Target user group expansion method, electronic equipment and storage medium

Abstract

The invention relates to the technical field of computers, in particular to a target user group expansion method, electronic equipment and a storage medium, and the method comprises the steps: extracting a plurality of product keywords corresponding to a first number of product description texts corresponding to a given user according to a product recommendation sequence of a target app of the given user, respectively matching the plurality of product keywords corresponding to the given user with a plurality of preset keywords in a preset keyword library to determine a target product sub-tag corresponding to the given user, and when the number of users corresponding to the target product sub-tag is smaller than a user number threshold value corresponding to the target product sub-tag, sending the target product sub-tag to a server; training a preset logistic regression model, inputting a plurality of product keywords corresponding to each given user into a trained target logistic regression model, selecting a plurality of given users not smaller than a user number threshold according to an output result, and determining the given users as target users; according to the invention, accurate expansion of the target user group can be completed.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method for expanding a target user group, an electronic device and a storage medium. Background Art

[0002] When recommending products in the target app, in order to improve the accuracy of product recommendations, it is necessary to set corresponding product preference labels for each device, that is, each user. The product labels of existing apps are generally set based on the user's usage behavior data for the app, but the current product label division is not detailed enough. For example, the mother and baby label is divided into multiple age groups as sub-labels. Each sub-label covers a large age range, which affects the accurate recommendation of products. After the product labels are subdivided, the user level of some sub-labels will not meet the current market demand. Therefore, it is urgent to provide a reasonable method for expanding the user base. Summary of the invention

[0003] In response to the above technical problems, the present invention provides a method for expanding the target user group, an electronic device and a storage medium, which can improve the accuracy of matching preset product sub-labels with users while also increasing the number of users corresponding to each preset product sub-label, thereby achieving accurate and reasonable expansion of the user group.

[0004] According to a first aspect of the present invention, a method for expanding a target user group is provided, comprising the following steps:

[0005] According to the product recommendation order of the target app of each given user, the first first number of product description texts corresponding to each given user are obtained, and a number of product keywords corresponding to each product description text are extracted.

[0006] For any given user, several product keywords corresponding to the given user are matched with several preset keywords in the preset keyword library respectively, and based on the preset product sub-tags corresponding to the matched several preset keywords respectively, the target product sub-tag corresponding to the given user is determined; the preset keyword library is a library composed of several preset keywords corresponding to several preset product sub-tags respectively.

[0007] When the number of users corresponding to the target product sub-tag is less than the threshold number of users corresponding to the target product sub-tag, according to the given users corresponding to each target product sub-tag, positive samples and negative samples are divided from several product description texts corresponding to several given users to train the preset logistic regression model to obtain the target logistic regression model.

[0008] Several product keywords corresponding to each given user are input into the target logistic regression model, and the number of given users corresponding to each preset product sub-label is output, and the confidence corresponding to each given user is output at the same time.

[0009] According to the user quantity threshold corresponding to each preset product sub-label, a number of given users whose number is not less than the user quantity threshold are selected in descending order of confidence and are all determined as target users to complete the expansion of the target user group.

[0010] According to a second aspect of the present invention, a non-transitory computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the above-mentioned method for expanding the target user group.

[0011] According to a third aspect of the present invention, there is provided an electronic device comprising a processor and the above-mentioned non-transitory computer-readable storage medium.

[0012] The present invention has at least the following beneficial effects:

[0013] The present invention provides a method for expanding a target user group. First, according to the product recommendation order of a target app of a given user, a plurality of product keywords corresponding to the first number of product description texts corresponding to the given user are extracted, the plurality of product keywords corresponding to the given user are respectively matched with a plurality of preset keywords in a preset keyword library, and a target product sub-label corresponding to the given user is determined. By determining the target product sub-label, the accuracy of matching the preset product sub-label with the user can be improved. When the number of users corresponding to the target product sub-label is less than a threshold value of the number of users corresponding to the target product sub-label, a preset logistic regression model is trained. By dividing positive samples and negative samples from the product description text corresponding to the given user, keyword expansion can be achieved. The expanded keywords are learned through the preset logistic regression model, so that the obtained target logistic regression model can screen out more users for each preset product sub-label, thereby achieving reasonable and accurate expansion of the user group. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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.

[0015] Figure 1 A flow chart of a method for expanding a target user group provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0017] This embodiment provides a method for expanding the number of target user groups, such as Figure 1 As shown, the method comprises the following steps:

[0018] S100, according to the product recommendation order of the target app of each given user, obtain the first first number of product description texts corresponding to each given user, and extract several product keywords corresponding to each product description text; technical personnel in this field set the first number according to actual needs, for example, 20.

[0019] Specifically, the product description text refers to the title text corresponding to any product in the first first number of products corresponding to a given user, for example, the product title on the page after clicking on any product in a shopping software.

[0020] Furthermore, the product keyword is any keyword obtained after word segmentation processing is performed on the product description text according to a word segmentation tool.

[0021] As described above, since the products that the user is interested in will be recommended to the front based on the user's usage behavior of the target app, when evaluating the user's product sub-tags, obtaining the description texts of the first number of recommended products can make the subsequent determination of the target product sub-tags for a given user more accurate and reliable.

[0022] S200, for any given user, several product keywords corresponding to the given user are matched with several preset keywords in the preset keyword library respectively, and based on the preset product sub-tags corresponding to the matched several preset keywords respectively, the target product sub-tag corresponding to the given user is determined.

[0023] Specifically, the preset keyword library is a library composed of several preset keywords corresponding to several preset product sub-labels; it can be understood that each preset product sub-label corresponds to several preset keywords. In an implementable scenario, the maternal and child label is divided into several product sub-labels, for example, pregnancy, newborn, 1-6 months, 6-12 months, 1-2 years old, 2-3 years old, etc. The preset keywords corresponding to pregnancy are pregnant women, folic acid, delivery bag, maternity pad, fetal heart monitor, etc. The keywords corresponding to 1-2 years old are 3-stage milk powder, toddler shoes, baby picture books, etc., which will not be repeated here.

[0024] Furthermore, several preset keywords corresponding to each preset product sub-label are obtained through the following steps:

[0025] S10, obtaining the first several target products under the target label category; the target label is the upper level label of several preset product sub-labels; it can be understood that: under the target label, several preset product sub-labels are further divided, for example, the target label is the above-mentioned mother and baby label.

[0026] S20, performing word segmentation processing on the title text of each target product to obtain a number of initial keywords corresponding to each target product.

[0027] S30, selecting a number of preset keywords from all the initial keywords according to the TF-IDF value corresponding to each initial keyword and a preset TF-IDF threshold.

[0028] S40, receiving the preset product sub-tag corresponding to each preset keyword to obtain a number of preset keywords corresponding to each preset product sub-tag; it can be understood as: according to the manual judgment result, receiving the preset product sub-tag belonging to each preset keyword sent by the background user.

[0029] In a specific embodiment, the step of determining the target product sub-tag corresponding to a given user based on the preset product sub-tags corresponding to the matched plurality of preset keywords includes the following steps:

[0030] S201, according to the preset priority corresponding to each preset product sub-tag, obtain the tag priority corresponding to each preset product sub-tag in the preset product sub-tags corresponding to the matched plurality of preset keywords. In a specific implementation, for example, the younger the corresponding child age is, the higher the priority of the product sub-tag is.

[0031] S202: Determine a preset product sub-tag corresponding to the highest tag priority as a target product sub-tag corresponding to a given user.

[0032] As mentioned above, after each given user matches several preset product sub-tags, considering that at the current stage, the user will search for and purchase products for the next growth stage as a backup, but will not purchase products for the previous stage, the preset product sub-tags are prioritized according to age groups, and the preset product sub-tag with the highest priority is used as the target product sub-tag, which can improve the accuracy of matching product sub-tags with users and is conducive to subsequent accurate product recommendations to users.

[0033] In another specific embodiment, the step of determining the target product sub-tag corresponding to a given user based on the preset product sub-tags corresponding to the matched plurality of preset keywords further includes the following steps:

[0034] S210, according to the preset tag priority corresponding to each preset product sub-tag, the preset product sub-tags corresponding to the matched plurality of preset keywords are prioritized to obtain a ranking result.

[0035] S220, according to a preset order-score comparison table, obtain the order score corresponding to each preset product sub-tag in the sorting result; it can be understood that: the higher the priority corresponding to the preset product sub-tag, the higher its corresponding order score.

[0036] S230, obtaining the number of recommended products corresponding to each preset product sub-tag in the sorting result from the first number of recommended products corresponding to the given user. For example, among the first 20 recommended products for the given user, 5 products all match a preset product sub-tag, then the number of recommended products corresponding to the preset product sub-tag is 5.

[0037] S240, calculating a weighted sum based on the order score, the number of recommended products, and the preset weights corresponding to the order score and the number of recommended products, respectively, to obtain a final score corresponding to each preset product sub-tag in the sorting result.

[0038] Furthermore, the preset weight corresponding to the order score is greater than the preset weight corresponding to the number of product recommendations; since the priority corresponding to the preset product sub-tag plays a dominant role in the classification result, a higher weight is set for the order score.

[0039] S250: Determine the preset product sub-tag corresponding to the maximum final score as the target product sub-tag corresponding to the given user.

[0040] As mentioned above, since the priority corresponding to the preset product sub-label has a greater impact on the classification results, but since the number of product recommendations is also recommended based on user behavior, it also reflects the user's current product needs to a certain extent. Therefore, the target product sub-label determined by the combination of the two is more suitable for a given user, further improving the accuracy of labeling for users.

[0041] S300, when the number of users corresponding to the target product sub-tag is less than the threshold number of users corresponding to the target product sub-tag, according to the given users corresponding to each target product sub-tag, positive samples and negative samples are divided from the product description texts corresponding to the given users to train the preset logistic regression model to obtain the target logistic regression model; it can be understood that: the output layer of the preset logistic regression model is constructed according to the preset product sub-tags to achieve the purpose of multi-classification.

[0042] Specifically, in step S300, the target logistic regression model is obtained through the following steps:

[0043] S301, for any target product sub-tag, randomly select a second number of given users from the given users corresponding to the target product sub-tag, and divide the product description texts of the given users into a training set and a test set according to a preset ratio. Those skilled in the art set the second number according to actual needs, for example, for each preset product sub-tag, 320,000 pieces of data are selected.

[0044] Furthermore, the ratio of training set to test set is 9:1.

[0045] S302, product keywords in the product description text corresponding to the target product sub-tag in the training set are determined as positive samples, and the product keywords in the remaining product description texts are determined as negative samples; in the specific implementation, before determining the product keywords as positive samples and negative samples, the product keywords need to be deduplicated.

[0046] S303, input the feature vectors corresponding to the positive samples and the feature vectors corresponding to the negative samples into a preset logistic regression model for training to obtain an initial logistic regression model. For example, when the target product sub-label is pregnancy, the positive samples include cribs, baby, etc., and the negative samples include public sales bureau, teether tip, etc.

[0047] S304, inputting several product keywords corresponding to each product description text in the test set into the initial logistic regression model, and adjusting the parameters of the initial logistic regression model according to the output results to obtain a target logistic regression model. The parameter adjustment process of the model adopts the existing technology, or adjusts the parameters according to the actual needs of those skilled in the art.

[0048] As mentioned above, the target product sub-label assigned to a given user is already relatively accurate. However, in the actual implementation process, there is still a situation where the magnitude of the number of users corresponding to the preset product sub-label does not meet the business needs. Therefore, the present invention can achieve keyword expansion by dividing positive samples and negative samples from the product description text corresponding to the given user, and learn the expanded keywords through a preset logistic regression model, so that the obtained target logistic regression model can screen out more users for each preset product sub-label, thereby achieving reasonable and accurate expansion of the user group.

[0049] S400, inputting a number of product keywords corresponding to each given user into the target logistic regression model, outputting the number of given users corresponding to each preset product sub-label, and outputting the confidence corresponding to each given user at the same time. In a specific embodiment, since the target logistic regression model is a multi-classification model constructed based on a number of preset product sub-labels, it will output the preset product sub-label assigned to each given user, and at the same time output the classification confidence of each given user.

[0050] Furthermore, after outputting the given number of users corresponding to each preset product sub-label, the method further includes the following steps:

[0051] S410, when the number of given users corresponding to a preset product sub-tag is less than the threshold number of users corresponding to the preset product sub-tag, return to the execution step to divide positive samples and negative samples from a number of product description texts corresponding to all given users according to the given users corresponding to each target product sub-tag, so as to train the preset logistic regression model and obtain the target logistic regression model, until the number of given users corresponding to each preset product sub-tag outputted is not less than the threshold number of users corresponding to the preset product sub-tag.

[0052] S420, selecting a number of given users whose number is not less than the corresponding user quantity threshold in order of confidence from high to low and determining them as target users, so as to complete the expansion of the target user group.

[0053] As mentioned above, after expanding the keywords, if the number of users corresponding to the preset product sub-labels still does not meet the requirements, the keywords need to be expanded again and the logistic regression model needs to be retrained. The final number of users can be made to meet the business needs through iteration, so as to accurately recommend the product to more users.

[0054] S500, according to the user quantity threshold corresponding to each preset product sub-label, select a number of given users whose number is not less than the user quantity threshold in order of confidence from high to low and determine them as target users to complete the expansion of the target user group. Those skilled in the art set the user quantity threshold according to actual needs. For example, the user quantity threshold corresponding to the newborn label is 720,000, the user quantity threshold corresponding to the 1-6 month label is 4.32 million, and the user quantity threshold corresponding to the 1-2 year old label is 9.42 million.

[0055] In summary, the method for expanding the target user group described in the present invention first extracts a number of product keywords corresponding to the first number of product description texts corresponding to the given user according to the product recommendation order of the target app of the given user, matches the several product keywords corresponding to the given user with a number of preset keywords in the preset keyword library respectively, and determines the target product sub-tag corresponding to the given user. By determining the target product sub-tag, the accuracy of matching the preset product sub-tag with the user can be improved. When the number of users corresponding to the target product sub-tag is less than the threshold number of users corresponding to the target product sub-tag, the preset logistic regression model is trained. By dividing the positive samples and negative samples from the product description text corresponding to the given user, the keyword expansion can be achieved. The preset logistic regression model is used to learn the expanded keywords, so that the obtained target logistic regression model can screen out more users for each preset product sub-tag, thereby achieving reasonable and accurate expansion of the user group.

[0056] An embodiment of the present invention also provides a non-transitory computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a method in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided in the above embodiment.

[0057] An embodiment of the present invention further provides an electronic device, comprising a processor and the aforementioned non-transitory computer-readable storage medium.

[0058] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will also be appreciated by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A method for expanding a target user group, characterized in that: The method comprises the following steps: According to the product recommendation order of the target app of each given user, obtain the first number of product description texts corresponding to each given user, and extract a number of product keywords corresponding to each product description text; For any given user, several product keywords corresponding to the given user are matched with several preset keywords in a preset keyword library, and based on the preset product sub-tags respectively corresponding to the matched several preset keywords, a target product sub-tag corresponding to the given user is determined; the preset keyword library is a library composed of several preset keywords respectively corresponding to several preset product sub-tags; When the number of users corresponding to the target product sub-tag is less than the threshold number of users corresponding to the target product sub-tag, according to the given users corresponding to each target product sub-tag, positive samples and negative samples are divided from the product description texts corresponding to the given users to train the preset logistic regression model to obtain the target logistic regression model; Input a number of product keywords corresponding to each given user into the target logistic regression model, output the number of given users corresponding to each preset product sub-label, and at the same time output the confidence corresponding to each given user; According to the user quantity threshold corresponding to each preset product sub-label, a number of given users whose number is not less than the user quantity threshold are selected in descending order of confidence and are all determined as target users to complete the expansion of the target user group.

2. The method for expanding the target user group according to claim 1, characterized in that: Get several preset keywords corresponding to each preset product sub-tag through the following steps: Obtain the first several target products under the target tag category; the target tag is the parent tag of several preset product subtags; Perform word segmentation on the title text of each target product to obtain several initial keywords corresponding to each target product; According to the TF-IDF value corresponding to each initial keyword and the preset TF-IDF threshold, a number of preset keywords are selected from all the initial keywords; The preset product sub-tag corresponding to each preset keyword is received to obtain a plurality of preset keywords corresponding to each preset product sub-tag.

3. The method for expanding the target user group according to claim 1, characterized in that: The method of determining the target product sub-tag corresponding to a given user based on the preset product sub-tags corresponding to the matched plurality of preset keywords comprises the following steps: According to the preset priority corresponding to each preset product sub-tag, obtaining the tag priority corresponding to each preset product sub-tag among the preset product sub-tags corresponding to the matched plurality of preset keywords; A preset product sub-tag corresponding to the highest tag priority is determined as a target product sub-tag corresponding to a given user.

4. The method for expanding the target user group according to claim 1, characterized in that: The step of determining the target product sub-tag corresponding to the given user based on the preset product sub-tags corresponding to the matched plurality of preset keywords respectively includes the following steps: According to the preset tag priority corresponding to each preset product sub-tag, the preset product sub-tags corresponding to the matched plurality of preset keywords are prioritized to obtain a ranking result; According to a preset order-score comparison table, obtain the order score corresponding to each preset product sub-label in the sorting result; Obtaining the number of recommended products corresponding to each preset product sub-label in the sorting result from the first first number of recommended products corresponding to the given user; A weighted sum is calculated based on the order score, the number of recommended products, and the preset weights corresponding to the order score and the number of recommended products, to obtain a final score corresponding to each preset product sub-label in the ranking result; The preset product sub-tag corresponding to the maximum final score is determined as the target product sub-tag corresponding to the given user.

5. The method for expanding the target user group according to claim 4, characterized in that: The preset weight corresponding to the order score is greater than the preset weight corresponding to the number of product recommendations.

6. The method for expanding the target user group according to claim 1, characterized in that: The target logistic regression model is obtained through the following steps: For any target product sub-tag, randomly select a second number of given users from the given users corresponding to the target product sub-tag, and divide the product description texts of the given users into a training set and a test set according to a preset ratio; The product keywords of the product description text corresponding to the target product sub-label in the training set are all determined as positive samples, and the product keywords of the remaining product description texts are all determined as negative samples; Input the feature vector corresponding to the positive sample and the feature vector corresponding to the negative sample into the preset logistic regression model for training to obtain an initial logistic regression model; Several product keywords corresponding to each product description text in the test set are input into the initial logistic regression model, and the parameters of the initial logistic regression model are adjusted according to the output results to obtain the target logistic regression model.

7. The method for expanding the target user group according to claim 1, characterized in that: After outputting the given number of users corresponding to each preset product sub-label, the method further includes the following steps: When the number of given users corresponding to the preset product sub-tag is less than the threshold number of users corresponding to the preset product sub-tag, return to the step of dividing positive samples and negative samples from a number of product description texts corresponding to all given users according to the given users corresponding to each target product sub-tag, so as to train the preset logistic regression model and obtain the target logistic regression model, until the number of given users corresponding to each preset product sub-tag is not less than the threshold number of users corresponding to the preset product sub-tag; A number of given users whose number is not less than the corresponding user quantity threshold are selected in order of confidence from high to low and are all determined as target users to complete the expansion of the target user group.

8. A non-transitory computer-readable storage medium, wherein at least one instruction or at least one program is stored in the storage medium, characterized in that: The at least one instruction or the at least one program is loaded and executed by the processor to implement the method for expanding the target user group as described in any one of claims 1-7.

9. An electronic device, characterized in that: The invention comprises a processor and the non-transitory computer-readable storage medium as claimed in claim 8.

Images