User portrait generation method and system for Internet advertisement putting

Abstract

The invention relates to the technical field of data processing, and provides a user portrait generation method and system for Internet advertisement delivery, and the method comprises the steps: collecting the behavior data of a user, and obtaining a user portrait, the operation data of a label in the user portrait, a conversion weight, and a single-day conversion weight; marking the collection natural day of the operation data, obtaining the adjacent day duration and the operation data ratio of the label, and obtaining the stability of the label in combination with the conversion weight of the label and the adjacent day duration; according to the difference between the single-day conversion weights of the labels in all the natural days and the difference between the operation data ratios of the same labels, obtaining association indexes of the labels of the user; and obtaining an updated user portrait according to the difference between the single-day conversion weights of the same labels of all the users on the same natural day, the association indexes of the labels of the users and the stability of the labels. According to the invention, the accuracy of the updated user portrait can be improved, and the Internet advertisement putting effect is ensured.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for generating user portraits for Internet advertising delivery. Background Art

[0002] The accuracy of Internet advertising depends on the accurate construction of user portraits, which is generally achieved using user portrait generation technology. User portrait generation technology can accurately capture and analyze user interests, behaviors, demographic attributes and other characteristics through user related information and behavior data, achieve target audience positioning, improve advertising delivery effects, increase advertising click-through rates and conversion rates, and reduce the cost of ineffective delivery. Due to the highly dynamic characteristics of user attributes and behaviors, it is necessary to update and optimize user portraits in a timely manner based on user related information and behavior data, improve the accuracy of user tag positioning, and thus ensure the accuracy of Internet advertising.

[0003] In the process of updating and optimizing user portraits based on users' relevant information and behavioral data, there are differences in the effectiveness of different types of user data, and the degree of influence of different types of user data on users' various tags is also different. Therefore, directly extracting and updating user portraits based on users' relevant information and behavioral data within a time period will result in inaccurate updated user portraits and delayed updates, resulting in poor results for Internet advertising. Summary of the invention

[0004] The present invention provides a method and system for generating a user portrait for Internet advertising delivery, so as to solve the problem that the updated user portrait is inaccurate due to the different changes in the attributes of the user's behavior data and the different association methods, resulting in poor Internet advertising delivery effects. The technical solutions adopted are as follows: In a first aspect, an embodiment of the present invention provides a method for generating a user portrait for Internet advertising delivery, the method comprising the following steps: Collect user behavior data and obtain user portraits. User portraits include user tags. Based on the correspondence between behavior data and tags, obtain tag operation data, conversion weights, and daily conversion weights. The natural day corresponding to the generation time of the tag's operation data is recorded as the collection natural day of the tag's operation data. The continuous degree of the tag on adjacent days is obtained based on the difference between the single-day conversion weights of the tag on adjacent collection natural days. The operation data ratio of the tag is obtained based on the number of types of the tag's operation data and the number of types of all operation data. The stability of the tag is obtained by combining the conversion weight of the tag and the continuous degree of adjacent days. Obtain the user's tag relevance index based on the difference between the daily conversion weights of tags on all natural days and the difference between the operation data ratios of the same tags; According to the difference between the daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tags and the stability of the tags, an updated user portrait is obtained.

[0005] Furthermore, the specific method of obtaining the operation data, conversion weight and daily conversion weight of the tag includes: The behavior data corresponding to the same tag is recorded as the tag operation data; The number of users who click on the ads and make purchases in the ads placed by the platform for the tag is recorded as the conversion number of the tag, the number of all users with the same tag is recorded as the potential conversion number of the tag, and the ratio of the conversion number of the tag to the potential conversion number is recorded as the conversion weight of the tag; For each natural day, the ratio of the number of conversions of the label in the natural day to the number of potential conversions of the label in the same natural day is recorded as the single-day conversion weight of the label in the natural day.

[0006] Furthermore, the specific method for obtaining the continuation of the tag on adjacent days is as follows: Take any collection natural day as the target collection natural day; For the same label, the square of the difference between the label's daily conversion weight on the target collection day and the collection day next to the target collection day is recorded as the label's daily conversion weight difference on the target collection day; The continuous degree of the tag on adjacent days is determined according to the difference in the daily conversion weights of the tag on all the collected natural days. The continuous degree of the tag on adjacent days is negatively correlated with the difference in the daily conversion weights of the tag on all the collected natural days.

[0007] Furthermore, the specific method of obtaining the operation data ratio of the tag is as follows: For any tag of a user included in the user profile, the ratio of the number of types of operation data of the tag to the number of types of all operation data is recorded as the operation data ratio of the tag.

[0008] Furthermore, the stability of the label is obtained by: The normalized value of the product of the tag's operation data ratio, the tag's conversion weight, and the duration of adjacent days is recorded as the tag's stability.

[0009] Furthermore, the method of obtaining the association index of the user's tag according to the difference between the daily conversion weights of the tag on all natural days and the difference between the operation data ratios of the same tag includes the following specific methods: The normalized value of the standard deviation of the daily conversion weight of the label on all natural days is recorded as the daily conversion weight difference of the label, and the difference between the second preset threshold and the daily conversion weight difference of the label is recorded as the daily conversion weight stability of the label; Any user is recorded as a target user, the average of the operation data ratios of the same label of all other users except the target user is recorded as the standard operation data ratio of the target user, the absolute value of the difference between the operation data ratio of the label of the target user and the standard operation data ratio is recorded as the operation data ratio difference of the label of the target user, and the difference between the second preset threshold and the operation data ratio difference of the label is recorded as the operation data ratio stability of the label of the target user; The association index of the target user's tag is determined based on the stability of the target user's tag's daily conversion weight and the stability of the operation data ratio.

[0010] Furthermore, the method of determining the relevance index of the target user's tag according to the stability of the daily conversion weight and the stability of the operation data ratio of the target user's tag includes the following specific methods: The average of the daily conversion weight stability and the operation data ratio stability of the target user's label is recorded as the association index of the target user's label.

[0011] Furthermore, the method of obtaining an updated user portrait according to the difference between the daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tags and the stability of the tags includes the following specific methods: The average of the daily conversion weights of the same label of all other users except the target user on the same natural day is recorded as the standard daily conversion weight of the label on the natural day; the difference between the daily conversion weight of the target user's label on the natural day and the standard daily conversion weight is recorded as the daily conversion weight deviation of the target user's label on the natural day; the product of the target user's label association index, the daily conversion weight deviation of the user's label on the natural day, and the stability of the label is recorded as the first product of the target user's label on the natural day; the sum of the daily conversion weight of the target user's label on the natural day and the first product of the user's label on the natural day is recorded as the update weight of the target user's label on the natural day; Get updated user portraits based on the update weights of user tags on natural days.

[0012] Furthermore, the updated user portrait is obtained according to the updated weight of the user's tag on a natural day, including the following specific methods: When the update weight of the tag on a natural day is greater than or equal to a third preset threshold, the user's tag is continued to be retained; When the update weight of the tag on a natural day is less than a third preset threshold, deleting the user's tag; All remaining tags of the user are used as the updated user profile.

[0013] In a second aspect, an embodiment of the present invention further provides a user portrait generation system for Internet advertising delivery, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0014] The beneficial effects of the present invention are: The present invention determines the conversion weight and daily conversion weight of the user tag according to the characteristics determined by the joint influence of multiple types of behavioral data on each tag of the user, according to the behavior of attracting users to purchase through advertisements placed on the Internet, and analyzes the stability of the operation data of the tag according to the characteristics that the effectiveness of the data types corresponding to different behavioral data is different during the updating of the user portrait of the user, and determines the stability of the tag according to the differences in the behaviors of attracting users to purchase through advertisements on adjacent natural days of collection and the extensiveness of the distribution of the types of operation data corresponding to the tag. The stability of the tag is an evaluation of the comprehensiveness, objectivity and reliability of the tag's evaluation of user characteristics; further, according to the association between other users and users corresponding to each user's tag, the association index of the user's tag is obtained; finally, according to the difference between the daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tag and the stability of the tag, an updated user portrait is obtained, so as to solve the problem that the updated user portrait is inaccurate due to the changes in the attributes of the user's behavioral data and the different association methods, resulting in poor Internet advertising delivery effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative labor.

[0016] Figure 1 A schematic diagram of a flow chart of a method for generating a user portrait for Internet advertising delivery provided by an embodiment of the present invention; Figure 2 A stability acquisition flow chart provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 , which shows a flow chart of a method for generating a user portrait for Internet advertising delivery provided by an embodiment of the present invention, the method comprising the following steps: Step S001, collect the user's behavior data and obtain the user's user portrait. The user portrait includes the user's label. According to the correspondence between the behavior data and the label, obtain the label's operation data, conversion weight and single-day conversion weight.

[0019] Extract all the user's behavior data on the platform, use Flink to process all the user's behavior data, and update the user's user profile in real time. The user profile contains multiple tags of the user.

[0020] Among them, Apache Flink is a stream processing framework that can update user profiles in real time. Using Flink to update user profiles in real time is a well-known technology and will not be described in detail.

[0021] It can be understood that user tags include user attribute tags, user behavior tags, user behavior tags, etc. For example, user attribute tags include the user's age, such as 18-25 years old, 26-35 years old, etc., user behavior tags include the user's order-related behavior, such as frequent purchases, occasional purchases, etc., and user behavior tags include the user's consumption data, such as purchasing cotton clothes.

[0022] Preferably, as an embodiment of the present application, the behavior data includes basic information, consumption behavior data, behavior trajectory data, social interaction data and device usage data, wherein the basic information includes age, gender, occupation and educational background, the consumption behavior data includes purchase history, purchase frequency, average consumption amount and consumption preference, the behavior trajectory data includes browsing history, click behavior, search keywords and page dwell time, the social interaction data includes social media activities, posting content, interaction frequency and social network information, and the device usage data includes permanent residence, work place and travel history; all the user's behavior data are processed and stored using Hadoop; this embodiment sets the collection time period length of all the user's behavior data to 6 months. Among them, Hadoop is an open source framework that allows data storage and processing across distributed computing environments. Using Hadoop to process and store behavior data is a well-known technology and will not be repeated.

[0023] In the actual application process, as another implementation method, Spark can also be used to process and store all the user's behavior data, and this application does not make special restrictions. Spark is a fast and universal open source cluster computing system and data processing platform. Using Spark to process and store behavior data is a well-known technology and will not be repeated. In the actual application process, as another implementation method, the data type of the appropriate behavior data can also be selected according to the needs, and this application does not make special restrictions.

[0024] It is understandable that each user's tag is affected by multiple types of behavioral data, and different tags are not isolated, but are determined by the joint influence of multiple types of behavioral data. Different behavioral data have different enhancing or weakening effects on different tags. Advertisements placed on the Internet are intended to attract users who browse the ads to browse the ads and pay for the products recommended by the ads. Therefore, it is necessary to determine the possibility that each tag's corresponding advertisement can stimulate user purchases based on the ads corresponding to the tags.

[0025] The behavior data corresponding to the same tag is recorded as the tag operation data. The number of users who click on the ads and make purchases in the ads placed by the platform for the tag is recorded as the tag conversion number, the number of all users with the same tag is recorded as the tag potential conversion number, and the ratio of the tag conversion number to the potential conversion number is recorded as the tag conversion weight.

[0026] For each natural day, the ratio of the number of conversions of the label in the natural day to the number of potential conversions of the label in the same natural day is recorded as the single-day conversion weight of the label in the natural day.

[0027] It is understandable that different users may have the same tags. For example, the consumption data of two different users in the user behavior tags are both more than 10,000 yuan for the day. However, the operation data used by two different users to obtain the same tag may be different. For example, one user bought 20,000 yuan of furniture, and the other user bought 15,000 yuan of computers. Therefore, the operation data corresponding to the same tags of different users may be different, and can be different. Furthermore, when calculating the daily conversion weight of a tag on a natural day, the potential conversion number of the tag within the natural day and the potential conversion number of the tag within the same natural day are required. Therefore, the daily conversion weight of the same tag of different users on the same natural day may be different.

[0028] At this point, the tag's operation data, conversion weight, and daily conversion weight are obtained.

[0029] Step S002, record the natural day corresponding to the generation time of the label's operation data as the collection natural day of the label's operation data, obtain the label's adjacent day continuity based on the difference between the label's single-day conversion weights on adjacent collection natural days, obtain the label's operation data ratio based on the number of types of the label's operation data and the number of types of all operation data, and obtain the label's stability by combining the label's conversion weight and adjacent day continuity.

[0030] In the process of updating the user's user profile, the effectiveness of the data types corresponding to different behavioral data varies. Behavioral data with high stability will remain unchanged or change slowly for a long time, but behavioral data with low stability will change greatly. Therefore, it is necessary to analyze the stability of the tag's operation data.

[0031] The natural diary corresponding to the generation time of the tag operation data is the collection natural day of the operation data, and any collection natural day is used as the target collection natural day. This embodiment takes any collection natural day within 6 months of collection as an example for explanation. It can be understood that the latest collection natural day within 6 months of collection does not have a subsequent collection natural day and is not analyzed.

[0032] The continuous degree of a tag on consecutive days is obtained based on the difference between the daily conversion weights of the tag on consecutive natural days of collection.

[0033] For the same label, the square of the difference between the label's daily conversion weight on the target collection day and the next adjacent collection day after the target collection day is recorded as the label's daily conversion weight difference on the target collection day.

[0034] The continuous degree of the tag on adjacent days is determined according to the difference in the daily conversion weights of the tag on all the collected natural days. The continuous degree of the tag on adjacent days is negatively correlated with the difference in the daily conversion weights of the tag on all the collected natural days.

[0035] Preferably, as an embodiment of the present application, the sum of the average of the daily conversion weight differences of the label on all collected natural days and the first preset threshold is used as the denominator, and the number 1 is used as the fractional value of the numerator, which is recorded as the adjacent day continuity of the label.

[0036] In this embodiment, the value of the first preset threshold is 0.01, and the function of the first preset threshold is to prevent the denominator from being zero.

[0037] In actual application, as another implementation method, the exponential power of the natural constant is used as the base and the opposite of the average of the daily conversion weight difference of the label on all collected natural days is used as the value of the exponential power, which is recorded as the adjacent day continuity of the label.

[0038] The greater the difference in the daily conversion weights between two adjacent natural days of collection, the greater the difference between the operation data of the labels in different natural days of collection, the weaker the stability of the operation data corresponding to the labels, the weaker the comprehensiveness, objectivity and reliability of the labels in evaluating user characteristics, and at this time, the smaller the continuity of the labels on adjacent days.

[0039] The stability of the label is obtained based on the label's conversion weight, the duration of adjacent days, the number of types of operation data, and the number of types of all operation data.

[0040] Preferably, as an embodiment of the present application, for any label of the user included in the user portrait, the ratio of the number of types of operation data of the label to the number of types of all operation data is recorded as the operation data ratio of the label, and the normalized value of the operation data ratio of the label and the product of the conversion weight of the label and the continuity of adjacent days is recorded as the stability of the label.

[0041] It should be noted that this embodiment uses the Z-Score standard normalization method to calculate the normalized value. In actual application, the implementer may use other methods in the prior art such as the maximum and minimum value normalization method to calculate the normalized value, which is not limited here.

[0042] When the operation data ratio of the label is larger, the more types of data affect the operation data of the label, and the more operation data reflects the user characteristic evaluation of the label. At the same time, when the conversion weight and adjacent day continuity of the label are larger, the comprehensiveness, objectivity and reliability of the label for user characteristic evaluation are more significant, and the significance of the label needing to continue to be retained is greater. At this time, the stability of the label is greater.

[0043] It is understandable that different users’ labels may have the same situation. For example, the consumption data of the users in the user behavior labels of two different users both show that the annual consumption amount exceeds 50,000 yuan. However, the operation data used by two different users to obtain the same label may be different. For example, one user purchased furniture worth 60,000 yuan, and the other user purchased computers and other electronic equipment worth 55,000 yuan. Therefore, the types of operation data corresponding to the same labels of different users may be different, and can be different. In other words, the number of types of operation data corresponding to the same labels of different users may also be different. Furthermore, when calculating the operation data ratio of the label, the number of types of operation data of the label is required, so the operation data ratio of the labels of different users may be different.

[0044] At this point, the stability of each tag is obtained. The stability acquisition flow chart is as follows: Figure 2 shown.

[0045] Step S003, obtaining the relevance index of the user's tag according to the difference between the daily conversion weights of the tags on all natural days and the difference between the operation data ratios of the same tags.

[0046] Furthermore, according to the correlation between each user's tag and other users corresponding to the user, the user's tags are further analyzed.

[0047] The association index of the user's tag is obtained based on the difference between the daily conversion weights of the tags on all natural days and the difference between the operation data ratios of the same tags.

[0048] Preferably, as an embodiment of the present application, for any tag of a user, the normalized value of the standard deviation of the daily conversion weight of the tag on all natural days is recorded as the daily conversion weight difference of the tag, and the difference between the second preset threshold and the daily conversion weight difference of the tag is recorded as the daily conversion weight stability of the tag. Any user is recorded as the target user, the mean of the operation data ratios of the same tag of all other users except the target user is recorded as the standard operation data ratio of the target user, the absolute value of the difference between the operation data ratio of the target user's tag and the standard operation data ratio is recorded as the operation data ratio difference of the target user's tag, and the difference between the second preset threshold and the operation data ratio difference of the tag is recorded as the operation data ratio stability of the target user's tag. The mean of the daily conversion weight stability of the target user's tag and the operation data ratio stability is recorded as the association index of the target user's tag.

[0049] In this embodiment, the value of the second preset threshold is 1.

[0050] Generally, different users with the same tag show the same features corresponding to the tag, so there is similarity between the operation data of different users corresponding to the same tag. When the difference between the daily conversion weights of tags on all natural days is smaller and the difference between the operation data ratios of the same tag is smaller, the correlation between other users with the same tag as the user is more obvious, and at this time, the correlation index of the user's tag is greater.

[0051] At this point, the relevance index of the user's tag is obtained.

[0052] Step S004, obtaining an updated user portrait based on the difference between the daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tag and the stability of the tag.

[0053] The update weight of a user's tag on a natural day is determined based on the difference between the daily conversion weights of the same tag of all users on the same natural day, the association index of the user's tag and the stability of the tag.

[0054] Preferably, as an embodiment of the present application, the average of the daily conversion weights of the same label of all other users except the target user on the same natural day is recorded as the standard daily conversion weight of the label on the natural day, the difference between the daily conversion weight of the target user's label on the natural day and the standard daily conversion weight is recorded as the daily conversion weight deviation of the target user's label on the natural day, the product of the association index of the target user's label, the daily conversion weight deviation of the user's label on the natural day and the stability of the label is recorded as the first product of the target user's label on the natural day, and the sum of the daily conversion weight of the target user's label on the natural day and the first product of the user's label on the natural day is recorded as the update weight of the target user's label on the natural day.

[0055] When the update weight of the tag on a natural day is greater than or equal to a third preset threshold, the user's tag is retained; when the update weight of the tag on a natural day is less than the third preset threshold, the user's tag is deleted.

[0056] In this embodiment, the value of the third preset threshold is 1.

[0057] All tags of the deleted user are used as the updated user portrait, and the user portrait generation result for Internet advertising is obtained.

[0058] Based on the same inventive concept as the above method, an embodiment of the present invention also provides a user portrait generation system for Internet advertising delivery, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above-mentioned user portrait generation methods for Internet advertising delivery are implemented.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating a user portrait for Internet advertising, characterized in that: The method comprises the following steps: Collect user behavior data and obtain user portraits. User portraits include user tags. Based on the correspondence between behavior data and tags, obtain tag operation data, conversion weights, and daily conversion weights. The natural day corresponding to the generation time of the tag's operation data is recorded as the collection natural day of the tag's operation data. The continuous degree of the tag on adjacent days is obtained based on the difference between the single-day conversion weights of the tag on adjacent collection natural days. The operation data ratio of the tag is obtained based on the number of types of the tag's operation data and the number of types of all operation data. The stability of the tag is obtained by combining the conversion weight of the tag and the continuous degree of adjacent days. Obtain the user's tag relevance index based on the difference between the daily conversion weights of tags on all natural days and the difference between the operation data ratios of the same tags; An updated user portrait is obtained based on the difference in daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tags, and the stability of the tags.

2. The method for generating a user portrait for Internet advertising delivery according to claim 1, characterized in that: The specific method of obtaining the operation data, conversion weight and daily conversion weight of the tag includes: The behavior data corresponding to the same tag is recorded as the tag operation data; The number of users who click on the ads and make purchases in the ads placed by the platform for the tag is recorded as the conversion number of the tag, the number of all users with the same tag is recorded as the potential conversion number of the tag, and the ratio of the conversion number of the tag to the potential conversion number is recorded as the conversion weight of the tag; For each natural day, the ratio of the number of conversions of the label in the natural day to the number of potential conversions of the label in the same natural day is recorded as the single-day conversion weight of the label in the natural day.

3. The method for generating a user portrait for Internet advertising delivery according to claim 1, characterized in that: The specific method for obtaining the continuation of the tag on adjacent days is as follows: Take any collection natural day as the target collection natural day; For the same label, the square of the difference between the label's daily conversion weight on the target collection day and the collection day next to the target collection day is recorded as the label's daily conversion weight difference on the target collection day; The continuous degree of the tag on adjacent days is determined according to the difference in the daily conversion weights of the tag on all the collected natural days. The continuous degree of the tag on adjacent days is negatively correlated with the difference in the daily conversion weights of the tag on all the collected natural days.

4. The method for generating a user portrait for Internet advertising delivery according to claim 1, characterized in that: The specific method of obtaining the operation data ratio of the tag is as follows: For any tag of a user included in the user profile, the ratio of the number of types of operation data of the tag to the number of types of all operation data is recorded as the operation data ratio of the tag.

5. The method for generating a user portrait for Internet advertising delivery according to claim 1, characterized in that: The specific method for obtaining the stability of the label is as follows: The normalized value of the product of the tag's operation data ratio, the tag's conversion weight, and the duration of adjacent days is recorded as the tag's stability.

6. The method for generating a user portrait for Internet advertising delivery according to claim 1, characterized in that: The specific method of obtaining the association index of the user's tag according to the difference between the daily conversion weights of the tags on all natural days and the difference between the operation data ratios of the same tags is as follows: The normalized value of the standard deviation of the daily conversion weight of the label on all natural days is recorded as the daily conversion weight difference of the label, and the difference between the second preset threshold and the daily conversion weight difference of the label is recorded as the daily conversion weight stability of the label; Any user is recorded as a target user, the average of the operation data ratios of the same label of all other users except the target user is recorded as the standard operation data ratio of the target user, the absolute value of the difference between the operation data ratio of the label of the target user and the standard operation data ratio is recorded as the operation data ratio difference of the label of the target user, and the difference between the second preset threshold and the operation data ratio difference of the label is recorded as the operation data ratio stability of the label of the target user; The association index of the target user's tag is determined based on the stability of the target user's tag's daily conversion weight and the stability of the operation data ratio.

7. The method for generating a user portrait for Internet advertising delivery according to claim 6, characterized in that: The specific method of determining the relevance index of the target user's tag according to the stability of the daily conversion weight and the stability of the operation data ratio of the target user's tag is as follows: The average of the daily conversion weight stability and the operation data ratio stability of the target user's label is recorded as the association index of the target user's label.

8. The method for generating a user portrait for Internet advertising delivery according to claim 7, characterized in that: The method of obtaining an updated user portrait based on the difference between the daily conversion weights of the same tags of all users on the same natural day, the association index of the user's tags and the stability of the tags includes: The average of the daily conversion weights of the same label of all other users except the target user on the same natural day is recorded as the standard daily conversion weight of the label on the natural day; the difference between the daily conversion weight of the target user's label on the natural day and the standard daily conversion weight is recorded as the daily conversion weight deviation of the target user's label on the natural day; the product of the target user's label association index, the daily conversion weight deviation of the user's label on the natural day, and the stability of the label is recorded as the first product of the target user's label on the natural day; the sum of the daily conversion weight of the target user's label on the natural day and the first product of the user's label on the natural day is recorded as the update weight of the target user's label on the natural day; Get updated user portraits based on the update weights of user tags on natural days.

9. The method for generating a user portrait for Internet advertising delivery according to claim 8, characterized in that: The specific method of obtaining the updated user portrait according to the update weight of the user's tag on a natural day includes: When the update weight of the tag on a natural day is greater than or equal to a third preset threshold, the user's tag is continued to be retained; When the update weight of the tag on a natural day is less than a third preset threshold, deleting the user's tag; All remaining tags of the user are used as the updated user profile.

10. A user portrait generation system for Internet advertising delivery, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images