Recommendation method and system based on big data tags

Abstract

The recommendation method and system based on the big data tag comprises the steps that vehicle-mounted burying point data is acquired, the vehicle-mounted burying point data comprises vehicle-mounted application data and vehicle-mounted information machine data, and a behavior tag is generated according to the vehicle-mounted burying point data by adopting a machine learning algorithm; constructing a preference database according to the behavior tags, the user tags, the vehicle tags and the vehicle travel data, wherein the behavior tags are in one-to-one correspondence with the user tags and the vehicle tags; constructing a preference model according to the preference database and the vehicle-mounted burying point data, wherein the preference model is used for generating preference information; judging a user intention according to the user voice by adopting a preset arbitration model, and when the user intention is matched with a preset arbitration result, acquiring vehicle-mounted real-time data according to the user intention; a preference model is called according to the user intention to obtain preference information, and the preset recommendation model generates recommendation information according to the preference information and the vehicle-mounted real-time data.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology for in-vehicle interactive information, and more specifically, to a recommendation method and system based on big data tags. Background Technology

[0002] With the increasing intelligence and widespread adoption of connected vehicles, users are becoming more reliant on them, making user behavior data a core asset for automakers to understand user needs. Building user tags (such as age and user type), vehicle tags (such as vehicle purpose and battery health), and behavioral tags (such as music listening habits and navigation habits) based on this data has become crucial for user profiling and precision marketing. However, traditional tagging systems rely on manual rules or shallow models, making it difficult to dynamically capture implicit correlations and long-term preference shifts in user behavior. Especially in complex scenarios, they struggle to support fine-grained, real-time personalized service needs.

[0003] In recent years, groundbreaking advancements in Large Language Models (LLM) and AI agents have provided a new paradigm for deeply mining user behavior patterns. With their massive parameter scale (hundreds of billions) and self-attention mechanisms, LLMs can extract semantic features from massive amounts of unstructured data and capture non-linear relationships in behavioral sequences through contextual reasoning. Simultaneously, LLMs support natural language interaction and dynamic knowledge injection, enabling them to possess both logical interpretability and scenario adaptability in personalized recommendation scenarios. However, current technologies have not yet fully integrated the fine-grained features and event-triggered mechanisms of embedded data, leaving room for improvement in recommendation timeliness and scenario relevance. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, embodiments of the present invention provide a recommendation method based on big data tags, comprising: acquiring in-vehicle tracking data, including in-vehicle application data and vehicle infotainment system data; generating behavioral tags based on the in-vehicle tracking data using a machine learning algorithm; constructing a preference database based on the behavioral tags, user tags, vehicle tags, and vehicle trip data, wherein the behavioral tags correspond one-to-one with the user tags and vehicle tags; constructing a preference model based on the preference database and the in-vehicle tracking data, wherein the preference model is used to generate preference information; determining user intent based on user voice using a preset arbitration model, and when the user intent matches the preset arbitration result, acquiring real-time in-vehicle data based on the user intent; invoking the preference model to obtain preference information based on the user intent; and generating recommendation information based on the preference information and the real-time in-vehicle data using a preset recommendation model.

[0005] Preferably, the in-vehicle application data includes navigation application data, music application data, and weather application data; the vehicle system data includes interface interaction control data, adjustment control data, and setting control data; the behavior tags include music behavior tags, navigation behavior tags, vehicle control behavior tags, and weather behavior tags; and the preference models include music preference models, navigation preference models, vehicle control preference models, and weather preference models.

[0006] Preferably, the step of generating behavior labels based on vehicle-mounted data points using machine learning algorithms further includes: preprocessing the acquired vehicle-mounted data points, including cleaning, deduplication, and normalization; extracting feature vectors from the preprocessed vehicle-mounted data points; training a behavior classification model based on the feature vectors using machine learning algorithms; and generating behavior labels based on the vehicle-mounted data points using the behavior classification model.

[0007] Preferably, the steps of constructing a music recommendation model based on the preference database and in-vehicle data include: using a large model to generate user music preferences based on music behavior tags and music application data, where music behavior tags include frequently listened music and listening time periods, and music application data includes music attribute information; retrieving target music data through music applications based on user music preferences; and fine-tuning the large model using the target music data and music behavior tags to obtain a music recommendation model.

[0008] Preferably, the preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to a received preset weather signal.

[0009] Preferably, the recommendation information also includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

[0010] On the other hand, a recommendation system based on big data tags is provided, including: an in-vehicle infotainment system (IVR) including navigation applications, music applications, weather applications, and an interactive interface; a preference database that obtains in-vehicle embedded data through the IVR, including in-vehicle application data and IVR data, and uses machine learning algorithms to generate behavioral tags based on the in-vehicle embedded data, and binds the behavioral tags with user tags and vehicle tags; and a recommendation device that is communicatively connected to both the IVR and the preference database. The recommendation device includes a preference module, an arbitration module, and a recommendation module. The preference module includes multiple preference units, each of which includes a preference model. The preference model is configured to generate user preferences based on behavioral tags and user embedded data. The arbitration module includes a preset arbitration model, which is configured to determine user intent based on user voice. When the user intent matches the preset arbitration result, real-time in-vehicle data is obtained based on the user intent. The recommendation module is configured to call the preference model to obtain user preferences based on the user intent. The recommendation module also includes a preset recommendation model, which generates recommendation information based on preference information and real-time in-vehicle data. The recommendation module outputs the recommendation information through the IVR.

[0011] Preferably, the preference model includes a music preference model, a navigation preference model, a vehicle control preference model, and a weather preference model.

[0012] Preferably, the preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to a received preset weather signal.

[0013] Preferably, the recommendation information also includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

[0014] The big data tag-based recommendation method and system of this invention have the following technical effects: Through a multi-dimensional data fusion and extraction framework, data from multiple embedded points such as vehicle controllers and in-vehicle apps is extracted. Based on Flink real-time streaming technology, a dynamically updated three-dimensional tag system of user-vehicle-behavior is constructed. A domain-specific intelligent agent driven by a large model is introduced, which intelligently extracts user behavior preferences by combining user behavior tags and maps user tags to the semantic space. Finally, an event-triggered adaptive recommendation mechanism is developed. When the system detects a specific behavior (voice trigger) or external signal (such as abnormal weather signals), it automatically calls the large model to generate personalized recommendation suggestions in real time, providing behavioral recommendations and customized suggestion responses via voice. This solution effectively improves the intelligence and real-time performance of personalized recommendations, constructing a customized recommendation assistant tailored to each individual user. Attached Figure Description

[0015] To better understand the above and other objects, features, advantages, and functions of the present invention, reference can be made to the embodiments shown in the accompanying drawings. The same reference numerals in the drawings refer to the same parts. Those skilled in the art should understand that the drawings are intended to schematically illustrate preferred embodiments of the invention and do not limit the scope of the invention in any way; the parts in the drawings are not drawn to scale.

[0016] Figure 1 A schematic diagram illustrating an application scenario of a recommendation method based on big data tags according to an embodiment of the present invention is shown.

[0017] Figure 2 A block diagram of a big data tag-based recommendation system according to an embodiment of the present invention is shown. Detailed Implementation

[0018] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0019] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0020] To at least partially address one or more of the above-mentioned problems and other potential problems, embodiments of this disclosure propose an embodiment of the present invention that provides a recommendation method based on big data tags, including: step S1, acquiring in-vehicle embedded data, the in-vehicle embedded data including in-vehicle application data and vehicle system data, and using a machine learning algorithm to generate behavioral tags based on the in-vehicle embedded data.

[0021] Specifically, the cloud connects to the vehicle's infotainment system through vehicle networking technology, embedding data collection points in the application interface, functional modules, and interaction processes of the infotainment system to collect vehicle embedded data in real time, and the cloud stores the acquired vehicle embedded data through a database.

[0022] Based on the aforementioned multi-dimensional vehicle tracking data, deep tagging data (such as frequently visited locations and vehicle usage) is extracted using Flink real-time stream processing and machine learning technologies. Apache Flink is a high-performance, low-latency real-time stream processing framework widely used in big data real-time processing scenarios. Its core features include real-time performance, high throughput, low latency, consistency guarantees, fault tolerance, and flexible event-time processing capabilities. Machine learning algorithms generate behavioral tags based on vehicle tracking data stored in a database, including historical and current tracking data. Real-time collection of vehicle tracking data allows the machine learning algorithm to update the generated behavioral tags in a timely manner.

[0023] In some embodiments, the step of generating behavioral labels based on vehicle-mounted data sets using machine learning algorithms further includes: step S11, preprocessing the acquired vehicle-mounted data, the preprocessing including cleaning, deduplication and normalization.

[0024] Specifically, data preprocessing is a key step, which requires cleaning, deduplication, and normalization of the collected raw vehicle-mounted data points to ensure data quality and consistency.

[0025] Step S12: Extract feature vectors based on the preprocessed vehicle-mounted data.

[0026] Specifically, feature extraction transforms user behavior data and user attribute data into feature vectors that describe user characteristics. For example, feature vectors such as a user's listening frequency and style preferences can be extracted from a user's music search history and music playback history data. Furthermore, statistical methods, similarity calculation algorithms, classification and clustering algorithms, and other techniques can be used to further mine the data.

[0027] Step S13: Use machine learning algorithms to train a behavior classification model based on feature vectors.

[0028] Specifically, machine learning algorithms are used to model feature vectors, thereby training a behavior classification model. Commonly used machine learning algorithms include decision trees, logistic regression, support vector machines, neural networks, and deep learning.

[0029] Step S14: Use a behavior classification model to generate behavior labels based on vehicle-mounted data.

[0030] Specifically, user behavior is categorized and labeled based on the model, generating user behavior tags. Different in-vehicle tracking data has different tags; therefore, machine learning is performed separately on the tracking data of different in-vehicle applications to obtain tags for each application. Examples include tags for music behavior, navigation behavior, vehicle model behavior, and weather behavior.

[0031] Step S2: Construct a preference database based on behavior tags, user tags, vehicle tags, and vehicle trip data, with each behavior tag corresponding to a user tag and a vehicle tag.

[0032] Specifically, the preference database is a cloud-based database that stores multi-dimensional data sets composed of user tags, vehicle tags, behavioral tags, and vehicle data. This enables data support for multi-user and multi-vehicle application scenarios. User tags are the registered user IDs, and vehicle tags are the vehicle VIN codes. By mapping behavioral tags to user tags and vehicle tags one-to-one, the database binds users, vehicles, and behavioral tags, thereby providing accurate data for the target vehicle.

[0033] Step S3: Construct a preference model based on the preference database and vehicle-mounted data points. The preference model is used to generate preference information.

[0034] Specifically, a preference model is built based on a large language model. For example, in the music field, the large model serves as the decision-making brain, connecting tools such as search plugins and music plugins. It can combine user music behavior and personal attributes to extract user music preferences (such as time period-song type preference, playlist recommendations, etc.).

[0035] In some embodiments, the step of constructing a music recommendation model based on a preference database and in-vehicle data includes: step S31, using a large model to generate user music preferences based on music behavior tags and music application data, where music behavior tags include frequently listened music and listening time periods, and music application data includes music attribute information.

[0036] Specifically, by combining user behavior tags (such as frequently listened music and listening times) extracted from in-vehicle data tracking in music applications, and simultaneously obtaining frequently listened music attributes (style, era, etc.) from music search plugins, these are all injected into a large model for analysis and decision-making to summarize user music preferences. The input prompt templates for the large model are preset instruction content, which includes task information, personality information, etc.

[0037] Step S32: Retrieve target music data through a music application based on the user's music preferences.

[0038] Specifically, based on the music preferences output by the large model, the system retrieves the target music data obtained by the in-vehicle music application.

[0039] Step S33: Fine-tune the large model using target music data and music behavior tags to obtain a music recommendation model.

[0040] Specifically, the target music data corresponds to the current music preferences. The retrieved target music data and music behavior tags are then input back into the large model to fine-tune it. In another embodiment, steps S32 and S33 can be repeated to correct and summarize the large model, obtaining the final user preferences and recommended music content after multiple iterations.

[0041] In some implementations of the fine-tuning steps for music preference models, music behavior tags and information obtained from deep retrieval are input into a large model. The large model outputs music preferences, such as rock music, classical music, or frequently listening to a particular artist during work hours. These music preferences are then input into a search plugin, which searches the overall music database to retrieve target information that corresponds to the current music preference. The retrieved target information and music behavior tags are then input back into the fine-tuned large model. This process is repeated multiple times to obtain the final preference.

[0042] In some embodiments, in-vehicle application data includes navigation application data, music application data, and weather application data; vehicle system data includes interface interaction control data, adjustment control data, and setting control data; behavior tags include music behavior tags, navigation behavior tags, vehicle control behavior tags, and weather behavior tags; and preference models include music preference models, navigation preference models, vehicle control preference models, and weather preference models.

[0043] Specifically, data points are also embedded in various applications installed in the vehicle system, such as music, navigation, vehicle control, and weather, to collect information such as the types of music users frequently listen to, navigation destinations, vehicle control operations, and weather information they are interested in.

[0044] Data is collected in various functional modules of the human-machine interface, such as voice recognition, data management, and electronic and electrical control. This includes collecting user adjustment data for the air conditioning, seats, and steering wheel, vehicle function settings data, and interaction commands with the vehicle's infotainment system.

[0045] The cloud database extracts behavioral tags from the in-vehicle data points and vehicle infotainment system data obtained by various in-vehicle applications.

[0046] The navigation preference model, vehicle control preference model, and weather preference model within the preference model each employ a large language model. For example, a navigation preference model is constructed by setting navigation preference prompt templates, enabling the large model to generate navigation preferences based on navigation behavior tags and navigation application tracking data. The navigation preference prompt templates include task information and user profile information. The navigation preference model is configured to generate navigation preferences based on these prompt templates, using navigation behavior tags and navigation application tracking data. For instance, navigation behavior tags include navigation distance and destination type, while navigation application tracking data includes navigation address and time. Navigation preferences can be used to determine a user's preferred navigation destination, route, etc.

[0047] The vehicle control preference model employs a large model to construct a vehicle control preference prompt template. Based on this template, the large model generates vehicle control preferences using vehicle control behavior tags and in-vehicle data from embedded data points. For example, vehicle control behavior tags include window opening time periods, temperature when windows are open, air conditioning start time, and air conditioning start temperature. Embedded data from the in-vehicle system includes in-vehicle system status (including window status, wiper status, etc.) and scene information (including time, temperature, road conditions, etc.). The vehicle control preference prompt template includes user profile information and task information, enabling the large model to output vehicle control preferences. The vehicle control preference model determines user preferences such as window opening preferences and air conditioning preferences.

[0048] The weather preference model employs a large-scale model to construct a weather preference prompt template. Based on this template, the large-scale model generates vehicle control preferences using vehicle control behavior tags and in-vehicle data from embedded data points. For example, vehicle control behavior tags include weather when windows are open and weather when air conditioning is on; embedded data points include vehicle status (window status, air conditioning status) and weather application data; and the vehicle control preference prompt template includes user profile information and task information, enabling the large-scale model to output weather preferences. The weather preference model determines user preferences such as window opening preferences and air conditioning preferences.

[0049] Step S4: Use a preset arbitration model to determine the user's intent based on the user's voice. When the user's intent matches the preset arbitration result, obtain real-time vehicle data based on the user's intent.

[0050] Specifically, the preset arbitration model adopts a large language model. The large language model recognizes user intent through voice. The preset arbitration results include playing music, adjusting the air conditioning, opening the car window, and navigation recommendations. For example, when the user intent recognized by the large language model matches any of the preset arbitration results, the corresponding current in-vehicle real-time data is obtained according to the user intent. For example, when the user intent is to play music, the in-vehicle real-time data is the current music application data retrieved and output from the music application according to music preferences.

[0051] Step S5: Based on the user's intent, the preference model is invoked to obtain preference information, and the preset recommendation model generates recommendation information based on the preference information and real-time vehicle data.

[0052] Specifically, the corresponding preference model is invoked based on the user's intent. For example, when the identified user intent is to play music, the music preference model is invoked to obtain the user's music preferences; when the identified user intent is to adjust the air conditioning (or other vehicle control intent), the vehicle control preference model is invoked to obtain the vehicle control preferences; and when the identified user intent is to recommend navigation, the weather preference model is invoked to obtain the weather preferences.

[0053] The preset recommendation model adopts a large-scale model based on existing technology. This model includes preset recommendation prompt templates, which incorporate task information and character settings. By using preference information and real-time in-vehicle data obtained in the previous step as input to the large recommendation model, the model outputs recommendation information.

[0054] In some embodiments, the preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to a received preset weather signal.

[0055] Specifically, the system acquires weather signals in real time. When the weather signal matches the preset weather signal, the preset recommendation model calls the weather preference model to obtain weather preferences and acquires weather application data from the vehicle's real-time data. Based on the preset recommendation template, the system uses a large recommendation model to generate recommendation information according to the weather preferences and the vehicle's real-time data. For example, it can generate recommendation information such as turning on fog lights in smoggy weather.

[0056] In some embodiments, the recommendation information further includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

[0057] Specifically, the recommended instructions in the recommendation information include turning on the air conditioner. The in-vehicle air conditioning application is configured to turn on the air conditioner in response to the air conditioner turning-on information in the recommendation information and adjust the air conditioner temperature according to the air conditioner temperature in the recommendation information.

[0058] On the other hand, a recommendation system based on big data tags is provided, including: an in-vehicle infotainment system, which includes navigation applications, music applications, weather applications, and interactive interfaces.

[0059] Specifically, the in-vehicle infotainment system is a vehicle that supports the Internet of Vehicles. The in-vehicle infotainment system is equipped with navigation applications, music applications, weather applications and their corresponding operating interfaces. The in-vehicle infotainment system also includes interactive interfaces that support user voice input and manual operation, such as operation panels such as smart displays.

[0060] The preference database obtains in-vehicle tracking data through the vehicle's infotainment system. This data includes in-vehicle application data and vehicle infotainment system data. Machine learning algorithms are used to generate behavioral tags based on the in-vehicle tracking data, and these behavioral tags are then bound to user tags and vehicle tags.

[0061] Specifically, the preference database is located in the cloud. The cloud database also includes user tags and vehicle tags. The user tag is the user's registration ID, and the vehicle tag is the vehicle's VI code. The cloud database is used to provide data support for multiple users and multiple vehicles. The user tags and vehicle tags are used to distinguish different vehicle tracking data in the preference database. For example, vehicle tracking data of different users in the same vehicle, vehicle tracking data of the same user in different vehicles, and vehicle tracking data of different users in different vehicles can be accurately distinguished through user tags and vehicle tags.

[0062] The cloud database determines the corresponding vehicle tracking data based on user tags and vehicle tags respectively, then generates behavior tags based on the vehicle tracking data, and binds and stores the behavior tags, user tags, and vehicle tags in the preference database.

[0063] The recommendation device is communicatively connected to both the vehicle infotainment system and the preference database. The device includes a preference module, an arbitration module, and a recommendation module. The preference module comprises multiple preference units, each of which includes a preference model. The preference model is configured to generate user preferences based on behavioral tags and user tracking data. The arbitration module includes a preset arbitration model, configured to determine user intent based on user voice. When the user intent matches the preset arbitration result, real-time vehicle data is obtained based on the user intent. The recommendation module is configured to retrieve user preferences by calling the preference model based on the user intent. The recommendation module also includes a preset recommendation model, which generates recommendation information based on preference information and real-time vehicle data. The recommendation module outputs the recommendation information through the vehicle infotainment system.

[0064] Specifically, the recommendation device is a mobile device with a memory and a processor. The recommendation device has an interface for data connection and communication with the vehicle's infotainment system. The memory of the recommendation device includes computer program instructions. When the recommendation device is connected to the vehicle's infotainment system through the interface, the computer program instructions stored in the recommendation device are executed by the processor to perform a recommendation method based on big data tags.

[0065] In some embodiments, the recommendation device further includes a communication module, which, when connected to the vehicle's infotainment system, is used to call up existing large-scale models online to enable preference analysis, user intent arbitration, and generation of recommendation information for each module.

[0066] In some embodiments, the preference model includes a music preference model, a navigation preference model, a vehicle control preference model, and a weather preference model.

[0067] Specifically, the preference module includes multiple units such as music preference unit, navigation preference unit, vehicle control preference unit, and weather preference unit.

[0068] The music preference unit includes a music preference model, which generates user preferences based on music behavior tags and in-vehicle tracking data. In some embodiments, the music preference model employs a large language model. The music preference unit includes a preset music preference prompt template. Based on the music preference prompt template, the large model generates music preferences according to music behavior tags and music application data in real-time in-vehicle data. The music preference prompt template includes task information and personality information, etc.

[0069] The navigation preference unit includes a navigation preference model, which generates user preferences based on navigation behavior tags and in-vehicle data. In some embodiments, the navigation preference model employs a large language model. The navigation preference unit includes a preset navigation preference prompt template. Based on the prompt template, the large model generates navigation preferences using navigation behavior tags and navigation application data from real-time in-vehicle data. The navigation preference prompt template includes task information and user profile information. The navigation preference unit allows users to select navigation route preferences and / or traffic condition preferences. For example, users may prefer long, highway routes and secondarily choose short, low-speed routes.

[0070] The vehicle control preference unit includes a vehicle control preference model, which is used to generate user preferences based on vehicle control behavior tags and in-vehicle data. In some embodiments, the vehicle control preference model adopts a large language model. The vehicle control preference unit includes a preset vehicle control preference prompt template. Based on the vehicle control preference prompt template, the large model generates vehicle control preferences based on vehicle control behavior tags and vehicle control data in real-time in-vehicle data. The vehicle control preference prompt template includes task information and personality information, etc.

[0071] The weather preference unit includes a weather preference model, which is used to generate user preferences based on weather behavior tags and vehicle-mounted data. In some embodiments, the weather preference model employs a large language model. The weather preference unit includes a preset weather preference prompt template. Based on the weather preference prompt template, the large model generates weather preferences according to weather behavior tags and weather application data in real-time vehicle data. The weather preference prompt template includes task information and personality information, etc.

[0072] The arbitration module includes a pre-set arbitration model that uses a large model. When the recommendation device is connected to the vehicle's infotainment system, the arbitration module receives voice information collected by the system in real time. The pre-set arbitration model is configured to recognize user intent based on voice information. The arbitration module also includes a pre-set arbitration result. When the recognized user intent matches the pre-set arbitration result, the recommendation module of the recommendation device is activated.

[0073] The recommendation module receives user intent through the arbitration module and calls the corresponding preference model based on the user intent to obtain user preferences. The preset recommendation model adopts a large model of existing technology. The recommendation module includes preset recommendation prompt templates, which include task information and character settings, etc. By using preference information and real-time vehicle data obtained in the previous step as input to the large recommendation model, the large model outputs recommendation information.

[0074] In some embodiments, the preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to a received preset weather signal.

[0075] Specifically, the recommendation module is activated in response to a preset weather signal, and uses the weather preference model to obtain weather preferences based on the preset weather signal. At the same time, it obtains weather application data from real-time vehicle data, so that the preset recommendation model can generate recommendation information.

[0076] In some embodiments, the recommendation information further includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

[0077] Specifically, the vehicle's infotainment system is configured to receive recommendation information via a recommendation module and perform corresponding operations based on that information. For example, the system's voice interaction module might activate the speaker in response to the received recommendation information and play the recommended information through the speaker. The vehicle control module is configured to perform corresponding operations in response to recommended instructions within the received recommendation information. For instance, if the recommendation information includes a suggestion to turn on the windshield wipers, the vehicle control module would then activate the windshield wipers in response to this suggestion.

[0078] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand this document.

Claims

1. A recommendation method based on big data tags, characterized in that, include: Acquire in-vehicle data, which includes in-vehicle application data and vehicle infotainment data, and use machine learning algorithms to generate behavioral tags based on the in-vehicle data. A preference database is constructed based on behavioral tags, user tags, vehicle tags, and vehicle trip data, with a one-to-one correspondence between behavioral tags and user and vehicle tags. A preference model is constructed based on the preference database and vehicle-mounted data points. The preference model is used to generate preference information. A preset arbitration model is used to determine the user's intent based on the user's voice. When the user's intent matches the preset arbitration result, real-time vehicle data is obtained based on the user's intent. The user's intention is used to retrieve preference information by calling the preference model. The preset recommendation model generates recommendation information based on the preference information and real-time vehicle data.

2. The method according to claim 1, characterized in that, In-vehicle application data includes navigation application data, music application data, and weather application data; vehicle system data includes interface interaction control data, adjustment control data, and setting control data; behavior tags include music behavior tags, navigation behavior tags, vehicle control behavior tags, and weather behavior tags; and preference models include music preference models, navigation preference models, vehicle control preference models, and weather preference models.

3. The method according to claim 2, characterized in that, The steps of generating behavioral labels based on vehicle-mounted data points using machine learning algorithms also include: The acquired vehicle-mounted data points are preprocessed, including cleaning, deduplication, and normalization. Feature vectors are extracted from the preprocessed vehicle-mounted data points. A behavior classification model is trained using machine learning algorithms based on feature vectors; A behavior classification model is used to generate behavior labels based on vehicle-mounted data.

4. The method according to claim 2, characterized in that, The steps involved in building a music recommendation model based on preference databases and in-vehicle tracking data include: A large model is used to generate user music preferences based on music behavior tags and music application data. Music behavior tags include frequently listened music and listening times, while music application data includes music attribute information. Based on the user's music preferences, target music data is retrieved through music applications. A music recommendation model is obtained by fine-tuning the large model using target music data and music behavior tags.

5. The method according to claim 1, characterized in that, The preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to received preset weather signals.

6. The method according to claim 1, characterized in that, The recommendation information also includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

7. A recommendation system based on big data tags, characterized in that, include: The in-vehicle infotainment system includes navigation applications, music applications, weather applications, and a user interface. The preference database obtains in-vehicle data through the vehicle system, which includes in-vehicle application data and vehicle system data. Machine learning algorithms are used to generate behavioral tags based on the in-vehicle data, and these behavioral tags are then bound to user tags and vehicle tags. The recommendation device is communicatively connected to both the vehicle infotainment system and the preference database. The device includes a preference module, an arbitration module, and a recommendation module. The preference module comprises multiple preference units, each of which includes a preference model. The preference model is configured to generate user preferences based on behavioral tags and user tracking data. The arbitration module includes a preset arbitration model, configured to determine user intent based on user voice. When the user intent matches the preset arbitration result, real-time vehicle data is obtained based on the user intent. The recommendation module is configured to retrieve user preferences by calling the preference model based on the user intent. The recommendation module also includes a preset recommendation model, which generates recommendation information based on preference information and real-time vehicle data. The recommendation module outputs the recommendation information through the vehicle infotainment system.

8. The system according to claim 7, characterized in that, Preference models include music preference model, navigation preference model, vehicle control preference model and weather preference model.

9. The system according to claim 7, characterized in that, The preset recommendation model is configured to generate recommendation information based on real-time vehicle data in response to received preset weather signals.

10. The system according to claim 7, characterized in that, The recommendation information also includes recommendation instructions, which are configured to control the vehicle system to perform corresponding recommended operations.

Images