Digital marketing method, system, storage medium and computer

Abstract

The application provides a digital marketing method, system, storage medium and computer, and the marketing method comprises the following steps: acquiring multi-dimensional data of a customer group and performing encryption processing, and integrating the encrypted data; performing feature extraction on the integrated data based on a distributed big data calculation framework; setting labels for customers in the customer group based on the extracted features, forming a to-be-screened customer group, performing intelligent matching of products according to the customers in the to-be-screened customer group, and forming a matching white list of product and customer adaptation; constructing a marketing model, evaluating and analyzing the matching white list according to the marketing model, quantifying the purchase intention of the customers, and screening out intended customers for product marketing recommendation. The digital marketing method provided by the application can accurately acquire target intended customers and improve the marketing conversion rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital marketing, and particularly relates to a digital marketing method, system, storage medium and computer. BACKGROUND

[0002] Marketing is the process of creating, delivering and exchanging products or value to achieve organizational goals. Its core is to deliver products or services from producers to consumers through effective communication and exchange, and to establish long-term customer relationships.

[0003] At present, traditional marketing includes static marketing and dynamic marketing; static marketing mainly includes passive marketing methods such as advertisement placement, which is difficult to respond to market changes in real time, and the advertisement placement efficiency is low, and the marketing is over-reliant on manual work, the marketing efficiency is low and the cost is high, the target intention customers cannot be accurately obtained, and the conversion rate is low; dynamic marketing cannot comprehensively capture customer cross-platform behavior characteristics by obtaining single platform data such as e-commerce sales records of customers, it is difficult to analyze multi-industry customer behavior information, and it is difficult to accurately obtain target intention customers. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a digital marketing method, system and storage medium to solve the technical problems in the prior art.

[0005] The present application provides a digital marketing method, which comprises the following steps:

[0006] Based on the multi-source data engine, multi-dimensional data of the customer group is obtained, and the multi-dimensional data is dynamically desensitized and encrypted;

[0007] The encrypted multi-dimensional data is integrated, and the integrated data is feature extracted based on a distributed big data computing framework to capture key information and dynamic characteristics in the multi-dimensional data;

[0008] Based on the key information and dynamic characteristics in the multi-dimensional data, the customers in the customer group are labeled to form a to-be-screened customer group, and the customers in the to-be-screened customer group are used for intelligent matching of products to form a matching white list of product and customer adaptation;

[0009] A marketing model is constructed, the matching white list is evaluated and analyzed according to the marketing model to quantify the purchase intention of the customers, and the intention customers are screened out for product marketing recommendation.

[0010] Preferably, the step of obtaining multi-dimensional data of the customer group based on the multi-source data engine and dynamically desensitizing and encrypting the multi-dimensional data comprises:

[0011] The multi-source data engine is used for incremental extraction and full extraction in the data source to obtain multi-dimensional data of the customer group, and the multi-dimensional data at least includes regional data, time data, brand category data, online sales data and offline sales data.

[0012] An agent model is constructed, and the agent model at least includes an agent end and a server end.

[0013] Based on the multi-dimensional data, a request message is sent to the agent end, and the agent end forwards the request message to the server end after receiving the request message.

[0014] Sensitive information is queried in the server end, and a result set message containing sensitive information is returned to the agent end, the agent end queries the nearest record from the session time table according to the current time, and obtains the corresponding customer name according to the session ID in the record.

[0015] The agent end obtains the desensitization scheme corresponding to the current customer name from the desensitization configuration table according to the corresponding customer name, and performs data desensitization, assembles the desensitized data to form a result set message, and completes dynamic desensitization and encryption of data.

[0016] Preferably, the step of integrating the encrypted multi-dimensional data, extracting features of the integrated data based on a distributed big data computing framework, and capturing key information and dynamic features in the multi-dimensional data comprises:

[0017] The encrypted multi-dimensional data is de-duplicated, filled, and abnormal value tested to complete data cleaning, and the cleaned data is standardized, discretized, and layered to complete data conversion.

[0018] The converted data is batch-loaded to a distributed storage area based on an ETL tool, features corresponding to the multi-dimensional data are extracted and integrated, and the integrated multi-dimensional data is visually processed to obtain a visual image corresponding to the key information and dynamic features in the multi-dimensional data.

[0019] Preferably, the step of setting a label for a customer in the customer group based on the key information and dynamic features in the multi-dimensional data to form a to-be-screened customer group, and intelligently matching a product according to the customer in the to-be-screened customer group comprises:

[0020] The visual image is input into an IKG module, the visual image is received by an image encoder in the IKG module and features are extracted to generate corresponding serialized image features.

[0021] An image decoder predicts from a starting sequence in the serialized image features, gradually predicts a text description sequence corresponding to the visual image based on a greedy algorithm, obtains a token sequence text, and cleanses the token sequence text to obtain a to-be-selected vocabulary.

[0022] Encode all word-meta information in the candidate vocabulary based on the text encoder in the IKG module, perform image-text similarity calculation on the encoded word-meta information and the serialized image features, screen out several relevant keywords and their weights, and use the several keywords as labels corresponding to customers in the customer group to form a customer group to be screened;

[0023] Obtain the product tag library, calculate the similarity between all keywords corresponding to customers in the customer group to be screened and the tag library, obtain the similarity matrix, add the weights corresponding to the keywords to the similarity matrix, obtain the probability distribution between products and customers, and determine the intelligent matching results based on the probability distribution.

[0024] Preferably, the expression of the serialized image feature is:

[0025]

[0026] Where, Represent the global and local features of the visualized image, respectively. represents the image encoder, A visualization image representing the input;

[0027] The expression for step-by-step prediction of the greedy algorithm is:

[0028]

[0029]

[0030]

[0031] Where, Indicates the The output of the image decoder after the linear layer and normalization is predicted. Indicates the The result after the self-attention layer and residual connection when the image decoder predicts, Indicates the When predicting an image decoder, it fuses the global features of the visual image and the intermediate results of the cross-attention layer; represents a multi-head self-attention layer, represents a multi-head cross attention layer, represents a multilayer perceptron, Represents a normalization layer.

[0032] The present invention also proposes a digital marketing system, comprising:

[0033] An encryption module is used to obtain multi-dimensional data of customer groups based on a multi-source data engine and perform dynamic desensitization and encryption processing on the multi-dimensional data;

[0034] The capture module is configured to integrate the encrypted multi-dimensional data, extract features of the integrated data based on a distributed big data computing framework, and capture key information and dynamic features in the multi-dimensional data.

[0035] The matching module is configured to set labels for customers in the customer group based on the key information and dynamic features in the multi-dimensional data, form a to-be-screened customer group, perform intelligent product matching on customers in the to-be-screened customer group, and form a matching whitelist of product and customer adaptation.

[0036] The recommendation module is configured to build a marketing model, evaluate and analyze the matching whitelist based on the marketing model, quantify the purchase intention of the customers, and recommend product marketing to the customers with the intention.

[0037] Preferably, the encryption module comprises:

[0038] The extraction unit is configured to perform incremental extraction and full-quantity extraction in the data source based on the multi-source data engine to obtain multi-dimensional data of the customer group, wherein the multi-dimensional data at least includes regional data, time data, brand type data, online sales data, and offline sales data.

[0039] The construction unit is configured to build a proxy model, wherein the proxy model at least includes a proxy end and a service end.

[0040] The sending unit is configured to send a request message to the proxy end based on the multi-dimensional data, and the proxy end forwards the request message to the service end after receiving the request message.

[0041] The query unit is configured to query sensitive information in the service end, return a result set message containing the sensitive information to the proxy end, query the nearest record from a session time table according to the proxy end, and obtain the corresponding customer name according to the record.

[0042] The desensitization unit is configured to obtain a desensitization scheme corresponding to the current customer name from a desensitization configuration table based on the proxy end, perform data desensitization according to the desensitization scheme, assemble the desensitized data to form a result set message, and complete dynamic desensitization and encryption of the data.

[0043] Preferably, the capture module comprises:

[0044] The processing unit is configured to perform de-duplication, filling, and outlier detection on the encrypted multi-dimensional data to complete data cleaning, and perform standardization, discretization, and layering on the cleaned data to complete data conversion.

[0045] The visualization unit is used for loading the converted data to a distributed storage area in batches based on the ETL tool, extracting and integrating features corresponding to the multi-dimensional data respectively, and performing visualization processing on the integrated multi-dimensional data to obtain a visualization image corresponding to key information and dynamic characteristics in the multi-dimensional data.

[0046] The application further provides a storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the digital marketing method.

[0047] The application further provides a computer comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the digital marketing method when executing the computer program.

[0048] Compared with the prior art, the application provides a dynamic active marketing method, first, multi-dimensional data of a customer group is obtained through a multi-source data engine, and the multi-dimensional data is subjected to desensitization and encryption processing, a security line is built from a data storage source, security protection is provided for data assets of a business system in a whole life cycle, and safe circulation and application of data are ensured under a compliance framework; the encrypted multi-dimensional data is integrated, multi-dimensional data can capture customer cross-platform behavior characteristics comprehensively, and target customer selection accuracy is improved; customers in the customer group are labeled through the integrated data, then the customers with the set labels are subjected to intelligent product matching, and a matching white list of product and customer adaptation is formed; an intended customer is accurately positioned through the matching white list, a customer volume of subsequent marketing is greatly reduced, subsequent marketing reach rate and conversion rate are improved, a marketing model is constructed, the matching white list is evaluated and analyzed according to the marketing model, purchase intention of customers is quantified, and intended customers are screened out for product marketing recommendation; the digital marketing method provided by the application is accurate in product-customer matching, and has high marketing reach rate and conversion rate; and the application is suitable for wide range of popularization.

[0049] Additional aspects and advantages of the application will be given, partially in the following description, partially will become obvious from the following description, or will be learned by practicing the application. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A flowchart of the digital marketing method in the first embodiment of the application;

[0051] Figure 2 A structure block diagram of the computer in the fourth embodiment of the application.

[0052] The following specific embodiments will further illustrate the application in combination with the above drawings. DETAILED DESCRIPTION

[0053] For the purpose of promoting an understanding of the application, the application will now be described in greater detail with reference to the figures. Several embodiments of the application are depicted in the figures. The application may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0055] Embodiment one

[0056] Referring to Figure 1 , a digital marketing method in the first embodiment of the application is shown, and specifically, the digital marketing method specifically includes steps S10 to S40:

[0057] S10, obtaining multi-dimensional data of a customer group based on a multi-source data engine, and performing dynamic desensitization and encryption processing on the multi-dimensional data;

[0058] Optionally, the step of obtaining multi-dimensional data of a customer group based on a multi-source data engine and performing dynamic desensitization and encryption processing on the multi-dimensional data includes:

[0059] The multi-source data engine performs incremental extraction and full-amount extraction in the data source to obtain multi-dimensional data of a customer group, and the multi-dimensional data at least includes regional data, time data, brand type data, online sales data, and offline sales data;

[0060] An agent model is constructed, and the agent model at least includes an agent end and a service end;

[0061] A request message is sent to the agent end based on the multi-dimensional data, and the agent end forwards the request message to the service end after receiving the request message;

[0062] Sensitive information is queried in the service end, and a result set message containing sensitive information is returned to the agent end, the agent end queries the record closest to the current time from a session time table, and acquires the corresponding customer name according to the record;

[0063] The agent end acquires the desensitization scheme corresponding to the current customer name from a desensitization configuration table according to the corresponding customer name to perform data desensitization, assembles the desensitized data to form a result set message, and completes dynamic desensitization and encryption of the data.

[0064] Optionally, the multi-source data fusion engine independently developed by the application is used to extract data in different dimensions such as regions, e-commerce and fast consumption, and a proxy model, which can be a MySQL proxy model, is used to monitor, analyze or transmit communication between data, and the data is communicated and desensitized through the proxy end and the service end; specifically, the proxy end queries the session ID from the session time table according to the nearest record to the current time, and queries the corresponding customer name in the session time table according to the session ID, judges whether the current customer needs to be desensitized, if yes, obtains the desensitization scheme corresponding to the customer in the desensitization configuration table to perform data desensitization, and replaces the desensitized data with the original data for storage; the data dynamic desensitization encryption can build a security line from the data storage source, provide full life cycle security protection for the data assets of the business system, and ensure the safe circulation and application of data under the compliance framework.

[0065] S20, integrating the encrypted multi-dimensional data, extracting features of the integrated data based on a distributed big data computing framework, and capturing key information and dynamic characteristics in the multi-dimensional data;

[0066] The step of integrating the encrypted multi-dimensional data, extracting features of the integrated data based on a distributed big data computing framework, and capturing key information and dynamic characteristics in the multi-dimensional data includes:

[0067] The encrypted multi-dimensional data is de-duplicated, filled, and tested for abnormal values to complete data cleaning, and the cleaned data is standardized, discretized, and layered to complete data conversion;

[0068] The converted data is loaded in batches to a distributed storage area based on an ETL tool, features corresponding to the multi-dimensional data are extracted and integrated, the integrated multi-dimensional data is visually processed, and a visual image corresponding to the key information and dynamic characteristics in the multi-dimensional data is obtained.

[0069] Due to different dimensions of data, the data source is complex, and before data integration, data processing is needed to extract the data required for precision marketing from the complex data original format. Data cleaning is to solve the quality problems of data integrity, legality and consistency, so that the data is more suitable for mining; data conversion is to make the data more consistent and easier to be processed by the model through standardization, discretization and stratification; based on the ETL (Extract-Transform-Load) tool, the converted data is loaded into the distributed storage area in batches; the data is stored in partitioned segments to improve the analysis and response capabilities; for example, structured, semi-structured and unstructured data are stored in partitioned segments, and data visualization is to convert data into visual images displayed on the screen using computer graphics and image processing technology. The main features of data visualization are: interactivity, multidimensionality, and visualization; visualization can help quickly analyze customer dynamic changes; in this embodiment, by obtaining data of different dimensions, customer cross-platform behavior characteristics can be captured comprehensively, and the accuracy of target customer selection can be improved. The integrated data is analyzed through a distributed big data computing framework, which can realize real-time feature extraction and accurately capture key information and dynamic trends in various fields, providing solid data support and strong technical power for subsequent data mining, analysis and business decision-making.

[0070] S30, based on the key information and dynamic characteristics in the multi-dimensional data, label setting is performed on the customers in the customer group to form a to-be-screened customer group, and intelligent product matching is performed according to the customers in the to-be-screened customer group to form a matching white list of product and customer adaptation;

[0071] The step of setting labels on the customers in the customer group based on the key information and dynamic characteristics in the multi-dimensional data to form a to-be-screened customer group, and intelligently matching products according to the customers in the to-be-screened customer group includes:

[0072] The visual image is input into the IKG module, the image encoder in the IKG module receives the visual image and performs feature extraction, and generates corresponding serialized image features;

[0073] The image decoder predicts from the starting sequence in the serialized image features, gradually predicts the text description sequence corresponding to the visual image based on the greedy algorithm, obtains a token sequence text, cleanses the token sequence text, and obtains a to-be-selected vocabulary;

[0074] All token information in the to-be-selected vocabulary is encoded based on the text encoder in the IKG module, the encoded token information and the serialized image features are subjected to graph-text similarity calculation, a plurality of related keywords and their weights are screened out, the plurality of keywords are used as labels corresponding to the customers in the customer group, and a to-be-screened customer group is formed.

[0075] A label library of the product is acquired, similarity between all keywords corresponding to the customers in the customer group to be screened and the label library is calculated to obtain a similarity matrix, weights corresponding to the keywords are weighted into the similarity matrix to obtain a probability distribution between the product and the customers, and an intelligent matching result is determined according to the probability distribution.

[0076] The expression of the serialized image feature is:

[0077]

[0078] In the formula, respectively represent global features and local features of the visual image, represents an image encoder, represents an input visual image;

[0079] The expression of the step-by-step prediction of the greedy algorithm is:

[0080]

[0081]

[0082]

[0083] In the formula, represents an output of a linear layer and a standardization layer after prediction of an i-th image decoder, represents a result after a self-attention layer and a residual connection after prediction of an i-th image decoder, represents a result after a cross-attention layer after prediction of an i-th image decoder, represents an intermediate result of fusion of global features of a visual image and a cross-attention layer after prediction of an i-th image decoder; represents a multi-head self-attention layer, represents a multi-head cross-attention layer, represents a multi-layer perception, represents a standardization layer.

[0084] ​​Optionally, the visual image of the customer is a dynamic customer portrait model, the visual image is input into an image keyword generation (IKG) module, in the IKG module, a contrastive language-image pre-training (CLIP) model is used as an image encoder for image coding to generate corresponding serialized image features; a bidirectional encoder based on Transfrmer is used as an image decoder, and a greedy algorithm is used to gradually predict the corresponding text description sequence of the visual image, and then stop words, invalid words and redundant words in the token sequence text are removed to obtain a candidate vocabulary; all token information is encoded by a text encoder in the CLIP model, in the CLIP model, the visual and text encoders are fully aligned, so that image-text similarity calculation can be performed to filter out k most relevant keywords and their weights, and the customer is labeled by the keywords and weights to generate a multi-dimensional accurate customer label system; a label library of the product is obtained, the similarity calculation method is used to improve the accuracy of target customer identification, and intelligent matching of the product and the customer is realized; and a customer whitelist is obtained.

[0085] S40, a marketing model is constructed, the matching whitelist is evaluated and analyzed according to the marketing model, and the purchase intention of the customer is quantified, and an intended customer is screened out for product marketing recommendation.

[0086] After obtaining the customer whitelist, the customer can be called by the intelligent AI outbound system according to the preset script and strategy, and efficient communication and interaction with the customer can be performed; the intention degree model is used to classify the intention degree of the customer, and optionally, the intention label can be divided into the following categories: A level (with clear intention), B level (may have intention), C level (clearly refuses), D level (user busy), E level (dialing failure), F level (invalid customer) and the like, and the purchase intention of the customer is scientifically quantitatively analyzed, and a high-intention customer is accurately screened out; the AI outbound system greatly improves the outgoing volume and communication efficiency compared with the traditional manual outbound, and the intention degree model evaluation helps the bank focus on high-value customer resources and optimize marketing resource allocation; the AI customer service robot is introduced, which can provide 24-hour uninterrupted consulting services for customers, answer common questions and collect customer feedback; optionally, the marketing model can also include marketing tools such as short message, operator-level telephone line, business process outsourcing (BPO), and the like, the operator-level telephone line guarantees 99% connection rate, and the short message channel has industry-leading direct rate; target customers are efficiently reached.

[0087] In summary, the application provides a dynamic active marketing method. First, multi-dimensional data of a customer group is obtained through a multi-source data engine, and the multi-dimensional data is desensitized and encrypted, thereby building a security line from the data storage source to provide a full life cycle security guarantee for the data assets of the business system and ensuring the safe circulation and application of the data under the compliance framework. The encrypted multi-dimensional data is integrated to comprehensively capture the cross-platform behavior characteristics of customers and improve the accuracy of target customer selection. The customers in the customer group are labeled through the integrated data, and then the customers with the set labels are intelligently matched with products to form a matching whitelist of product and customer adaptation. The intended customers are accurately positioned through the matching whitelist, which greatly reduces the customer volume of subsequent marketing, is conducive to improving the subsequent marketing reach rate and conversion rate, builds a marketing model, evaluates and analyzes the matching whitelist according to the marketing model to quantify the purchase intention of customers, and selects intended customers for product marketing recommendation. The digital marketing method provided by the application has accurate product-customer matching, high marketing reach rate and conversion rate, and is suitable for wide promotion.

[0088] Embodiment two

[0089] The embodiment provides a digital marketing system, comprising:

[0090] An encryption module is configured to obtain multi-dimensional data of a customer group based on a multi-source data engine and dynamically desensitize and encrypt the multi-dimensional data.

[0091] A capture module is configured to integrate the encrypted multi-dimensional data, extract features from the integrated data based on a distributed big data computing framework, and capture key information and dynamic characteristics in the multi-dimensional data.

[0092] A matching module is configured to label customers in the customer group based on the key information and dynamic characteristics in the multi-dimensional data, form a to-be-screened customer group, intelligently match products with customers according to the customers in the to-be-screened customer group, and form a matching whitelist of product and customer adaptation.

[0093] A recommendation module is configured to build a marketing model, evaluate and analyze the matching whitelist according to the marketing model, quantify the purchase intention of customers, and select intended customers for product marketing recommendation.

[0094] Preferably, the step of obtaining multi-dimensional data of a customer group based on a multi-source data engine and dynamically desensitizing and encrypting the multi-dimensional data comprises:

[0095] Incremental extraction and full extraction are performed in the data source based on the multi-source data engine to obtain multi-dimensional data of the customer group, and the multi-dimensional data at least includes regional data, time data, brand type data, online sales data, and offline sales data.

[0096] constructing an agent model, the agent model comprising at least an agent end and a service end;

[0097] sending a request message to the agent end based on multi-dimensional data, the agent end forwarding the request message to the service end after receiving the request message;

[0098] querying sensitive information in the service end, returning a result set message containing sensitive information to the agent end, the agent end querying the nearest record from the session time table according to the current time, and obtaining the corresponding customer name according to the record;

[0099] the agent end obtains the desensitization scheme corresponding to the current customer name from the desensitization configuration table according to the corresponding customer name, and performs data desensitization, assembles the desensitized data to form a result set message, and completes the dynamic desensitization and encryption of data.

[0100] Preferably, the step of integrating the encrypted multi-dimensional data, and extracting features from the integrated data based on a distributed big data computing framework, to capture key information and dynamic features in the multi-dimensional data comprises:

[0101] de-duplicating, filling, and outlier testing the encrypted multi-dimensional data to complete data cleaning, and standardizing, discretizing, and layering the cleaned data to complete data conversion;

[0102] loading the converted data in batches to a distributed storage area based on an ETL tool, extracting corresponding features of the multi-dimensional data and integrating them, and visualizing the integrated multi-dimensional data to obtain a visual image corresponding to the key information and dynamic features in the multi-dimensional data.

[0103] Preferably, the step of setting labels for customers in a customer group based on the key information and dynamic features in the multi-dimensional data to form a to-be-screened customer group, and intelligently matching products according to customers in the to-be-screened customer group comprises:

[0104] inputting the visual image into an IKG module, receiving the visual image and extracting features based on an image encoder in the IKG module to generate corresponding serialized image features;

[0105] the image decoder predicts from the starting sequence in the serialized image features, gradually predicts a text description sequence corresponding to the visual image based on a greedy algorithm, obtains a token sequence text, and cleanses the token sequence text to obtain a to-be-selected vocabulary;

[0106] Encode all word-meta information in the candidate vocabulary based on the text encoder in the IKG module, perform image-text similarity calculation on the encoded word-meta information and the serialized image features, screen out several relevant keywords and their weights, and use the several keywords as labels corresponding to customers in the customer group to form a customer group to be screened;

[0107] Obtain the product tag library, calculate the similarity between all keywords corresponding to customers in the customer group to be screened and the tag library, obtain the similarity matrix, add the weights corresponding to the keywords to the similarity matrix, obtain the probability distribution between products and customers, and determine the intelligent matching results based on the probability distribution.

[0108] Preferably, the expression of the serialized image feature is:

[0109]

[0110] Where, Represent the global and local features of the visualized image, respectively. represents the image encoder, A visualization image representing the input;

[0111] The expression for step-by-step prediction of the greedy algorithm is:

[0112]

[0113]

[0114]

[0115] Where, Indicates the The output of the image decoder after the linear layer and normalization is predicted. Indicates the The result after the self-attention layer and residual connection when the image decoder predicts, Indicates the When predicting an image decoder, it fuses the global features of the visual image and the intermediate results of the cross-attention layer; represents a multi-head self-attention layer, represents a multi-head cross attention layer, represents a multilayer perceptron, Represents a normalization layer.

[0116] Example 3

[0117] This embodiment provides a storage medium having a computer program stored thereon, which implements the digital marketing method as described above when executed by a processor.

[0118] Example 4

[0119] The application further provides a computer, which is shown in Figure 2 The computer comprises a memory 10, a processor 20 and a computer program 30 stored in the memory 10 and capable of running on the processor 20, and the processor 20 implements the digital marketing method when running the computer program 30.

[0120] The memory 10 comprises at least one type of storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 10 can be an internal storage unit of the computer, for example, a hard disk of the computer. In other embodiments, the memory 10 can also be an external storage device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 10 can comprise both an internal storage unit and an external storage device of the computer. The memory 10 can be used not only for storing application software and various data installed in the computer, but also for temporarily storing data that has been output or will be output.

[0121] The processor 20 can be an electronic control unit (ECU), a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip in some embodiments, and is used for running program codes or processing data stored in the memory 10, for example, executing an access restriction program.

[0122] It should be noted that Figure 2 The structure shown does not constitute a limitation on the computer, and in other embodiments, the computer can comprise fewer or more components than shown, or combine certain components, or have different component arrangements.

[0123] Those skilled in the art will appreciate that the logic and / or steps represented in the flow diagrams, or otherwise described herein, can be embodied in

[0124] More specific examples (a non-exhaustive list) of the computer readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0125] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be employed: a discrete logic circuit having logic gates for implementing logic functions upon data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), or the like.

[0126] The technical features of the above-described embodiments can be combined in any manner, and in order to make the description concise, not all possible combinations of the technical features in the above-described embodiments are described, however, as long as the combinations of the technical features do not contradict each other, they should be considered to be within the scope of the present specification.

[0127] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A digital marketing method, characterized in that include: Acquire multi-dimensional data of customer groups based on a multi-source data engine, and perform dynamic desensitization and encryption processing on the multi-dimensional data; Integrate the encrypted multi-dimensional data and extract features from the integrated data based on the distributed big data computing framework to capture key information and dynamic features in the multi-dimensional data; Based on the key information and dynamic features in the multi-dimensional data, customers in the customer group are labeled to form a customer group to be screened. Products are intelligently matched based on the customers in the customer group to form a matching whitelist of products and customers. Build a marketing model and evaluate and analyze the matching whitelist based on the marketing model to quantify customers' purchasing intentions and screen out potential customers for product marketing recommendations; The steps of integrating the encrypted multi-dimensional data, extracting features from the integrated data based on a distributed big data computing framework, and capturing key information and dynamic features in the multi-dimensional data include: Perform deduplication, padding, and outlier detection on encrypted multi-dimensional data to complete data cleaning, and perform standardization, discretization, and stratification on the cleaned data to complete data conversion; Using ETL tools, the converted data is batch loaded into a distributed storage area. The features corresponding to the multi-dimensional data are extracted and integrated. The integrated multi-dimensional data is visualized to obtain key information in the multi-dimensional data and visualization images corresponding to the dynamic features. The steps of labeling customers in the customer group based on key information and dynamic features in the multi-dimensional data to form a customer group to be screened, and performing intelligent product matching based on the customers in the customer group to be screened include: The visualization image is input into the IKG module. The image encoder in the IKG module receives the visualization image and performs feature extraction to generate corresponding serialized image features. The image decoder predicts the starting sequence in the serialized image features, and gradually predicts the text description sequence corresponding to the visual image based on a greedy algorithm to obtain a word sequence text. The word sequence text is cleaned to obtain candidate words; Encode all word-meta information in the candidate vocabulary based on the text encoder in the IKG module, perform image-text similarity calculation on the encoded word-meta information and the serialized image features, screen out several relevant keywords and their weights, and use the several keywords as labels corresponding to customers in the customer group to form a customer group to be screened; Obtain the product tag library, calculate the similarity between all keywords corresponding to customers in the customer group to be screened and the tag library, obtain the similarity matrix, add the weights corresponding to the keywords to the similarity matrix, obtain the probability distribution between products and customers, and determine the intelligent matching results based on the probability distribution.

2. The digital marketing method according to claim 1, characterized in that: The step of obtaining multi-dimensional data of customer groups based on a multi-source data engine and performing dynamic desensitization and encryption processing on the multi-dimensional data includes: Perform incremental and full extractions from data sources based on a multi-source data engine to obtain multi-dimensional data on customer groups, including at least regional data, time data, brand data, online sales data, and offline sales data. Constructing a proxy model, wherein the proxy model includes at least a proxy end and a server end; Send a request message to the proxy based on multi-dimensional data. The proxy receives the request message and forwards it to the server. The server queries for sensitive information and returns the result set containing the sensitive information to the agent. The agent then searches the session timeline for the record closest to the current time and obtains the corresponding customer name based on the session ID in the record. The agent obtains the desensitization scheme corresponding to the current customer name from the desensitization configuration table according to the corresponding customer name to perform data desensitization, assembles the desensitized data to form a result set message, and completes the dynamic desensitization encryption of the data.

3. The digital marketing method according to claim 1, characterized in that: The expression of the serialized image feature is: Where, Represent the global and local features of the visualized image, respectively. represents the image encoder, A visualization image representing the input; The expression for step-by-step prediction of the greedy algorithm is: Where, Indicates the The output of the image decoder after the linear layer and normalization is predicted. Indicates the The result after the self-attention layer and residual connection when the image decoder predicts, Indicates the When predicting an image decoder, it fuses the global features of the visual image and the intermediate results of the cross-attention layer; represents a multi-head self-attention layer, represents a multi-head cross attention layer, represents a multilayer perceptron, Represents a normalization layer.

4. A digital marketing system, characterized in that: include; An encryption module is used to obtain multi-dimensional data of customer groups based on a multi-source data engine and perform dynamic desensitization and encryption processing on the multi-dimensional data; The capture module is used to integrate the encrypted multi-dimensional data, extract features from the integrated data based on the distributed big data computing framework, and capture key information and dynamic features in the multi-dimensional data; The matching module is used to label customers in the customer group based on key information and dynamic features in multi-dimensional data to form a customer group to be screened. It then performs intelligent product matching based on the customers in the customer group to form a matching whitelist of products and customers. The recommendation module is used to build a marketing model, evaluate and analyze the matching whitelist based on the marketing model, quantify the customer's purchase intention, and screen out potential customers for product marketing recommendations; The capture module includes: The processing unit is used to perform deduplication, padding, and outlier detection on the encrypted multi-dimensional data to complete data cleaning, and to perform standardization, discretization, and stratification on the cleaned data to complete data conversion; The visualization unit is used to load the converted data into a distributed storage area in batches based on the ETL tool, extract the features corresponding to the multi-dimensional data and integrate them, and visualize the integrated multi-dimensional data to obtain key information in the multi-dimensional data and visualization images corresponding to the dynamic features; The matching module is specifically used for: The visualization image is input into the IKG module. The image encoder in the IKG module receives the visualization image and performs feature extraction to generate corresponding serialized image features. The image decoder predicts the starting sequence in the serialized image features, and gradually predicts the text description sequence corresponding to the visual image based on a greedy algorithm to obtain a word sequence text. The word sequence text is cleaned to obtain candidate words; Encode all word-meta information in the candidate vocabulary based on the text encoder in the IKG module, perform image-text similarity calculation on the encoded word-meta information and the serialized image features, screen out several relevant keywords and their weights, and use the several keywords as labels corresponding to customers in the customer group to form a customer group to be screened; Obtain the product tag library, calculate the similarity between all keywords corresponding to customers in the customer group to be screened and the tag library, obtain the similarity matrix, add the weights corresponding to the keywords to the similarity matrix, obtain the probability distribution between products and customers, and determine the intelligent matching results based on the probability distribution.

5. The digital marketing system according to claim 4, characterized in that: The encryption module includes: An extraction unit, configured to perform incremental extraction and full extraction from a data source based on a multi-source data engine to obtain multi-dimensional data of a customer base, wherein the multi-dimensional data includes at least regional data, time data, brand data, online sales data, and offline sales data; A construction unit, configured to construct a proxy model, wherein the proxy model includes at least a proxy end and a server end; The sending unit is used to send a request message to the agent based on the multi-dimensional data. The agent forwards the request message to the server after receiving it. The query unit is used to query sensitive information on the server side and return the result set message containing sensitive information to the agent side. The agent side queries the record closest to the current time from the session time table and obtains the corresponding customer name based on the session ID in the record; The desensitizing unit is used to obtain the desensitizing scheme corresponding to the current customer name from the desensitizing configuration table based on the corresponding customer name on the agent side to perform data desensitization, assemble the desensitized data to form a result set message, and complete the dynamic desensitization encryption of the data.

6. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the digital marketing method according to any one of claims 1 to 3 is implemented.

7. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the digital marketing method according to any one of claims 1 to 3 is implemented.

Images