User intelligent recommendation system based on information flow delivery

Abstract

The invention belongs to the field of advertisement pushing, relates to a data analysis technology, and provides an intelligent user recommendation system based on information flow putting, which can perform pushing analysis by adopting multiple modes and dynamically switch recommendation modes in combination with a pushing effect, in particular to an intelligent user recommendation system based on information flow putting. Comprising a basic recommendation module, a conversion evaluation module and a recommendation optimization module, and the basic recommendation module, the conversion evaluation module and the recommendation optimization module are in communication connection in sequence; the basic recommendation module is used for carrying out basic recommendation analysis on the advertisement putting users: screening a batch of platform users and marking the platform users as analysis objects, and carrying out primary screening analysis on the analysis objects by adopting a basic recommendation mode to obtain a basic sequence; according to the method, basic recommendation analysis can be carried out on the advertisement putting user, the basic recommendation priority of the analysis object is evaluated according to the basic coefficient, the basic recommendation object is screened out for primary pushing, and the conversion rate of primary pushing is ensured.

Description

Technical Field

[0001] The present invention belongs to the field of advertisement push, and relates to data analysis technology, and specifically to a user intelligent recommendation system based on information flow delivery. Background Art

[0002] The user intelligent recommendation method based on information flow delivery mainly relies on big data and artificial intelligence technology. By collecting and analyzing multi-dimensional data of users, such as browsing history, search history, behavioral data, etc., it can accurately recommend advertising content that better meets the needs of users. These methods not only improve the click-through rate and conversion rate of advertisements, but also significantly improve the user experience.

[0003] The user intelligent recommendation system in the prior art can only use a single recommendation method to push user screening, but cannot use multiple modes for push analysis and dynamically switch the recommendation mode based on the push effect, resulting in the inability to guarantee the advertising push effect and conversion rate.

[0004] In view of the above technical problems, this application proposes a solution. Summary of the invention

[0005] The purpose of the present invention is to provide a user intelligent recommendation system based on information flow delivery, which is used to solve the problem that the prior art cannot use multiple modes for push analysis and dynamically switch the recommendation mode based on the push effect;

[0006] The technical problem to be solved by the present invention is: how to provide a user intelligent recommendation system based on information flow delivery that can adopt multiple modes for push analysis and dynamically switch the recommendation mode based on the push effect.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] The user intelligent recommendation system based on information flow delivery includes a basic recommendation module, a conversion evaluation module and a recommendation optimization module, wherein the basic recommendation module, the conversion evaluation module and the recommendation optimization module are sequentially connected in communication;

[0009] The basic recommendation module is used to perform basic recommendation analysis on advertising delivery users: randomly screen a batch of platform users and mark them as analysis objects, use the basic recommendation mode to perform initial screening analysis on the analysis objects and obtain a basic sequence; intercept L2 analysis objects from the basic sequence and mark them as basic recommendation objects, and push advertisements to the clients of the basic recommendation objects;

[0010] The conversion evaluation module is used to evaluate and analyze the advertising push conversion rate of the basic recommendation objects: after completing the advertising push to the client of the basic recommendation objects, a conversion cycle with a duration of L3 hours is generated. At the end moment of the conversion cycle, the overall conversion rate of the basic recommendation objects is obtained and marked as the conversion value. The conversion value is compared with the preset conversion threshold: if the conversion value is less than the conversion threshold, it is determined that the conversion effect of the basic recommendation objects does not meet the requirements, a re-screening signal is generated and sent to the basic recommendation module. After receiving the re-screening signal, the basic recommendation module re-screens a batch of platform users from the database as the analysis objects; if the conversion value is greater than or equal to the conversion threshold, it is determined that the conversion effect of the basic recommendation objects meets the requirements, and in-depth screening and analysis are performed on the analysis objects.

[0011] The recommendation optimization module is used to optimize and analyze the advertising push method.

[0012] Further, the specific process of the initial screening and analysis of the analysis objects using the basic recommendation mode includes: obtaining the browsing data LL, search data SS, and favorite data SC of the analysis objects and performing numerical calculations to obtain the basic coefficient JC of the analysis objects; arranging all the analysis objects in descending order of the numerical value of the basic coefficient JC to obtain the basic sequence.

[0013] Further, the process of obtaining the browsing data LL includes: obtaining the type of the product corresponding to the advertisement and marking it as the reference type, marking all the products corresponding to the reference type as the reference products, and obtaining the number of times the analysis objects browse the reference products within the most recent L1 months and marking it as the browsing data LL; the process of obtaining the search data SS includes: obtaining all the keywords corresponding to the product corresponding to the advertisement during retrieval and marking them as the reference words, and marking the number of times the analysis objects enter the reference words into the search box within L1 months as the search data SS; the process of obtaining the favorite data SC includes: obtaining the shopping cart product information of the analysis objects, and marking the number of the reference products in the shopping cart of the analysis objects as the favorite data SC.

[0014] Further, the specific process of the in-depth screening and analysis of the analysis objects includes: randomly selecting a screening method to perform in-depth screening on the remaining analysis objects to obtain the in-depth recommendation objects, performing advertising push to the clients of the in-depth recommendation objects, and then generating a conversion cycle with a duration of L3 hours again. At the end moment of the conversion cycle, the conversion effect of the in-depth recommendation objects is evaluated again: if the conversion effect meets the requirements, repeat the in-depth screening and analysis of the analysis objects until the conversion effect does not meet the requirements; if the conversion effect does not meet the requirements, generate a re-screening signal and send the re-screening signal to the basic recommendation module.

[0015] Further, the screening methods include linear screening and portrait screening. The specific process of using linear screening to screen deep recommendation objects is as follows: Remove the analysis objects for which the advertisement push has been completed from the basic sequence, and mark the top L2 analysis objects in the basic sequence after removal as deep recommendation objects.

[0016] Further, the specific process of using portrait screening to screen deep recommendation objects is as follows: Mark the analysis objects that have completed conversion in the previous advertisement push as feature objects, obtain the basic feature information of the feature objects. The basic feature information includes age, education level, income, and location information. Generate the corresponding portrait comparison range according to the basic feature information. Mark the analysis objects whose basic feature information meets the portrait comparison range and for which the advertisement push has not been completed as candidate objects. Arrange the candidate objects in descending order according to the value of the basic coefficient JC to obtain a portrait sequence, and mark the top L2 candidate objects in the portrait sequence as deep recommendation objects.

[0017] Further, the specific process of the recommendation optimization module for optimizing the advertisement push method is as follows: When the number of analysis objects for which the advertisement push has been completed reaches M1, retrieve the conversion value corresponding to each advertisement push process. The conversion range is composed of the minimum value and the maximum value of the conversion value. Divide the conversion range into several conversion intervals and mark the priority methods for the conversion intervals. When performing a deep screening analysis on the analysis objects later, first obtain the conversion value of the previous advertisement push process, retrieve the priority method of the conversion interval to which the conversion value belongs, and directly use the priority method to screen the deep recommendation objects.

[0018] Further, the process of obtaining the priority method for the conversion interval is as follows: Mark the advertisement push process in which the conversion value is within the conversion interval as the matching process for the conversion interval. If the next advertisement push process of the matching process uses linear screening, mark the matching process as a linear optimization process; if the next advertisement push process of the matching process uses portrait screening, mark the matching process as a portrait optimization process; Sum and average the conversion values corresponding to the next advertisement push processes of all linear optimization processes within the conversion interval to obtain a linear priority value, and sum and average the conversion values corresponding to the next advertisement push processes of all portrait optimization processes within the conversion interval to obtain a portrait priority value; Compare the linear priority value with the portrait priority value: If the linear priority value is greater than or equal to the portrait priority value, mark linear screening as the priority method for the conversion interval; if the linear priority value is less than the portrait priority value, mark portrait screening as the priority method for the conversion interval.

[0019] The present invention has the following beneficial effects:

[0020] 1. Through the basic recommendation module, basic recommendation analysis can be performed on advertising users. The push tendency data of the analysis object is collected and calculated to obtain the basic coefficient. According to the basic coefficient, the basic recommendation priority of the analysis object is evaluated, and the basic recommendation objects are screened out for the initial push to ensure the conversion rate of the initial push.

[0021] 2. The conversion evaluation module is used to evaluate and analyze the advertising push conversion rate of the basic recommendation objects. According to the conversion evaluation results, processing decision analysis is carried out. When the conversion evaluation effect is qualified, different methods are used for in-depth screening analysis to provide data support for the recommendation optimization process.

[0022] 3. Through the recommendation optimization module, the advertising push method can be optimized and analyzed. The priority methods for each conversion interval are marked. When in-depth screening analysis is carried out later, the corresponding optimization methods can be directly used for in-depth object screening to ensure the conversion effect of each advertising push. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is the system block diagram of Embodiment 1 of the present invention;

[0025] Figure 2 It is the method flow chart of Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0027] Embodiment 1: As Figure 1 shown, the user intelligent recommendation system based on information flow delivery includes a basic recommendation module, a conversion evaluation module, and a recommendation optimization module. The basic recommendation module, the conversion evaluation module, and the recommendation optimization module are sequentially connected for communication.

[0028] The basic recommendation module is used to conduct basic recommendation analysis on advertising users: randomly select a batch of platform users and mark them as analysis objects, and use the basic recommendation mode to conduct initial screening analysis on the analysis objects: obtain the browsing data LL, search data SS, and favorite data SC of the analysis objects.

[0029] The process of obtaining the browsing data LL includes: obtaining the type of the product corresponding to the advertisement and marking it as the benchmark type, marking all products corresponding to the benchmark type as benchmark products, and obtaining the number of times the analysis object browses the benchmark products within the most recent L1 months and marking it as the browsing data LL;

[0030] The process of obtaining the search data SS includes: obtaining all keywords corresponding to the product corresponding to the advertisement during retrieval and marking them as benchmark words, and marking the number of times the analysis object enters the benchmark words into the search box within L1 months as the search data SS;

[0031] The process of obtaining the favorite data SC includes: obtaining the shopping cart product information of the analysis object, and marking the number of benchmark products in the analysis object's shopping cart as the favorite data SC; obtaining the basic coefficient JC of the analysis object through the formula JC = k1×SC + k2×SS + k3×LL, where k1, k2, and k3 are all proportionality coefficients, and k1 > k2 > k3 > 1; arranging all analysis objects in descending order of the value of the basic coefficient JC to obtain the basic sequence; intercepting L2 analysis objects from the basic sequence and marking them as basic recommendation objects, and pushing advertisements to the clients of the basic recommendation objects. L1 and L2 are both numerical constants, and the specific values of L1 and L2 are set by the management staff; collecting and calculating various push tendency data of the analysis objects to obtain the basic coefficient, evaluating the basic recommendation priority of the analysis objects according to the basic coefficient, and screening out the basic recommendation objects for the initial push to ensure the conversion rate of the initial push.

[0032] The conversion evaluation module is used to evaluate and analyze the advertisement push conversion rate of the basic recommendation objects: after completing the advertisement push to the clients of the basic recommendation objects, generate a conversion period of L3 hours. L3 is a numerical constant, and the specific value of L3 is set by the management staff themselves. Obtain the overall conversion rate of the basic recommendation objects at the end of the conversion period and mark it as the conversion value, and compare the conversion value with the preset conversion threshold: if the conversion value is less than the conversion threshold, it is determined that the conversion effect of the basic recommendation objects does not meet the requirements, generate a re-screening signal and send the re-screening signal to the basic recommendation module. After receiving the re-screening signal, the basic recommendation module re-selects a batch of platform users from the database as analysis objects; if the conversion value is greater than or equal to the conversion threshold, it is determined that the conversion effect of the basic recommendation objects meets the requirements, and conduct in-depth screening analysis on the analysis objects:

[0033] Randomly select a screening method to deeply screen the remaining analysis objects to obtain deeply recommended objects, push advertisements to the clients of the deeply recommended objects, and then generate a conversion cycle with a duration of L3 hours again. At the end of the conversion cycle, evaluate the conversion effect of the deeply recommended objects again: if the conversion effect meets the requirements, repeat the deep screening analysis of the analysis objects until the conversion effect does not meet the requirements; if the conversion effect does not meet the requirements, generate a re-screening signal and send the re-screening signal to the basic recommendation module; evaluate and analyze the advertisement push conversion rate of the basic recommended objects, conduct processing decision analysis based on the conversion evaluation results, and adopt different methods for deep screening analysis when the conversion evaluation effect is qualified to provide data support for the recommendation optimization process.

[0034] The screening methods include linear screening and portrait screening. The specific process of using linear screening to screen deeply recommended objects includes: removing the analysis objects that have completed advertisement pushing from the basic sequence, and marking the top L2 analysis objects in the basic sequence after removal as deeply recommended objects;

[0035] The specific process of using portrait screening to screen deeply recommended objects includes: marking the analysis objects that have completed conversion in the previous advertisement push as feature objects, obtaining the basic feature information of the feature objects. The basic feature information includes age, education level, income, and location information. Generate the corresponding portrait comparison range according to the basic feature information, mark the analysis objects whose basic feature information meets the portrait comparison range and has not completed advertisement pushing as candidate objects, arrange the candidate objects in descending order of the basic coefficient JC value to obtain a portrait sequence, and mark the top L2 candidate objects in the portrait sequence as deeply recommended objects; the portrait comparison range is generated after data cleaning of each item of basic feature information, and it includes an age range, an education level range, an income range, and a geographical location range;

[0036] Exemplarily, the specific process of obtaining the age range by data cleaning the age can include: taking the age values in all the basic feature information to form an age set, calculating the variance of all the elements in the age set to obtain an age feature value, and comparing the age feature value with a preset age feature threshold: if the age feature value is less than the age feature threshold, the age range is composed of the largest and smallest elements in the age set; if the age feature value is greater than or equal to the age feature threshold, the largest and smallest elements in the age set are removed at the same time, and then the age feature value is recalculated, and so on, until the age feature value is less than the age feature threshold; the income range can also be obtained in the same way as the age range; then mark the basic feature information whose age and income meet the age range and income range respectively as screening information, and the education level range is composed of the lowest and highest education levels in the screening information, and the geographical location range is composed of all the location information in the screening information.

[0037] The recommendation optimization module is used to optimize and analyze the advertising push method: when the number of analysis objects that have completed advertising push reaches M1, M1 is a numerical constant, and the specific value of M1 is set by the management personnel; the conversion value corresponding to each advertising push process is retrieved, and the minimum and maximum values ​​of the conversion values ​​constitute a conversion range, and the conversion range is divided into several conversion intervals, and the advertising push process with a conversion value within the conversion interval is marked as a matching process of the conversion interval;

[0038] If the next advertisement push process of the matching process adopts linear screening, the matching process is marked as a linear optimization process;

[0039] If the next ad push process of the matching process adopts portrait screening, the matching process is marked as a portrait optimization process;

[0040] The conversion values ​​corresponding to the next ad push process of all linear optimization processes within the conversion interval are summed up and averaged to obtain the linear priority value, and the conversion values ​​corresponding to the next ad push process of all portrait optimization processes within the conversion interval are summed up and averaged to obtain the portrait priority value; the linear priority value is compared with the portrait priority value: if the linear priority value is greater than or equal to the portrait priority value, the linear screening is marked as the priority method of the conversion interval; if the linear priority value is less than the portrait priority value, the portrait screening is marked as the priority method of the conversion interval; when performing in-depth screening analysis on the analysis object, the conversion value of the previous ad push process is first obtained, the priority method of the conversion interval to which the conversion value belongs is retrieved, and the priority method is directly used to screen the in-depth recommended objects; the advertising push method is optimized and analyzed, and the priority method of each conversion interval is marked. When performing in-depth screening analysis later, the corresponding optimization method can be directly used for in-depth object screening to ensure the conversion effect of each advertising push.

[0041] Embodiment 2: Figure 2 As shown, the user intelligent recommendation method based on information flow delivery includes the following steps:

[0042] Step 1: Conduct basic recommendation analysis on advertising users: Randomly select a batch of platform users and mark them as analysis objects, use the basic recommendation model to conduct initial screening and analysis on the analysis objects and obtain basic recommendation objects;

[0043] Step 2: Evaluate and analyze the conversion rate of the advertisement push of the basic recommendation object: after the advertisement push is completed to the client of the basic recommendation object, a conversion cycle of L3 hours is generated. At the end of the conversion cycle, the overall conversion rate of the basic recommendation object is obtained and marked as the conversion value. The conversion value is used to determine whether the conversion effect of the basic recommendation object meets the requirements;

[0044] Step 3: Optimize and analyze the advertisement push method: When the number of analysis objects for which advertisement pushing is completed reaches M1, retrieve the conversion value corresponding to each advertisement pushing process. The conversion range is formed by the minimum and maximum values of the conversion values. Divide the conversion range into several conversion intervals and mark the priority methods of the conversion intervals.

[0045] In a user intelligent recommendation system based on information flow delivery, during operation, a batch of platform users is randomly selected and marked as analysis objects. The basic recommendation mode is used to conduct a preliminary screening analysis on the analysis objects to obtain basic recommended objects. After completing the advertisement push to the clients of the basic recommended objects, a conversion cycle with a duration of L3 hours is generated. At the end moment of the conversion cycle, the overall conversion rate of the basic recommended objects is obtained and marked as the conversion value. It is determined whether the conversion effect of the basic recommended objects meets the requirements through the conversion value. When the number of analysis objects for which advertisement pushing is completed reaches M1, retrieve the conversion value corresponding to each advertisement pushing process. The conversion range is formed by the minimum and maximum values of the conversion values. Divide the conversion range into several conversion intervals and mark the priority methods of the conversion intervals.

[0046] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology make various modifications or supplements to the described specific embodiments or use similar methods for substitution. As long as it does not deviate from the structure of the invention or exceed the scope defined by this claim book, it shall fall within the protection scope of the present invention.

[0047] The above formulas are all obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value. The coefficients in the formula are set by those skilled in the art according to the actual situation. For example: the formula JC = k1×SC + k2×SS + k3×LL; those skilled in the art collect multiple groups of sample data and set corresponding basic coefficients for each group of sample data; substitute the set basic coefficients and the collected sample data into the formula. Any three formulas form a system of linear equations with three variables. Screen the calculated coefficients and take the average value to obtain the values of k1, k2, and k3 as 4.25, 2.81, and 2.59 respectively.

[0048] The magnitude of the coefficient is a specific value obtained by quantifying each parameter for subsequent comparison. Regarding the magnitude of the coefficient, it depends on the amount of sample data and the basic coefficients initially set by those skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantified value, for example, the basic coefficient is directly proportional to the value of the collection data.

[0049] In the description of this specification, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0050] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific implementation manners. Obviously, many modifications and variations can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A user intelligent recommendation system based on information flow delivery, characterized by: It includes a basic recommendation module, a conversion evaluation module and a recommendation optimization module, wherein the basic recommendation module, the conversion evaluation module and the recommendation optimization module are sequentially connected in communication; The basic recommendation module is used to perform basic recommendation analysis on advertising delivery users: randomly screen a batch of platform users and mark them as analysis objects, use the basic recommendation mode to perform initial screening analysis on the analysis objects and obtain a basic sequence; Extract L2 analysis objects from the basic sequence and mark them as basic recommendation objects, and push advertisements to the clients of the basic recommendation objects; The conversion evaluation module is used to evaluate and analyze the conversion rate of the advertisement push of the basic recommendation object: after completing the advertisement push to the client of the basic recommendation object, a conversion cycle with a duration of L3 hours is generated, and at the end of the conversion cycle, the overall conversion rate of the basic recommendation object is obtained and marked as the conversion value, and the conversion value is compared with the preset conversion threshold: if the conversion value is less than the conversion threshold, it is determined that the conversion effect of the basic recommendation object does not meet the requirements, and a re-screening signal is generated and sent to the basic recommendation module. After receiving the re-screening signal, the basic recommendation module re-screens a batch of platform users from the database as analysis objects; if the conversion value is greater than or equal to the conversion threshold, it is determined that the conversion effect of the basic recommendation object meets the requirements, and an in-depth screening analysis is performed on the analysis object; The recommendation optimization module is used to optimize and analyze the advertisement push method.

2. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The specific process of using the basic recommendation model to conduct initial screening analysis on the analysis object includes: obtaining the browsing data LL, search data SS and collection data SC of the analysis object and performing numerical calculations to obtain the basic coefficient JC of the analysis object; arranging all analysis objects in descending order according to the basic coefficient JC values ​​to obtain a basic sequence.

3. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The process of obtaining browsing data LL includes: obtaining the type of the product corresponding to the advertisement and marking it as the benchmark type, marking all products corresponding to the benchmark type as benchmark products, obtaining the number of times the analysis object browses the benchmark product in the last L1 months and marking it as browsing data LL; the process of obtaining search data SS includes: obtaining all keywords corresponding to the product corresponding to the advertisement when searching and marking them as benchmark words, marking the number of times the analysis object enters the benchmark word into the search box in L1 months as search data SS; the process of obtaining collection data SC includes: obtaining the shopping cart product information of the analysis object, and marking the number of benchmark products in the shopping cart of the analysis object as collection data SC.

4. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The specific process of deep screening and analysis of the analysis object includes: randomly selecting a screening method to perform deep screening on the remaining analysis objects to obtain deep recommendation objects, pushing advertisements to the clients of the deep recommendation objects, and then generating a conversion cycle with a duration of L3 hours again, and evaluating the conversion effect of the deep recommendation object again at the end of the conversion cycle: if the conversion effect meets the requirements, repeat the deep screening and analysis of the analysis object until the conversion effect does not meet the requirements; if the conversion effect does not meet the requirements, generate a re-screening signal and send the re-screening signal to the basic recommendation module.

5. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The screening methods include linear screening and portrait screening. The specific process of using linear screening for deep recommendation object screening includes: removing the analysis objects that have completed advertising push from the basic sequence, and marking the top L2 analysis objects in the basic sequence after removal as deep recommendation objects.

6. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The specific process of using portrait screening to screen deep recommendation objects includes: marking the analysis objects that have completed conversion in the last advertising push as feature objects, obtaining basic feature information of the feature objects, the basic feature information includes age, education, income and location information, generating a corresponding portrait comparison range based on the basic feature information, marking the analysis objects whose basic feature information meets the portrait comparison range and have not completed advertising push as candidates, arranging the candidates in descending order according to the basic coefficient JC value to obtain a portrait sequence, and marking the top L2 candidates in the portrait sequence as deep recommendation objects.

7. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The specific process of optimizing the advertising push method by the recommendation optimization module includes: when the number of analysis objects that have completed advertising push reaches M1, the conversion value corresponding to each advertising push process is retrieved, and the conversion range is composed of the minimum and maximum values ​​of the conversion value, and the conversion range is divided into several conversion intervals, and the priority method of the conversion interval is marked; when performing in-depth screening and analysis on the analysis object, the conversion value of the previous advertising push process is first obtained, and the priority method of the conversion interval to which the conversion value belongs is retrieved, and the in-depth recommendation object is directly screened using the priority method.

8. The user intelligent recommendation system based on information flow delivery according to claim 1 is characterized in that: The process of obtaining the priority method of the conversion interval includes: marking the advertising push process whose conversion value is within the conversion interval as the matching process of the conversion interval; if the next advertising push process of the matching process adopts linear screening, the matching process is marked as a linear optimization process; if the next advertising push process of the matching process adopts portrait screening, the matching process is marked as a portrait optimization process; summing up and averaging the conversion values ​​corresponding to the next advertising push process of all linear optimization processes in the conversion interval to obtain a linear priority value, summing up and averaging the conversion values ​​corresponding to the next advertising push process of all portrait optimization processes in the conversion interval to obtain a portrait priority value; comparing the linear priority value with the portrait priority value: if the linear priority value is greater than or equal to the portrait priority value, linear screening is marked as the priority method of the conversion interval; if the linear priority value is less than the portrait priority value, portrait screening is marked as the priority method of the conversion interval.

Images