Information flow advertisement matching system and method based on user browsing and consuming behaviors

Abstract

The invention discloses an information flow advertisement matching system and method based on user browsing and consuming behaviors, and relates to the technical field of advertisement matching, and the matching method comprises the following steps: obtaining any user behavior record, and carrying out the effectiveness evaluation of the user behavior record; carrying out feature extraction by adopting a corresponding feature extraction strategy based on the effectiveness evaluation result of any user behavior record; performing influence degree analysis on each extracted feature, and establishing a behavior feature evaluation model to perform feature evaluation on the user behavior record; adjusting the behavior characteristic evaluation model according to the difference condition presented by the two evaluation results to obtain a characteristic threshold for judging the validity of each user behavior record; feature information extraction is carried out on advertisement information presented by any advertisement, and the matching effectiveness of the advertisement is evaluated; and judging whether to recommend based on an evaluation result, and carrying out priority ranking on the recommended advertisements.

Description

Technical Field

[0001] The present invention relates to the field of advertisement matching technology, and in particular to an information flow advertisement matching system and method based on user browsing and consumption behavior. Background Art

[0002] Information flow ads are a form of advertising embedded in the content that users browse daily, usually appearing in places such as social media, news websites or video platforms. Its main feature is that it is highly integrated with the surrounding content, giving users a natural browsing experience.

[0003] On all major platforms, users' browsing and consumption behaviors can be collected to screen similar products and recommend advertisements to users. Under this recommendation mechanism, although the browsing selectivity of users can be increased to a certain extent, there are still problems, because the platform will be full of various advertisements. When users browse advertisements on various platforms, they may easily click on unnecessary advertisements by mistake, thereby generating invalid browsing records. However, the platform cannot identify such invalid records, but will continue to push related advertisements based on the invalid records, which will affect the browsing and consumption burden of users. Summary of the invention

[0004] The purpose of the present invention is to provide an information flow advertisement matching system and method based on user browsing and consumption behavior to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: an information flow advertisement matching method based on user browsing and consumption behavior, the matching method comprising the following steps:

[0006] Step S100: Constructing a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after authorization by the user, and generate a number of user behavior records; obtaining the user feedback recorded in any user behavior record, and evaluating the effectiveness of the user behavior record;

[0007] Step S200: setting different feature extraction strategies for different effectiveness evaluation results, and taking corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and performing influence degree analysis on each extracted feature based on the distribution of each feature among each user behavior record;

[0008] Step S300: Based on each feature extracted from any user behavior record, a behavior feature evaluation model is established to perform feature evaluation on the user behavior record; the difference between two evaluation results of the user behavior record is analyzed, the behavior feature evaluation model is adjusted, and a feature threshold for judging the validity of each user behavior record is obtained;

[0009] Step S400: extract several features from each user behavior record, extract feature information of the advertising information presented by any advertisement, compare the similarity with each feature, and evaluate the matching effectiveness of the advertisement; determine whether to recommend the advertisement based on the evaluation result, and prioritize the recommended advertisements.

[0010] Furthermore, step S100 includes the following steps:

[0011] Step S101: Whenever a user enters an information detail page through an advertisement presented on any platform, or whenever a user performs a consumption behavior, a corresponding user behavior record is generated in the behavior record management system; wherein a user behavior record including a browsing behavior is set as a browsing behavior record, and a user behavior record including a consumption behavior is set as a consumption behavior record;

[0012] Step S102: randomly selecting a browsing behavior record, obtaining the information details page where the user is located in the browsing behavior record, and extracting the browsing data of the user in the information details page; presetting evaluation rules of several dimensions, extracting corresponding browsing data from the browsing data according to each dimension, and obtaining the effective value presented under any dimension, and accumulating to obtain the effective value of the browsing behavior record; the browsing data includes data such as click rate, dwell time, and scroll depth;

[0013] Step S103: randomly select a consumption behavior record, obtain the commodity information recorded in the consumption behavior record, compare the commodity information with the commodity information recorded in any browsing behavior record, and if there are two sets of identical commodity information, then feature mark the compared browsing behavior records; the user's consumption behavior can directly indicate that the user's browsing of the corresponding commodity information is a valid operation, so the effective value of the browsing record of the commodity corresponding to the consumption behavior can be used as a basis for distinguishing whether each user behavior record is valid, which is conducive to distinguishing the validity of the records later;

[0014] Step S104: Obtain the valid value of each browsing behavior record with a characteristic mark, and select the valid value with the smallest numerical value as the valid threshold for judging whether any user behavior record is valid; randomly select a user behavior record, and if the valid value of the user behavior record exceeds the valid threshold, set the user behavior record as a valid behavior record; if it is lower than the valid threshold, set it as an invalid behavior record.

[0015] Further, step S200 includes the following steps:

[0016] Step S201: randomly selecting a user behavior record, if the user behavior record is a valid behavior record, extracting the recorded commodity information from the user behavior record, extracting a number of commodity features, and obtaining each valid feature of the user behavior record; summarizing each valid feature in all valid behavior records to obtain a valid feature set;

[0017] Step S202: If the user behavior record is an invalid behavior record, the recorded product information is also extracted for product features to obtain several invalid features of the user behavior record; an invalid feature is randomly selected for similarity comparison with any valid feature in the valid feature set, and if the similarity exceeds the set similarity threshold, the invalid feature is removed; after invalid feature screening is performed on each invalid behavior record, an invalid feature set is generated; the product information recorded in the invalid behavior record does not mean that all features of the product are invalid, because a valid record requires that all features meet user requirements, but an invalid record only requires one invalid feature to meet the requirements, so it is necessary to further distinguish each feature in the invalid record in order to obtain an accurate invalid feature set;

[0018] Step S203: Summarize the valid feature set and the invalid feature set to obtain a feature set, and select the ith feature at random from the feature set to obtain the valid feature value of the ith feature as T i =Ju(Flag i =1), where Ju() is the judgment function, Flag i Is the validity flag of the i-th feature. If the i-th feature is a valid feature, then Flag i =1, if the i-th feature is an invalid feature, then Flag i =0, if Flag i =1, then T i =Ju(Flag i =1)=1, if Flag i =0, then T i =Ju(Flag i =1)=0;

[0019] Step S204: respectively obtain m valid behavior records containing the i-th feature i1 and the number of invalid behavior records containing the i-th feature is m i2 , set the number of valid behavior records stored in the behavior record management system to a and the number of invalid behavior records to b; according to the formula:

[0020]

[0021] Calculate the influence Y of the i-th feature i The influence of effective features is to positively improve the effective value of each user behavior record, and the influence of invalid features is to negatively improve the effective value. The frequency of occurrence of each feature can effectively reflect the influence of each feature, which is conducive to the subsequent judgment of the effectiveness of each behavior record through the influence of the feature.

[0022] Further, step S300 includes the following steps:

[0023] Step S301: randomly select a user behavior record and extract all the features in the user behavior record, wherein the influence degree of the i-th feature is set to Y i ; Establish behavioral characteristics evaluation model:

[0024]

[0025] Wherein, x is the number of features of the user behavior record, Ju() is a judgment function, Record is a valid mark of the user behavior record, if the user behavior record is a valid behavior record, then Ju(Record)=1, if it is an invalid behavior record, then Ju(Record)=-1, k and c are both constant coefficients; calculate and obtain the feature value Z of the user behavior record;

[0026] Step S302: extract the characteristic value Z' of any valid behavior record and the characteristic value Z", of any invalid behavior record respectively, if Z'<Z", adjust the constant coefficients k and c so that Z'≥Z", train the behavior characteristic evaluation model so that the characteristic value of any valid behavior record is greater than the characteristic value of any invalid behavior record, thereby determining the values ​​of the constant coefficients k and c;

[0027] Step S303: Obtain the largest characteristic value among all invalid behavior records as Z max The smallest eigenvalue among all valid behavior records is Z min , get the feature threshold Z to determine whether each user behavior record is valid th =(Z max +Z min ) / 2.

[0028] Furthermore, step S400 includes the following steps:

[0029] Step S401: randomly selecting an advertisement, acquiring the advertisement information presented in the advertisement, and extracting feature information from the advertisement information to obtain a feature information set of the advertisement; randomly selecting feature information, performing similarity comparison between the feature information and each feature to obtain the similarity between the feature and any feature, and selecting a feature with the largest value as a target feature of the feature information;

[0030] Step S402: Obtain the influence degree of the target feature corresponding to the pth feature information of the advertisement as Y p , set a similarity threshold S th , according to the formula:

[0031]

[0032] Where u is the number of characteristic information of the advertisement, S p is the similarity between the pth feature information and the corresponding target feature, IF() is the judgment function, if S p <S th , then IF(S p <S th )=1, otherwise, IF(S p <S th )=0; calculate the matching effective value E of the advertisement; for feature information with high similarity, directly use the influence degree of the target feature, and for feature information with low similarity, adjust the influence degree according to the occurrence probability, so as to obtain a more accurate matching effective value;

[0033] Step S403: Obtain feature threshold Z th , if E<Z th , then the advertisement is not recommended; if E ≥ Z th , then the advertisement is marked as recommended; the matching effective value of each recommended advertisement is obtained, and the matching effective value is sorted from large to small according to the value, and the recommendation priority order of each recommended advertisement is generated.

[0034] In order to better implement the above method, an information flow advertising matching system is also proposed, which includes a user behavior screening module, a feature information processing module, a user behavior evaluation module and a real-time advertising push module;

[0035] The user behavior screening module is used to build a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; obtain the user feedback recorded in any user behavior record, and evaluate the effectiveness of the user behavior record;

[0036] The feature information processing module is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and analyze the influence of each extracted feature based on the distribution of each feature among each user behavior record;

[0037] A user behavior evaluation module is used to establish a behavior feature evaluation model based on each feature extracted from any user behavior record to perform feature evaluation on the user behavior record; analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain a feature threshold for judging the validity of each user behavior record;

[0038] The real-time advertising push module is used to extract several features from each user behavior record, extract feature information of the advertising information presented by any advertisement, compare the similarity with each feature, and evaluate the matching effectiveness of the advertisement; determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

[0039] Further, the user behavior screening module includes a behavior record setting unit and a valid behavior evaluation unit;

[0040] The behavior record setting unit is used to build a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; the effective behavior evaluation unit is used to obtain the user feedback recorded in any user behavior record and evaluate the effectiveness of the user behavior record.

[0041] Furthermore, the feature information processing module includes a feature information extraction unit and an influence degree calculation unit;

[0042] The feature information extraction unit is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies based on the effectiveness evaluation results of any user behavior records to extract features; the influence degree calculation unit is used to analyze the influence degree of each extracted feature based on the distribution of each feature among each user behavior record.

[0043] Further, the user behavior assessment module includes a behavior feature assessment unit and a feature threshold determination unit;

[0044] The feature evaluation unit is used to establish a behavior feature evaluation model based on the various features extracted from any user behavior record to perform feature evaluation on the user behavior record; the feature threshold determination unit is used to analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain the feature threshold for judging the validity of each user behavior record.

[0045] Furthermore, the real-time advertisement push module includes an advertisement matching evaluation unit and a push sequence setting unit;

[0046] The advertisement matching evaluation unit is used to extract a number of features from each user behavior record, extract feature information of the advertisement information presented by any advertisement, and compare the similarity with each feature to evaluate the matching effectiveness of the advertisement; the push order setting unit is used to determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The present invention distinguishes the behavior records generated by users by judging the validity of each behavior record of users, helps to exclude invalid records, and provides users with more accurate advertising recommendations, thereby improving users' browsing selectivity and effectively reducing users' browsing burden;

[0049] 2. The present invention can deeply explore the user's desired product features by extracting features from each product information and performing effectiveness analysis, so as to make more accurate recommendations for each advertisement. On the basis of the original conventional recommendation strategy, it can provide more accurate recommendation suggestions and check for advertisements that do not actually meet the requirements, thereby reducing the user's browsing burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of the steps of a method for matching information flow advertisements based on user browsing and consumption behaviors;

[0051] Figure 2 This is a structural diagram of the information flow advertising matching system based on user browsing and consumption behavior. DETAILED DESCRIPTION

[0052] 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.

[0053] Example: Figure 1 to Figure 2 As shown, the present invention provides an information flow advertisement matching method based on user browsing and consumption behavior, and the matching method includes the following steps:

[0054] Step S100: Constructing a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after authorization by the user, and generate a number of user behavior records; obtaining the user feedback recorded in any user behavior record, and evaluating the effectiveness of the user behavior record;

[0055] Wherein, step S100 includes the following steps:

[0056] Step S101: Whenever a user enters an information detail page through an advertisement presented on any platform, or whenever a user performs a consumption behavior, a corresponding user behavior record is generated in the behavior record management system; wherein a user behavior record including a browsing behavior is set as a browsing behavior record, and a user behavior record including a consumption behavior is set as a consumption behavior record;

[0057] Step S102: randomly selecting a browsing behavior record, obtaining the information details page where the user is located in the browsing behavior record, and extracting the browsing data of the user in the information details page; presetting evaluation rules of several dimensions, extracting corresponding browsing data from the browsing data according to each dimension, and obtaining the effective value presented under any dimension, and accumulating them to obtain the effective value of the browsing behavior record;

[0058] Step S103: randomly selecting a consumption behavior record, obtaining the product information recorded in the consumption behavior record, and comparing the product information with the product information recorded in any browsing behavior record. If there are two sets of identical product information, the compared browsing behavior records are marked with features;

[0059] Step S104: Obtain the valid value of each browsing behavior record with a characteristic mark, and select the valid value with the smallest numerical value as the valid threshold for judging whether any user behavior record is valid; randomly select a user behavior record, and if the valid value of the user behavior record exceeds the valid threshold, set the user behavior record as a valid behavior record; if it is lower than the valid threshold, set it as an invalid behavior record.

[0060] Step S200: setting different feature extraction strategies for different effectiveness evaluation results, and taking corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and performing influence degree analysis on each extracted feature based on the distribution of each feature among each user behavior record;

[0061] Wherein, step S200 includes the following steps:

[0062] Step S201: randomly selecting a user behavior record, if the user behavior record is a valid behavior record, extracting the recorded commodity information from the user behavior record, extracting a number of commodity features, and obtaining each valid feature of the user behavior record; summarizing each valid feature in all valid behavior records to obtain a valid feature set;

[0063] Step S202: If the user behavior record is an invalid behavior record, the recorded product information is also extracted for product features to obtain a number of invalid features of the user behavior record; an invalid feature is randomly selected for similarity comparison with any valid feature in the valid feature set, and if the similarity exceeds a set similarity threshold, the invalid feature is removed; after invalid feature screening is performed on each invalid behavior record, an invalid feature set is generated;

[0064] Step S203: Summarize the valid feature set and the invalid feature set to obtain a feature set, and select the ith feature at random from the feature set to obtain the valid feature value of the ith feature as T i =Ju(Flag i =1), where Ju() is the judgment function, Flag i Is the validity flag of the i-th feature. If the i-th feature is a valid feature, then Flag i =1, if the i-th feature is an invalid feature, then Flag i =0, if Flag i =1, then T i =Ju(Flag i =1)=1, if Flag i =0, then T i =Ju(Flag i =1)=0;

[0065] Step S204: respectively obtain m valid behavior records containing the i-th feature i1 and the number of invalid behavior records containing the i-th feature is m i2 , set the number of valid behavior records stored in the behavior record management system to a and the number of invalid behavior records to b; according to the formula:

[0066]

[0067] Calculate the influence Y of the i-th feature i ;

[0068] Example 1: Assume that the number of valid behavior records containing the ith feature is 10, the number of invalid behavior records is 5, and the number of valid behavior records stored in the behavior record management system is 40 and the number of invalid behavior records is 10. Assume that the ith feature is a valid feature, so T is obtained. i =1,Y i =10 / 15×(1+1 / 4-1 / 2)×1=50%.

[0069] Step S300: Based on each feature extracted from any user behavior record, a behavior feature evaluation model is established to perform feature evaluation on the user behavior record; the difference between two evaluation results of the user behavior record is analyzed, the behavior feature evaluation model is adjusted, and a feature threshold for judging the validity of each user behavior record is obtained;

[0070] Wherein, step S300 includes the following steps:

[0071] Step S301: randomly select a user behavior record and extract all the features in the user behavior record, wherein the influence degree of the i-th feature is set to Y i ; Establish behavioral characteristics evaluation model:

[0072]

[0073] Wherein, x is the number of features of the user behavior record, Ju() is a judgment function, Record is a valid mark of the user behavior record, if the user behavior record is a valid behavior record, then Ju(Record)=1, if it is an invalid behavior record, then Ju(Record)=-1, k and c are both constant coefficients; calculate and obtain the feature value Z of the user behavior record;

[0074] Example 2: Assuming that there are three features in the user behavior record, and the influence of each feature is 70%, 50% and -20% respectively, k=50%, c=0, the feature value of the user behavior record is Z=1.7×1.5×0.8+0.5=2.54;

[0075] Step S302: extract the characteristic value Z' of any valid behavior record and the characteristic value Z", of any invalid behavior record respectively, if Z'<Z", adjust the constant coefficients k and c so that Z'≥Z", train the behavior characteristic evaluation model so that the characteristic value of any valid behavior record is greater than the characteristic value of any invalid behavior record, thereby determining the values ​​of the constant coefficients k and c;

[0076] Step S303: Obtain the largest characteristic value among all invalid behavior records as Z maxThe smallest eigenvalue among all valid behavior records is Z min , get the feature threshold Z to determine whether each user behavior record is valid th =(Z max +Z min ) / 2.

[0077] Step S400: extracting several features from each user behavior record, extracting feature information from the advertising information presented by any advertisement, and performing similarity comparison with each feature, and evaluating the matching effectiveness of the advertisement; judging whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritizing the recommended advertisements;

[0078] Wherein, step S400 includes the following steps:

[0079] Step S401: randomly selecting an advertisement, acquiring the advertisement information presented in the advertisement, and extracting feature information from the advertisement information to obtain a feature information set of the advertisement; randomly selecting feature information, performing similarity comparison between the feature information and each feature to obtain the similarity between the feature and any feature, and selecting a feature with the largest value as a target feature of the feature information;

[0080] Step S402: Obtain the influence degree of the target feature corresponding to the pth feature information of the advertisement as Y p , set a similarity threshold S th , according to the formula:

[0081]

[0082] Where u is the number of characteristic information of the advertisement, S p is the similarity between the pth feature information and the corresponding target feature, IF() is the judgment function, if S p <S th , then IF(S p <S th )=1, otherwise, IF(S p <S th )=0; calculate and obtain the matching effective value E of the advertisement;

[0083] Step S403: Obtain feature threshold Z th , if E<Z th , then the advertisement is not recommended; if E ≥ Z th , then the advertisement is marked as recommended; the matching effective value of each recommended advertisement is obtained, and the matching effective value is sorted from large to small according to the value, and the recommendation priority order of each recommended advertisement is generated.

[0084] Information flow advertising matching system, the matching system includes a user behavior screening module, a feature information processing module, a user behavior evaluation module and a real-time advertising push module;

[0085] The user behavior screening module is used to build a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; obtain the user feedback recorded in any user behavior record, and evaluate the effectiveness of the user behavior record;

[0086] The feature information processing module is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and analyze the influence of each extracted feature based on the distribution of each feature among each user behavior record;

[0087] A user behavior evaluation module is used to establish a behavior feature evaluation model based on each feature extracted from any user behavior record to perform feature evaluation on the user behavior record; analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain a feature threshold for judging the validity of each user behavior record;

[0088] The real-time advertising push module is used to extract several features from each user behavior record, extract feature information of the advertising information presented by any advertisement, compare the similarity with each feature, and evaluate the matching effectiveness of the advertisement; determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

[0089] Among them, the user behavior screening module includes a behavior record setting unit and an effective behavior evaluation unit;

[0090] The behavior record setting unit is used to build a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; the effective behavior evaluation unit is used to obtain the user feedback recorded in any user behavior record and evaluate the effectiveness of the user behavior record.

[0091] Among them, the feature information processing module includes a feature information extraction unit and an influence degree calculation unit;

[0092] The feature information extraction unit is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies based on the effectiveness evaluation results of any user behavior records to extract features; the influence degree calculation unit is used to analyze the influence degree of each extracted feature based on the distribution of each feature among each user behavior record.

[0093] Wherein, the user behavior evaluation module includes a behavior feature evaluation unit and a feature threshold determination unit;

[0094] The feature evaluation unit is used to establish a behavior feature evaluation model based on the various features extracted from any user behavior record to perform feature evaluation on the user behavior record; the feature threshold determination unit is used to analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain the feature threshold for judging the validity of each user behavior record.

[0095] Among them, the real-time advertisement push module includes an advertisement matching evaluation unit and a push sequence setting unit;

[0096] The advertisement matching evaluation unit is used to extract a number of features from each user behavior record, extract feature information of the advertisement information presented by any advertisement, and compare the similarity with each feature to evaluate the matching effectiveness of the advertisement; the push order setting unit is used to determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

[0097] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and range of equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An information flow advertisement matching method based on user browsing and consumption behavior, characterized in that: The matching method comprises the following steps: Step S100: Constructing a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after authorization by the user, and generate a number of user behavior records; obtaining the user feedback recorded in any user behavior record, and evaluating the effectiveness of the user behavior record; Step S200: setting different feature extraction strategies for different effectiveness evaluation results, and taking corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and performing influence degree analysis on each extracted feature based on the distribution of each feature among each user behavior record; Step S300: Based on each feature extracted from any user behavior record, a behavior feature evaluation model is established to perform feature evaluation on the user behavior record; the difference between two evaluation results of the user behavior record is analyzed, the behavior feature evaluation model is adjusted, and a feature threshold for judging the validity of each user behavior record is obtained; Step S400: extract several features from each user behavior record, extract feature information of the advertising information presented by any advertisement, compare the similarity with each feature, and evaluate the matching effectiveness of the advertisement; determine whether to recommend the advertisement based on the evaluation result, and prioritize the recommended advertisements.

2. The information flow advertisement matching method based on user browsing and consumption behavior according to claim 1 is characterized by: The step S100 includes the following steps: Step S101: Whenever a user enters an information detail page through an advertisement presented on any platform, or whenever a user performs a consumption behavior, a corresponding user behavior record is generated in the behavior record management system; wherein a user behavior record including a browsing behavior is set as a browsing behavior record, and a user behavior record including a consumption behavior is set as a consumption behavior record; Step S102: randomly selecting a browsing behavior record, obtaining the information details page where the user is located in the browsing behavior record, and extracting the browsing data of the user in the information details page; presetting evaluation rules of several dimensions, extracting corresponding browsing data from the browsing data according to each dimension, and obtaining the effective value presented under any dimension, and accumulating them to obtain the effective value of the browsing behavior record; Step S103: randomly selecting a consumption behavior record, obtaining the product information recorded in the consumption behavior record, and comparing the product information with the product information recorded in any browsing behavior record. If there are two sets of identical product information, the compared browsing behavior records are marked with features; Step S104: Obtain the valid value of each browsing behavior record with a characteristic mark, and select the valid value with the smallest numerical value as the valid threshold for judging whether any user behavior record is valid; randomly select a user behavior record, and if the valid value of the user behavior record exceeds the valid threshold, set the user behavior record as a valid behavior record; if it is lower than the valid threshold, set it as an invalid behavior record.

3. The information flow advertisement matching method based on user browsing and consumption behavior according to claim 2 is characterized by: The step S200 includes the following steps: Step S201: randomly selecting a user behavior record, if the user behavior record is a valid behavior record, extracting the recorded commodity information from the user behavior record, extracting a number of commodity features, and obtaining each valid feature of the user behavior record; summarizing each valid feature in all valid behavior records to obtain a valid feature set; Step S202: If the user behavior record is an invalid behavior record, the recorded product information is also extracted for product features to obtain a number of invalid features of the user behavior record; an invalid feature is randomly selected for similarity comparison with any valid feature in the valid feature set, and if the similarity exceeds a set similarity threshold, the invalid feature is removed; after invalid feature screening is performed on each invalid behavior record, an invalid feature set is generated; Step S203: Summarize the valid feature set and the invalid feature set to obtain a feature set, and select the ith feature at random from the feature set to obtain the valid feature value of the ith feature as T i =Ju(Flag i =1), where Ju() is the judgment function, Flag i is the validity flag of the i-th feature. If the i-th feature is a valid feature, then Flag i =1, if the i-th feature is an invalid feature, then Flag i =0, if Flag i =1, then T i =Ju(Flag i =1)=1, if Flag i =0, then T i =Ju(Flag i =1)=0; Step S204: respectively obtain m valid behavior records containing the i-th feature i1 and the number of invalid behavior records containing the i-th feature is m i2 , set the number of valid behavior records stored in the behavior record management system to a and the number of invalid behavior records to b; according to the formula: Calculate the influence Y of the i-th feature i .

4. The information flow advertisement matching method based on user browsing and consumption behavior according to claim 3 is characterized by: The step S300 includes the following steps: Step S301: randomly select a user behavior record and extract all the features in the user behavior record, wherein the influence degree of the i-th feature is set to Y i ; Establish behavioral characteristics evaluation model: Wherein, x is the number of features of the user behavior record, Ju() is a judgment function, Record is a valid mark of the user behavior record, if the user behavior record is a valid behavior record, then Ju(Record)=1, if it is an invalid behavior record, then Ju(Record)=-1, k and c are both constant coefficients; calculate and obtain the feature value Z of the user behavior record; Step S302: extract the characteristic value Z' of any valid behavior record and the characteristic value Z", of any invalid behavior record respectively, if Z'<Z", adjust the constant coefficients k and c so that Z'≥Z", train the behavior characteristic evaluation model so that the characteristic value of any valid behavior record is greater than the characteristic value of any invalid behavior record, thereby determining the values ​​of the constant coefficients k and c; Step S303: Obtain the largest characteristic value among all invalid behavior records as Z max The smallest eigenvalue among all valid behavior records is Z min , get the feature threshold Z to determine whether each user behavior record is valid th =(Z max +Z min ) / 2.

5. The information flow advertisement matching method based on user browsing and consumption behavior according to claim 4 is characterized by: The step S400 includes the following steps: Step S401: randomly selecting an advertisement, acquiring the advertisement information presented in the advertisement, and extracting feature information from the advertisement information to obtain a feature information set of the advertisement; randomly selecting feature information, performing similarity comparison between the feature information and each feature to obtain the similarity between the feature and any feature, and selecting a feature with the largest value as a target feature of the feature information; Step S402: Obtain the influence degree of the target feature corresponding to the pth feature information of the advertisement as Y p , set a similarity threshold S th , according to the formula: Where u is the number of characteristic information of the advertisement, S p is the similarity between the pth feature information and the corresponding target feature, IF() is the judgment function, if S p <S th , then IF(S p <S th )=1, otherwise, IF(S p <S th )=0; calculate and obtain the matching effective value E of the advertisement; Step S403: Obtain feature threshold Z th , if E<Z th , then the advertisement is not recommended; if E ≥ Z th , then the advertisement is marked as recommended; the matching effective value of each recommended advertisement is obtained, and the matching effective value is sorted from large to small according to the value, and the recommendation priority order of each recommended advertisement is generated.

6. An information flow advertisement matching system, used to execute the information flow advertisement matching method based on user browsing and consumption behavior as described in any one of claims 1 to 5, characterized in that: The matching system includes a user behavior screening module, a feature information processing module, a user behavior evaluation module and a real-time advertisement push module; The user behavior screening module is used to construct a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; obtain the user feedback recorded in any user behavior record, and evaluate the effectiveness of the user behavior record; The feature information processing module is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies to extract features based on the effectiveness evaluation results of any user behavior records; and analyze the influence of each extracted feature based on the distribution of each feature among each user behavior record; The user behavior evaluation module is used to establish a behavior feature evaluation model based on each feature extracted from any user behavior record to perform feature evaluation on the user behavior record; analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain a feature threshold for judging the validity of each user behavior record; The real-time advertising push module is used to extract several features from each user behavior record, extract feature information of the advertising information presented by any advertisement, and compare the similarity with each feature to evaluate the matching effectiveness of the advertisement; determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

7. The information flow advertisement matching system according to claim 6, characterized in that: The user behavior screening module includes a behavior record setting unit and an effective behavior evaluation unit; The behavior record setting unit is used to construct a behavior record management system to collect the user's browsing behavior and consumption behavior on all platforms after the user's authorization, and generate a number of user behavior records; the effective behavior evaluation unit is used to obtain the user feedback recorded in any user behavior record and evaluate the effectiveness of the user behavior record.

8. The information flow advertisement matching system according to claim 6, characterized in that: The feature information processing module includes a feature information extraction unit and an influence degree calculation unit; The feature information extraction unit is used to set different feature extraction strategies for different effectiveness evaluation results, and adopt corresponding feature extraction strategies based on the effectiveness evaluation results of any user behavior records to perform feature extraction; the impact degree calculation unit is used to analyze the impact degree of each extracted feature based on the distribution of each feature among each user behavior record.

9. The information flow advertisement matching system according to claim 6, characterized in that: The user behavior evaluation module includes a behavior feature evaluation unit and a feature threshold determination unit; The feature evaluation unit is used to establish a behavior feature evaluation model based on each feature extracted from any user behavior record to perform feature evaluation on the user behavior record; the feature threshold determination unit is used to analyze the difference between the two evaluation results of the user behavior record, adjust the behavior feature evaluation model, and obtain the feature threshold for judging the validity of each user behavior record.

10. The information flow advertisement matching system according to claim 6, characterized in that: The real-time advertisement push module includes an advertisement matching evaluation unit and a push sequence setting unit; The advertisement matching evaluation unit is used to extract a number of features from each user behavior record, extract feature information of the advertisement information presented by any advertisement, and compare the similarity with each feature to evaluate the matching effectiveness of the advertisement; the push order setting unit is used to determine whether to recommend the advertisement based on the evaluation result of the advertisement, and prioritize the recommended advertisements.

Images