiser signing rate.

[0102] Optionally, calculating the Bayesian smoothing coefficient according to the advertiser signing rate mean and the advertiser signing rate variance includes: using the following formulas to respectively calculate a first Bayesian smoothing coefficient and a second Bayesian smoothing coefficient:

[0103] α=mean*((mean*(1-mean)) / (var*var)-1)

[0104] β=(1-mean)*((mean*(1-mean)) / (var*var)-1)

[0105] Among them, mean is the mean of advertiser signing rate, var is the variance of advertiser signing rate, α is the first Bayesian smoothing coefficient, and β is the second Bayesian smoothing coefficient.

[0106] Optionally, recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient includes: calculating the advertiser mining rate of the target mining personnel using the following formula:

[0107] M=(S+α) / (P+α+β)

[0108] Among them, S is the number of advertisers signed, and P is the number of merchants visited.

[0109] For example, if 100 merchants are assigned to mining personnel A and 20 merchants are signed, then the advertiser signing rate of mining personnel A is 0.2; if 10 merchants are assigned to mining personnel B and 3 merchants are signed, then the advertiser signing rate of mining personnel B is 0.3; Bayesian smoothing includes the following processes:

[0110] 1) Calculate the mean of advertiser signing rate: mean = (0.2 + 0.3) / 2 = 0.25

[0111] 2) Calculate the variance of advertiser signing rate:

[0112] var=((0.2-0.25)^2+(0.3-0.25)^2) / 2=0.0025

[0113] 3) Calculate Bayesian smoothing coefficients α and β

[0114] α=mean*((mean*(1-mean)) / (var*var)-1)

[0115] =0.25*((0.25*0.75) / (0.0025*0.0025)-1)=7499.75

[0116] β=(1-mean)*((mean*(1-mean)) / (var*var)-1)=22499.25

[0117] 4) Recalculate the advertiser signing rate using the Bayesian smoothing coefficient:

[0118] A Excavator:

[0119] (Number of advertisers signed + α) / (Number of merchants visited + α + β) =

[0120] (20+7499.75) / (100+7499.75+22499.25)=0.2498

[0121] B Excavation personnel:

[0122] (Number of advertisers signed + α) / (Number of merchants visited + α + β) =

[0123] (3+7499.75) / (10+7499.75+22499.25)=0.25

[0124] It can be seen that from the recalculated advertiser signing, B has decreased compared with before, and A has increased relatively compared with before. Therefore, the gap between AB is not as large as at the beginning. When the features are weighted, there is more confidence, which helps to improve the calculation accuracy of the advertiser's conversion rate.

[0125] Step 102: Process the dimensional data of the target merchant into dimensional features.

[0126] In this step, the dimensional data of the target merchant is processed into dimensional features through feature engineering. Optionally, step 102 may include: using multi-hot encoding to process the dimensional data of the target merchant into dimensional features; using a pipeline mechanism to serialize the dimensional features of each dimension and aggregate them into a total dimensional feature. Compared with the one-hot encoding method, the multi-hot encoding method can describe the multi-faceted dimensional data of the merchant in more detail. The dimensional features obtained by the multi-hot encoding method have values ​​in multiple positions, such as [0,1,0,0,0,1], and then the pipeline mechanism is used to serialize all the feature engineering processing mechanisms, and finally aggregate them into a total dimensional feature.

[0127] Step 103, input the dimensional features of the target merchant into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is obtained by training with a neural network model.

[0128] After feature processing is completed, the dimensional features of the target merchant are input into the advertiser conversion rate model. The pre-trained advertiser conversion rate model will output the advertiser conversion rate of the target merchant.

[0129] The dimensional characteristics of each target merchant can be input into the advertiser conversion rate model, which outputs the advertiser conversion rate of each target merchant. The advertiser conversion rates of these target merchants are then ranked, and a list of merchants with relatively high advertiser conversion rates is issued according to the visiting capabilities of each region. Regional mining personnel visit high-potential merchants, which effectively saves manpower and improves the accuracy of merchant conversion.

[0130] The mining personnel will provide feedback on the visit records, including merchants that have converted, merchants that the model predicts will convert but did not convert, and their reasons. Combined with the reasons for the visits, this will provide other dimensional support for the next optimization of the model, continuously improve the model's prediction accuracy, accurately assess whether merchants have the potential to become advertisers, and enable mining personnel to make accurate visits.

[0131] According to the various embodiments described above, it can be seen that the embodiment of the present invention uses the industrial and commercial registration information, main category information, year-on-year information, and the advertiser mining rate of the target mining personnel corresponding to the target merchant as feature data input into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant, and adopts the technical means of Bayesian smoothing to process the advertiser mining rate, thereby solving the technical problem of low accuracy of advertiser potential mining in the prior art. The embodiment of the present invention uses the industrial and commercial registration information, main category information, year-on-year information, and the advertiser mining rate of the target mining personnel corresponding to the target merchant as feature data and adopts Bayesian smoothing to process the advertiser mining rate, and accurately calculates the probability of each merchant becoming an advertiser through the advertiser conversion rate model, thereby providing a judgment basis for mining personnel to mine high-potential advertisers, allowing mining personnel to make precise visits.

[0132] Figure 2 is a flow chart of a method for determining a merchant advertiser conversion rate according to a reference embodiment of the present invention. As another embodiment of the present invention, Figure 2 As shown, the method for determining the merchant advertiser conversion rate may include:

[0133] Step 201 , extracting the dimensional data of each positive sample merchant and the dimensional data of each negative sample merchant according to a preset ratio.

[0134] The dimensional data may include business registration information, main business category information, year-on-year and month-on-month information, and the advertiser mining rate of the mining personnel corresponding to the merchant; wherein the business registration information mainly includes the legal representative's appointment information, business status, establishment date, registered capital, paid-in capital, business term, taxpayer qualifications, enterprise type, industry, staff size, number of insured persons, business scope, etc. The main business category information may include primary categories, secondary categories, tertiary categories, etc., and the advertiser mining rate of the mining personnel is calculated using Bayesian smoothing.

[0135] The steps of obtaining the dimensional data of each positive sample merchant and the dimensional data of each negative sample merchant are similar to step 101 and will not be repeated herein.

[0136] For example, advertiser merchants from July 2021 to July 2022 are obtained and marked as positive samples; for merchants who have not become advertisers, they are sampled according to a certain proportion, and the merchants sampled in proportion are marked as negative samples, so that the positive and negative samples are within a reasonable distribution.

[0137] Step 202: Process the dimensional data of each positive sample merchant and the dimensional data of each negative sample merchant into dimensional features.

[0138] Specifically, the dimensional data of each positive sample merchant and the dimensional data of each negative sample merchant are processed into dimensional features using a multi-hot encoding method, and then the dimensional features of each dimension are serialized using a pipeline mechanism, and the dimensional features of each sample merchant are summarized into a total dimensional feature.

[0139] Step 203: Input the dimensional features of each positive sample merchant and the dimensional features of each negative sample merchant into a neural network model for supervised training, thereby training to obtain an advertiser conversion rate model.

[0140] After the pipeline mechanism serializes the features, it is combined with the labeled positive and negative examples and input into the neural network model, such as DeepFM, Wide&Deep, Xgboost, GBDT, LightGBM and other neural networks. After supervised training, the advertiser conversion rate model is trained. Taking DeepFM as an example, the model can be abstracted into a three-layer network: input layer, hidden layer and output layer; among them, the input layer is mainly divided into the Deep end and the FM end. The Deep end mainly performs automatic crossover and mining of features of order 3 and above, and the FM end performs low-order feature crossover, thereby realizing multi-dimensional detection and intrinsic mining of merchant data; the hidden layer can realize data sharing after feature embedding, so as to input it into the lower output layer for algorithm training. The output layer is a single node for binary classification prediction. DeepFM uses nonlinear activation and feature crossover. The nonlinear activation function mainly uses the sigmoid function, which can transform the original linear features into nonlinear features after the activation function, realize the nonlinear change of features, and observe the dimensional relationship of feature hidden from multiple perspectives.

[0141] The input data for model training is divided into in-set data and out-of-set data according to the time dimension. The in-set data is divided into training set and test set according to 7:3. The training set is used to train the model, and the test set is used to evaluate the model effect. The training set and test set of the in-set data belong to the same data distribution, and there is a certain correlation between the features, which is used to judge the prediction effect of the model. The out-of-set data and the in-set data are completely isolated, and the real prediction effect can be predicted and evaluated.

[0142] You can also further tune the model parameters, such as the maximum number of iterations, the choice of linear activation function, the choice of convergence method, the setting of convergence threshold, and other parameter tuning methods.

[0143] The model evaluation indicators mainly include: the model's AUC indicator and the precision, recall and F1 value of positive and negative samples. Taking the positive sample as an example, the main evaluation methods include the following:

[0144] Precision = data predicted as positive samples and actually positive samples / all data predicted as positive samples.

[0145] Recall rate = data predicted to be positive samples and actually positive samples / all positive sample data.

[0146] F1 value = 2*precision*recall / (recall+precision).

[0147] The calculation method of negative sample and positive sample evaluation indicators is the same, except that the statistical caliber is changed from positive samples to negative samples.

[0148] Step 204, obtaining dimension data of the target merchant.

[0149] Step 205: Process the dimensional data of the target merchant into dimensional features.

[0150] Step 206: Input the dimensional features of the target merchant into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant.

[0151] Step 204 to step 206 are similar to step 101 to step 103 and will not be described in detail.

[0152] The embodiment of the present invention inputs the dimensional features of merchants into the advertiser conversion rate model. The advertiser conversion rate model outputs the probability value of the merchant being converted into an advertiser. Mining personnel can give priority to visiting and mining merchants with relatively high predicted probabilities, thereby saving manpower and realizing AI's precise selection of high-potential merchants for advertisers.

[0153] In addition, the specific implementation content of the method for determining the merchant advertiser conversion rate in a reference embodiment of the present invention has been described in detail in the method for determining the merchant advertiser conversion rate described above, so the repeated content will not be described here.

[0154] Figure 3 is a schematic diagram of an apparatus for determining a merchant advertiser conversion rate according to an embodiment of the present invention. Figure 3As shown, the device 300 for determining the advertiser conversion rate of a merchant includes an acquisition module 301, a processing module 302 and a calculation module 303; wherein, the acquisition module 301 is used to acquire the dimensional data of the target merchant, and the dimensional data includes industrial and commercial registration information, main business category information, year-on-year information and the advertiser mining rate of the target mining personnel corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated using Bayesian smoothing processing; the processing module 302 is used to process the dimensional data of the target merchant into dimensional features; the calculation module 303 is used to input the dimensional features of the target merchant into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is trained using a neural network model.

[0155] Optionally, the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing, including:

[0156] Calculate the advertiser signing rate of each mining personnel in the past period of time, and then calculate the mean and variance of the advertiser signing rate;

[0157] Calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate;

[0158] The advertiser mining rate of the target mining personnel is recalculated according to the Bayesian smoothing coefficient.

[0159] Optionally, the advertiser signing rate of each mining personnel in the past period of time is calculated, so as to calculate the advertiser signing rate mean and the advertiser signing rate variance, including:

[0160] For each mining personnel, the number of advertiser contracts signed by the mining personnel in the past period of time is divided by the number of merchants visited, thereby obtaining the advertiser contract signing rate of the mining personnel in the past period of time;

[0161] Calculate the average of the advertiser signing rates according to the advertiser signing rates of each mining personnel in the past period of time;

[0162] The advertiser signing rate variance is calculated based on the advertiser signing rate of each mining personnel in the past period of time and the average of the advertiser signing rate.

[0163] Optionally, calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate includes:

[0164] The first Bayesian smoothing coefficient and the second Bayesian smoothing coefficient are calculated using the following formulas:

[0165] α=mean*((mean*(1-mean)) / (var*var)-1)

[0166] β=(1-mean)*((mean*(1-mean)) / (var*var)-1)

[0167] Among them, mean is the mean of advertiser signing rate, var is the variance of advertiser signing rate, α is the first Bayesian smoothing coefficient, and β is the second Bayesian smoothing coefficient.

[0168] Optionally, recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient includes:

[0169] The advertiser mining rate of the target mining personnel is calculated using the following formula:

[0170] M=(S+α) / (P+α+β)

[0171] Among them, S is the number of advertisers signed, and P is the number of merchants visited.

[0172] Optionally, the processing module 302 is further used for:

[0173] Processing the dimensional data of the target merchant into dimensional features by using a multi-hot encoding method;

[0174] The pipeline mechanism is used to serialize the dimensional features of each dimension and aggregate them into a total dimensional feature.

[0175] Optionally, a training module is also included, for:

[0176] Extracting the dimension data of each positive sample merchant and the dimension data of each negative sample merchant according to a preset ratio; wherein the same period last year information includes daily period last year information, weekly period last year information, and monthly period last year information;

[0177] Processing the dimensional data of each of the positive sample merchants and the dimensional data of each of the negative sample merchants into dimensional features;

[0178] The dimensional features of each positive sample merchant and the dimensional features of each negative sample merchant are input into a neural network model for supervised training, so as to obtain an advertiser conversion rate model through training.

[0179] It should be noted that the specific implementation content of the device for determining the merchant advertiser conversion rate of the present invention has been described in detail in the method for determining the merchant advertiser conversion rate described above, so the repeated content will not be described here.

[0180] Figure 4An exemplary system architecture 400 is shown to which the method for determining a merchant advertiser conversion rate or the apparatus for determining a merchant advertiser conversion rate according to the embodiments of the present invention can be applied.

[0181] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, 403, network 404 and server 405. Network 404 is used to provide a medium for communication links between terminal devices 401, 402, 403 and server 405. Network 404 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0182] Users can use terminal devices 401, 402, 403 to interact with server 405 through network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, 403, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0183] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0184] Server 405 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using terminal devices 401, 402, and 403. The backend management server may analyze and process the received data such as item information query requests, and feed back the processing results to the terminal device.

[0185] It should be noted that the method for determining the merchant-advertiser conversion rate provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the device for determining the merchant-advertiser conversion rate is generally disposed in the server 405 .

[0186] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0187] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer system 500 of a terminal device suitable for implementing an embodiment of the present invention. Figure 5 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0188] like Figure 5As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage part 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0189] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that a computer program read therefrom is installed into the storage section 508 as needed.

[0190] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present invention are executed.

[0191] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0192] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0193] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor, for example, it may be described as follows: a processor includes an acquisition module, a processing module, and a calculation module, wherein the names of these modules do not constitute a limitation on the modules themselves in some cases.

[0194] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device implements the following method: obtaining the dimensional data of the target merchant, the dimensional data including the industrial and commercial registration information, the main category information, the same period information and the advertiser mining rate of the target mining personnel corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing; processing the dimensional data of the target merchant into dimensional features; inputting the dimensional features of the target merchant into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is trained by a neural network model.

[0195] As another aspect, an embodiment of the present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method described in any of the above embodiments is implemented.

[0196] According to the technical solution of the embodiment of the present invention, the industrial and commercial registration information, main category information, year-on-year information and the advertiser mining rate of the target mining personnel corresponding to the target merchant are all input as feature data into the advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant, and the technical means of Bayesian smoothing of the advertiser mining rate are adopted, so as to overcome the technical problem of low accuracy of advertiser potential mining in the prior art. The embodiment of the present invention uses the industrial and commercial registration information, main category information, year-on-year information and the advertiser mining rate of the target mining personnel corresponding to the target merchant as feature data and adopts Bayesian smoothing to process the advertiser mining rate, and accurately calculates the probability of each merchant becoming an advertiser through the advertiser conversion rate model, thereby providing a judgment basis for mining personnel to mine high-potential advertisers, so that mining personnel can make accurate visits.

[0197] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for determining a merchant advertiser conversion rate, characterized in that: include: Acquire dimensional data of the target merchant, the dimensional data including business registration information, main category information, year-on-year and month-on-month information, and the advertiser mining rate of the target mining personnel corresponding to the target merchant; wherein the advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing processing; Processing the dimensional data of the target merchant into dimensional features; The dimensional features of the target merchant are input into an advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is obtained by training with a neural network model.

2. The method according to claim 1, characterized in that: The advertiser mining rate of the target mining personnel is calculated by Bayesian smoothing, including: Calculate the advertiser signing rate of each mining personnel in the past period of time, and then calculate the mean and variance of the advertiser signing rate; Calculating a Bayesian smoothing coefficient according to the mean of the advertiser signing rate and the variance of the advertiser signing rate; The advertiser mining rate of the target mining personnel is recalculated according to the Bayesian smoothing coefficient.

3. The method according to claim 2, characterized in that Calculate the advertiser signing rate of each mining personnel in the past period of time, so as to calculate the mean and variance of the advertiser signing rate, including: For each mining personnel, the number of advertiser contracts signed by the mining personnel in the past period of time is divided by the number of merchants visited, thereby obtaining the advertiser contract signing rate of the mining personnel in the past period of time; Calculate the average of the advertiser signing rates according to the advertiser signing rates of each mining personnel in the past period of time; The advertiser signing rate variance is calculated based on the advertiser signing rate of each mining personnel in the past period of time and the average of the advertiser signing rate.

4. The method according to claim 2, characterized in that: Calculating a Bayesian smoothing coefficient according to the advertiser signing rate mean and the advertiser signing rate variance includes: The first Bayesian smoothing coefficient and the second Bayesian smoothing coefficient are calculated using the following formulas: α=mean*((mean*(1-mean)) / (var*var)-1) β=(1-mean)*((mean*(1-mean)) / (var*var)-1) Among them, mean is the mean of advertiser signing rate, var is the variance of advertiser signing rate, α is the first Bayesian smoothing coefficient, and β is the second Bayesian smoothing coefficient.

5. The method according to claim 4, characterized in that Recalculating the advertiser mining rate of the target mining personnel according to the Bayesian smoothing coefficient includes: The advertiser mining rate of the target mining personnel is calculated using the following formula: M=(S+α) / (P+α+β) Among them, S is the number of advertisers signed, and P is the number of merchants visited.

6. The method according to claim 1, characterized in that Processing the target merchant's dimensional data into dimensional features includes: Processing the dimensional data of the target merchant into dimensional features by using a multi-hot encoding method; The pipeline mechanism is used to serialize the dimensional features of each dimension and aggregate them into a total dimensional feature.

7. The method according to claim 1, characterized in that Before obtaining the dimension data of the target merchant, it also includes: Extracting the dimension data of each positive sample merchant and the dimension data of each negative sample merchant according to a preset ratio; wherein the same period last year information includes daily period last year information, weekly period last year information, and monthly period last year information; Processing the dimensional data of each of the positive sample merchants and the dimensional data of each of the negative sample merchants into dimensional features; The dimensional features of each positive sample merchant and the dimensional features of each negative sample merchant are input into a neural network model for supervised training, so as to obtain an advertiser conversion rate model through training.

8. A device for determining a merchant advertiser conversion rate, characterized in that: include: An acquisition module is used to acquire dimension data of a target merchant, wherein the dimension data includes business registration information, main business category information, year-on-year and month-on-month information, and an advertiser mining rate of a target mining person corresponding to the target merchant; wherein the advertiser mining rate of the target mining person is calculated by Bayesian smoothing processing; A processing module, used for processing the dimension data of the target merchant into dimension features; A calculation module is used to input the dimensional features of the target merchant into an advertiser conversion rate model, thereby outputting the advertiser conversion rate of the target merchant; wherein the advertiser conversion rate model is obtained by training a neural network model.

9. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.