Marketing recommendation method based on behavior data analysis, electronic device, and program product

Abstract

The application provides a marketing recommendation method based on behavior data analysis, an electronic device and a program product, and relates to the technical field of computer processing. The method comprises the following steps: acquiring behavior data representing the interaction behavior of a user and a scenic spot; constructing a strategy through a preset portrait according to the behavior data, and generating a user portrait representing the travel demand preference of the user; performing group processing on the user according to the user portrait through a preset group strategy, obtaining a plurality of user groups and a group label corresponding to each user group; generating a marketing strategy corresponding to the user portrait through a preset marketing scheme prediction strategy according to the user portrait, the user group and the group label; and pushing the marketing strategy to the user corresponding to the user portrait. In this way, the problem of insufficient flexibility and insufficient effectiveness of marketing strategy pushing in the traditional marketing recommendation mode can be improved.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and more specifically, to a marketing recommendation method, electronic device, and program product based on behavioral data analysis. Background Technology

[0002] In the field of digital marketing, personalized recommendations based on user behavior have become a core means of improving business conversion rates. Especially in the tourism industry, user interactions with attractions (such as search frequency, dwell time, and collection behavior) contain a wealth of potential demand information. Currently, mainstream platforms generally use algorithms such as collaborative filtering and content recommendation to generate marketing strategies, but these methods often treat user behavior as homogeneous data input, making it difficult to effectively distinguish the differentiated impact of different behavior types on demand preferences.

[0003] A typical existing technical solution involves collecting users' historical behavior (such as order records and tourist attraction browsing data) to build static user profiles, and then matching these profiles with pre-set generic marketing templates (such as discount coupon combinations). This approach relies on pre-set marketing strategies, but in practice, different types of users have different focuses and needs, so the generation of marketing strategies also needs to be customized to meet user requirements. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a marketing recommendation method, electronic device and program product based on behavioral data analysis, which can improve the problems of insufficient flexibility and insufficient effectiveness of marketing strategy push in traditional marketing recommendation methods.

[0005] To achieve the above technical objectives, the technical solution adopted in this application is as follows:

[0006] In a first aspect, embodiments of this application provide a marketing recommendation method based on user behavior data analysis, the method comprising:

[0007] Acquire behavioral data that characterizes user interactions with attractions;

[0008] Based on the behavioral data, a user profile representing the user's travel needs and preferences is generated through a preset profile building strategy.

[0009] Based on the user profile, users are segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group.

[0010] Based on the user profile, the user group, and the group tag, a marketing strategy corresponding to the user profile is generated through a preset marketing plan prediction strategy;

[0011] The marketing strategy is pushed to users corresponding to the user profile.

[0012] In conjunction with the first aspect, in some optional implementations, behavioral data characterizing user interactions with attractions is acquired, including:

[0013] Obtain initial behavioral data representing user interactions with attractions;

[0014] Feature extraction is performed on the initial behavior to obtain behavioral features corresponding to different types of initial behavior data, and the set of behavioral features corresponding to all types of initial behavior data is taken as the behavioral data.

[0015] In conjunction with the first aspect, in some optional implementations, based on the behavioral data, a user profile representing the user's travel needs and preferences is generated through a preset profile building strategy, including:

[0016] Based on the behavioral data, determine the attention weight corresponding to each type of behavioral data:

[0017]

[0018] In the formula, Indicates user The Attention weights corresponding to behavioral data. This represents a collection of all behavioral data types. , , This represents the trainable attention weight matrix, bias, and attention parameter vector. Indicates user The Behavioral data;

[0019] The user profile is determined based on the attention weights and the behavioral data:

[0020]

[0021] In the formula, Representing user profiles, This represents the L2 normalization function.

[0022] In conjunction with the first aspect, in some optional implementations, based on the user profile, users are segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group, including:

[0023] Based on the user profile, determine the collaboration matrix for the current iteration round:

[0024]

[0025] In the formula, This represents the collaboration matrix for the current iteration round. This represents the collaboration matrix of the previous iteration in the current iteration. Indicates the time decay factor. Representing user profiles, Represents the set of all users. Indicates the outer product operation;

[0026] Based on the collaboration matrix and the user profile, users are segmented to obtain the multiple user groups:

[0027]

[0028]

[0029]

[0030] In the formula, Indicates the first Round-robin grouping User groups This represents the user profile of user u. This represents the dynamic radius of each user group. Indicates user group In the middle, the standard deviation of the user profiles of all users, Scaling factor Represents the spectral clustering algorithm. Indicates the preset number of clusters;

[0031] Based on the user groups and the user profiles, determine the tag influence matrix corresponding to each user group:

[0032]

[0033]

[0034] In the formula, Indicates the first Tag influence matrix for each user group Represents the average eigenvector of the group. This represents the pre-trained label weight matrix. Indicates the first Number of users in each user group This represents the user profile of user u. This represents the preset dimension projection matrix;

[0035] Based on the tag influence matrix, the user groups, and the user profiles, determine the group tags corresponding to each user group:

[0036]

[0037] In the formula, Indicates the first Group tags corresponding to each user group Indicates reservation The maximum value, This represents the softmax activation function.

[0038] In conjunction with the first aspect, in some optional implementations, based on the user profile, the user group, and the group tag, a marketing strategy corresponding to the user profile is generated through a preset marketing plan prediction strategy, including:

[0039] Based on the user profile, the user group, and the group tag, predict the user needs corresponding to the user profile;

[0040] The marketing strategy is generated based on the user needs and the group tags.

[0041] In conjunction with the first aspect, in some optional implementations, predicting user needs corresponding to the user profile based on the user profile, the user group, and the group tag includes:

[0042] Based on the user profile and the tag influence matrix corresponding to the user group, determine the modified features corresponding to the user profile:

[0043]

[0044] In the formula, Indicates a modified feature, Represents the ReLU activation function. This indicates a normalization operation. This represents the user profile of user u. This represents the tag influence matrix of the user group to which user u belongs;

[0045] Based on the user profile, the modified features, and the group tags, predict the user's needs:

[0046]

[0047]

[0048] In the formula, This represents the user needs of user u. This represents the sigmoid activation function. , This represents the trainable prediction matrix and prediction bias. , Indicates that the user group referred to by user u is in the 1st... and The corresponding group label in each iteration round.

[0049] In conjunction with the first aspect, in some optional implementations, the marketing strategy is generated based on the user needs and the group tags, including:

[0050] Based on the user requirements and the group tags, determine the tag constraints:

[0051]

[0052] In the formula, Indicates label constraints, This indicates element-wise multiplication. This represents the range constraint function. Indicates label Confidence weights Indicates label The embedding vector;

[0053] Determine the feedback coefficient based on the user's historical marketing strategies:

[0054]

[0055] In the formula, Indicates the feedback coefficient. This indicates the preset attenuation coefficient. This indicates the time since the last policy push. This indicates the marketing strategy previously generated for user u;

[0056] The marketing strategy is generated based on the label constraints and the feedback coefficients:

[0057]

[0058] In the formula, This indicates a marketing strategy.

[0059] In conjunction with the first aspect, in some alternative implementations, the method further includes:

[0060] The collaboration matrix is ​​updated based on the user profile and the marketing strategy:

[0061]

[0062] In the formula, This represents the updated collaboration matrix. Indicates the learning rate. Indicates the preset effect threshold. Represents the set of feedback users. Indicates marketing strategy, Represents a symbolic function. This represents the user profile of user u. Find the outer product.

[0063] Secondly, embodiments of this application also provide an electronic device, which includes a processor and a memory coupled to each other. The memory stores a computer program, and when the computer program is executed by the processor, the electronic device performs the above-described method.

[0064] Thirdly, embodiments of this application also provide a computer program product, including a computer program that implements the above-described method when executed by a processor.

[0065] The invention employing the above technical solution has the following advantages:

[0066] The technical solution provided in this application first acquires behavioral data representing user interactions with tourist attractions. Based on this behavioral data, a user profile representing user travel needs and preferences is generated through a pre-set profile building strategy. Then, based on the user profile, users are segmented using a pre-set segmentation strategy to obtain multiple user groups and corresponding group tags for each user group. Next, based on the user profile, user groups, and group tags, a marketing strategy corresponding to each user profile is generated through a pre-set marketing plan prediction strategy. Finally, the marketing strategy is pushed to the users corresponding to the user profiles. In this way, user profiles representing user needs are generated based on multimodal user behavior data. Users are then segmented based on these user profiles to obtain user groups and group tags. The generation of marketing strategies is then constrained by the group tags, avoiding the generation of a large number of ineffective marketing strategies and improving upon the problems of insufficient flexibility and ineffective marketing strategy delivery in traditional marketing recommendation methods. Attached Figure Description

[0067] This application can be further illustrated by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate some embodiments of this application and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without any inventive effort.

[0068] Figure 1 A structural block diagram of an electronic device provided in an embodiment of this application.

[0069] Figure 2 This is one of the flowcharts illustrating a marketing recommendation method based on user behavior data analysis provided in an embodiment of this application. Figure 3This is the second flowchart illustrating the marketing recommendation method based on user behavior data analysis provided in this application embodiment. Figure 4 A flowchart illustrating a marketing recommendation method based on user behavior data analysis provided in this application embodiment.

[0070] Icons: 100 - Electronic device; 101 - Processor; 102 - Memory. Detailed Implementation

[0071] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts are referred to by the same reference numerals in the drawings or description. Implementations not shown or described in the drawings are forms known to those skilled in the art. In the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0072] Please refer to Figure 1 This application provides an electronic device 100 that may include a processor 101 and a memory 102. The memory 102 stores a computer program, which, when executed by the processor 101, enables the electronic device 100 to perform corresponding steps in the marketing recommendation method based on user behavior data analysis described below.

[0073] In this embodiment, the processor 101 can be an integrated circuit chip with signal processing capabilities. The processor 101 can be a general-purpose processor. For example, the processor 101 can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0074] The memory 102 can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc. In this embodiment, the memory 102 can be used to store behavioral data, preset profile building strategies, user profiles, preset grouping strategies, user groups, group tags, preset marketing plan prediction strategies, marketing strategies, updated collaboration matrices, etc. Of course, the memory 102 can also be used to store programs, which the processor 101 executes after receiving an execution instruction.

[0075] Understandable, Figure 1 The electronic device 100 shown is only a schematic diagram; the electronic device 100 may also include components that are more... Figure 1 More components are shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0076] In this embodiment, the electronic device 100 can be a personal computer, laptop, etc. It is used to acquire behavioral data representing user interactions with tourist attractions, and based on this behavioral data, generate user profiles representing user travel needs and preferences using a preset profile building strategy. Then, based on the user profiles, users are segmented using a preset grouping strategy to obtain multiple user groups and corresponding group tags for each user group. Next, based on the user profiles, user groups, and group tags, a marketing strategy corresponding to each user profile is generated using a preset marketing plan prediction strategy. Finally, the marketing strategy is pushed to the users corresponding to the user profiles.

[0077] Please refer to Figure 2 This application also provides a marketing recommendation method based on user behavior data analysis, which can be applied to electronic device 100 and executed or implemented by electronic device 100. The marketing recommendation method based on user behavior data analysis may include the following steps:

[0078] Step 210: Obtain behavioral data representing user interaction with attractions;

[0079] Step 220: Based on the behavioral data, generate a user profile representing the user's travel needs and preferences using a preset profile building strategy;

[0080] Step 230: Based on the user profile, the user is segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group.

[0081] Step 240: Based on the user profile, the user group, and the group tag, generate a marketing strategy corresponding to the user profile using a preset marketing plan prediction strategy;

[0082] Step 250: Push the marketing strategy to the users corresponding to the user profile.

[0083] In the above implementation, behavioral data representing user interactions with attractions is first acquired. Based on this data, a user profile representing user travel needs and preferences is generated using a pre-defined profile building strategy. Then, based on the user profile, users are segmented using a pre-defined grouping strategy, resulting in multiple user groups and corresponding group tags for each group. Next, based on the user profile, user groups, and group tags, a marketing strategy corresponding to each user profile is generated using a pre-defined marketing plan prediction strategy. Finally, the marketing strategy is pushed to the users corresponding to the user profiles. In this way, user profiles representing user needs are generated based on multimodal user behavior data. Users are then segmented based on these profiles to obtain user groups and group tags. The generation of marketing strategies is then constrained by the group tags, avoiding the generation of numerous ineffective marketing strategies and improving upon the inflexibility and effectiveness issues of traditional marketing recommendation methods.

[0084] The following section will elaborate on each step of the marketing recommendation method based on user behavior data analysis:

[0085] In step 210, the behavioral data may include online behavioral data and offline behavioral data. Online behavioral data may include search behavior, browsing behavior, interaction behavior, consumption behavior, device location information, etc., based on the scenic area's APP; offline behavioral data may include scenic area gate access records, WIFI probe data, smart device interaction records, hotel check-in records, etc. Behavioral data can be represented as follows:

[0086] In this embodiment, obtaining behavioral data representing user interaction with attractions may include:

[0087] Obtain initial behavioral data representing user interactions with attractions;

[0088] Feature extraction is performed on the initial behavior to obtain behavioral features corresponding to different types of initial behavior data, and the set of behavioral features corresponding to all types of initial behavior data is taken as the behavioral data.

[0089] In this embodiment, the initial behavioral data can be obtained by recording the online behavioral data of the scenic area's linked APP, as well as the offline behavioral data of the various machines and equipment set up in the scenic area (such as turnstiles, navigation panels, self-service ticketing machines, WIFI, hotel front desk registration computers, etc.) and the user. This data is stored in the memory 102 of the aforementioned electronic device 100, or in a server communicatively connected to the electronic device 100. Then, in subsequent user profile construction, user segmentation processing, and marketing strategy generation and push, the data is invoked based on instructions issued by the user through the processor 101. Alternatively, the initial behavioral data can also be obtained during the testing phase of the technical solution of this application, based on data packets input by the user in real time, to simulate the user's online and offline behavioral data. The specific method for obtaining the initial behavioral data is not limited here.

[0090] The initial behavioral data can be represented as follows:

[0091] (1)

[0092] In the formula, This represents the user's initial behavioral data. Indicates user identifier, Indicates the type of behavior (e.g., search behavior, consumption behavior, etc.). Represents a timestamp. This indicates contextual information (including device ID, location information, transaction amount, etc.).

[0093] In this embodiment, feature extraction is performed on the initial behavior to obtain behavioral features corresponding to different types of initial behavior data:

[0094] (2)

[0095] In the formula, Indicates user The Class behavioral characteristics, This represents the activation function. Representation type The weight matrix, Representation type The bias vector, Representation type The feature transformation function.

[0096] After feature extraction of each type of initial behavioral data is completed, the set of behavioral features corresponding to all types of initial behavioral data is taken as behavioral data.

[0097] In this way, by extracting features from the raw behavioral data, the utilization rate of sparse behavioral data is improved. Discrete behaviors are transformed into multi-dimensional feature vectors, providing high-quality input for user profile construction and avoiding feature bias caused by directly using raw data.

[0098] In step 220, based on the behavioral data, a user profile representing the user's travel needs and preferences is generated through a preset profile building strategy, which may include:

[0099] Based on the behavioral data, determine the attention weight corresponding to each type of behavioral data:

[0100] (3)

[0101] In the formula, Indicates user The Attention weights corresponding to behavioral data. This represents a collection of all behavioral data types. , , This represents the trainable attention weight matrix, bias, and attention parameter vector. Indicates user The Behavioral data;

[0102] The user profile is determined based on the attention weights and the behavioral data:

[0103] (4)

[0104] In the formula, Representing user profiles, This represents the L2 normalization function.

[0105] Thus, by introducing attention weights, the contribution weight of different types of behavioral data to user profiles is dynamically allocated. For example, high-frequency consumption behaviors receive higher weights, while low-frequency browsing behaviors receive lower weights. This design, replacing manual weighting, allows user profiles to accurately capture users' core needs and characteristics, improving the accuracy of subsequent user segmentation and marketing strategy recommendations.

[0106] In step 230, based on the user profile, users are segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group, which may include:

[0107] Based on the user profile, determine the collaboration matrix for the current iteration round:

[0108] (5)

[0109] In the formula, This represents the collaboration matrix for the current iteration round. This represents the collaboration matrix of the previous iteration in the current iteration. This represents the time decay factor, with a default value of 0.85. Representing user profiles, Represents the set of all users. Indicates the outer product operation;

[0110] Based on the collaboration matrix and the user profile, users are segmented to obtain the multiple user groups:

[0111] (6)

[0112] (7)

[0113] (8)

[0114] In the formula, Indicates the first Round-robin grouping User groups This represents the user profile of user u. This represents the dynamic radius of each user group. Indicates user group In the middle, the standard deviation of the user profiles of all users, Scaling factor Represents the spectral clustering algorithm. Indicates the preset number of clusters;

[0115] Based on the user groups and the user profiles, determine the tag influence matrix corresponding to each user group:

[0116] (9)

[0117] (10)

[0118] In the formula, Indicates the first Tag influence matrix for each user group Represents the average eigenvector of the group. This represents the pre-trained label weight matrix. Indicates the first Number of users in each user group This represents the user profile of user u. This represents the preset dimension projection matrix;

[0119] Based on the tag influence matrix, the user groups, and the user profiles, determine the group tags corresponding to each user group:

[0120] (11)

[0121] In the formula, Indicates the first Group tags corresponding to each user group Indicates reservation The maximum value, The default value is 5. This represents the softmax activation function.

[0122] In this embodiment, the step size of the iteration rounds can be flexibly set according to user needs, such as iterating once a day, or iterating once after pushing marketing strategies to a preset number of users. That is, the iteration of the collaborative matrix (similar to cluster processing) can be based on time or on user-preset events.

[0123] In this embodiment, It is an adjustable scaling factor, usually By adjusting We can control the group radius relative to the group's dispersion. For example:

[0124] when When the value is larger, the group radius is larger, and the group contains more users (more relaxed clustering).

[0125] when When the size is smaller, the group radius is smaller, and the group contains fewer users (more compact clustering).

[0126] Thus, through a three-tiered technical design involving iterative updates of the collaborative matrix, dynamic radius spectrum clustering, and a tag influence matrix, adaptive optimization of user groups is achieved. The adaptive clustering radius enhances the clarity of user group boundaries; simultaneously, the tag influence matrix quantifies the preferences of different user groups, providing a basis for marketing strategy generation.

[0127] In step 240, based on the user profile, the user group, and the group tag, a marketing strategy corresponding to the user profile is generated using a preset marketing plan prediction strategy, which may include:

[0128] Based on the user profile, the user group, and the group tag, predict the user needs corresponding to the user profile;

[0129] The marketing strategy is generated based on the user needs and the group tags.

[0130] In this embodiment, predicting user needs corresponding to the user profile based on the user profile, the user group, and the group tag may include:

[0131] Based on the user profile and the tag influence matrix corresponding to the user group, determine the modified features corresponding to the user profile:

[0132] (12)

[0133] In the formula, Indicates a modified feature, Represents the ReLU activation function. This indicates a normalization operation. This represents the user profile of user u. This represents the tag influence matrix of the user group to which user u belongs;

[0134] Based on the user profile, the modified features, and the group tags, predict the user's needs:

[0135] (13)

[0136] (14)

[0137] In the formula, This represents the user needs of user u. This represents the sigmoid activation function. , This represents the trainable prediction matrix and prediction bias. , Indicates that the user group referred to by user u is in the 1st... and The corresponding group label in each iteration round.

[0138] Thus, a predictive chain of "user profile - group tags - user needs" is established, transforming abstract user profiles into marketing elements. Through a hierarchical prediction mechanism (i.e., needs first, strategy later), the mechanical nature of directly matching preset marketing templates is avoided, making subsequent marketing strategy generation more scenario-adaptable. Furthermore, by utilizing modified features and cross-period tag changes... It captures the dynamic evolution of user needs. Through normalization and feature concatenation, it addresses the prediction fluctuations caused by sudden changes in user behavior, thereby improving the stability of user demand prediction.

[0139] In this embodiment, generating the marketing strategy based on the user needs and the group tags may include:

[0140] Based on the user requirements and the group tags, determine the tag constraints:

[0141] (15)

[0142] In the formula, Indicates label constraints, This indicates element-wise multiplication. This represents the range constraint function. Indicates label Confidence weights Indicates label The embedding vector;

[0143] Determine the feedback coefficient based on the user's historical marketing strategies:

[0144] (16)

[0145] (17)

[0146] In the formula, Indicates the feedback coefficient. This indicates the preset attenuation coefficient. This indicates the time since the last policy push. This indicates the marketing strategy generated for user u in the last instance. Describing the L1 norm, This represents the initial attenuation coefficient (default 0.1). This represents the sensitivity coefficient, typically set to a value of 1.2. This indicates the user rejection rate (usually calculated over 7 days).

[0147] The marketing strategy is generated based on the label constraints and the feedback coefficients:

[0148] (18)

[0149] In the formula, This indicates a marketing strategy.

[0150] In this way, by using tag constraints and feedback coefficients based on historical marketing strategies, the rationality and timeliness of marketing strategies are improved. Through confidence weights and decay coefficients, it is ensured that marketing strategies align with the core tags of the user group while avoiding excessive push notifications to inactive users, significantly reducing user disturbance rates.

[0151] In step 250, after generating the marketing strategy, the marketing strategy is pushed to the terminal where the user is logged in.

[0152] Please refer to Figure 3 As an optional implementation, the method may further include:

[0153] Step 260: Update the collaboration matrix based on the user profile and the marketing strategy.

[0154] (19)

[0155] In the formula, This represents the updated collaboration matrix. Indicates the learning rate. The default value is 0.05. Indicates the preset effect threshold. The default value is 0.6. Represents the set of feedback users. Indicates marketing strategy, Represents a symbolic function. This represents the user profile of user u. Find the outer product.

[0156] This collaborative matrix online update mechanism constructs a closed loop for strategy effectiveness feedback. By using effectiveness thresholds and feedback user sets to filter effective strategy data, the matrix continuously optimizes the user association model, avoiding situations where lagging model iteration leads to insufficient accuracy in the final marketing strategy generation and low delivery efficiency.

[0157] In summary, please refer to Figure 4 This application provides a marketing recommendation method based on user behavior data analysis. First, raw behavioral data (i.e., initial behavioral data, including users' online and offline behaviors) is acquired. Then, user profiles are constructed based on the raw behavioral data, and dynamic collaborative grouping processing is performed on these user profiles to obtain multiple user groups, as well as a tag influence matrix and group tags for each user group. Next, the user profiles are feature-corrected based on the tag influence matrix to obtain corrected features. Then, demand prediction is performed based on the corrected features and group tags to obtain user needs. Finally, strategy constraints (i.e., tag constraints) are determined using user needs and group tags, thereby generating a marketing strategy, which is then pushed to the user's logged-in terminal (i.e., the user terminal). Finally, user feedback (such as the percentage of users who accept the information (e.g., booking hotels or buying tickets through the pushed links, or browsing for more than the preset time) and the percentage of users who reject the information) is collected. Then, according to the user's preset iteration step (e.g., once a day, once every 7 days, or when the number of feedback messages exceeds the preset number, etc., time and event conditions can all be used as an iteration step), the collaboration matrix is ​​updated based on the user feedback information, and user segmentation is performed again to ensure the continuous operation, real-time performance, and effectiveness of the technical solution and subsequent marketing strategy generation.

[0158] This application also provides a computer program product, including a computer program that, when executed by processor 101, implements the above-described marketing recommendation method based on behavioral data analysis.

[0159] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This software product can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0160] In summary, this application provides a marketing recommendation method, electronic device, and program product based on behavioral data analysis. In this technical solution, behavioral data representing user interactions with attractions is first acquired. Based on this behavioral data, a user profile representing user travel needs and preferences is generated through a preset profile construction strategy. Then, based on the user profile, users are segmented using a preset grouping strategy to obtain multiple user groups and corresponding group tags for each user group. Next, based on the user profile, user groups, and group tags, a marketing strategy corresponding to the user profile is generated through a preset marketing plan prediction strategy. Finally, the marketing strategy is pushed to the users corresponding to the user profile. Thus, user profiles representing user needs are generated based on multimodal user behavior data. Users are then segmented based on these user profiles to obtain user groups and group tags. The generation of marketing strategies is then constrained by the group tags, avoiding the generation of a large number of ineffective marketing strategies and improving upon the problems of insufficient flexibility and ineffective marketing strategy delivery in traditional marketing recommendation methods.

[0161] In the embodiments provided in this application, it should be understood that the disclosed methods can also be implemented in other ways. The method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code, which includes one or more executable instructions for implementing a specified logical function. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0162] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A marketing recommendation method based on user behavior data analysis, characterized in that, The method includes: Acquire behavioral data that characterizes user interactions with attractions; Based on the behavioral data, a user profile representing the user's travel needs and preferences is generated through a preset profile building strategy. Based on the user profile, users are segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group. Based on the user profile, the user group, and the group tag, a marketing strategy corresponding to the user profile is generated through a preset marketing plan prediction strategy; The marketing strategy is pushed to users corresponding to the user profile.

2. The method according to claim 1, characterized in that, Obtain behavioral data representing user interactions with attractions, including: Obtain initial behavioral data representing user interactions with attractions; Feature extraction is performed on the initial behavior to obtain behavioral features corresponding to different types of initial behavior data, and the set of behavioral features corresponding to all types of initial behavior data is taken as the behavioral data.

3. The method according to claim 1, characterized in that, Based on the behavioral data, a user profile representing the user's travel needs and preferences is generated through a preset profile building strategy, including: Based on the behavioral data, determine the attention weight corresponding to each type of behavioral data: ; In the formula, Indicates user The Attention weights corresponding to behavioral data This represents a collection of all behavioral data types. , , This represents the trainable attention weight matrix, bias, and attention parameter vector. Indicates user The Behavioral data; The user profile is determined based on the attention weights and the behavioral data: ; In the formula, Representing user profiles, This represents the L2 normalization function.

4. The method according to claim 1, characterized in that, Based on the user profile, users are segmented using a preset segmentation strategy to obtain multiple user groups and group tags corresponding to each user group, including: Based on the user profile, determine the collaboration matrix for the current iteration round: ; In the formula, This represents the collaboration matrix for the current iteration round. This represents the collaboration matrix of the previous iteration in the current iteration. Indicates the time decay factor. Representing user profiles, Represents the set of all users. Indicates the outer product operation; Based on the collaboration matrix and the user profile, users are segmented to obtain the multiple user groups: ; ; ; In the formula, Indicates the first Round-robin grouping User groups This represents the user profile of user u. This represents the dynamic radius of each user group. Indicates user group In the middle, the standard deviation of the user profiles of all users, Scaling factor Represents the spectral clustering algorithm. Indicates the preset number of clusters; Based on the user groups and the user profiles, determine the tag influence matrix corresponding to each user group: ; ; In the formula, Indicates the first Tag influence matrix for each user group Represents the average eigenvector of the group. This represents the pre-trained label weight matrix. Indicates the first Number of users in each user group This represents the user profile of user u. This represents the preset dimension projection matrix; Based on the tag influence matrix, the user groups, and the user profiles, determine the group tags corresponding to each user group: ; In the formula, Indicates the first Group tags corresponding to each user group Indicates reservation The maximum value, This represents the softmax activation function.

5. The method according to claim 4, characterized in that, Based on the user profile, the user group, and the group tags, a marketing strategy corresponding to the user profile is generated using a preset marketing plan prediction strategy, including: Based on the user profile, the user group, and the group tag, predict the user needs corresponding to the user profile; The marketing strategy is generated based on the user needs and the group tags.

6. The method according to claim 5, characterized in that, Based on the user profile, the user group, and the group tag, predict the user needs corresponding to the user profile, including: Based on the user profile and the tag influence matrix corresponding to the user group, determine the modified features corresponding to the user profile: ; In the formula, Indicates a modified feature, Represents the ReLU activation function. This indicates a normalization operation. This represents the user profile of user u. This represents the tag influence matrix of the user group to which user u belongs; Based on the user profile, the modified features, and the group tags, predict the user's needs: ; ; In the formula, This represents the user needs of user u. This represents the sigmoid activation function. , This represents the trainable prediction matrix and prediction bias. , Indicates that the user group referred to by user u is in the 1st... and The corresponding group label in each iteration round.

7. The method according to claim 6, characterized in that, Based on the user needs and the group tags, the marketing strategy is generated, including: Based on the user requirements and the group tags, determine the tag constraints: ; In the formula, Indicates label constraints, This indicates element-wise multiplication. This represents the range constraint function. Indicates label Confidence weights Indicates label The embedding vector; Determine the feedback coefficient based on the user's historical marketing strategies: ; In the formula, Indicates the feedback coefficient. Indicates the preset attenuation coefficient. This indicates the time since the last policy push. This indicates the marketing strategy previously generated for user u; The marketing strategy is generated based on the label constraints and the feedback coefficients: ; In the formula, This indicates a marketing strategy.

8. The method according to claim 4, characterized in that, The method further includes: The collaboration matrix is ​​updated based on the user profile and the marketing strategy: ; In the formula, This represents the updated collaboration matrix. Indicates the learning rate. Indicates the preset effect threshold. This represents the set of feedback users. Indicates marketing strategy, Represents a symbolic function. This represents the user profile of user u. Find the outer product.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory coupled together, the memory storing a computer program that, when executed by the processor, causes the electronic device to perform the method as described in any one of claims 1-8.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-8.

Images