Automobile marketing method based on AI model

Abstract

The invention discloses an automobile marketing method based on an AI model, relates to the technical field of artificial intelligence, and can accurately identify an automobile purchase intention of a user by collecting multi-source heterogeneous data of the user and performing deep analysis by using the AI model, and push personalized marketing content based on the intention, thereby realizing accurate marketing, avoiding harassment of invalid information, and improving the user experience. And the user experience is improved. By introducing a purchase stage identification model, whether a user is in a cognitive period, a car selection period or a decision period can be judged, and marketing contents of corresponding types are pushed according to different stages, so that accurate guidance of the whole life cycle of car purchase of the user is realized, and the matching degree of the marketing contents and the user conversion rate are greatly improved; the AI technology and the automobile marketing scene are deeply fused, the pain point of a traditional marketing mode is solved, and an efficient, intelligent and accurate marketing solution is provided for the automobile industry.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to an AI-based automotive marketing method. Background Technology

[0002] As the automotive market matures and competition intensifies, automotive marketing is shifting from the traditional broad-based approach to a more precise and personalized one. However, existing automotive marketing methods still have many shortcomings.

[0003] First, marketing precision is insufficient. Many car manufacturers or dealers still rely on basic demographic information of users (such as age, gender, and region) for extensive marketing, failing to accurately understand users' true and dynamic car-buying intentions. For example, pushing sedan ads to a user who has recently been frequently watching SUV review videos is clearly ineffective; this not only wastes marketing resources but also degrades the user experience.

[0004] Secondly, the user experience is poor. Due to the inability to accurately identify user needs, users often receive a large amount of marketing information that is not of interest, such as SMS bombardment and irrelevant advertising pushes. This can easily cause user resentment and even create a negative impression of the brand.

[0005] Secondly, marketing conversion rates are low. Traditional marketing methods struggle to provide effective information at crucial junctures in a user's car-buying decision-making process. The car-buying process is a complex, multi-stage journey, including initial brand awareness, mid-stage model comparison, and final purchase decision. At each stage, users focus on vastly different information. If content isn't tailored to the user's specific stage, it's difficult to effectively guide them to complete the purchase, resulting in low marketing conversion rates. Summary of the Invention

[0006] This application aims to provide an AI-based automotive marketing method that addresses the problems of low accuracy, poor user experience, and low conversion efficiency in existing automotive marketing technologies.

[0007] This application provides an AI-based automotive marketing method, including: Collect multi-source heterogeneous data related to users, and preprocess the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data; An AI model is used to identify the preprocessed multi-source heterogeneous data and obtain the user's corresponding car purchase intention tag; Based on the user's corresponding car purchase intention tags, a first personalized car marketing content is generated and pushed to the user to complete the car marketing based on the AI ​​model.

[0008] In one possible implementation, collecting user-related multi-source heterogeneous data includes: Based on a preset data sampling period, collect user-related content browsing data, vehicle interaction data, and service behavior interaction data to obtain the user-related multi-source heterogeneous data. The content browsing data includes the vehicle model code corresponding to the browsed information, the dwell time, likes, comments, shares and / or favorites; the vehicle interaction data includes the vehicle model code of the virtual interaction, the configuration code of the viewed information, the viewing frequency and the viewing duration; and the service behavior interaction data includes the vehicle model code for calculating the purchase price, the number of calculations, the vehicle model code for scheduling a test drive, and the number of schedulings.

[0009] In one possible implementation, preprocessing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data includes: normalizing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data.

[0010] In one possible implementation, before employing an AI model to identify the preprocessed multi-source heterogeneous data, the method further includes: AI models are built using convolutional neural networks or long short-term memory networks; Based on historical multi-source heterogeneous data and users' real car purchase intention tags, the AI ​​model is trained using an intelligent optimization algorithm to obtain a trained AI model, which is then used in subsequent car marketing processes.

[0011] In one possible implementation, based on the user's corresponding car purchase intention tag, corresponding first personalized car marketing content is generated, and the first personalized car marketing content is pushed to the user, including: Based on the user's corresponding car purchase intention tag, the target car model with the car purchase intention tag is determined; each car model is pre-set with at least one car purchase intention tag; Randomly collect comparative reviews and / or owner testimonials related to the target car model to obtain the first personalized car marketing content, and push the randomly collected first personalized car marketing content to users.

[0012] In one possible implementation, it also includes: A pre-set purchase stage identification model is used to identify the pre-processed multi-source heterogeneous data to obtain the predicted purchase stage corresponding to the user; wherein, the predicted purchase stage includes the awareness period, the car selection period, and the decision-making period; Based on the user's predicted purchase stage, a second personalized car marketing content is generated and pushed to the user.

[0013] In one possible implementation, the pre-setting method for the purchase stage identification model includes: A purchase phase identification model can be constructed using convolutional neural networks or long short-term memory networks; Based on historical multi-source heterogeneous data and user's actual purchase stage labels, an intelligent optimization algorithm is used to train the purchase stage identification model to obtain a pre-set purchase stage identification model.

[0014] In one possible implementation, if the user's predicted purchase stage is the awareness period, then brand introduction information and / or technical analysis information are randomly collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0015] In one possible implementation, if the user's predicted purchase stage is the car selection period, then random comparative evaluations and / or owner reviews are collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0016] In one possible implementation, if the user's predicted purchase stage is the decision-making period, then financial plan information, preferential policy information, and / or vehicle delivery process information are randomly collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0017] Beneficial effects: This application provides an AI-based automotive marketing method. By collecting multi-source heterogeneous user data and using AI models for in-depth analysis, it can accurately identify users' car purchase intentions and push personalized marketing content based on these intentions, achieving precise marketing, avoiding the nuisance of invalid information, and improving user experience. By introducing a purchase stage recognition model, it can determine whether a user is in the awareness stage, the car selection stage, or the decision-making stage, and push corresponding types of marketing content according to different stages, achieving precise guidance for users throughout the entire car purchase lifecycle, greatly improving the relevance of marketing content and user conversion rate. By deeply integrating AI technology with automotive marketing scenarios, it solves the pain points of traditional marketing methods and provides the automotive industry with an efficient, intelligent, and precise marketing solution. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an AI-based car marketing method proposed in one embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] like Figure 1 As shown in the figure, this application provides an AI model-based car marketing method, including: S101. Collect multi-source heterogeneous data related to users, and preprocess the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data.

[0022] To facilitate understanding of the embodiments of this application, examples of potentially multi-source heterogeneous data are provided as follows: Content browsing data: This reflects users' consumption behavior regarding automotive information content. For example, which articles or videos a user viewed, what the corresponding car model codes are, the duration of the user's stay on each page, and actions such as liking, commenting, sharing, or saving content can reveal a user's initial interest in a specific car model.

[0023] Vehicle interaction data reflects the depth of user interaction with specific vehicle information pages. For example, it includes virtual car viewing, 360° surround view, color switching, and corresponding vehicle model codes; detailed configuration information such as engine, transmission, and interior materials; and the frequency and total duration of these configuration viewings. This type of data reveals users' true interests and focus more effectively than simple browsing.

[0024] Service behavior interaction data: This reflects user behaviors indicating a strong purchase intention. For example, if a user uses a car calculator to calculate the price and monthly payment of a certain car model, the corresponding car model code and the number of calculations will be recorded; if a user books a test drive online, the corresponding car model code and the number of bookings will be recorded. This type of data is the most direct reflection of a user's car purchase intention.

[0025] S102. Use an AI model to identify the preprocessed multi-source heterogeneous data and obtain the user's corresponding car purchase intention tag.

[0026] In this embodiment, the purchase intention tag is a category tag used to describe the user's car purchase preferences, such as "family practicality," "sporty handling," "economical and fuel-efficient," "luxury business," and "off-road adventure." Each car model is pre-labeled with one or more such tags in the system. It is worth noting that for a single brand, the purchase intention tag can also be set for various different car models, thus identifying which car model the user prefers to buy and obtaining the user's corresponding purchase intention tag.

[0027] Before using an AI model for recognition, it is necessary to build and train the model first. The specific process is as follows: Model Construction: Deep learning models can be used, such as convolutional neural networks, long short-term memory networks, or a combination of both. CNNs excel at extracting spatial features from data and are suitable for scenarios where user behavior data is constructed into a two-dimensional matrix (such as a user-behavior matrix). LSTMs are adept at processing sequential data and can capture the evolution of user behavior over time, such as the shift in user interest from browsing A-class cars to gradually focusing on B-class cars.

[0028] Model Training: Prepare a historical dataset containing a large amount of heterogeneous data from multiple sources from historical users, along with their actual car purchase models (or their actual purchase intentions obtained through surveys). Use the intention labels corresponding to the actual car models as supervisory signals. Train the constructed AI model using intelligent optimization algorithms (such as variations of gradient descent algorithms like Adam and SGD), adjusting the model parameters to ensure that the predicted car purchase intention labels output by the model, after inputting the heterogeneous data from multiple sources from historical users, are as consistent as possible with the actual labels. After training, a usable trained AI model is obtained.

[0029] During real-time marketing, the preprocessed multi-source heterogeneous data of the current user in step S101 is input into the trained AI model, and the model can output the most likely car purchase intention tag for the user, such as "family practical".

[0030] S103. Based on the user's corresponding car purchase intention tag, generate corresponding first personalized car marketing content, and push the first personalized car marketing content to the user to complete the car marketing based on the AI ​​model.

[0031] For example, based on the "family practical" tag obtained by the user, all models tagged with this tag can be filtered from the vehicle database, such as model A (MPV), model B (compact SUV), and model C (mid-size sedan). These models are the target models. To further assist users in choosing among these target models, relevant content that can aid in decision-making can be randomly collected. This content mainly includes comparative reviews (such as a space comparison between model A and model B) and owner testimonials (such as real-world user experiences of model C owners). This content is highly valuable for users in the car selection stage. Finally, the generated first personalized automotive marketing content is precisely pushed to the user through channels they frequently use (such as in-app messages, SMS, and WeChat official account template messages).

[0032] This application provides an AI-based automotive marketing method. By collecting multi-source heterogeneous user data and using AI models for in-depth analysis, it can accurately identify users' car purchase intentions and push personalized marketing content based on these intentions, achieving precise marketing, avoiding the nuisance of invalid information, and improving user experience. By introducing a purchase stage recognition model, it can determine whether a user is in the awareness stage, the car selection stage, or the decision-making stage, and push corresponding types of marketing content according to different stages, achieving precise guidance for users throughout the entire car purchase lifecycle, greatly improving the relevance of marketing content and user conversion rate. By deeply integrating AI technology with automotive marketing scenarios, it solves the pain points of traditional marketing methods and provides the automotive industry with an efficient, intelligent, and precise marketing solution.

[0033] In one possible implementation, collecting user-related multi-source heterogeneous data includes: Based on a preset data sampling period, collect user-related content browsing data, vehicle interaction data, and service behavior interaction data to obtain the user-related multi-source heterogeneous data. The content browsing data includes the vehicle model code corresponding to the browsed information, dwell time, likes (set to 1 if the user likes the browsed information, otherwise 0), comments (set to 1 if the user comments on the browsed information, otherwise 0), sharing data (set to 1 if the user shares the browsed information, otherwise 0), and / or favorites data (set to 1 if the user shares the browsed information, otherwise 0). The vehicle interaction data includes the vehicle model code of the virtual interaction, the configuration code of the viewed information (such as the code for viewing the interior, the code for viewing engine parameters, the code for viewing vehicle dimensions), the viewing frequency (i.e., the number of times the user views the corresponding configuration page), and the viewing duration (i.e., the total dwell time of the user on the corresponding configuration page). The service behavior interaction data includes the vehicle model code for calculating the purchase price, the number of calculations, the vehicle model code for scheduling a test drive, and the number of schedulings. It is worth noting that, for ease of identification, all codes involved in this application embodiment are numerical codes.

[0034] For AI (Artificial Intelligence) models, the input is generally required to improve the accuracy of data recognition. Therefore, based on a preset data sampling period, the latest N pieces of content browsing data corresponding to users can be collected, along with vehicle interaction data and service behavior interaction data within that sampling period, thereby obtaining multi-source heterogeneous data related to the user.

[0035] In one possible implementation, preprocessing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data includes: normalizing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data.

[0036] In one possible implementation, before employing an AI model to identify the preprocessed multi-source heterogeneous data, the method further includes: AI models are built using convolutional neural networks or long short-term memory networks; Based on historical multi-source heterogeneous data and users' genuine car purchase intention tags, an intelligent optimization algorithm is used to train the AI ​​model, resulting in a trained AI model. This trained AI model is then used in subsequent car marketing processes. It's worth noting that the historical multi-source heterogeneous data and users' genuine car purchase intention tags are provided by staff. Furthermore, the historical multi-source heterogeneous data can be pre-processed. This AI model-based prediction and recommendation approach effectively improves marketing efficiency and accuracy.

[0037] Existing technologies for optimizing AI model parameters are prone to getting trapped in local optima, causing the AI ​​model to fail to accurately identify the data relationship between historical multi-source heterogeneous data and users' actual car purchase intention tags. This results in low prediction accuracy for car marketing by the trained AI model. Therefore, this application proposes a novel intelligent optimization algorithm to address the shortcomings of existing technologies and improve the accuracy of car marketing.

[0038] Optionally, the AI ​​model is trained using an intelligent optimization algorithm to obtain a trained AI model, including: The model parameters of the AI ​​model are initialized as follows:

[0039] in, Indicates the first i The first model parameter vector d dimensional hyperparameters, d =1,2,...,D, where D represents the total dimension of hyperparameters in the model parameter vector; when i When =1, then Denotes the first hyperparameter of a randomly initialized hyperparameter vector. d dimensional hyperparameters, Indicates the first i +1 model parameter vectors of the first d dimensional hyperparameters, Represents pi (π). Represents the first random number between (0,1). This represents the second random number between (0,1). This represents a third random number between (0,1). Represents the sine function. Represents the modulo function. Denotes the first constant term, and , This represents the chaos initialization control factor.

[0040] Optionally, random initialization includes: performing random initialization between the upper limit and the lower limit of model parameters, and encoding the randomly initialized model parameters into a vector to obtain a randomly initialized hyperparameter vector.

[0041] The above initialization method can effectively avoid the problem of uneven distribution of initial solutions in the solution space caused by random initialization, thereby improving the algorithm's ability to find the global optimal solution and accelerating the optimization speed of the algorithm.

[0042] Based on historical multi-source heterogeneous data and users' real car purchase intention tags, the loss function value corresponding to each model parameter vector is obtained, and the vector with the smallest loss function value is determined as the optimal vector. For example, historical multi-source heterogeneous data can be used as input, and the user's actual car purchase intention label can be used as the expected output. The cross-entropy loss function can be used to obtain the loss function value corresponding to the model parameter vector.

[0043] Based on the optimal vector, position information cross-learning is performed on the model parameter vector to obtain the model parameter vector after position information cross-learning:

[0044]

[0045]

[0046] in, Indicates the first t During the second optimization process, the first n A model parameter vector, Represents the optimal vector. n =1,2,…,N, where N represents the total number of model parameter vectors. Indicates the first n The model parameter vector after cross-learning of location information. Represents the natural constant. Represents the cosine function. Represents pi (π). This represents the inertial weight that increases with the number of training iterations. This represents the position disturbance factor, which can be set as a constant term between (0.1, 0.25); This represents the second constant term, which can be set to 1; This represents the third constant term, which can be set to 0.45; This represents the fourth random number between (0,1), and T represents the preset maximum number of training iterations. b Indicates the adaptive spiral shape factor. It represents the random spiral direction factor between (-1, 1).

[0047] The aforementioned cross-learning of positional information allows the model parameter vector to cross with the optimal vector in the early stages of the algorithm. This cross-learning process is non-linear, resulting in a completely new position that is closer to the currently known optimal position in the solution space. Then, by using non-linear learning to learn the information of the optimal vector, the algorithm's training speed is improved, and the exploration rate of the solution space and the ability to escape local optima are further enhanced. As the algorithm progresses, the training accuracy also gradually increases, ensuring the training accuracy of the algorithm.

[0048] After cross-learning the location information, the model parameter vector is subjected to basic information learning, resulting in the following model parameter vector after basic information learning:

[0049]

[0050]

[0051] in, Indicates the first t During the training process, the first m The model parameter vector after cross-learning of location information. Indicates the first m The model parameter vector after learning basic information. Indicates the basic learning control factor. Represents (0,2) π Random angles between ) π Represents pi (π). This represents the upper bound vector composed of the upper bounds of the parameters. This represents the lower bound vector composed of the lower bounds of the parameters. This represents the fifth random number between (0,1). This represents the sixth random number between (0,1). Represents the seventh random number between (0, 1); This represents the maximum value of the basic learning control factor, which can be set to 0.85; This represents the minimum value of the basic learning control factor, which can be set to 0.01; Indicates control parameters, This represents the eighth random number between (0,1).

[0052] The aforementioned basic information learning can be combined with boundary information to achieve neighborhood search, and random information can be used for search in the early stage of the algorithm to ensure diversity in the training process, which is more conducive to finding the global optimal solution. In the later stage of the algorithm, it tends to use the information of the optimal vector for search, which helps to improve the algorithm's random search ability in the early stage and the convergence ability in the later stage.

[0053] A global search is performed on the model parameter vector after learning the basic information, resulting in the following model parameter vector:

[0054]

[0055] in, Indicates the first t During the training process, the first kThe model parameter vector after learning basic information. Indicates the first k The model parameter vector after a global search. This represents the ninth random number between (0,1). Indicates the global search control factor. This represents the model parameter vector after learning random basic information. express and The Euclidean distance between them This represents the global search coefficient, and is set as a constant term between 3 and 5.

[0056] The aforementioned global search can effectively improve the algorithm's global search capability. Furthermore, as the algorithm progresses, the global search capability weakens, thus preventing the algorithm's training effect from stagnating or deteriorating.

[0057] Optionally, simulated annealing can be used to improve the global search, ensuring convergence in later stages. After each exploration of the model parameter vector, out-of-bounds handling can be performed on the parameters in the model parameter vector to ensure the validity of the parameters.

[0058] Determine if the current number of training iterations is greater than or equal to the preset maximum number of training iterations. If so, redetermine the optimal vector based on the model parameter vector after global search. Otherwise, return to the step of determining the vector with the minimum loss function value as the optimal vector.

[0059] The model parameters in the newly determined optimal vector are used as the final parameters of the AI ​​model, resulting in the trained AI model.

[0060] The intelligent optimization algorithm provided in this application embodiment can better train the model parameters of the AI ​​model (such as the connection weights between network layers), thereby finding the optimal combination of model parameters in the high-dimensional solution space, so that the trained AI model can better achieve prediction and improve the accuracy of automobile marketing.

[0061] In one possible implementation, based on the user's corresponding car purchase intention tag, corresponding first personalized car marketing content is generated, and the first personalized car marketing content is pushed to the user, including: Based on the user's corresponding car purchase intention tag, the target car model with the car purchase intention tag is determined; each car model is pre-set with at least one car purchase intention tag; Randomly collect comparative reviews and / or owner testimonials related to the target car model to obtain the first personalized car marketing content, and push the randomly collected first personalized car marketing content to users.

[0062] In one possible implementation, it also includes: A pre-set purchase stage identification model is used to identify the pre-processed multi-source heterogeneous data to obtain the predicted purchase stage corresponding to the user; wherein, the predicted purchase stage includes the awareness period, the car selection period, and the decision-making period; Based on the user's predicted purchase stage, a second personalized car marketing content is generated and pushed to the user.

[0063] In one possible implementation, the pre-setting method for the purchase stage identification model includes: A purchase phase identification model can be constructed using convolutional neural networks or long short-term memory networks; Purchase phase identification models built using convolutional neural networks, long short-term memory networks, or a combination of both can learn deep behavioral patterns and time-series dependencies from seemingly fragmented "multi-source heterogeneous data" of users.

[0064] Convolutional neural networks excel at extracting spatial features from data. For example, they can identify the behavior combination of "frequently comparing the configurations of three cars, A, B, and C" and regard it as a strong feature of the "car selection period".

[0065] Long Short-Term Memory (LSTM) networks excel at processing time-series data. For example, they can identify the evolution of user behavior paths: from "browsing information on multiple car models or technologies (cognition period)" -> "focusing on 2-3 specific car models for in-depth comparison (car selection period)" -> "frequently calculating financing options and scheduling test drives (decision period)".

[0066] This model can predict a user's current psychological state and decision-making progress, not just their superficial interests. This makes marketing pushes no longer blind, but precisely target the moment when the user "most needs" the information. For example, pushing "financial solutions" to a user still in the "awareness phase" is not only ineffective but may also cause resentment; while pushing "brand history" to a user who has entered the "decision-making phase" has too little information value. This solution solves the timing problem in the core marketing challenge of "pushing the right content at the right time through the right channel," effectively improving marketing precision.

[0067] Based on historical multi-source heterogeneous data and user's actual purchase stage labels, an intelligent optimization algorithm is used to train the purchase stage identification model to obtain a pre-set purchase stage identification model.

[0068] It is worth noting that the intelligent optimization algorithm provided in the embodiments of this application can be used to train the purchase stage identification model to obtain a pre-set purchase stage identification model.

[0069] In one possible implementation, if the user's predicted purchase stage is the awareness period, then brand introduction information and / or technical analysis information are randomly collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0070] When users are at this stage, their needs are for understanding and learning. Pushing in-depth content such as brand stories and core technologies (e.g., hybrid principles, intelligent driving systems) can satisfy their thirst for knowledge and establish a professional and trustworthy brand image. This is more effective than directly selling products in reducing user defensiveness, establishing an initial emotional connection and brand preference, and laying a solid foundation for subsequent conversions.

[0071] In one possible implementation, if the user's predicted purchase stage is the car selection period, then random comparative evaluations and / or owner reviews are collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0072] At this stage, the core task for users is comparison and selection. Pushing professional horizontal / vertical reviews helps users quickly understand the strengths and weaknesses of their target model compared to competitors in the same class; pushing genuine owner testimonials provides highly persuasive third-party verification, effectively dispelling user doubts and enhancing their confidence in a particular model. This content acts as a "professional advisor," guiding users to make choices that are beneficial to the brand.

[0073] In one possible implementation, if the user's predicted purchase stage is the decision-making period, then financial plan information, preferential policy information, and / or vehicle delivery process information are randomly collected to obtain second personalized car marketing content, and the second personalized car marketing content is pushed to the user.

[0074] When users reach this stage, it signifies a very strong purchase intention, and their focus shifts to cost and process. Pushing flexible financing options (such as low down payments and long-term loans) and limited-time offers can directly lower the purchase threshold and stimulate impulsive buying. Pushing information about the vehicle delivery process allows users to intuitively experience the joy of taking delivery and the smoothness of the process, eliminating their concerns and uncertainties about the final step, thereby effectively promoting the final transaction.

[0075] Each time a second personalized car marketing message is pushed, a fixed number of such messages can be pushed. Through precise marketing, marketing efficiency and conversion rates have been significantly improved. At the same time, the brand image has been optimized, user loyalty has been enhanced, and the entire marketing system has been driven towards automation and intelligence.

[0076] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0077] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0080] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0081] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0082] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An AI model-based car marketing method, characterized by, include: Collect multi-source heterogeneous data related to users, and preprocess the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data; An AI model is used to identify the preprocessed multi-source heterogeneous data and obtain the user's corresponding car purchase intention tag; Based on the user's corresponding car purchase intention tags, a first personalized car marketing content is generated and pushed to the user to complete the car marketing based on the AI ​​model. 2.The AI model-based automobile marketing method of claim 1, wherein, Collect multi-source heterogeneous data related to users, including: Based on a preset data sampling period, collect user-related content browsing data, vehicle interaction data, and service behavior interaction data to obtain the user-related multi-source heterogeneous data. The content browsing data includes the vehicle model code corresponding to the browsed information, the dwell time, likes, comments, shares and / or favorites; the vehicle interaction data includes the vehicle model code of the virtual interaction, the configuration code of the viewed information, the viewing frequency and the viewing duration; and the service behavior interaction data includes the vehicle model code for calculating the purchase price, the number of calculations, the vehicle model code for scheduling a test drive, and the number of schedulings. 3.The AI model-based automobile marketing method of claim 1, wherein, Preprocessing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data includes: normalizing the multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data. 4.The AI model-based automobile marketing method of claim 1, wherein, Before using an AI model to identify the preprocessed multi-source heterogeneous data, the process also includes: AI models are built using convolutional neural networks or long short-term memory networks; Based on historical multi-source heterogeneous data and users' real car purchase intention tags, the AI ​​model is trained using an intelligent optimization algorithm to obtain a trained AI model, which is then used in subsequent car marketing processes. 5.The AI model-based automobile marketing method of claim 1, wherein, Based on the user's corresponding car purchase intention tags, generate corresponding first personalized car marketing content, and push the first personalized car marketing content to the user, including: Based on the user's corresponding car purchase intention tag, the target car model with the car purchase intention tag is determined; each car model is pre-set with at least one car purchase intention tag; Randomly collect comparative reviews and / or owner testimonials related to the target car model to obtain the first personalized car marketing content, and push the randomly collected first personalized car marketing content to users. 6.The AI model-based automobile marketing method of claim 1, wherein, Also includes: A pre-set purchase stage identification model is used to identify the pre-processed multi-source heterogeneous data to obtain the predicted purchase stage corresponding to the user; wherein, the predicted purchase stage includes the awareness period, the car selection period, and the decision-making period; Based on the user's predicted purchase stage, a second personalized car marketing content is generated and pushed to the user. 7.The AI model-based automobile marketing method of claim 6, wherein, The pre-setting method for the purchase stage identification model includes: A purchase phase identification model can be constructed using convolutional neural networks or long short-term memory networks; The purchase stage recognition model is trained by using an intelligent optimization algorithm based on historical multi-source heterogeneous data and real purchase stage labels of users, so as to obtain a pre-set purchase stage recognition model. 8.The AI model-based automobile marketing method of claim 6, wherein, In the case that the predicted purchase stage corresponding to the user is the cognition stage, brand introduction information and / or technical analysis information are randomly collected to obtain second personalized automobile marketing content, and the second personalized automobile marketing content is pushed to the user. 9.The AI model-based automobile marketing method of claim 6, wherein, In the case that the predicted purchase stage corresponding to the user is the vehicle selection stage, comparison and / or evaluation and / or owner reputation are randomly collected to obtain second personalized automobile marketing content, and the second personalized automobile marketing content is pushed to the user. 10.The AI model-based automobile marketing method of claim 6, wherein, In the case that the predicted purchase stage corresponding to the user is the decision stage, financial scheme information, preferential policy information and / or vehicle delivery process information are randomly collected to obtain second personalized automobile marketing content, and the second personalized automobile marketing content is pushed to the user.

Images