Media information intelligent analysis system and method based on big data

Abstract

The invention discloses a media information intelligent analysis system and method based on big data, and relates to the technical field of data analysis. Media data related to target advertisement putting are obtained from multiple channels, format recognition and processing are carried out, and to-be-analyzed data in a unified format are formed; extracting a putting record of the target advertisement, analyzing a putting strategy of the target advertisement, identifying actual user group characteristics, comparing the actual user group characteristics with target user group characteristics, and evaluating an advertisement putting effect; performing clustering analysis on the user behavior data by using a K-means algorithm, identifying an actual user group from the user behavior data, and calculating feature difference and correlation between the user behavior data and the actual user group; and analyzing and optimizing an advertisement putting strategy according to the putting effect evaluation index, and proposing an adjustment suggestion. According to the method, the advertisement putting efficiency and effect are improved, the decision making process is more scientific and reasonable through the data driving method, market changes can be better adapted, and user requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a system and method for intelligent analysis of media information based on big data. Background Art

[0002] In today's era of information explosion, media advertising has become an indispensable part of corporate marketing strategies. With the popularization of the Internet and mobile devices, advertising channels are becoming increasingly diverse, including social media, search engines, video platforms, etc. In order to improve the efficiency and effectiveness of advertising, companies urgently need to use big data technology to conduct intelligent analysis of media information. The media data involved in advertising comes from a wide range of sources, including online advertising, social platform interactions, user behavior logs, etc. These data usually exist in different formats and structures, which brings challenges to data integration and analysis.

[0003] However, although the existing media information analysis system has basic data collection and processing capabilities, it still has some shortcomings, which are specifically reflected in the evaluation of advertising effects. For example, the existing media information analysis system focuses on basic indicators in advertising effect evaluation, such as click-through rate and exposure rate. Such a single evaluation method cannot fully reflect the advertising effect and limits the optimization space of the delivery strategy. Summary of the invention

[0004] The purpose of the present invention is to provide a media information intelligent analysis system and method based on big data to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: The method for intelligent analysis of media information based on big data includes the following steps: Step S100. Acquire media data related to target advertisement delivery in different formats from different data channels, perform format recognition on the acquired media data, and process the media data accordingly according to the result of the format recognition, so as to integrate the media data into a unified format, thereby obtaining data to be analyzed; Step S200. Obtain the delivery record corresponding to the target advertisement, extract the delivery strategy of the target advertisement from the delivery record, and obtain the target user group characteristics according to the delivery strategy of the target advertisement; analyze the data to be analyzed, thereby identifying the actual user group; Step S300: extract the characteristics of the actual user group and compare and analyze them with the characteristics of the target user group to obtain the difference between the two; combine the differences between the corresponding characteristics of the target user group and the actual user group to evaluate the delivery effect of the target advertising delivery strategy; Step S400. Analyze the optimization demand coefficient of the delivery strategy corresponding to the target advertisement according to the delivery effect evaluation index of the target advertisement; analyze the corresponding delivery strategy according to the optimization demand coefficient, and output corresponding prompt information based on the analysis result.

[0006] Furthermore, step S100 includes: S101. Obtain media data related to target advertising in different formats from different data channels, wherein the formats of the media data include text, image, audio and video formats, and the data format corresponding to each piece of media data is not unique; for example, a social media post may have text description, image, audio and video; for each piece of media data Di, perform format identification, where i represents the media data number, which is a positive integer; according to the result of the format identification, divide the current media data, so that the media data Di is represented as: Di={d1,d2,...,dn}i, where d1 represents the first data segment of the media data Di, d2 represents the second data segment of the media data Di, and so on, dn represents the nth data segment of the media data Di, where n represents the data format number corresponding to the media data Di, which is 1 to 4, representing text, image, audio and video formats respectively; S102. For each data segment in the image, audio and video format in the media data Di, it is converted into text format, and the converted text format data replaces the corresponding data segment in the image, audio and video format, so that the media data is represented as Di', and the data proportion b of each data segment of the media data Di' is calculated, and b=N_dj / N_Di', wherein N_dj represents the number of bytes of the j-th data segment in the media data Di', and N_Di' represents the number of bytes of the media data Di'; semantic analysis is performed on each data segment of the media data Di', so as to form a corresponding semantic feature vector set Vi, and V={v1,v2,...,vn}i, similarly, v1 represents the semantic feature vector of the first data segment, v2 represents the semantic feature vector of the second data segment, and vn represents the semantic feature vector of the n-th data segment, and each semantic feature vector contains multi-dimensional information, such as keywords, themes, etc.; S103. According to the semantic feature vector set Vi of the media data Di', the semantic association index L between each data segment is calculated. The specific calculation formula is: Ljk=(vj·vk) / |vj||vk|, where Ljk represents the semantic association index between the j-th data segment and the k-th data segment, vj represents the semantic feature vector corresponding to the j-th data segment, vk represents the semantic feature vector corresponding to the k-th data segment, and j and k are both 1 to n, j≠k; for each data segment di, combined with the data proportion and the semantic association index, calculate Calculate the corresponding semantic association strength G, and Gi=b×∑j∈[1,n],|Lij|; sort the data segments of each media data Di according to the size relationship of the semantic association strength G, take the data segment with the largest semantic association strength G as the main data segment, and take the data segment with the semantic association strength G not equal to 0 as the secondary data segment; take the data segment with the semantic association strength G equal to 0 as irrelevant data segment and filter it out, so as to obtain media data D; organize the media data D into structured data for storage, so as to obtain the data to be analyzed.

[0007] Further, step S200 includes: S201. Obtain the delivery records corresponding to the target advertisement from the database, and sort out the target advertisement delivery record data set F, and F{f1,f2,...,fm}, where f1 represents the first delivery record of the target advertisement, f2 represents the second delivery record of the target advertisement, and so on, fm represents the mth delivery record of the target advertisement; each delivery record of the target advertisement corresponds to only one delivery strategy, and each delivery record contains multiple features, such as delivery time, delivery platform, delivery cost, delivery area, user interaction data, etc.; for each delivery record fe, extract the corresponding delivery strategy therefrom, and define the extracted delivery strategy as the strategy data set Se, so as to obtain the target user group characteristics, and Se={s1,s2,s3,s4}, where s1 represents the numerical representation of the delivery time, s2 represents the regional code, s3 represents the target user group characteristics, and s4 represents the budget; S202. Obtain user behavior data from the data to be analyzed, and extract features from the user behavior data to construct a feature matrix U, and the dimension of the feature matrix U is q×r, where q represents the number of users and r represents the number of features; use the K-means algorithm to cluster users to identify different user groups, and represent the identified user groups as Pu, where u represents the number of the user group; for each user group, obtain the corresponding number of users, and define the user group with the largest number of users as the actual user group.

[0008] Furthermore, the specific process of clustering users using the K-means algorithm is as follows: The feature matrix U is standardized, k0 cluster centers are selected, and the cluster center is recorded as c. For each user q, the distance from each user q to all cluster centers is calculated, and it is assigned to the nearest cluster. The calculation formula for the distance from each user q to all cluster centers is: d(uh,cp)={Σw∈[1,r],(uh_w-cp_w)}^(1 / 2), Among them, uh represents the feature vector of the h-th user, cp represents the p-th cluster center, uh_w represents the w-th feature value of the h-th user, and cp_w represents the w-th feature value of the p-th cluster center. After all users are assigned, the center C of each cluster is recalculated, and the update formula is: Cp=(1 / |Qp|)Σuh∈Qp,uh, Where Qp represents the feature vectors of all users assigned to cluster p, |Qp| is the number of users in cluster p; Repeat the above until the preset number of iterations is reached.

[0009] Furthermore, step S300 includes: S301. Extracting actual user group characteristics and target user extraction characteristics, standardizing the actual user group characteristics and target user group characteristics, performing correlation analysis on the actual user group characteristics and target user group characteristics, respectively calculating the Pearson correlation coefficient ρ between the actual user group characteristics and the target user group characteristics and the preset advertising delivery effect indicators, comparing the calculated Pearson correlation coefficient ρ with the absolute value of the threshold ρ0, and screening out the actual user group characteristics and the target user group characteristics whose absolute values ​​are greater than the threshold ρ0; S302. Calculate the difference measurement index CY between the actual user group characteristics after screening and the target user group characteristics. The specific calculation formula is: CY=Σb∈[1,g],|x_b-y_b|, Where x_b and y_b represent the feature values ​​of the bth actual user group and the target user group respectively, and g represents the maximum number of the actual user group features and the target user group features after screening; S303. Based on the difference indicators between the corresponding characteristics of the target user group and the actual user group, the delivery effect of the target advertising delivery strategy is evaluated, and the delivery effect evaluation index TF of the target advertising delivery strategy is calculated. The specific calculation formula is: TF=CY×R, where R represents the original advertising effect evaluation function, which can be indicators such as click-through rate and conversion rate, depending on the specific situation.

[0010] Furthermore, step S400 includes: S401. Summarize the delivery effect evaluation index TF corresponding to all delivery strategies of the target advertisement to form a delivery effect evaluation index set M, where M={TF1, TF2, ...., TFm}, where TF1 represents the delivery effect evaluation index corresponding to the first delivery strategy of the target advertisement, TF2 represents the delivery effect evaluation index corresponding to the second delivery strategy of the target advertisement, and so on, TFm represents the delivery effect evaluation index corresponding to the mth delivery strategy of the target advertisement; according to the delivery effect evaluation index set M, calculate the average value TF0 and standard deviation σ of the delivery effect evaluation index of all delivery strategies, and for each delivery strategy of the target advertisement, calculate the corresponding optimization demand coefficient O, and the specific calculation formula is: O=|TF-TF0| / σ; S402. Compare the size relationship between the optimization demand coefficient O corresponding to each delivery strategy and the threshold. When O>T, mark the corresponding delivery strategy as the delivery strategy to be adjusted, and output the number corresponding to the delivery strategy to be adjusted to the relevant personnel, who will make the desired adjustments; when O≤T, mark the corresponding delivery strategy as the target delivery strategy for storage.

[0011] The intelligent analysis system of media information based on big data includes: data collection and processing module, user group feature analysis module, feature comparison and difference analysis module, delivery effect evaluation module and optimization suggestion module; The data collection and processing module obtains media data related to the target advertising delivery from different data channels, identifies and processes the format of the obtained media data, and integrates it into a unified format; the user group feature analysis module obtains the delivery record of the target advertisement, extracts the delivery strategy and features, and forms a target user group feature data set; extracts user behavior features from the data to be analyzed, constructs a feature matrix, uses the K-means algorithm to cluster users, identifies the actual user group and extracts features; the feature comparison and difference analysis module standardizes the actual user group features and the target user group features, calculates the correlation between the actual user group features and the target user group features, and screens out important features; calculates the difference measure between the screened features, and evaluates the feature differences between the actual user group and the target user group; the delivery effect evaluation module calculates the delivery effect evaluation index based on the feature difference index and the original advertising effect evaluation function, summarizes the effect evaluation index of all delivery strategies, and calculates the average and standard deviation; and calculates the optimization demand coefficient of each delivery strategy to evaluate the optimization demand of the delivery strategy; the optimization suggestion module compares the optimization demand coefficient with the set threshold, marks the delivery strategy to be adjusted and the target delivery strategy, outputs the number of the strategy to be adjusted to the relevant personnel, and stores the target delivery strategy.

[0012] Further, the data acquisition and processing module includes a data acquisition unit, a data format recognition unit, and a data integration and storage unit; The data acquisition unit is responsible for acquiring media data related to target advertising from different data channels, including text, image, audio and video formats; the data format recognition unit identifies the format of the acquired media data, analyzes the type of each data segment, and converts it into a unified format; the data integration and storage unit organizes the processed media data into structured data, stores it, and forms a data set to be analyzed.

[0013] Furthermore, the user group feature analysis module includes a delivery record extraction unit and a user behavior data analysis unit; The delivery record extraction unit obtains the delivery record of the target advertisement from the database, extracts the delivery strategy and the characteristics of the target user group, and forms a target characteristic data set; the user behavior data analysis unit analyzes the user behavior data in the data to be analyzed, constructs a characteristic matrix, and uses the K-means algorithm to cluster users and identify the actual user group; The feature comparison and difference analysis module includes a correlation analysis unit and a difference measurement calculation unit; The correlation analysis unit normalizes the actual user group characteristics and the target user group characteristics, and calculates the Pearson correlation coefficient between them and the advertising effect index to screen out important characteristics; the difference measurement calculation unit calculates the difference measurement index between the screened actual user group characteristics and the target user group characteristics.

[0014] Furthermore, the delivery effect evaluation module includes an effect evaluation index calculation unit and an optimization demand analysis unit; The effect evaluation index calculation unit calculates the corresponding effect evaluation index based on the difference measurement index, summarizes the effect evaluation index of each delivery strategy, and calculates the average value and standard deviation; the optimization demand analysis unit calculates the optimization demand coefficient of each delivery strategy based on the comparison between the effect evaluation index and the threshold, and marks the delivery strategy that needs to be adjusted: The optimization suggestion module includes an adjustment strategy suggestion unit and a strategy storage unit; The adjustment strategy suggestion unit outputs the label of the delivery strategy to be adjusted to relevant personnel according to the optimization demand analysis result; the strategy storage unit stores the record marked as the target delivery strategy.

[0015] Compared with the prior art, the present invention has the following beneficial effects: the present invention can process media data in various formats (text, images, audio and video) and integrate them into a unified format. This multi-dimensional data integration can more comprehensively capture the contextual information of advertising delivery and enhance the depth and breadth of analysis; by performing semantic analysis on media data and calculating the semantic association strength, the present invention can identify the semantic connection between each data segment, so as to better understand the user's real reaction to the advertising content. This semantic association analysis can provide empirical evidence for optimizing the advertising content; the K-means algorithm is used to cluster users, which can effectively identify different user groups, making advertising The delivery strategy can be matched to actual users in a more targeted manner, thereby improving the precise delivery effect of advertisements; the present invention extracts the characteristics of the actual user group and compares them with the characteristics of the target user group, thereby identifying differences in user behavior and providing a more comprehensive evaluation of the advertising effect; the present invention provides a scientific evaluation model by calculating the difference measurement index and the delivery effect evaluation index, which can clearly reflect the effect of each delivery strategy and provide a basis for subsequent delivery strategy adjustments; by calculating the optimization demand coefficient, the delivery strategy that needs to be adjusted can be identified in time to ensure that the advertising delivery is always kept in the optimal state. This dynamic adjustment mechanism enhances the flexibility and effectiveness of advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 It is a schematic diagram of the module of the media information intelligent analysis system based on big data of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 , the present invention provides a technical solution: The intelligent analysis system of media information based on big data includes: data collection and processing module, user group feature analysis module, feature comparison and difference analysis module, delivery effect evaluation module and optimization suggestion module; The data collection and processing module obtains media data related to the target advertising delivery from different data channels, identifies and processes the format of the obtained media data, and integrates it into a unified format; the user group feature analysis module obtains the delivery record of the target advertisement, extracts the delivery strategy and features, and forms a target user group feature data set; extracts user behavior features from the data to be analyzed, constructs a feature matrix, uses the K-means algorithm to cluster users, identifies the actual user group and extracts features; the feature comparison and difference analysis module standardizes the actual user group features and the target user group features, calculates the correlation between the actual user group features and the target user group features, and screens out important features; calculates the difference measure between the screened features, and evaluates the feature differences between the actual user group and the target user group; the delivery effect evaluation module calculates the delivery effect evaluation index based on the feature difference index and the original advertising effect evaluation function, summarizes the effect evaluation index of all delivery strategies, and calculates the average and standard deviation; and calculates the optimization demand coefficient of each delivery strategy to evaluate the optimization demand of the delivery strategy; the optimization suggestion module compares the optimization demand coefficient with the set threshold, marks the delivery strategy to be adjusted and the target delivery strategy, outputs the number of the strategy to be adjusted to the relevant personnel, and stores the target delivery strategy.

[0019] The data acquisition and processing module includes a data acquisition unit, a data format recognition unit, and a data integration and storage unit; The data acquisition unit is responsible for acquiring media data related to target advertising from different data channels, including text, image, audio and video formats; the data format recognition unit identifies the format of the acquired media data, analyzes the type of each data segment, and converts it into a unified format; the data integration and storage unit organizes the processed media data into structured data, stores it, and forms a data set to be analyzed.

[0020] The user group feature analysis module includes a delivery record extraction unit and a user behavior data analysis unit; The delivery record extraction unit obtains the delivery record of the target advertisement from the database, extracts the delivery strategy and the characteristics of the target user group, and forms a target characteristic data set; the user behavior data analysis unit analyzes the user behavior data in the data to be analyzed, constructs a characteristic matrix, and uses the K-means algorithm to cluster users and identify the actual user group; The feature comparison and difference analysis module includes a correlation analysis unit and a difference measurement calculation unit; The correlation analysis unit normalizes the actual user group characteristics and the target user group characteristics, and calculates the Pearson correlation coefficient between them and the advertising effect index to screen out important characteristics; the difference measurement calculation unit calculates the difference measurement index between the screened actual user group characteristics and the target user group characteristics.

[0021] The delivery effect evaluation module includes an effect evaluation index calculation unit and an optimization demand analysis unit; The effect evaluation index calculation unit calculates the corresponding effect evaluation index based on the difference measurement index, summarizes the effect evaluation index of each delivery strategy, and calculates the average value and standard deviation; the optimization demand analysis unit calculates the optimization demand coefficient of each delivery strategy based on the comparison between the effect evaluation index and the threshold, and marks the delivery strategy that needs to be adjusted: The optimization suggestion module includes an adjustment strategy suggestion unit and a strategy storage unit; The adjustment strategy suggestion unit outputs the label of the delivery strategy to be adjusted to relevant personnel according to the optimization demand analysis result; the strategy storage unit stores the record marked as the target delivery strategy.

[0022] The method for intelligent analysis of media information based on big data includes the following steps: Step S100. Acquire media data related to target advertisement delivery in different formats from different data channels, perform format recognition on the acquired media data, and process the media data accordingly according to the result of the format recognition, so as to integrate the media data into a unified format, thereby obtaining data to be analyzed; Step S200. Obtain the delivery record corresponding to the target advertisement, extract the delivery strategy of the target advertisement from the delivery record, and obtain the target user group characteristics according to the delivery strategy of the target advertisement; analyze the data to be analyzed, thereby identifying the actual user group; Step S300: extract the characteristics of the actual user group and compare and analyze them with the characteristics of the target user group to obtain the difference between the two; combine the differences between the corresponding characteristics of the target user group and the actual user group to evaluate the delivery effect of the target advertising delivery strategy; Step S400. Analyze the optimization demand coefficient of the delivery strategy corresponding to the target advertisement according to the delivery effect evaluation index of the target advertisement; analyze the corresponding delivery strategy according to the optimization demand coefficient, and output corresponding prompt information based on the analysis result.

[0023] Step S100 includes: S101. Obtain media data related to target advertising in different formats from different data channels, wherein the formats of the media data include text, image, audio and video formats, and the data format corresponding to each piece of media data is not unique; for example, a social media post may have text description, image, audio and video; for each piece of media data Di, perform format identification, where i represents the media data number, which is a positive integer; according to the result of the format identification, divide the current media data, so that the media data Di is represented as: Di={d1,d2,...,dn}i, where d1 represents the first data segment of the media data Di, d2 represents the second data segment of the media data Di, and so on, dn represents the nth data segment of the media data Di, where n represents the data format number corresponding to the media data Di, which is 1 to 4, representing text, image, audio and video formats respectively; S102. For each data segment in the image, audio and video format in the media data Di, it is converted into text format, and the converted text format data replaces the corresponding data segment in the image, audio and video format, so that the media data is represented as Di', and the data proportion b of each data segment of the media data Di' is calculated, and b=N_dj / N_Di', wherein N_dj represents the number of bytes of the j-th data segment in the media data Di', and N_Di' represents the number of bytes of the media data Di'; semantic analysis is performed on each data segment of the media data Di', so as to form a corresponding semantic feature vector set Vi, and V={v1,v2,...,vn}i, similarly, v1 represents the semantic feature vector of the first data segment, v2 represents the semantic feature vector of the second data segment, and vn represents the semantic feature vector of the n-th data segment, and each semantic feature vector contains multi-dimensional information, such as keywords, themes, etc.; S103. According to the semantic feature vector set Vi of the media data Di', the semantic association index L between each data segment is calculated. The specific calculation formula is: Ljk=(vj·vk) / |vj||vk|, where Ljk represents the semantic association index between the j-th data segment and the k-th data segment, vj represents the semantic feature vector corresponding to the j-th data segment, vk represents the semantic feature vector corresponding to the k-th data segment, and j and k are both 1 to n, j≠k; for each data segment di, combined with the data proportion and the semantic association index, calculate Calculate the corresponding semantic association strength G, and Gi=b×∑j∈[1,n],|Lij|; sort the data segments of each media data Di according to the size relationship of the semantic association strength G, take the data segment with the largest semantic association strength G as the main data segment, and take the data segment with the semantic association strength G not equal to 0 as the secondary data segment; take the data segment with the semantic association strength G equal to 0 as irrelevant data segment and filter it out, so as to obtain media data D; organize the media data D into structured data for storage, so as to obtain the data to be analyzed.

[0024] In this embodiment, an advertising company wants to analyze the advertising effectiveness on a social media platform, and the acquired data includes text posts, pictures, audio comments, and video ads.

[0025] Multiple social media channels (such as Twitter, Instagram, Facebook) capture advertising-related data. The format of each piece of media data Di may include: Text: descriptive text, Image: advertising image, Audio: user comment recording, Video: advertising video clip; Analyze each piece of data through the format recognition tool and identify the respective data segments, for example: D1={d1,d2,d3}1; Convert the image and audio data segments into text format to obtain D1', and convert d2 into descriptive text and d3 into transcribed text; Calculate the number of bytes in each data segment and the proportion b of each data segment: For example, b1=N_d1 / N_D1'; Perform semantic analysis on each data segment to generate a feature vector set V1, where V1={v1,v2,v3}1; Calculate the semantic association index L, for example: L12=(v1·v2) / |v1||v2|; Calculate the semantic association strength G: Combine the proportion and correlation index to calculate the strength of each data segment, for example: G1=b1×(∣L11∣+∣L12∣+∣L13∣).

[0026] Sort all data segments by G value: determine the main data segments, secondary data segments and irrelevant data segments.

[0027] The primary and secondary data segments are organized into structured data and saved as database tables or CSV files to obtain the data to be analyzed.

[0028] Step S200 includes: S201. Obtain the delivery records corresponding to the target advertisement from the database, and sort out the target advertisement delivery record data set F, and F{f1,f2,...,fm}, where f1 represents the first delivery record of the target advertisement, f2 represents the second delivery record of the target advertisement, and so on, fm represents the mth delivery record of the target advertisement; each delivery record of the target advertisement corresponds to only one delivery strategy, and each delivery record contains multiple features, such as delivery time, delivery platform, delivery cost, delivery area, user interaction data, etc.; for each delivery record fe, extract the corresponding delivery strategy therefrom, and define the extracted delivery strategy as the strategy data set Se, so as to obtain the target user group characteristics, and Se={s1,s2,s3,s4}, where s1 represents the numerical representation of the delivery time, s2 represents the regional code, s3 represents the target user group characteristics, and s4 represents the budget; S202. Obtain user behavior data from the data to be analyzed, and extract features from the user behavior data to construct a feature matrix U, and the dimension of the feature matrix U is q×r, where q represents the number of users and r represents the number of features; use the K-means algorithm to cluster users to identify different user groups, and represent the identified user groups as Pu, where u represents the number of the user group; for each user group, obtain the corresponding number of users, and define the user group with the largest number of users as the actual user group.

[0029] The specific process of clustering users using the K-means algorithm is as follows: The feature matrix U is standardized, k0 cluster centers are selected, and the cluster center is recorded as c. For each user q, the distance from each user q to all cluster centers is calculated, and it is assigned to the nearest cluster. The calculation formula for the distance from each user q to all cluster centers is: d(uh,cp)={Σw∈[1,r],(uh_w-cp_w)}^(1 / 2), Among them, uh represents the feature vector of the h-th user, cp represents the p-th cluster center, uh_w represents the w-th feature value of the h-th user, and cp_w represents the w-th feature value of the p-th cluster center. After all users are assigned, the center C of each cluster is recalculated, and the update formula is: Cp=(1 / |Qp|)Σuh∈Qp,uh, Where Qp represents the feature vectors of all users assigned to cluster p, |Qp| is the number of users in cluster p; Repeat the above until the preset number of iterations is reached.

[0030] Step S300 includes: S301. Extracting actual user group characteristics and target user extraction characteristics, standardizing the actual user group characteristics and target user group characteristics, performing correlation analysis on the actual user group characteristics and target user group characteristics, respectively calculating the Pearson correlation coefficient ρ between the actual user group characteristics and the target user group characteristics and the preset advertising delivery effect indicators, comparing the calculated Pearson correlation coefficient ρ with the absolute value of the threshold ρ0, and screening out the actual user group characteristics and the target user group characteristics whose absolute values ​​are greater than the threshold ρ0; S302. Calculate the difference measurement index CY between the actual user group characteristics after screening and the target user group characteristics. The specific calculation formula is: CY=Σb∈[1,g],|x_b-y_b|, Where x_b and y_b represent the feature values ​​of the bth actual user group and the target user group respectively, and g represents the maximum number of the actual user group features and the target user group features after screening; S303. Based on the difference indicators between the corresponding characteristics of the target user group and the actual user group, the delivery effect of the target advertising delivery strategy is evaluated, and the delivery effect evaluation index TF of the target advertising delivery strategy is calculated. The specific calculation formula is: TF=CY×R, where R represents the original advertising effect evaluation function, which can be indicators such as click-through rate and conversion rate, depending on the specific situation.

[0031] In this embodiment, it is assumed that the original advertising effect evaluation function is composed of the following indicators: click-through rate CTR, conversion rate CVR and return on investment ROI. According to the above indicators, a weighted comprehensive effect evaluation function R is defined, and R=w1×CTR+w2×CVR+w3×ROI, where w1, w2 and w3 represent the weights of each indicator. Therefore, the calculation formula of the delivery effect evaluation index TF of the target advertising delivery strategy is: TF=CY×R=CY×(w1×CTR+w2×CVR+w3×ROI).

[0032] Step S400 includes: S401. Summarize the delivery effect evaluation index TF corresponding to all delivery strategies of the target advertisement to form a delivery effect evaluation index set M, where M={TF1, TF2, ...., TFm}, where TF1 represents the delivery effect evaluation index corresponding to the first delivery strategy of the target advertisement, TF2 represents the delivery effect evaluation index corresponding to the second delivery strategy of the target advertisement, and so on, TFm represents the delivery effect evaluation index corresponding to the mth delivery strategy of the target advertisement; according to the delivery effect evaluation index set M, calculate the average value TF0 and standard deviation σ of the delivery effect evaluation index of all delivery strategies, and for each delivery strategy of the target advertisement, calculate the corresponding optimization demand coefficient O, and the specific calculation formula is: O=|TF-TF0| / σ; S402. Compare the size relationship between the optimization demand coefficient O corresponding to each delivery strategy and the threshold. When O>T, mark the corresponding delivery strategy as the delivery strategy to be adjusted, and output the number corresponding to the delivery strategy to be adjusted to the relevant personnel, who will make the desired adjustments; when O≤T, mark the corresponding delivery strategy as the target delivery strategy for storage.

[0033] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0034] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent analysis method of media information based on big data, characterized by: The method comprises the following steps: Step S100. Acquire media data related to target advertisement delivery in different formats from different data channels, perform format recognition on the acquired media data, and process the media data accordingly according to the result of the format recognition, so as to integrate the media data into a unified format, thereby obtaining data to be analyzed; Step S200. Obtain the delivery record corresponding to the target advertisement, extract the delivery strategy of the target advertisement from the delivery record, and obtain the target user group characteristics according to the delivery strategy of the target advertisement; analyze the data to be analyzed, thereby identifying the actual user group; Step S300: extract the characteristics of the actual user group and compare and analyze them with the characteristics of the target user group to obtain the difference between the two; combine the differences between the corresponding characteristics of the target user group and the actual user group to evaluate the delivery effect of the target advertising delivery strategy; Step S400. Analyze the optimization demand coefficient of the delivery strategy corresponding to the target advertisement according to the delivery effect evaluation index of the target advertisement; analyze the corresponding delivery strategy according to the optimization demand coefficient, and output corresponding prompt information based on the analysis result.

2. The method for intelligent analysis of media information based on big data according to claim 1, characterized in that: The step S100 includes: S101. Obtain media data related to target advertising in different formats from different data channels, wherein the formats of the media data include text, image, audio and video formats, and the data format corresponding to each piece of media data is not unique; perform format identification on each piece of media data Di, wherein i represents the media data number, which is a positive integer; divide the current media data according to the result of the format identification, so that the media data Di is represented as: Di={d1,d2,...,dn}i, wherein d1 represents the first data segment of the media data Di, d2 represents the second data segment of the media data Di, and so on, dn represents the nth data segment of the media data Di, wherein n represents the data format number corresponding to the media data Di, which is 1 to 4, representing text, image, audio and video formats respectively; S102. For each data segment in the image, audio and video format in the media data Di, convert them into text format, replace the corresponding data segment in the image, audio and video format with the converted text format data, so as to represent the media data as Di', calculate the data proportion b of each data segment of the media data Di', and b=N_dj / N_Di', wherein N_dj represents the number of bytes of the j-th data segment in the media data Di', and N_Di' represents the number of bytes of the media data Di'; perform semantic analysis on each data segment of the media data Di', so as to form a corresponding semantic feature vector set Vi, and V={v1,v2,...,vn}i, similarly, v1 represents the semantic feature vector of the first data segment, v2 represents the semantic feature vector of the second data segment, vn represents the semantic feature vector of the n-th data segment, and each semantic feature vector contains multi-dimensional information; S103. According to the semantic feature vector set Vi of the media data Di', the semantic association index L between each data segment is calculated. The specific calculation formula is: Ljk=(vj·vk) / |vj||vk|, where Ljk represents the semantic association index between the j-th data segment and the k-th data segment, vj represents the semantic feature vector corresponding to the j-th data segment, vk represents the semantic feature vector corresponding to the k-th data segment, and j and k are both 1 to n, j≠k; for each data segment di, combined with the data proportion and the semantic association index, calculate Calculate the corresponding semantic association strength G, and Gi=b×∑j∈[1,n],|Lij|; sort the data segments of each media data Di according to the size relationship of the semantic association strength G, take the data segment with the largest semantic association strength G as the main data segment, and take the data segment with the semantic association strength G not equal to 0 as the secondary data segment; take the data segment with the semantic association strength G equal to 0 as irrelevant data segment and filter it out, so as to obtain media data D; organize the media data D into structured data for storage, so as to obtain the data to be analyzed.

3. The method for intelligent analysis of media information based on big data according to claim 1, characterized in that: The step S200 includes: S201. Obtain the delivery records corresponding to the target advertisement from the database, and sort out the target advertisement delivery record data set F, and F{f1,f2,...,fm}, where f1 represents the first delivery record of the target advertisement, f2 represents the second delivery record of the target advertisement, and so on, fm represents the mth delivery record of the target advertisement; each delivery record of the target advertisement corresponds to only one delivery strategy, and each delivery record contains multiple features; for each delivery record fe, extract the corresponding delivery strategy therefrom, and define the extracted delivery strategy as the strategy data set Se, so as to obtain the target user group characteristics, and Se={s1,s2,s3,s4}, where s1 represents the numerical representation of the delivery time, s2 represents the regional code, s3 represents the target user group characteristics, and s4 represents the budget; S202. Obtain user behavior data from the data to be analyzed, and extract features from the user behavior data to construct a feature matrix U, and the dimension of the feature matrix U is q×r, where q represents the number of users and r represents the number of features; use the K-means algorithm to cluster users to identify different user groups, and represent the identified user groups as Pu, where u represents the number of the user group; for each user group, obtain the corresponding number of users, and define the user group with the largest number of users as the actual user group.

4. The method for intelligent analysis of media information based on big data according to claim 3 is characterized in that: The specific process of clustering users using the K-means algorithm is as follows: The feature matrix U is standardized, k0 cluster centers are selected, and the cluster center is recorded as c. For each user q, the distance from each user q to all cluster centers is calculated, and it is assigned to the nearest cluster. The calculation formula for the distance from each user q to all cluster centers is: d(uh,cp)={Σw∈[1,r],(uh_w-cp_w)}^(1 / 2), Among them, uh represents the feature vector of the h-th user, cp represents the p-th cluster center, uh_w represents the w-th feature value of the h-th user, and cp_w represents the w-th feature value of the p-th cluster center. After all users are assigned, the center C of each cluster is recalculated, and the update formula is: Cp=(1 / |Qp|)Σuh∈Qp,uh, Where Qp represents the feature vectors of all users assigned to cluster p, |Qp| is the number of users in cluster p; Repeat the above until the preset number of iterations is reached.

5. The method for intelligent analysis of media information based on big data according to claim 1, characterized in that: The step S300 includes: S301. Extracting actual user group characteristics and target user extraction characteristics, standardizing the actual user group characteristics and target user group characteristics, performing correlation analysis on the actual user group characteristics and target user group characteristics, respectively calculating the Pearson correlation coefficient ρ between the actual user group characteristics and the target user group characteristics and the preset advertising delivery effect indicators, comparing the calculated Pearson correlation coefficient ρ with the absolute value of the threshold ρ0, and screening out the actual user group characteristics and the target user group characteristics whose absolute values ​​are greater than the threshold ρ0; S302. Calculate the difference measurement index CY between the actual user group characteristics after screening and the target user group characteristics. The specific calculation formula is: CY=Σb∈[1,g],|x_b-y_b|, Where x_b and y_b represent the feature values ​​of the bth actual user group and the target user group respectively, and g represents the maximum number of the actual user group features and the target user group features after screening; S303. Based on the difference indicators between the corresponding characteristics of the target user group and the actual user group, the delivery effect of the target advertising delivery strategy is evaluated, and the delivery effect evaluation index TF of the target advertising delivery strategy is calculated. The specific calculation formula is: TF=CY×R, where R represents the original advertising effect evaluation function.

6. The method for intelligent analysis of media information based on big data according to claim 5, characterized in that: The step S400 includes: S401. Summarize the delivery effect evaluation index TF corresponding to all delivery strategies of the target advertisement to form a delivery effect evaluation index set M, where M={TF1, TF2, ...., TFm}, where TF1 represents the delivery effect evaluation index corresponding to the first delivery strategy of the target advertisement, TF2 represents the delivery effect evaluation index corresponding to the second delivery strategy of the target advertisement, and so on, TFm represents the delivery effect evaluation index corresponding to the mth delivery strategy of the target advertisement; according to the delivery effect evaluation index set M, calculate the average value TF0 and standard deviation σ of the delivery effect evaluation index of all delivery strategies, and for each delivery strategy of the target advertisement, calculate the corresponding optimization demand coefficient O, and the specific calculation formula is: O=|TF-TF0| / σ; S402. Compare the size relationship between the optimization demand coefficient O corresponding to each delivery strategy and the threshold. When O>T, mark the corresponding delivery strategy as the delivery strategy to be adjusted, and output the number corresponding to the delivery strategy to be adjusted to the relevant personnel, who will make the desired adjustments; when O≤T, mark the corresponding delivery strategy as the target delivery strategy for storage.

7. A media information intelligent analysis system based on big data, applied to the media information intelligent analysis based on big data as claimed in any one of claims 1 to 6, characterized in that: The system includes: a data collection and processing module, a user group feature analysis module, a feature comparison and difference analysis module, an advertising effect evaluation module, and an optimization suggestion module; The data acquisition and processing module obtains media data related to the target advertisement delivery from different data channels, performs format recognition and processing on the obtained media data, and integrates them into a unified format; the user group feature analysis module obtains the delivery record of the target advertisement, extracts the delivery strategy and features, and forms a target user group feature data set; extracts user behavior features from the data to be analyzed, constructs a feature matrix, clusters users using the K-means algorithm, identifies the actual user group and performs feature extraction; the feature comparison and difference analysis module performs standardization processing on the actual user group features and the target user group features, calculates the correlation between the actual user group features and the target user group features, and screens out important features; calculates the difference measure between the screened features, and evaluates the feature difference between the actual user group and the target user group; the delivery effect evaluation module calculates the delivery effect evaluation index according to the feature difference index and the original advertisement effect evaluation function, summarizes the effect evaluation index of all delivery strategies, calculates the average value and standard deviation; and calculates the optimization demand coefficient of each delivery strategy, and evaluates the optimization demand of the delivery strategy; the optimization suggestion module compares the optimization demand coefficient with the set threshold, marks the delivery strategy to be adjusted and the target delivery strategy, outputs the number of the strategy to be adjusted to the relevant personnel, and stores the target delivery strategy.

8. The intelligent media information analysis system based on big data according to claim 7 is characterized by: The data acquisition and processing module includes a data acquisition unit, a data format recognition unit, and a data integration and storage unit; The data acquisition unit is responsible for acquiring media data related to target advertising from different data channels, including text, image, audio and video formats; the data format identification unit identifies the format of the acquired media data, analyzes the type of each data segment, and converts it into a unified format; the data integration and storage unit organizes the processed media data into structured data, stores it, and forms a data set to be analyzed.

9. The intelligent media information analysis system based on big data according to claim 7, characterized in that: The user group feature analysis module includes a delivery record extraction unit and a user behavior data analysis unit; The delivery record extraction unit obtains the delivery record of the target advertisement from the database, extracts the delivery strategy and the target user group characteristics, and forms a target characteristic data set; the user behavior data analysis unit analyzes the user behavior data in the data to be analyzed, constructs a characteristic matrix, and uses the K-means algorithm to cluster the users to identify the actual user group; The feature comparison and difference analysis module includes a correlation analysis unit and a difference metric calculation unit; The correlation analysis unit normalizes the actual user group characteristics and the target user group characteristics, and calculates the Pearson correlation coefficient between them and the advertising delivery effect index to screen out important characteristics; the difference measurement calculation unit calculates the difference measurement index between the screened actual user group characteristics and the target user group characteristics.

10. The media information intelligent analysis system based on big data according to claim 7, characterized in that: The delivery effect evaluation module includes an effect evaluation index calculation unit and an optimization demand analysis unit; The effect evaluation index calculation unit calculates the corresponding effect evaluation index according to the difference measurement index, summarizes the effect evaluation index of each delivery strategy, and calculates the average value and standard deviation; the optimization demand analysis unit calculates the optimization demand coefficient of each delivery strategy according to the comparison between the effect evaluation index and the threshold value, and marks the delivery strategy that needs to be adjusted: The optimization suggestion module includes an adjustment strategy suggestion unit and a strategy storage unit; The adjustment strategy suggestion unit outputs the label of the delivery strategy to be adjusted to relevant personnel according to the optimization demand analysis result; and the strategy storage unit stores the record marked as the target delivery strategy.

Images