Product-oriented user outbound calling and product recommendation method and device

Abstract

The present application provides a product-oriented user outbound call and product recommendation method and device, which relates to the field of product promotion technology. The first demand degree, the second demand degree, and the third demand degree corresponding to each target candidate user can be input into the pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, so that the accuracy and reliability of the probability that each target candidate user has a purchase intention for the target product can be very high. The product introduction information and purchase link of the target product are pushed to the target candidate users whose probability of having a purchase intention for the target product is greater than the set probability threshold. Since the accuracy and reliability of the probability that the target candidate user has a purchase intention for the target product are very high, the probability that the target candidate user purchases the target product based on the product introduction information and the purchase link is also particularly high.

Description

Technical Field

[0001] The present application relates to the technical field of product promotion, and in particular to a method and device for product-oriented user outbound calling and product recommendation. Background Art

[0002] Currently, more and more businesses can collect the product types and personal attribute information (such as age, gender, etc.) purchased by users in the past to determine whether the user has a purchase demand for the target product (such as insurance, financial products, clothing, computers, mobile phones, etc.). When it is determined that the user has a purchase demand for the target product, a voice call is made to the user who is determined to have a purchase demand, and the target product is promoted to the user.

[0003] However, simply determining whether a user has a purchase demand for a target product based on the product types and personal attribute information (such as age, gender, etc.) purchased by the user in the past is unreliable. As a result, after the user answers the voice call made by the merchant to the user, they are not interested in the target product promoted by the voice call and hang up the phone directly within a short period of time, resulting in a low conversion rate of voice call promotion and causing certain troubles to the user. Summary of the invention

[0004] The present application provides a product-oriented user outbound calling and product recommendation method and device, which are used to solve the problem that the reliability of determining whether a user has a purchase demand for a target product in the prior art is low, resulting in the user not being interested in the target product promoted by the voice call after the user answers the voice call made by the merchant, and directly hanging up the phone within a short period of time, resulting in a low conversion rate of the voice call promotion and causing certain troubles to the user.

[0005] In a first aspect, the present application provides a product-oriented user outbound calling and product recommendation method, comprising:

[0006] Obtain a set of candidate users who have potential demand for the target product;

[0007] Determine, based on payment details data of each candidate user in the candidate user set in the payment application, a first ratio of the number of products purchased by each candidate user that are the same as the target product type to the total number of products purchased, a second ratio of the product amount of the products purchased that are the same as the target product type to the total amount of products purchased, a change rate of the first ratio in a recent target period, and a change rate of the second ratio, wherein the target period includes multiple periodic periods;

[0008] Determine a first demand degree of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user;

[0009] Determine the quantity distribution of each type of product purchased by each candidate user in the previous cycle period according to the payment details data of each candidate user in the candidate user set in the payment application, and obtain the quantity distribution of each type of product purchased by the historical users who purchased the target product in the previous cycle period;

[0010] Determine the mean and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period;

[0011] Determine the second demand degree of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the multiple differences, and the largest covariance and the smallest covariance among the multiple covariances;

[0012] Determining an estimated demand degree for the target product by each candidate user according to the first demand degree and the second demand degree;

[0013] Determine at least one candidate user whose estimated demand degree is greater than a set demand degree as a target candidate user;

[0014] Make an outbound voice call to each target candidate user, and conduct voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset words;

[0015] Determine the frequency of occurrence of different keywords that represent demand for the target product during voice interaction;

[0016] In the process of recognizing voice interaction, the semantic similarity between each question and answer sentence and each standard question and answer sentence in the preset standard expectation library that represents the user's demand for the target product is calculated according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question-answering sentence and each standard question-answering sentence in the standard expectation library, and n is the number of question-answering sentences;

[0017] Determine the third demand degree of each target candidate user for the target product based on the semantic similarity between the current voice interaction and the standard voice interaction of the user having demand for the target product, and the frequency of occurrence of different keywords having demand for the target product;

[0018] Inputting the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, wherein the user purchase intention determination model is obtained by inputting a plurality of training samples into an initial neural network training, and each training sample includes the historical first demand degree, the historical second demand degree, the historical third demand degree of the user for the target product and the corresponding historical actual probability that the user has a purchase intention for the target product;

[0019] The product introduction information and purchase link of the target product are pushed to the target candidate users whose probability of having purchase intention for the target product is greater than the set probability threshold.

[0020] In some implementations, determining the first demand of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user includes:

[0021] According to the formula G 1 =αk 1 +βk 2 +γk 3 +δk 4 , determine the first demand degree of the candidate user for the target product, where G 1 is the first demand degree of the candidate user for the target product, k 1 is the first proportion, k 2 is the second proportion, k 3 is the rate of change of the first proportion, k 4 is the change rate of the second proportion, α is the set first conversion factor, β is the set second conversion factor, γ is the set third conversion factor, and δ is the set fourth conversion factor.

[0022] In some implementations, determining the second demand for the target product of each candidate user according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the multiple differences, and the largest covariance and the smallest covariance among the multiple covariances includes:

[0023] According to the formula Determine the second demand degree of each candidate user for the target product, where G 2is the second demand degree, W is the average value of the differences corresponding to each candidate user, V is the covariance of the differences corresponding to each candidate user, max(W) is the average value of the maximum difference among the average values ​​of multiple differences, min(W) is the average value of the minimum difference among the average values ​​of multiple differences, max(V) is the largest covariance among multiple covariances, and min(V) is the smallest covariance among multiple covariances.

[0024] In some implementations, determining the estimated demand for the target product of each candidate user according to the first demand and the second demand includes:

[0025] According to the formula G 3 =m 1 G 1 +m 2 G 2 , determine the estimated demand of each candidate user for the target product, where m 1 is the first weighting coefficient set, m 2 is the second weighting coefficient, G 1 is the first demand degree, G 2 The second demand.

[0026] In some implementations, determining the frequency of occurrence of different keywords indicating demand for the target product during voice interaction includes:

[0027] Convert voice information in the voice interaction process into text information;

[0028] Segment the text information to obtain a voice interaction vocabulary set;

[0029] Extract keywords that represent the demand for the target product from the preset key word library from the voice interaction vocabulary set;

[0030] Record the frequency of occurrence of each extracted keyword.

[0031] In some implementations, determining the third demand degree of each target candidate user for the target product according to the semantic similarity between the current voice interaction and the standard voice interaction of the user having demand for the target product, and the frequency of occurrence of different keywords having demand for the target product, includes:

[0032] According to the formula G 4 =Qk 5 +σk 6 , determine the third demand degree of each target candidate user for the target product, where G 4 is the third demand degree of each target candidate user for the target product, k 5 is the preset fifth conversion factor, k 6is the preset sixth conversion factor, Q is the semantic similarity, and σ is the frequency of occurrence of different keywords that have demand for the target product.

[0033] In a second aspect, the present application also provides a product-oriented user outbound call and product recommendation device, including:

[0034] An information acquisition unit, used to acquire a set of candidate users who have potential demand for the target product;

[0035] A first data determination unit is used to determine, based on payment details data of each candidate user in the candidate user set in the payment application, a first ratio of the number of products purchased by each candidate user of the same target product type to the total number of products purchased, a second ratio of the product amount of the purchased products of the same target product type to the total amount of products purchased, a change rate of the first ratio in a recent target period, and a change rate of the second ratio, wherein the target period includes multiple periodic periods;

[0036] A first demand determination unit is used to determine a first demand of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion and the change rate of the second proportion corresponding to each candidate user;

[0037] A second data determination unit, configured to determine, based on payment details data of each candidate user in the candidate user set in the payment application, the quantity distribution of each type of product purchased by each candidate user in a previous cycle period;

[0038] The information acquisition unit is further used to acquire the quantity distribution of each type of products purchased by the historical users who have purchased the target product in the previous periodic period;

[0039] The second demand determination unit is used to determine the average value and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period; determine the second demand of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the differences, and the largest covariance and the smallest covariance among the multiple covariances;

[0040] An estimated demand determination unit, configured to determine an estimated demand for a target product of each candidate user according to the first demand and the second demand;

[0041] a target candidate user determination unit, configured to determine at least one candidate user whose estimated demand is greater than a set demand as a target candidate user;

[0042] A voice interaction unit is used to make an outbound voice call to each target candidate user, and perform voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset words;

[0043] The third demand determination unit is used to determine the frequency of occurrence of different keywords that represent the demand for the target product during the voice interaction; identify the semantic similarity between each question and answer sentence and each standard question and answer sentence in the preset standard expectation library that represents the user's demand for the target product during the voice interaction, and compare the semantic similarity between each question and answer sentence and the standard expectation library that represents the user's demand for the target product according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question-and-answer statement and each standard question-and-answer statement in the standard expectation library, and n is the number of question-and-answer statements; according to the semantic similarity between the current voice interaction and the standard voice interaction of users with demand for the target product, and the frequency of occurrence of different keywords with demand for the target product, determine the third demand degree of each target candidate user for the target product;

[0044] a purchase intention determination unit, for inputting the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, wherein the user purchase intention determination model is obtained by inputting a plurality of training samples into an initial neural network for training, each training sample including a historical first demand degree, a historical second demand degree and a historical third demand degree of a historical user for the target product and a corresponding historical actual probability that the user has a purchase intention for the target product;

[0045] The product recommendation unit is used to push product introduction information and purchase links of the target product to target candidate users whose probability of having purchase intention for the target product is greater than a set probability threshold.

[0046] In a third aspect, the present application also provides a background server, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the background server executes the method provided in the first aspect as described above.

[0047] In a fourth aspect, the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the computer executes the method provided in the first aspect.

[0048] The present application provides a product-oriented user outbound calling and product recommendation method and device, which can determine the first demand degree of the candidate user for the target product based on a first ratio of the number of purchased products of the same type as the target product to the total number of purchased products for each candidate user in a previous cycle period, a second ratio of the product amount of purchased products of the same type as the target product to the total amount of purchased products, the change rate of the first ratio in a most recent target period, and the change rate of the second ratio; in this way, the reliability of the determined first demand degree can be high.

[0049] Based on the quantity distribution of each type of product purchased by each candidate user in the previous cycle period, and obtaining the quantity distribution of each type of product purchased by historical users who purchased the target product in the previous cycle period; determine the average value and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by historical users who purchased the target product in the previous cycle period; determine the second demand degree of each candidate user for the target product based on the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of multiple differences, and the largest covariance and the smallest covariance among the multiple covariances. In this way, the reliability of the determined second demand degree can be high.

[0050] Furthermore, the reliability of determining the estimated demand for the target product of each candidate user based on the first demand and the second demand can be made higher.

[0051] Next, at least one candidate user whose estimated demand is greater than the set demand is determined as a target candidate user; a voice call is made to each target candidate user, and based on the target candidate user's answer and the preset words, a voice interaction is performed with the corresponding target candidate user; during the voice interaction, the frequency of occurrence of different keywords representing the demand for the target product is determined; during the voice interaction, the semantic similarity of each question and answer sentence with each standard question and answer sentence in the preset standard expectation library representing the user's demand for the target product is identified, and the semantic similarity is calculated according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question and answer sentence and each standard question and answer sentence in the standard expectation library, and n is the number of question and answer sentences; according to the semantic similarity between the current voice interaction and the standard voice interaction of users with demand for the target product, and the frequency of occurrence of different keywords with demand for the target product, the third demand degree of each target candidate user for the target product is determined. In this way, the reliability of the obtained third demand degree can also be high.

[0052] Furthermore, the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user are input into the pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product. Since the user purchase intention determination model is obtained by inputting multiple training samples into the initial neural network training, each training sample includes the historical first demand degree, the historical second demand degree, the historical third demand degree of the target product by the historical user and the historical actual probability that the corresponding user has a purchase intention for the target product. In this way, the accuracy and reliability of the probability that each target candidate user has a purchase intention for the target product can be made very high. The product introduction information and purchase link of the target product are pushed to the target candidate user whose probability of having a purchase intention for the target product is greater than the set probability threshold. Since the accuracy and reliability of the probability that the target candidate user has a purchase intention for the target product are very high, the probability that the target candidate user purchases the target product based on the product introduction information and the purchase link is also particularly high. Furthermore, the conversion rate of the promotion of the target product is improved, and the troubles brought to the user are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0054] Figure 1 A flowchart of a product-oriented user outbound calling and product recommendation method provided in an embodiment of the present application;

[0055] Figure 2 A block diagram of the functional modules of the product-oriented user outbound call and product recommendation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments made by ordinary technicians in this field under the enlightenment of the embodiments belong to the scope of protection of the present application.

[0057] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0058] The present application embodiment provides a product-oriented user outbound call and product recommendation method, which is applied to a backend server. Figure 1 As shown, the method provided in the embodiment of the present application includes:

[0059] S101: Acquire a set of candidate users who have potential demand for a target product.

[0060] In some embodiments, a set of users who have browsed product promotion web pages or live broadcast rooms associated with a target product (such as insurance, financial products, clothing, or software applications, etc.) within a previous preset period of time (such as the previous 3 days) can be obtained as a set of candidate users who have potential demand for the target product.

[0061] S102: Based on the payment details data of each candidate user in the candidate user set in the payment application, determine a first ratio of the number of purchased products of the same target product type as that of each candidate user in a previous cycle period to the total number of purchased products, a second ratio of the product amount of purchased products of the same target product type as that of each candidate user to the total amount of purchased products, a change rate of the first ratio in a most recent target period, and a change rate of the second ratio, wherein the target period includes multiple cycle periods.

[0062] It can be understood that the first proportion can represent the candidate user's demand for the target product in terms of purchase quantity proportion, the second proportion can represent the candidate user's demand for the target product in terms of purchase amount proportion, the change rate of the first proportion within the most recent target time period can represent the changing trend of the candidate user's demand for the target product in terms of purchase quantity proportion, and the change rate of the second proportion within the most recent target time period can represent the changing trend of the candidate user's demand for the target product in terms of purchase amount proportion.

[0063] The target period may be one year, and the cycle period may be one month.

[0064] S103: Determine a first demand degree of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user.

[0065] For example, according to the formula G 1 =αk 1 +βk 2 +γk 3 +δk 4 , determine the first demand degree of the candidate user for the target product. 1 is the first demand degree of the candidate user for the target product, k 1 is the first proportion (such as 0.2 or 0.3), k 2 is the second proportion (such as 0.2 or 0.3), k 3 is the rate of change of the first proportion (such as 2 or 3), k 4 is the change rate of the second proportion (such as 2 or 3), α is the set first conversion factor, β is the set second conversion factor, γ is the set third conversion factor, and δ is the set fourth conversion factor. The first conversion factor, the second conversion factor, the third conversion factor, and the fourth conversion factor are all pre-configured different constants.

[0066] S104: Determine the quantity distribution of each type of product purchased by each candidate user in the candidate user set in the payment application according to the payment details data, and obtain the quantity distribution of each type of product purchased by historical users who have purchased the target product in the previous cycle period.

[0067] The quantity distribution of each type of product purchased by each candidate user in the previous cycle period may be: the number of type A products purchased by the candidate user in the previous month A1, the number of type B products purchased B1, and the number of type C products purchased C1. The quantity distribution of each type of product purchased by the historical users who purchased the target product in the previous cycle period may be: the number of type A products purchased A2, the number of type B products purchased B2, and the number of type C products purchased C2 in the previous month.

[0068] S105: Determine the mean and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period.

[0069] For example, based on the above, the average value of the difference of each candidate user may be equal to [|A1-A2|+|B1-B2|+|B1-B2|] / 3.

[0070] S106: Determine the second demand degree of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of multiple differences, and the largest covariance and the smallest covariance among the multiple covariances.

[0071] In some embodiments, the formula Determine the second demand degree of each candidate user for the target product. 2 is the second demand degree, W is the average value of the differences corresponding to each candidate user, V is the covariance of the differences corresponding to each candidate user, max(W) is the average value of the maximum difference among the average values ​​of multiple differences, min(W) is the average value of the minimum difference among the average values ​​of multiple differences, max(V) is the largest covariance among multiple covariances, and min(V) is the smallest covariance among multiple covariances.

[0072] S107: Determine the estimated demand of each candidate user for the target product according to the first demand and the second demand.

[0073] For example, according to the formula G 3 =m 1 G 1 +m 2 G 2 , determine the estimated demand of each candidate user for the target product, where m 1 is the first weighting coefficient set, m 2 is the second weighting coefficient, G 1 is the first demand degree, G 2 is the second demand degree. For example, m 1 can be equal to 0.4, m 2 It can be equal to 0.6.

[0074] S108: Determine at least one candidate user whose estimated demand degree is greater than the set demand degree as a target candidate user.

[0075] S109: Make an outbound voice call to each target candidate user, and conduct voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset words.

[0076] S110: Determine the frequency of occurrence of different keywords representing the demand for the target product during the voice interaction process.

[0077] Specifically, S110 may include:

[0078] Step 1: Convert the voice information in the voice interaction process into text information.

[0079] Step 2: Segment the text information to obtain a voice interaction vocabulary set.

[0080] Step 3: Extract keywords that represent the demand for the target product and are present in the preset keyword library from the voice interaction vocabulary set.

[0081] Step 4: Record the frequency of occurrence of each extracted keyword.

[0082] For example, the keywords include "interested", "have demand", "want to buy", "willing to", and "preferably like". "Interested" appears 2 times, "have demand" appears 2 times, "want to buy" appears 1 time, "willing to" appears 0 times, and "preferably like" appears 1 time. The frequency of occurrence of each different keyword is 6 times.

[0083] S111: In the process of recognizing voice interaction, the semantic similarity between each question and answer sentence and each standard question and answer sentence in the preset standard expectation library representing the user's demand for the target product is calculated according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which the user has demand for the target product.

[0084] Among them, S i is the highest semantic similarity between the i-th question and answer sentence and each standard question and answer sentence in the standard expectation library, and n is the number of question and answer sentences.

[0085] S112: Determine the third demand degree of each target candidate user for the target product based on the semantic similarity between the current voice interaction and the standard voice interaction of the user having demand for the target product, and the frequency of occurrence of different keywords having demand for the target product.

[0086] For example, according to the formula G 4 =Qk 5 +σk 6 , determine the third demand degree of each target candidate user for the target product.

[0087] Among them, G 4 is the third demand degree of each target candidate user for the target product, k 5 is the preset fifth conversion factor, k 6 is the preset sixth conversion factor, Q is the semantic similarity, and σ is the frequency of occurrence of different keywords that have demand for the target product. The fifth conversion factor and the sixth conversion factor may be different constants.

[0088] S113: Inputting the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product.

[0089] Among them, the user purchase intention determination model is obtained by inputting multiple training samples into the initial neural network training, and each training sample includes the historical first demand degree, historical second demand degree, historical third demand degree of the user for the target product and the corresponding historical actual probability of the user having the purchase intention for the target product.

[0090] S114: Push product introduction information and a purchase link of the target product to target candidate users whose probability of having a purchase intention for the target product is greater than a set probability threshold.

[0091] In summary, the embodiments of the present application provide a product-oriented user outbound calling and product recommendation method, which can determine the first demand degree of the candidate user for the target product based on a first ratio of the number of purchased products of the same type as the target product to the total number of purchased products for each candidate user in a previous cycle period, a second ratio of the product amount of purchased products of the same type as the target product to the total amount of purchased products, the rate of change of the first ratio in a most recent target period, and the rate of change of the second ratio; in this way, the reliability of the determined first demand degree can be made high.

[0092] Based on the quantity distribution of each type of product purchased by each candidate user in the previous cycle period, and obtaining the quantity distribution of each type of product purchased by historical users who purchased the target product in the previous cycle period; determine the average value and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by historical users who purchased the target product in the previous cycle period; determine the second demand degree of each candidate user for the target product based on the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of multiple differences, and the largest covariance and the smallest covariance among the multiple covariances. In this way, the reliability of the determined second demand degree can be high.

[0093] Furthermore, the reliability of determining the estimated demand for the target product of each candidate user based on the first demand and the second demand can be made higher.

[0094] Next, at least one candidate user whose estimated demand is greater than the set demand is determined as a target candidate user; a voice call is made to each target candidate user, and based on the target candidate user's answer and the preset words, a voice interaction is performed with the corresponding target candidate user; during the voice interaction, the frequency of occurrence of different keywords representing the demand for the target product is determined; during the voice interaction, the semantic similarity of each question and answer sentence with each standard question and answer sentence in the preset standard expectation library representing the user's demand for the target product is identified, and the semantic similarity is calculated according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question and answer sentence and each standard question and answer sentence in the standard expectation library, and n is the number of question and answer sentences; according to the semantic similarity between the current voice interaction and the standard voice interaction of users with demand for the target product, and the frequency of occurrence of different keywords with demand for the target product, the third demand degree of each target candidate user for the target product is determined. In this way, the reliability of the obtained third demand degree can also be high.

[0095] Furthermore, the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user are input into the pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product. Since the user purchase intention determination model is obtained by inputting multiple training samples into the initial neural network training, each training sample includes the historical first demand degree, the historical second demand degree, the historical third demand degree of the target product by the historical user and the historical actual probability that the corresponding user has a purchase intention for the target product. In this way, the accuracy and reliability of the probability that each target candidate user has a purchase intention for the target product can be made very high. The product introduction information and purchase link of the target product are pushed to the target candidate user whose probability of having a purchase intention for the target product is greater than the set probability threshold. Since the accuracy and reliability of the probability that the target candidate user has a purchase intention for the target product are very high, the probability that the target candidate user purchases the target product based on the product introduction information and the purchase link is also particularly high. Furthermore, the conversion rate of the promotion of the target product is improved, and the troubles brought to the user are reduced.

[0096] See also Figure 2 , the embodiment of the present application also provides a product-oriented user outbound call and product recommendation device. It should be noted that the basic principle and technical effects of the product-oriented user outbound call and product recommendation device provided in the embodiment of the present application are the same as those in the above embodiment. For the sake of brief description, for the parts not mentioned in the embodiment of the present application, please refer to the corresponding contents in the above embodiment. Including:

[0097] An information acquisition unit, used to acquire a set of candidate users who have potential demand for the target product;

[0098] A first data determination unit is used to determine, based on payment details data of each candidate user in the candidate user set in the payment application, a first ratio of the number of products purchased by each candidate user of the same target product type to the total number of products purchased, a second ratio of the product amount of the purchased products of the same target product type to the total amount of products purchased, a change rate of the first ratio in a recent target period, and a change rate of the second ratio, wherein the target period includes multiple periodic periods;

[0099] A first demand determination unit is used to determine a first demand of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion and the change rate of the second proportion corresponding to each candidate user;

[0100] A second data determination unit, configured to determine, based on payment details data of each candidate user in the candidate user set in the payment application, the quantity distribution of each type of product purchased by each candidate user in a previous cycle period;

[0101] The information acquisition unit is further used to acquire the quantity distribution of each type of products purchased by the historical users who have purchased the target product in the previous periodic period;

[0102] The second demand determination unit is used to determine the average value and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period; determine the second demand of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the differences, and the largest covariance and the smallest covariance among the multiple covariances;

[0103] An estimated demand determination unit, configured to determine an estimated demand for a target product of each candidate user according to the first demand and the second demand;

[0104] a target candidate user determination unit, configured to determine at least one candidate user whose estimated demand is greater than a set demand as a target candidate user;

[0105] A voice interaction unit is used to make an outbound voice call to each target candidate user, and perform voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset words;

[0106] The third demand determination unit is used to determine the frequency of occurrence of different keywords that represent the demand for the target product during the voice interaction; identify the semantic similarity between each question and answer sentence and each standard question and answer sentence in the preset standard expectation library that represents the user's demand for the target product during the voice interaction, and compare the semantic similarity between each question and answer sentence and the standard expectation library that represents the user's demand for the target product according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which the user has demand for the target product.

[0107] Among them, S i is the highest semantic similarity between the ith question and answer statement and each standard question and answer statement in the standard expectation library, and n is the number of question and answer statements; according to the semantic similarity between the current voice interaction and the standard voice interaction in which users have demand for the target product, and the frequency of occurrence of different keywords that have demand for the target product, the third demand degree of each target candidate user for the target product is determined.

[0108] a purchase intention determination unit, for inputting the first demand degree, the second demand degree and the third demand degree corresponding to each target candidate user into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, wherein the user purchase intention determination model is obtained by inputting a plurality of training samples into an initial neural network for training, each training sample including a historical first demand degree, a historical second demand degree and a historical third demand degree of a historical user for the target product and a corresponding historical actual probability that the user has a purchase intention for the target product;

[0109] The product recommendation unit is used to push product introduction information and purchase links of the target product to target candidate users whose probability of having purchase intention for the target product is greater than a set probability threshold.

[0110] In addition, an embodiment of the present application also provides a background server, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the background server executes the method provided in the above embodiment.

[0111] In addition, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the computer executes the method provided in the above embodiment.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A product-oriented user outbound calling and product recommendation method, characterized in that: The method comprises: Obtain a set of candidate users who have potential demand for the target product; Determine, according to the payment details data of each candidate user in the candidate user set in the payment application, a first ratio of the number of products purchased by each candidate user of the same target product type to the total number of products purchased, a second ratio of the product amount of the purchased products of the same target product type to the total amount of products purchased, a change rate of the first ratio in a recent target period, and a change rate of the second ratio, wherein the target period includes a plurality of the cycle periods; Determining a first demand degree of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user; Determine, based on payment details of each candidate user in the candidate user set in the payment application, the quantity distribution of each type of product purchased by each candidate user in a previous cycle period, and obtain the quantity distribution of each type of product purchased by historical users who have purchased the target product in a previous cycle period; Determine the mean and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period; Determine the second demand degree of each candidate user for the target product according to the average value and covariance of the difference values ​​corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the multiple difference values, and the largest covariance and the smallest covariance among the multiple covariances; Determining an estimated demand degree for the target product by each of the candidate users according to the first demand degree and the second demand degree; Determine at least one candidate user whose estimated demand degree is greater than a set demand degree as a target candidate user; Making an outbound voice call to each of the target candidate users, and performing voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset speech; Determine the frequency of occurrence of different keywords that represent demand for the target product during voice interaction; In the process of recognizing voice interaction, the semantic similarity between each question and answer sentence and each standard question and answer sentence in the preset standard expectation library that represents the user's demand for the target product is calculated according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question-answering sentence and each standard question-answering sentence in the standard expectation library, and n is the number of question-answering sentences; Determining a third demand degree of each target candidate user for the target product according to the semantic similarity between the current voice interaction and a standard voice interaction of a user having demand for the target product, and the frequency of occurrence of different keywords having demand for the target product; Inputting the first demand degree, the second demand degree and the third demand degree corresponding to each of the target candidate users into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, wherein the user purchase intention determination model is obtained by inputting a plurality of training samples into an initial neural network training, each of the training samples including a historical first demand degree, a historical second demand degree and a historical third demand degree of a historical user for a target product and a corresponding historical actual probability that the user has a purchase intention for the target product; The product introduction information and purchase link of the target product are pushed to the target candidate users whose probability of having purchase intention for the target product is greater than a set probability threshold.

2. The method according to claim 1, characterized in that The determining, according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user, a first demand degree of the candidate user for the target product includes: According to the formula G1=αk1+βk2+γk3+δk4, the first demand degree of the candidate user for the target product is determined, wherein G1 is the first demand degree of the candidate user for the target product, k1 is the first proportion, k2 is the second proportion, k3 is the change rate of the first proportion, k4 is the change rate of the second proportion, α is the set first conversion factor, β is the set second conversion factor, γ is the set third conversion factor, and δ is the set fourth conversion factor.

3. The method according to claim 1, characterized in that Determining the second demand degree of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the multiple difference values, and the largest covariance and the smallest covariance among the multiple covariances, includes: According to the formula Determine the second demand degree of each candidate user for the target product, where G2 is the second demand degree, W is the average value of the differences corresponding to each candidate user, V is the covariance of the differences corresponding to each candidate user, max(W) is the average value of the maximum difference among the average values ​​of multiple differences, min(W) is the average value of the minimum difference among the average values ​​of multiple differences, max(V) is the maximum covariance among multiple covariances, and min(V) is the minimum covariance among multiple covariances.

4. The method according to claim 1, characterized in that: Determining the estimated demand of each candidate user for the target product according to the first demand and the second demand includes: According to the formula G3=m1G1+m2G2, the estimated demand of each candidate user for the target product is determined, wherein m1 is the set first weighting coefficient, m2 is the set second weighting coefficient, G1 is the first demand, and G2 is the second demand.

5. The method according to claim 1, characterized in that In the process of determining the voice interaction, the frequency of occurrence of different keywords indicating demand for the target product includes: Convert voice information in the voice interaction process into text information; Segmenting the text information to obtain a voice interaction vocabulary set; Extracting keywords that represent the demand for the target product and are present in a preset keyword library from the voice interaction vocabulary set; Record the frequency of occurrence of each extracted keyword.

6. The method according to claim 1, characterized in that The determining, based on the semantic similarity between the current voice interaction and a standard voice interaction in which a user has a demand for the target product, and the frequency of occurrence of different keywords in which the user has a demand for the target product, a third demand degree of each target candidate user for the target product includes: According to the formula G4=Qk5+σk6, the third demand degree of each target candidate user for the target product is determined, wherein G4 is the third demand degree of each target candidate user for the target product, k5 is the preset fifth conversion factor, k6 is the preset sixth conversion factor, Q is the semantic similarity, and σ is the frequency of occurrence of different keywords that indicate demand for the target product.

7. The method according to claim 1, characterized in that The step of obtaining a set of candidate users who have potential demand for the target product includes: A set of users who have browsed a product promotion webpage or a live broadcast room associated with the target product within a previous preset time period is obtained as a set of candidate users with potential demand for the target product.

8. A product-oriented user outbound call and product recommendation device, characterized in that: The device comprises: An information acquisition unit, used to acquire a set of candidate users who have potential demand for the target product; A first data determination unit is used to determine, based on the payment details data of each candidate user in the candidate user set in the payment application, a first ratio of the number of products purchased by each candidate user of the same target product type to the total number of products purchased, a second ratio of the product amount of the purchased products of the same target product type to the total amount of products purchased, a change rate of the first ratio in a recent target period, and a change rate of the second ratio, wherein the target period includes a plurality of the cycle periods; A first demand determination unit is used to determine a first demand of the candidate user for the target product according to the first proportion, the second proportion, the change rate of the first proportion, and the change rate of the second proportion corresponding to each candidate user; A second data determination unit, configured to determine, based on payment details of each candidate user in the candidate user set in the payment application, the quantity distribution of each type of product purchased by each candidate user in a previous period; The information acquisition unit is further used to acquire the quantity distribution of each type of products purchased by historical users who have purchased the target product in the previous periodic period; A second demand determination unit is used to determine the average value and covariance of the difference between the quantity of each type of product purchased by each candidate user in the previous cycle period and the quantity of the corresponding type of product purchased by the historical users who purchased the target product in the previous cycle period; determine the second demand of each candidate user for the target product according to the average value and covariance of the difference corresponding to each candidate user, the average value of the largest difference and the average value of the smallest difference among the average values ​​of the differences, and the largest covariance and the smallest covariance among the multiple covariances; an estimated demand determination unit, configured to determine an estimated demand of each candidate user for the target product according to the first demand and the second demand; a target candidate user determination unit, configured to determine at least one candidate user whose estimated demand degree is greater than a set demand degree as a target candidate user; A voice interaction unit, used to make an outbound voice call to each of the target candidate users, and perform voice interaction with the corresponding target candidate user based on the target candidate user's answer and preset speech; The third demand determination unit is used to determine the frequency of occurrence of different keywords that represent the demand for the target product during the voice interaction; identify the semantic similarity between each question and answer statement and each standard question and answer statement in the preset standard expectation library that represents the user's demand for the target product during the voice interaction, and compare the semantic similarity between each question and answer statement and the standard expectation library that represents the user's demand for the target product according to the formula Determine the semantic similarity Q between the current voice interaction and the standard voice interaction in which users have demand for the target product, where S i is the highest semantic similarity between the i-th question-and-answer statement and each standard question-and-answer statement in the standard expectation library, and n is the number of question-and-answer statements; according to the semantic similarity between the current voice interaction and the standard voice interaction in which the user has a demand for the target product, and the frequency of occurrence of each different keyword in which the user has a demand for the target product, determine the third demand degree of each target candidate user for the target product; a purchase intention determination unit, configured to input the first demand degree, the second demand degree, and the third demand degree corresponding to each of the target candidate users into a pre-trained user purchase intention determination model to determine the probability that each corresponding target candidate user has a purchase intention for the target product, wherein the user purchase intention determination model is obtained by inputting a plurality of training samples into an initial neural network for training, each of the training samples comprising a historical first demand degree, a historical second demand degree, a historical third demand degree of a historical user for a target product, and a corresponding historical actual probability that the user has a purchase intention for the target product; The product recommendation unit is used to push product introduction information and purchase links of the target product to target candidate users whose probability of having purchase intention for the target product is greater than a set probability threshold.

9. A backend server, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the background server executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the computer is caused to perform the method according to any one of claims 1 to 7.

Images