Multi-platform advertisement putting data integration and analysis system

Abstract

The invention discloses a multi-platform advertisement putting data integration and analysis system, and relates to the technical field of data integration and analysis, and the system comprises a data layer which comprises a plurality of data collection modules and a plurality of data processing modules; each data acquisition module is matched with a group of platforms; the plurality of data processing modules are respectively used for receiving the original data of the corresponding data acquisition modules and processing the original data to obtain processed data with feature points; and the processing layer comprises a screening network module which corresponds to the data processing module of one group of platforms and is used for pre-grading the feature points and establishing a multi-layer screening network. According to the method, the received original data is subjected to duplicate removal and format standardization processing, the data consistency and standardization are ensured, feature point extraction is performed through the feature library, and a global replacement feature point library is established, so that the uniformity of feature point standards among platforms is ensured, and standardized storage and management of feature points of different platforms are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of data integration and analysis, and in particular to a multi-platform advertising data integration and analysis system. Background Art

[0002] With the rapid development of the Internet and the digital transformation of advertising, more and more advertisers have begun to place advertisements on multiple platforms to maximize their coverage and accuracy. In multi-platform advertising, data integration and analysis have become the key for advertisers to optimize advertising effects and improve user conversion rates. However, due to the differences in data formats, collection methods and user feature points of each platform, traditional advertising data analysis systems have great limitations in cross-platform data integration.

[0003] After searching, a Chinese patent (publication number: CN116805255B) discloses an automatic optimization and delivery system for advertisements based on user portrait analysis, including an intelligent push module, a user portrait module, a delivery optimization module and an effect analysis module. The user portrait module is used to label user information based on shopping information to obtain a user portrait model of the user. The intelligent push module is used to automatically deliver advertisements to the user based on the user portrait model. The effect analysis module is used to analyze the traffic effect of the product after the advertisement is delivered, and generate a poor benefit signal, a slow benefit signal or a normal benefit signal. The delivery optimization module is used to receive a delivery optimization instruction and optimize the delivery of the corresponding advertisement of the product.

[0004] In the prior art, due to the differences in user data of various platforms, the distribution of user portrait feature points is uneven, making it difficult to achieve unified analysis across platforms. In addition, the prior art usually relies on fixed screening rules and lacks the ability to dynamically adjust according to data distribution or target portraits, making it difficult to quickly adapt to the differences between different platforms. Therefore, the present invention proposes a multi-platform advertising data integration and analysis system. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-platform advertising data integration and analysis system to solve the problems mentioned in the above background technology.

[0006] The present invention can be implemented by the following technical solutions: a multi-platform advertising data integration and analysis system, including a data layer, a processing layer and a storage layer;

[0007] The data layer includes multiple data acquisition modules and multiple data processing modules;

[0008] Each of the data acquisition modules is matched with a group of platforms respectively, and obtains raw data from the corresponding platform API or platform database according to predetermined rules;

[0009] The plurality of data processing modules respectively correspond to a group of data acquisition modules, and are used to receive the original data of the corresponding data acquisition modules, which include a data cleaning unit, a feature library, and a feature extraction unit;

[0010] The cleaning unit is used to clean the received raw data, including deduplication and format standardization;

[0011] The feature extraction unit generates feature points for the cleaned raw data based on the feature library to obtain processed data;

[0012] The processing layer is used to receive and dynamically screen the processing data corresponding to each platform, and includes multiple screening network modules, statistical modules and analysis modules;

[0013] Each screening network module corresponds to a group of data processing modules of the platform, and the screening network module pre-classifies the feature points, establishes a multi-layer screening network to screen the processing data of the corresponding platform, where different levels correspond to different importance of feature points. When screening, the importance of feature points is gradually screened from high to low, and the successfully screened processing data is integrated to obtain a screening data set for subsequent statistics and analysis;

[0014] The statistical module receives the cleaned raw data from each cleaning unit and receives the screening data set from the corresponding screening network module;

[0015] After receiving the cleaned raw data and the screening data set, the statistical module performs correlation processing on them to obtain the screening rate of the corresponding screening network;

[0016] The analysis module integrates and analyzes the screening data sets that meet the screening rate requirements to obtain analysis data, including behavior patterns, interest trends, and user portraits of the corresponding platform;

[0017] The storage layer is used to store the original data and analysis data of each platform. By storing the original data of each platform, it ensures that the data has a complete record before being processed.

[0018] A further technical improvement of the present invention is that the screening network adopts a dynamic adjustment mechanism, and the adjustment method includes the following steps:

[0019] S1. The screening network sets weights for each feature point based on its importance, and sets multiple screening layers;

[0020] S2. After screening the data, each screening layer counts the actual screening rate and the number of screenings at each characteristic point, and uses the statistical results as feedback signals for subsequent dynamic adjustment of the screening standard;

[0021] S3. According to the feedback signal of the statistical results, the screening conditions of the subsequent layers are dynamically adjusted to ensure that the final screening network meets the target screening rate and maintains a balanced distribution of each feature point.

[0022] A further technical improvement of the present invention is that step S3 comprises:

[0023] A1. Calculation of screening rate deviation;

[0024] Calculate the difference between the actual screening rate and the target screening rate;

[0025] A2. Based on the number of feature point screenings fed back, adjust the weights of feature points at subsequent levels to ensure that key feature points are retained;

[0026] Its formula get;

[0027] w k,i+1 is the weight of feature point k in the i+1th layer;

[0028] N k,g is the expected data volume of the target feature point k;

[0029] N k,i is the actual data volume of feature point k in the i-th layer;

[0030] γ is the weight adjustment rate, η is a constant;

[0031] A3. Identify the screening rate of the corresponding feature point. When the corresponding screening rate deviation exceeds the preset threshold, introduce a feature point with the same weight to replace it to maintain the stability of screening.

[0032] A further technical improvement of the present invention is that the storage layer includes a global replacement feature point library, and the generation method thereof includes:

[0033] Q1. Global statistics;

[0034] Extract the distribution of feature points from the historical data of each platform, analyze the common values ​​of each feature point on the corresponding platform, and then determine the scope and standard of global feature points;

[0035] Q2, cluster analysis;

[0036] Similar feature points are classified by clustering algorithm to generate standard categories or intervals for each feature point. The value range and average weight of the center of each type of feature point constitute the replacement feature point;

[0037] Q3, manual tuning;

[0038] The global replacement feature points processed by Q2 are manually corrected, and the corrected replacement feature points are integrated to obtain a global replacement feature point library to ensure that the global replacement feature point library is consistent with actual business needs.

[0039] A further technical improvement of the present invention is that: the analysis module judges whether the screening network of each platform satisfies the output screening data set by pre-inputting user portrait and screening rate;

[0040] If the corresponding platform cannot meet the pre-input user portrait and screening rate, the corresponding screening network generates replacement feature points to replace the pre-input user portrait and screening rate feature points, and adjusts the screening network to assist users in performing integrated analysis between different platforms.

[0041] A further technical improvement of the present invention is that the analysis module's method for determining the screening network of each platform includes the following steps:

[0042] Z1. Determine platform adaptability, based on the actual data volume and feature point distribution of each platform, to determine whether it meets the pre-input user profile P in and screening rate R in , including the matching rate ratio and portrait similarity;

[0043] The adaptation rate ratio is determined by calculating the ratio of the actual data volume of the platform to the required data volume to determine whether the screening rate target is met. If the ratio of the actual data volume to the required data volume is less than 1, the platform cannot meet the screening rate requirement;

[0044] Z2. When the corresponding platform cannot meet the screening rate requirement, the corresponding screening network module generates replacement feature points similar to the pre-input user portrait feature points, and adjusts the user portrait feature points of the corresponding platform to make them close to or meet the pre-input screening rate and user portrait;

[0045] Its formula Determine the similarity of the replaced feature points;

[0046] Among them, T k is the feature point in the pre-input image, T′ k is the substitute feature point generated by the platform, M is the number of feature point samples, used to calculate T k and T′ k The distribution similarity of

[0047] Z3. Based on the replacement feature points, adjust the feature point weights of the screening network and redefine the screening conditions for each feature point to ensure that the screening process conforms to the pre-input user portrait;

[0048] It resets the weight w based on replacing the feature points k,new, adjust the influence of the feature point in the screening network, and the formula is: w k,new =w k ×(1-γ×(1-S′ k ));

[0049] where S′ k Indicates the replacement technical feature point T′ k Similarity with target feature points;

[0050] γ is the control replacement feature point T′ k The adjustment factor of

[0051] Z4. After re-adjusting the feature point weights and screening conditions, screen the platform's processed data and calculate the adjusted screening rate R new , to determine whether it meets the pre-input screening rate R in ;

[0052] in,

[0053] N p,new is the amount of data passed through the adjusted screening network;

[0054] N t is the total amount of input processing data;

[0055] δ is a constant;

[0056] If R new ≥R in , then the corresponding platform meets the pre-input screening rate requirements after adjustment;

[0057] Otherwise, the platform still marks it as a mismatch and reduces its data weight or marks it as a secondary reference during analysis;

[0058] Z5. If the screening rate of the platform still cannot be satisfied after adjustment, different replacement feature points are regenerated, and the feature point weights and screening conditions are continuously adjusted until the screening rate is the same as the pre-input screening rate;

[0059] If the screening rate standard cannot be met in the end, it will be recorded as "unsuitable platform" and marked accordingly to reduce the impact in subsequent analysis.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The present invention ensures data consistency and normalization by deduplicating and formatting the received raw data, extracting feature points through the feature library, and establishing a global feature point replacement library to ensure the uniformity of feature point standards between platforms, thus achieving standardized storage and management of feature points on different platforms. Even if the feature points are not completely consistent, the system can automatically select similar feature points for replacement, maintaining the consistency of cross-platform portrait analysis, and improving the adaptability and accuracy of the system;

[0062] In addition, the present invention adopts a multi-level dynamic screening network, supports hierarchical screening based on the importance of feature points, and realizes rapid screening through layer-by-layer screening and feedback mechanism. At the same time, the screening network is dynamically adjusted through feedback signals, so that the system can adapt to platform data in real time to meet the requirements of preset screening rate and user portrait. Even if the data of a certain platform cannot meet the initial conditions, the system can automatically adjust the screening conditions to ensure the adaptability of the screening network and provide efficient and flexible cross-platform data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0064] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0065] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0066] Example 1

[0067] See also Figure 1 As shown, the present invention provides a multi-platform advertising data integration and analysis system, including a data layer, a processing layer and a storage layer;

[0068] The data layer includes multiple data acquisition modules and multiple data processing modules;

[0069] Each data acquisition module is matched with a group of platforms and obtains raw data from the corresponding platform database according to predetermined rules;

[0070] The multiple data processing modules correspond to a group of data acquisition modules respectively, which are used to receive the original data of the corresponding data acquisition modules, and include a data cleaning unit, a feature library, and a feature extraction unit;

[0071] The cleaning unit is used to clean the received raw data, including deduplication and format standardization;

[0072] Deduplication of raw data includes unique identifier detection, field matching detection, similarity algorithm detection and time window deduplication;

[0073] Standardization of the format of raw data includes:

[0074] Field name standardization: uniformly map data field names from different sources;

[0075] Data type normalization: standardize the data type of the field. In this embodiment, the date field is standardized as "XXXX-YY-RR" and the currency is standardized as a decimal format to ensure the consistency of the precision of the numeric field.

[0076] Format unification: unify text format and numerical format into a unified standard;

[0077] Null value and outlier processing: set the missing data to fill in the value, using a fixed field or the average value;

[0078] Code conversion: unify the encoding formats of various original data into the same encoding;

[0079] The feature extraction unit generates feature points for the cleaned raw data based on the feature library to obtain processed data;

[0080] The processing layer is used to receive and dynamically screen the processing data corresponding to each platform, which includes multiple screening network modules, statistical modules and analysis modules;

[0081] Each screening network module corresponds to a group of data processing modules of the platform, and the screening network module pre-classifies the feature points, establishes a multi-layer screening network to screen the processing data of the corresponding platform, where different levels correspond to different importance of feature points. When screening, the importance of feature points is gradually screened from high to low, and the successfully screened processing data is integrated to obtain a screening data set for subsequent statistics and analysis;

[0082] When the missing data adopts a fixed field, the corresponding processed data is not screened when the corresponding feature points of the screening network are screened, and the screening process is performed;

[0083] When the missing data adopts the average value, the corresponding processing data is randomly allocated in the layer of feature points at the same level when it is in the layer of feature points corresponding to the screening network, and in the allocation of one cycle, the allocated feature points are removed until all feature points at the same level have been allocated, that is, the allocation of the next cycle begins;

[0084] The statistical module receives the cleaned raw data from each cleaning unit and receives the screening data set from the corresponding screening network module;

[0085] After receiving the cleaned raw data and the screening data set, the statistical module performs correlation processing on them to obtain the screening rate of the corresponding screening network;

[0086] The analysis module integrates and analyzes the screening data sets that meet the screening rate requirements to obtain analysis data, including behavior patterns, interest trends, and user portraits of the corresponding platforms;

[0087] The storage layer is used to store the original data and analysis data of each platform. By storing the original data of each platform, it ensures that the data is fully recorded before being processed.

[0088] The screening network adopts a dynamic adjustment mechanism, and its adjustment method includes the following steps:

[0089] S1. The screening network sets weights for each feature point based on its importance, and sets multiple screening layers;

[0090] S2. After screening the data, each screening layer counts the actual screening rate and the number of screenings at each characteristic point, and uses the statistical results as feedback signals for subsequent dynamic adjustment of the screening standard;

[0091] The actual screening rate is given by the formula get;

[0092] Among them, R i is the actual screening rate of the i-th layer, N p,i is the amount of data that passes the i-th layer screening, N t,i is the total amount of data entering the i-th layer, ∈ is a constant to prevent the denominator from being 0;

[0093] The number of screening for each feature point is calculated by the formula get;

[0094] Among them, N k,i is the amount of data containing feature point k in the i-th layer, d j is the jth data;

[0095] Ⅱ(·) is the indicator function, which is used to determine whether the corresponding processed data contains the feature point k;

[0096] S3. According to the feedback signal of the statistical results, the screening conditions of the subsequent layers are dynamically adjusted to ensure that the final screening network meets the target screening rate and maintains a balanced distribution of each feature point;

[0097] The S3 steps include:

[0098] A1, Screening rate deviation ΔR i calculate;

[0099] Calculate the difference between the actual screening rate and the target screening rate using the formula: ΔRi =R g,i -R i ;

[0100] R g,i is the target screening rate of the i-th layer, R i is the actual screening rate of the i-th layer;

[0101] A2. Based on the number of feature points screened in the feedback, adjust the weights of feature points in the subsequent layers to ensure that key feature points are retained. Key feature points include user information (age, gender, geographic location), interest tags (interest areas on the corresponding platform, including sports, fashion, technology, learning, life), purchase behavior (consumption frequency, consumption amount, shopping preferences), interactive behavior (click-through rate, browsing time, feedback actions), response behavior (conversion rate, exit rate, repeat visit behavior), channel characteristics (device type, device system) and time characteristics (visit time period, seasonal preference);

[0102] Its formula get;

[0103] w k,i+1 is the weight of feature point k in the i+1th layer;

[0104] N k,g is the expected data volume of the target feature point k;

[0105] N k,i is the actual data volume of feature point k in the i-th layer;

[0106] γ is the weight adjustment rate, η is a constant;

[0107] A3. Identify the screening rate of the corresponding feature point. When the corresponding screening rate deviation exceeds the preset threshold, introduce the feature point with the same weight to replace it to maintain the stability of screening. The adjustment formula used is:

[0108] θ k,i+1 =θ k′,i+1 +α×ΔR i × k,i+1 ;

[0109] θ k,i+1 is the screening threshold of feature point k at the i+1th layer;

[0110] θ k′,i+1 is the threshold value of feature point k′ in the i-th layer;

[0111] α is the threshold adjustment rate;

[0112] After each layer of screening, the actual screening rate and the number of feature points screened are counted to form a feedback signal and dynamically adjust the weight and importance threshold of the subsequent layers;

[0113] And through the feedback and dynamic adjustment mechanism of each layer, the screening network gradually generates a screening data set that meets the target screening rate and user profile.

[0114] Example 2

[0115] A multi-platform advertising data integration and analysis system, including a data layer, a processing layer and a storage layer;

[0116] The data layer includes multiple data acquisition modules and multiple data processing modules;

[0117] Each data collection module is matched with a set of platforms and obtains raw data from the corresponding platform API according to predetermined rules;

[0118] The multiple data processing modules correspond to a group of data acquisition modules respectively, which are used to receive the original data of the corresponding data acquisition modules, and include a data cleaning unit, a feature library, and a feature extraction unit;

[0119] The cleaning unit is used to clean the received raw data, including deduplication and format standardization;

[0120] Deduplication of raw data includes: unique identifier detection, field matching detection, similarity algorithm detection and time window deduplication;

[0121] Standardization of the format of raw data includes: field name standardization, data type normalization, format unification, null value and abnormal value processing, and encoding conversion;

[0122] The feature extraction unit generates feature points for the cleaned raw data based on the feature library to obtain processed data;

[0123] The processing layer is used to receive and dynamically screen the processing data corresponding to each platform, which includes multiple screening network modules, statistical modules and analysis modules;

[0124] Each screening network module corresponds to a group of data processing modules of the platform, and the screening network module pre-classifies the feature points, establishes a multi-layer screening network to screen the processing data of the corresponding platform, where different levels correspond to different importance of feature points. When screening, the importance of feature points is gradually screened from high to low, and the successfully screened processing data is integrated to obtain a screening data set for subsequent statistics and analysis;

[0125] The statistical module receives the cleaned raw data from each cleaning unit and receives the screening data set from the corresponding screening network module;

[0126] After receiving the cleaned raw data and the screening data set, the statistical module performs correlation processing on them to obtain the screening rate of the corresponding screening network;

[0127] The analysis module integrates and analyzes the screening data sets that meet the screening rate requirements to obtain analysis data, including behavior patterns, interest trends, and user portraits of the corresponding platforms;

[0128] The storage layer is used to store the original data and analysis data of each platform. By storing the original data of each platform, it ensures that the data is fully recorded before being processed.

[0129] The storage layer includes a global replacement feature point library, and the generation method thereof includes:

[0130] Q1. Global statistics;

[0131] Extract the distribution of feature points from the historical data of each platform, analyze the common values ​​of each feature point on the corresponding platform, and then determine the scope and standard of global feature points;

[0132] Q2, cluster analysis;

[0133] Similar feature points are classified by clustering algorithm to generate standard categories or intervals for each feature point. The value range and average weight of the center of each type of feature point constitute the replacement feature point;

[0134] Q3, manual tuning;

[0135] The global replacement feature points processed by Q2 are manually corrected, and the corrected replacement feature points are integrated to obtain a global replacement feature point library to ensure that the global replacement feature point library is consistent with actual business needs.

[0136] In this embodiment, the analysis module pre-inputs the user portrait and the screening rate to determine whether the screening network of each platform satisfies the output screening data set;

[0137] If the corresponding platform cannot meet the pre-input user portrait and screening rate, the corresponding screening network generates replacement feature points to replace the pre-input user portrait and screening rate feature points, and adjusts the screening network to assist users in performing integrated analysis between different platforms.

[0138] The analysis module's method for determining the screening network of each platform includes the following steps:

[0139] Z1. Determine platform adaptability, based on the actual data volume and feature point distribution of each platform, to determine whether it meets the pre-input user profile P in and screening rate R in , including the matching rate ratio and portrait similarity;

[0140] The adaptation rate ratio is determined by calculating the ratio of the actual data volume of the platform to the required data volume to determine whether the screening rate target is met. If the ratio of the actual data volume to the required data volume is less than 1, the platform cannot meet the screening rate requirement;

[0141] The similarity of the images is calculated by the formula If S exceeds the set threshold, the platform user profile is significantly different from the pre-input platform profile and needs to be adjusted;

[0142] Where K is the total number of feature points of the user portrait, that is, the number of key feature points contained in the pre-input user portrait;

[0143] w k is the weight of feature point k, indicating the importance of the feature point in the image;

[0144] T k is the ideal value of the kth feature point of the target image;

[0145] P k is the actual value of the kth feature point of the platform user portrait;

[0146] Z2. When the corresponding platform cannot meet the screening rate requirement, the corresponding screening network module generates replacement feature points similar to the pre-input user portrait feature points, and adjusts the user portrait feature points of the corresponding platform to make them close to or meet the pre-input screening rate and user portrait;

[0147] Its formula Determine the similarity of the replaced feature points;

[0148] Among them, T k is the feature point in the pre-input image, T′ k is the substitute feature point generated by the platform, M is the number of feature point samples, used to calculate T k and T′ k The distribution similarity of

[0149] Z3. Based on the replacement feature points, adjust the feature point weights of the screening network and redefine the screening conditions for each feature point to ensure that the screening process conforms to the pre-input user portrait;

[0150] It resets the weight w based on replacing the feature points k,new , adjust the influence of the feature point in the screening network, and the formula is: w k,new =w k ×(1-γ×(1-S′ k ));

[0151] where S′ k Indicates the replacement technical feature point T′ kSimilarity with target feature points;

[0152] γ is the control replacement feature point T′ k The adjustment factor of

[0153] Z4. After re-adjusting the feature point weights and screening conditions, screen the platform's processed data and calculate the adjusted screening rate R new , to determine whether it meets the pre-input screening rate R in ;

[0154] in,

[0155] N p,new is the amount of data passed through the adjusted screening network;

[0156] N t is the total amount of input processing data;

[0157] δ is a constant;

[0158] If R new ≥R in , then the corresponding platform meets the pre-input screening rate requirements after adjustment;

[0159] Otherwise, the platform still marks it as a mismatch and reduces its data weight or marks it as a secondary reference during analysis;

[0160] Z5. If the screening rate of the platform still cannot be satisfied after adjustment, different replacement feature points are regenerated, and the feature point weights and screening conditions are continuously adjusted until the screening rate is the same as the pre-input screening rate;

[0161] In this embodiment, a threshold of adjustment times is preset, and when the number of adjustments exceeds the threshold, the platform is recorded as an "unsuitable platform" to reduce resource consumption and computing overhead.

[0162] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A multi-platform advertising data integration and analysis system, characterized in that: include: The data layer includes a plurality of data acquisition modules and a plurality of data processing modules; Each of the data acquisition modules is matched with a group of platforms respectively, and obtains raw data from the corresponding platform API or platform database according to predetermined rules; The plurality of data processing modules respectively correspond to a group of data acquisition modules, which are used to receive the original data of the corresponding data acquisition modules and process the original data to obtain processed data with characteristic points; The processing layer receives and dynamically screens the processing data corresponding to each platform, including: The screening network module corresponds to a data processing module of a group of platforms, and the screening network module pre-classifies the feature points, establishes a multi-layer screening network, and screens the processing data of the corresponding platform, and obtains the screening data set through integration; The statistical module receives the cleaned raw data from each cleaning unit and receives the screening data set from the corresponding screening network module, and performs correlation processing to obtain the screening rate of the corresponding screening network; The analysis module integrates and analyzes the screening data sets that meet the screening rate requirements to obtain analytical data, including the behavior patterns, interest trends, and user portraits of the corresponding platforms; The storage layer stores the raw data and analysis data of each platform.

2. A multi-platform advertising data integration and analysis system according to claim 1, characterized in that: The data processing module includes a data cleaning unit, a feature library and a feature extraction unit; The cleaning unit cleans the received raw data, including deduplication and format standardization; The feature extraction unit generates feature points for the cleaned original data based on the feature library to obtain processed data.

3. The multi-platform advertising data integration and analysis system according to claim 1, characterized in that: Different levels of the screening network correspond to the importance of different feature points. During screening, the importance of feature points is gradually screened from high to low.

4. A multi-platform advertising data integration and analysis system according to claim 3, characterized in that: The screening network adopts a dynamic adjustment mechanism, and its adjustment method includes the following steps: S1. The screening network sets weights for each feature point based on its importance, and sets multiple screening layers; S2. After screening the data, each screening layer counts the actual screening rate and the number of screenings at each characteristic point, and uses the statistical results as feedback signals for subsequent dynamic adjustment of the screening standard; S3. Dynamically adjust the screening conditions of subsequent layers according to the feedback signal of the statistical results.

5. A multi-platform advertising data integration and analysis system according to claim 4, characterized in that: The S3 steps include: A1. Calculation of screening rate deviation; A2. Based on the number of feature point screenings fed back, adjust the weights of feature points at subsequent levels to ensure that key feature points are retained; A3. Identify the screening rate of the corresponding feature point. When the corresponding screening rate deviation exceeds the preset threshold, introduce a feature point with the same weight to replace it to maintain the stability of screening.

6. A multi-platform advertising data integration and analysis system according to claim 5, characterized in that: The formula used in step A2 is: Among them, w k,i+1 is the weight of feature point k in the i+1th layer; N k,g is the expected data volume of the target feature point k; N k,i is the actual data volume of feature point k in the i-th layer; γ is the weight adjustment rate and η is a constant.

7. The multi-platform advertising data integration and analysis system according to claim 1, characterized in that: The storage layer includes a global replacement feature point library, and the generation method thereof includes: Q1. Global statistics; Extract the distribution of feature points from the historical data of each platform, analyze the common values ​​of each feature point on the corresponding platform, and then determine the scope and standard of global feature points; Q2, cluster analysis; Similar feature points are classified by clustering algorithm to generate standard categories or intervals for each feature point. The value range and average weight of the center of each type of feature point constitute the replacement feature point; Q3, manual tuning; The global replacement feature points processed by Q2 are manually corrected, and the corrected replacement feature points are integrated to obtain a global replacement feature point library.

8. A multi-platform advertising data integration and analysis system according to claim 7, characterized in that: The analysis module determines whether the screening network of each platform satisfies the output screening data set by pre-inputting user portraits and screening rates; If the corresponding platform cannot satisfy the pre-input user portrait and screening rate, the corresponding screening network generates replacement feature points to replace the feature points of the pre-input user portrait and screening rate, and adjusts the screening network.

9. A multi-platform advertising data integration and analysis system according to claim 8, characterized in that: The analysis module determines the screening network of each platform, including the following steps: Z1. Determine the platform adaptability and whether it meets the pre-input user profile and screening rate based on the actual data volume and feature point distribution of each platform; Z2. When the corresponding platform cannot meet the screening rate requirement, the corresponding screening network module generates replacement feature points similar to the pre-input user portrait feature points, and adjusts the user portrait feature points of the corresponding platform to make them close to or meet the pre-input screening rate and user portrait; Z3. Based on the replacement feature points, adjust the feature point weights of the screening network and redefine the screening conditions for each feature point to ensure that the screening process conforms to the pre-input user portrait; Z4. After re-adjusting the feature point weights and screening conditions, the platform's processed data is screened, the adjusted screening rate is calculated, and it is determined whether it meets the pre-input screening rate; Z5. If the screening rate of the platform still cannot be met after adjustment, different replacement feature points are regenerated, and the feature point weights and screening conditions are continued to be adjusted.

Images