Scene mode construction method and electronic equipment

Abstract

The embodiment of the invention provides a scene mode construction method and electronic equipment, and the method comprises the steps: obtaining target feature data of a target vehicle, the target feature data being obtained by screening a plurality of first feature data according to the importance degree of the plurality of first feature data of the target vehicle for label learning; clustering the target feature data to obtain at least one clustering center, and determining a target label of the target feature data based on the at least one clustering center; mining the target feature data based on the target tag to obtain at least one execution action data and a configuration parameter corresponding to the at least one execution action data; and based on the at least one execution action data and the configuration parameters, constructing a scene mode corresponding to the target label. According to the invention, the technical problem of low construction accuracy of the scene mode in the prior art is solved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and particularly to a method for constructing a scene mode and an electronic device. Background Art

[0002] The current vehicle intelligent cockpit system is developing towards a more personalized and intelligent direction. To improve the user experience, many automobile manufacturers and technology companies are committed to understanding the preferences and behavior habits of car owners through data analysis.

[0003] However, most of the existing personalized construction technologies are based on artificially preset scenarios and rules. This passive scenario triggering often fails to comprehensively capture the personalized needs of car owners, resulting in a low construction accuracy of the scene mode in related technologies. Summary of the Invention

[0004] The embodiments of the present application provide a method for constructing a scene mode and an electronic device, aiming to improve the technical problem of low construction accuracy of the scene mode in related technologies.

[0005] According to one aspect of the embodiments of the present application, a method for constructing a scene mode is provided, including: obtaining target feature data of a target vehicle, where the target feature data is used to screen multiple first feature data according to the importance of the multiple first feature data for label learning; clustering the target feature data to obtain at least one clustering center, and determining a target label of the target feature data based on the at least one clustering center; mining the target feature data based on the target label to obtain at least one execution action data and at least one configuration parameter corresponding to the at least one execution action data; and constructing a scene mode corresponding to the target label based on the at least one execution action data and the configuration parameter.

[0006] In this process, based on the feature data of the car owner and the vehicle, a personalized and safe and controllable scene mode is constructed through clustering and mining technologies, thereby improving the construction accuracy of the scene mode.

[0007] Further, mining the target feature data based on the target label to obtain at least one execution action data and at least one configuration parameter corresponding to the at least one execution action data includes: performing action mining on the target feature data based on the target label to obtain at least one execution action data; and performing parameter mining on the target feature data based on the target label and the at least one execution action data to obtain a configuration parameter.

[0008] In this process, the action mining and parameter mining processes based on the target label are refined to ensure that the mined execution actions and configuration parameters are highly relevant to the target scenario and the user, thereby improving the construction accuracy and practicality of the scene mode.

[0009] Further, the target tags include an owner tag, a scenario tag, and a service tag. The owner tag is used to represent the owner information of the owner corresponding to the target vehicle. The scenario tag is used to represent the scenario information of the scenario where the target vehicle is located. The service tag is used to represent the service information of the service operation performed by the target vehicle. Action mining is performed on the target feature data based on the target tags to obtain at least one execution action data, including one of the following: filtering the target feature data based on the owner tag and the scenario tag to obtain the filtered feature data, and performing action mining on the filtered feature data based on the service tag to obtain at least one execution action data; performing action mining on the target feature data based on the service tag to obtain multiple initial execution action data, and filtering the multiple initial execution action data based on the owner tag and the scenario tag to obtain at least one execution action data.

[0010] In this process, not only is the driving experience of the owner significantly improved, the frequency of manual operations is reduced, and driving safety is enhanced, but also the deep interaction between the owner and the system is promoted, and the trust and satisfaction of the owner are improved.

[0011] Further, filtering the target feature data based on the owner tag and the scenario tag to obtain the filtered feature data includes: determining the correlation degree between the owner tag and the scenario tag; performing action mining on the target feature data based on the correlation degree to obtain the filtered feature data.

[0012] In this process, by combining the correlation degree with action mining, the system can achieve highly personalized service push. Not only is the driving experience of the owner significantly improved, the trust and dependence of the owner on the system are enhanced, but also the deep interaction between the system and the owner is promoted, and the market competitiveness of the system is improved.

[0013] Further, performing action mining on the filtered feature data based on the service tag to obtain at least one execution action data includes: based on the service tag, determining the feature data corresponding to different execution action data in the filtered feature data; based on the feature data corresponding to different execution action data, determining the action execution frequency of different execution action data; determining at least one execution action data from different execution action data based on the action execution frequency.

[0014] In this process, the frequency-based parameter mining method can more accurately reflect the preference settings and operation habits of the owner in a specific scenario, ensuring that the system can automatically provide the service that best meets the needs of the owner based on the high-frequency operation habits of the owner. Thus, not only is the operation burden of the owner reduced, but also the personalization and comfort of the driving experience are significantly improved, and the dependence and satisfaction of the owner on the system are enhanced.

[0015] Further, parameter mining is performed on the target feature data based on the target label and at least one execution action data to obtain configuration parameters, including: determining, based on the target label, action feature data in the target feature data corresponding to at least one execution action data; and performing parameter mining on the action feature data to obtain configuration parameters.

[0016] In this process, the system can accurately mine action feature data related to execution actions from the target feature data, and further refine the personalized configuration parameters of the vehicle owner, not only deepening the personalization degree of the system service, but also significantly improving the practicability of the service and user satisfaction.

[0017] Further, parameter mining is performed on the action feature data to obtain configuration parameters, including: determining the parameter setting frequencies of different parameters corresponding to at least one execution action data based on the action feature data; and determining configuration parameters from different parameters based on the parameter setting frequencies.

[0018] In this process, by automatically identifying and applying the parameter configuration that the vehicle owner is most accustomed to, the vehicle system can provide a more accurate and smooth service experience, reduce unnecessary operations, and enhance driving safety.

[0019] Further, determining the target label of the target feature data based on at least one clustering center includes: outputting at least one clustering center; receiving feedback information on the at least one clustering center; and naming the target feature data based on the feedback information to obtain the target label.

[0020] In this process, the personalization and intelligence of the service are enhanced through the user feedback mechanism. Also, through continuous interactive learning, the accuracy of the service and the user experience are continuously improved. At the same time, the transparency and interpretability of the system are enhanced through the feedback information, enabling users to understand the data logic behind the system recommendations and enhancing the trust in personalized services.

[0021] Further, obtaining the target feature data of the target vehicle includes: clustering a plurality of first feature data to obtain a first clustering result; performing a preset operation on at least one of the plurality of first feature data to obtain a plurality of second feature data; clustering the plurality of second feature data to obtain a second clustering result; and determining the target feature data from the plurality of first feature data based on the first clustering result and the second clustering result.

[0022] In this process, the target feature data is screened out through two rounds of clustering and preset operations, ensuring that the data used to construct the recommendation model is of high quality and closely related to the target scenario. At the same time, the data screening efficiency and accuracy are improved, and the interference of redundant data is reduced.

[0023] Further, determining target feature data from multiple first feature data based on the first clustering result and the second clustering result includes: determining the influence level of at least one feature data based on the first clustering result and the second clustering result, where the influence level is used to represent the influence degree of at least one feature data on clustering; determining the target feature data from at least one feature data based on the influence level.

[0024] In this process, it is ensured that the constructed scenario pattern focuses more on the key parameters that affect user behavior and scenario features, improving the practicality and effectiveness of the pattern.

[0025] Further, obtaining multiple first feature data of the target vehicle includes: collecting multi-dimensional data of the target vehicle; performing feature extraction on the multi-dimensional data to obtain multiple first feature data.

[0026] In this process, by collecting multi-dimensional data and performing feature extraction, a comprehensive and accurate data basis is provided for subsequent clustering and mining, ensuring that the constructed scenario pattern is based on sufficient and diverse data and can comprehensively reflect user needs and behavior patterns.

[0027] According to another aspect of the embodiments of the present application, there is also provided a device for constructing a scenario pattern, including: an acquisition module, configured to acquire target feature data of a target vehicle, where the target feature data is used to screen multiple first feature data according to the importance degree of the multiple first feature data of the target vehicle for label learning; a clustering module, configured to cluster the target feature data to obtain at least one clustering center, and determine a target label of the target feature data based on the at least one clustering center; a mining module, configured to mine the target feature data based on the target label to obtain at least one execution action data and configuration parameters corresponding to the at least one execution action data; a construction module, configured to construct a scenario pattern corresponding to the target label based on the at least one execution action data and the configuration parameters.

[0028] According to another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory storing an executable program; a processor configured to run the program, where when the program runs, it executes the above-mentioned method for constructing a scenario pattern.

[0029] According to another aspect of the embodiments of the present application, there is also provided a computer-readable storage medium, where the computer-readable storage medium includes a stored executable program, and when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the above-mentioned method for constructing a scenario pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flowchart of a method for constructing a scenario pattern according to an embodiment of the present application;

[0031] Figure 2 It is a schematic diagram of a framework mining process according to an embodiment of the present application;

[0032] Figure 3 It is a schematic diagram of a parameter mining process according to an embodiment of the present application;

[0033] Figure 4 It is a structural diagram of a construction device for a scenario mode according to an embodiment of the present application;

[0034] Figure 5 It is a structural diagram of an electronic device provided according to an embodiment of the present application. Specific embodiments

[0035] In order to make the technical problems, technical solutions and beneficial effects solved by the present application more clearly understood, the present application will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0036] Unsupervised learning: It refers to a machine learning method used to systematically understand and deconstruct the complexity of the behavior and preferences of vehicle owners in the system, and is committed to automatically extracting and characterizing the inherent patterns and structures from a large amount of unclassified or unlabeled vehicle owner data. Different from the traditional supervised learning method that relies on known labels or results, in the present application, unsupervised learning explores and reveals the behavior patterns of vehicle owners in different driving scenarios through steps such as clustering analysis, feature selection, and importance evaluation.

[0037] Clustering: It refers to an unsupervised learning method used to group samples in a dataset according to their feature similarities to form different "clusters" or "classes". The types of clustering methods can include but are not limited to the K-Means clustering algorithm (abbreviated as K-Means), the density-based clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, abbreviated as DBSCAN), and hierarchical clustering, etc.

[0038] Clustering algorithm: It refers to a data analysis tool under the unsupervised learning framework used to automatically discover and define groups of vehicle owners with similar behavior characteristics. By quantifying the similarity between vehicle owner data points, the clustering algorithm can group these data points into multiple clusters, and each cluster represents a group of vehicle owners with a common behavior pattern. In the present application, clustering is not limited to simply grouping based on the similarity of data points, but also incorporates a feature importance evaluation mechanism to ensure that the learned patterns have high relevance and interpretability for predicting vehicle owner behavior.

[0039] Association degree: It refers to a quantitative relationship between vehicle owner tags and scenario tags. The association degree can be obtained through association rule learning using the A Priori Algorithm (simply referred to as the Apriori algorithm). The association degree provides a means to quantify the connection between vehicle owner behaviors and scenario features for the system, enabling the system to perform more accurate action mining based on this connection.

[0040] A Priori Algorithm: It refers to an efficient algorithm for mining frequent item sets. The Apriori algorithm is based on a key assumption that if an item set is frequent, then all its subsets are also frequent. This assumption allows the algorithm to generate and check frequent item sets through layer-by-layer iteration, thereby significantly reducing the computational amount and improving the mining efficiency.

[0041] A method for constructing a scenario pattern provided by an embodiment of the present application includes: First, obtaining target feature data of a target vehicle; then clustering the target feature data to obtain at least one clustering center, and determining target tags of the target feature data based on the at least one clustering center; then mining the target feature data based on the target tags to obtain at least one execution action data and configuration parameters corresponding to the at least one execution action data; finally, constructing a scenario pattern corresponding to the target tag based on the at least one execution action data and configuration parameters.

[0042] The method for constructing the scenario pattern provided by the embodiment of the present application achieves the following technical effects: The present application first obtains the target feature data of the target vehicle; then, through the clustering algorithm, identifies vehicle owner groups with similar behavior patterns (i.e., at least one clustering center) based on the target feature data, and each clustering center is represented by a target tag; then mines the specific behaviors and service requirements of vehicle owners in each tag scenario in the target feature data to obtain execution action data and configuration parameters; finally, constructs a scenario pattern corresponding to each target tag according to the execution action data and configuration parameters, realizing the efficient construction from vehicle owner behavior data to personalized scenario patterns. In this process, determining the target tags based on the clustering center ensures the matching between the feature data and the tags, improving the interpretability of the data, and constructing the scenario pattern based on the mined execution action data and configuration parameters not only improves the construction accuracy of the scenario pattern, ensures that the service can accurately respond to the personalized needs of vehicle owners, but also takes into account the safety and controllability of the application of the scenario pattern, thereby solving the technical problem of relatively low construction accuracy of the scenario pattern in the related art.

[0043] Embodiment 1

[0044] An embodiment of the present application provides a method for constructing a scenario pattern, Figure 1 which is a flowchart of a method for constructing a scenario pattern according to an embodiment of the present application, asFigure 1 As shown, the method includes the following steps:

[0045] S110: Obtain target feature data of the target vehicle, where the target feature data is used to screen multiple first feature data according to the importance degree of the multiple first feature data of the target vehicle for label learning.

[0046] The above-mentioned target vehicle can refer to a specific vehicle selected for learning and analyzing data such as its driving habits and usage preferences. The types of target vehicles can include, but are not limited to, electric vehicles, hybrid vehicles, and vehicle fuel vehicles. The specific target vehicle needs to be determined according to the exploration target and actual situation, and is not limited here. The target vehicle is usually the object of key research in a fleet or user group to collect specific behavior data of the vehicle driver. The data collection of the target vehicle is the basis of intelligent scenario mining. By analyzing these data, owner labels and scenario labels can be constructed, so as to realize the customization and recommendation of personalized services.

[0047] The above-mentioned target feature data can refer to the data determined to have a key impact on the label learning process after importance evaluation and screening from the numerous first feature data collected from the target vehicle. The target feature data can include, but is not limited to, owner basic information, driving behavior data, vehicle usage environment data, and in-vehicle service usage data, etc. The specific target feature data needs to be determined according to the screening conditions and feature importance degree, and is not limited here. The target feature data can be used to avoid the influence of redundant information, reduce the waste of computing resources, and improve the efficiency and accuracy of the method, etc.

[0048] The above-mentioned first feature data can refer to the originally collected data points used to describe various behaviors, preferences, and environmental states of the target vehicle and its owner. The types of first feature data can include, but are not limited to, owner basic information, driving behavior data, vehicle usage environment data, and in-vehicle service usage data, etc. The specific first feature data needs to be determined according to the data acquisition situation, and is not limited here. The first feature data can be used as the original material for label learning. By analyzing the first feature data, specific behavior patterns and vehicle usage scenarios of the owner can be identified, and thus provide a basis for the development of personalized services.

[0049] In the first characteristic data, the basic information of the vehicle owner can be used to represent information reflecting the identity or habits of the vehicle owner. The driving behavior data can include, but is not limited to, driving time, frequently used routes, acceleration and braking habits, vehicle speed, etc. The vehicle usage environment data can include, but is not limited to, weather, temperature, humidity, road quality, and traffic conditions, etc. The in-vehicle service usage data can include, but is not limited to, air-conditioning temperature, wind speed setting, music type, volume, seat adjustment, navigation preference, window switch state, etc. The above basic information of the vehicle owner, driving behavior data, and vehicle usage environment data are only examples. The specific basic information of the vehicle owner, driving behavior data, and vehicle usage environment data need to be determined according to the data acquisition situation and are not limited here.

[0050] It should be noted that all the first characteristic data involved in this application are information and data that have been authorized by the user or fully authorized by all parties. And the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or reject.

[0051] The above-mentioned label learning can refer to a data analysis process aimed at extracting key information from the first characteristic data and forming labels that describe the habits of vehicle owners, vehicle usage scenarios, and service preferences. The types of labels in label learning can include, but are not limited to, vehicle owner labels, scenario labels, service labels, etc. The specific label learning and label types need to be determined according to the dimensions of the first characteristic data and the types of mining methods and are not limited here. Label learning can be used to simplify complex, multi-dimensional behavior data into a set of labels that are easy to understand and process, so that the system can recommend personalized services to vehicle owners based on these labels.

[0052] In an optional embodiment, first, the first characteristic data is collected from the target vehicle. The first characteristic data includes the basic information of the vehicle owner, driving behavior, vehicle usage environment, vehicle usage stage, and the preferences of the vehicle owner for vehicle control and services, etc., forming a database containing diverse user behavior patterns. Then, according to the importance of the characteristics for label learning, the target characteristic data is screened out from the first characteristic data. This screening process adopts a feature importance evaluation mechanism. By controllably deleting or replacing specific features in the dataset, it is determined which features have a higher contribution to label learning, and these features with high contribution are used as the target characteristic data, thereby effectively reducing the data dimension, avoiding data redundancy, enhancing the stability of the learning model, improving the interpretability of the labels, and ensuring that the generated labels can accurately reflect the behavior and preferences of the vehicle owner.

[0053] S120: Cluster the target characteristic data to obtain at least one cluster center, and determine the target label of the target characteristic data based on at least one cluster center.

[0054] The above clustering can refer to an unsupervised learning method used to group samples in a dataset according to their feature similarities, forming different "clusters" or "classes". The types of clustering methods can include, but are not limited to, the K-Means Clustering Algorithm (simply referred to as K-Means), the Density-Based Spatial Clustering of Applications with Noise (simply referred to as DBSCAN), and hierarchical clustering, etc. The specific clustering method needs to be determined according to actual requirements and is not limited here. Clustering can be used to group car owners with similar driving habits, vehicle usage preferences, or environmental requirements into one category, facilitating subsequent personalized service design.

[0055] The above clustering center can refer to the representative points formed during the cluster division process by the clustering algorithm. The clustering center represents the average features or patterns of the samples within the cluster and reflects the typical behavioral characteristics of car owners in the same category. The types of clustering centers can include, but are not limited to, the combined features of multiple dimensions such as the age distribution of car owners, driving time preferences, common routes, music type preferences, etc. The specific clustering center needs to be determined according to the target feature data and the clustering algorithm and is not limited here. The clustering center can be used to provide the center point of each cluster, facilitating the subsequent classification of new data points by the algorithm and also facilitating manual analysis and interpretation of the clustering results, providing a basis for generating target labels.

[0056] The above target label can refer to the label that describes the characteristics of the car owner group, which is manually analyzed and named based on the clustering results. The types of target labels can include, but are not limited to, car owner labels, scenario labels, and service labels, etc. The specific types need to be determined according to the results of the clustering analysis and the specific service direction of the system and are not limited here.

[0057] In an alternative embodiment, first, the target feature data is clustered by a clustering algorithm such as K-Means, hierarchical clustering, or DBSCAN to obtain at least one cluster center. Each cluster center reflects the typical behavior characteristics of the same group of vehicle owners. Then, based on the cluster centers, the target labels of the target feature data are determined. For example, the basic information of the vehicle owners in the target feature data can be clustered to determine the owner labels, the vehicle usage environment data in the target feature data can be clustered to determine the scenario labels, and the service operation data in the target feature data can be clustered to determine the service labels, thereby determining the vehicle usage scenarios and required services of the vehicle owners. The above process can dynamically adapt to the personalized needs of vehicle owners, avoid the monotony and inapplicability of recommended content, and significantly enhance the interaction experience and satisfaction between vehicle owners and the intelligent cockpit system. For example, cluster analysis may identify a group of vehicle owners who "like to go jogging in the park on weekend mornings". For this group, the system can intelligently recommend routes with fresh air, light music suitable for the morning, and a moderate air-conditioning temperature setting, thus creating a more considerate and user-friendly service experience.

[0058] S130: Mine the target feature data based on the target labels to obtain at least one execution action data and at least one configuration parameter corresponding to the at least one execution action data.

[0059] The above mining may refer to in-depth analysis of the target feature data based on the generated target labels to discover the rules or patterns of the execution actions and configuration parameters of vehicle owners in specific scenarios. The mining types may include but are not limited to framework mining and parameter mining. Among them, framework mining focuses on discovering the frequent execution actions of vehicle owners in specific scenarios, such as "liking to listen to music when parking", and parameter mining further refines and focuses on the specific configuration parameters of the execution actions, such as "volume size when listening to music", "air-conditioning temperature setting", etc. The above mining types and mining contents are only examples, and the specific mining method needs to be adjusted according to actual needs and is not limited here. Mining can be used to enable the system to learn the behaviors of vehicle owners under specific labels, reveal the specific actions performed by vehicle owner groups in different scenarios and the personalized configurations of these actions, and provide a direct basis for the intelligent cockpit system to provide personalized services.

[0060] The above execution action data may refer to various specific operations performed by vehicle owners when in the vehicle. The execution action data may include but is not limited to vehicle control operations such as adjusting the air conditioner, windows, seats, etc., infotainment operations such as music playback, radio selection, video playback, etc., navigation services such as route planning, location recommendation, etc., and other operations related to intelligent cockpit services. The specific types and contents of the execution action data need to be determined according to the actual operations of vehicle owners and are not limited here. The execution action data can be used as a direct trigger point for the system to provide personalized services to vehicle owners.

[0061] The above configuration parameters may refer to specific values or settings involved in the execution action data. The types of configuration parameters may include, but are not limited to, temperature setting parameters, volume size parameters, sound quality mode parameters, navigation preference parameters, seat position parameters, tilt angle parameters, etc. The specific configuration parameters need to be determined according to the execution action data and are not limited here. The configuration parameters can be used to enable the system to more accurately meet the personalized needs of the vehicle owner.

[0062] In an alternative embodiment, the system first identifies the target tags of the vehicle owner group, such as "music lover", "long-time driver", etc., and then mines the vehicle owner behavior data with these tags. For example, it mines vehicle owner information based on the vehicle owner tags, mines the execution actions in different scenarios based on the scenario tags, and mines the frequent actions and common configurations based on the service tags to discover the frequent actions and preferred configurations executed by the vehicle owner in different scenarios. The mined execution action data includes the operations most frequently performed by the vehicle owner in a specific scenario, such as adjusting the air conditioner temperature, playing music, turning on the navigation, etc., while the configuration parameters further refine the personalized settings of these operations, including the specific value of the air conditioner temperature, the type and volume of music, the preferred route of the navigation, etc. The above mining steps can generate accurate user behavior patterns and personalized service suggestions, greatly improving the user experience and satisfaction of the intelligent cockpit service. For example, for a vehicle owner with the tag "late-night driver", the system can identify that they prefer to play light music and set a lower volume when driving at night, and adjust the air conditioner temperature to a more comfortable level, so as to automatically provide corresponding services when the vehicle owner is driving at night, reducing the risk of fatigue driving and improving driving safety. For "long-time drivers", the system can mine their habit of turning on the massage mode during long-distance driving and the preferred configuration of the massage intensity and mode, so as to automatically adjust the seat to the massage state when the vehicle owner starts a long-distance drive and provide the most suitable driving comfort.

[0063] Through the above data mining based on target tags, the solution of this application achieves high-precision and personalization of the cockpit service, not only improving the driving experience of the vehicle owner, but also promoting the deep interaction and trust between the vehicle owner and the cockpit system.

[0064] S140: Construct a scenario mode corresponding to the target tag based on at least one execution action data and configuration parameter.

[0065] The above-mentioned scenario mode can refer to a highly generalized form that describes the driving habits and demand preferences of the vehicle owner in a specific situation or vehicle usage stage based on the vehicle owner's executed action data and configuration parameters. The types of scenario modes can include but are not limited to driving safety mode, comfort adjustment mode, entertainment customization mode, and customization modes based on specific time, location, or activity, etc. The specific scenario mode needs to be determined according to the executed action data and configuration parameters, which are not limited here. The scenario mode can be used to provide a clear instruction set for the system, guiding the system to automatically execute a series of predefined actions and their configuration parameters when recognizing the corresponding tags to meet the needs of the vehicle owner in a specific situation. The scenario mode can make the provision of personalized services more automated and accurate, reduce the frequency of manual operations by the vehicle owner, and improve the safety and convenience during driving.

[0066] In an alternative embodiment, the mined executed action data and configuration parameters are integrated into a higher-level abstract framework to form a personalized service instruction set that the intelligent cockpit system can understand and execute, that is, a scenario mode corresponding to the target tag is constructed based on at least one executed action data and configuration parameter. This process significantly improves the automation level and accuracy of personalized services, ensuring that the system can respond in a timely manner to the needs of the vehicle owner in different scenarios and provide appropriate services. For example, for a vehicle owner with the tag of "music lover", the scenario mode constructed by the system may include "automatically playing the music type preferred by the vehicle owner, such as light music, during driving and setting the volume to a moderate level according to the vehicle owner's preference". For the tag of "long-time driver", the scenario mode constructed by the system may include "automatically turning on the seat massage function when detecting a long-distance drive, adjusting it to the intensity and mode preferred by the vehicle owner, and at the same time adjusting the air-conditioning temperature according to the weather conditions". For the tag of "night driving preference", the scenario mode constructed by the system may include: "automatically adjusting the lighting in the cockpit to a soft mode during the period from evening to early morning, and at the same time making driving assistance settings according to the road conditions, such as enhancing the night vision function, to improve driving safety and comfort". For the scenario tag of "family travel", the scenario mode constructed by the system may include "automatically adjusting the seat layout to the maximum seating comfort when detecting the family travel mode on weekends and holidays, and at the same time recommending entertainment content suitable for the whole family according to the ages and preferences of family members". The construction of these scenario modes enables the intelligent cockpit system to more intelligently predict and respond to the needs of the vehicle owner, not only improving the driving experience but also enhancing the safety and convenience of the system.

[0067] For example, the scenario modes of the present application can include but are not limited to the scenario modes in Table 1:

[0068] Table 1

[0069]

[0070] As shown in Table 1, each scene mode has its corresponding scene description, and the system can also set personalized content for different scene modes. The scene modes may include but are not limited to late night greetings, driving suggestions, tunnel suggestions, parking suggestions, air purification, music recommendations, movie recommendations, long-distance driving recommendations, high temperature recommendations, and home recommendations. The corresponding scene description in the late-night greeting scene mode is that it recognizes that it is late at night, and the voice care "Go to bed early, you have worked hard"; the corresponding scene description in the driving suggestion scene mode is that it is recommended to turn on the headlights according to the night time and vehicle status; the corresponding scene description in the tunnel suggestion scene mode is that it is recommended to switch the air conditioner to internal circulation when passing through a tunnel; the corresponding scene description in the parking suggestion scene mode is that the driving route is planned and the parking location is recommended according to the vehicle position; the corresponding scene description in the air purification scene mode is that when the air in the car is moderately / severely polluted, it is recommended to turn on the air purification according to user preferences, and the personalized content further set on this basis is the purification time and parameters; the corresponding scene description in the music recommendation scene mode is that music is recommended according to user preferences for commuting or other driving scenarios, and the personalized content further set on this basis is music preferences (type: soothing, classical; singer: A, B). The correspondence between scene modes, scene descriptions and personalized content is as shown in the above description.

[0071] By clarifying the specific content of different scene modes, the system can generate personalized content in a targeted manner, thereby improving the owner's experience in different scene modes.

[0072] The present application first obtains the target feature data of the target vehicle; then, through a clustering algorithm, based on the target feature data, identifies a group of car owners with similar behavior patterns, that is, at least one cluster center, each cluster center is represented by a target label; then, based on the target label, the specific behavior and service needs of the car owner in each label scenario in the target feature data are mined to obtain execution action data and configuration parameters; finally, based on the execution action data and configuration parameters, a scene mode corresponding to each target label is constructed, thereby realizing efficient construction from car owner behavior data to personalized scene modes. In this process, determining the target label based on the cluster center ensures the matching between the feature data and the label, and improves the interpretability of the data. Constructing the scene mode based on the mined execution action data and configuration parameters not only improves the accuracy of scene mode construction, and ensures that the service can accurately respond to the personalized needs of the car owner, but also takes into account the safety and controllability of scene mode application, thereby solving the technical problem of low accuracy in scene mode construction in related technologies.

[0073] Step S130 includes: performing action mining on the target feature data based on the target label to obtain at least one execution action data; and performing parameter mining on the target feature data based on the target label and the at least one execution action data to obtain configuration parameters.

[0074] The above-mentioned action mining may refer to identifying the high-frequency operations performed by the vehicle owner in a specific scenario from the target feature data. The types of action mining may include, but are not limited to, vehicle control operations, infotainment operations, and navigation operations, etc. Among them, vehicle control operations may include adjusting the air conditioner, switching the driving mode, turning on the seat heating, etc.; infotainment operations may include playing music, listening to the radio, watching videos, etc.; navigation operations may include planning routes, setting destinations, etc. The above types and contents of action mining need to be determined according to the target feature data and are not limited here. Action mining can be used to clarify the specific behaviors of the vehicle owner in the vehicle and determine the execution action data, etc.

[0075] The above-mentioned parameter mining may refer to analyzing the configuration preferences in the execution action data to identify the specific values or settings preferred by the vehicle owner when performing a certain action. For example, in the music playback operation, parameter mining may include, but is not limited to, mining parameters such as volume size, sound quality mode, playlist, etc.; in the air conditioner adjustment operation, parameter mining may include, but is not limited to, mining parameters such as temperature setting, wind speed size, air circulation mode, etc. The above parameter mining is only an example, and the specific parameter mining needs to be determined according to the specific execution action and is not limited here. Parameter mining can be used to refine the results of action mining and provide more specific parameter guidance for the system.

[0076] In an optional embodiment, the action mining and parameter mining based on the target label are the key links for this application to achieve the personalized experience of the vehicle owner. Through action mining, the system can identify the frequent operations of the vehicle owner in a specific scenario and form a preliminary understanding of the vehicle owner's behavior; while parameter mining further refines this understanding and reveals the specific preferences of the vehicle owner when performing these operations. For example, for the label of "music lover", the system can not only identify the action of the vehicle owner frequently playing music during driving, but also understand the configuration parameters such as the preferred music type, volume size, playlist, etc. of the vehicle owner through parameter mining, so as to automatically play the favorite music of the vehicle owner and adjust it to the most suitable volume and sound quality mode when the vehicle owner starts the vehicle. This dual mining combining actions and parameters enables the system to provide more accurate and personalized services by accurately identifying and responding to the behavior preferences of the vehicle owner, thereby improving the driving experience of the vehicle owner.

[0077] Optionally, the target tags include an owner tag, a scenario tag, and a service tag. The owner tag is used to represent the owner information of the owner corresponding to the target vehicle. The scenario tag is used to represent the scenario information of the scenario where the target vehicle is located. The service tag is used to represent the service information of the service operation performed by the target vehicle. Based on the target tags, action mining is performed on the target feature data to obtain at least one execution action data, including one of the following: filtering the target feature data based on the owner tag and the scenario tag to obtain the filtered feature data, and performing action mining on the filtered feature data based on the service tag to obtain at least one execution action data; performing action mining on the target feature data based on the service tag to obtain multiple initial execution action data, and filtering the multiple initial execution action data based on the owner tag and the scenario tag to obtain at least one execution action data.

[0078] The above-mentioned owner tag can refer to a comprehensive description of the owner's personal behavior, preferences, and characteristics. The types of owner tags include, but are not limited to, the owner's age, gender, driving style, entertainment preferences, health status, etc. The specific owner tag needs to be determined according to the actual situation and is not limited here. The owner tag can be used to help the system better understand the owner's personalized needs and guide the system to provide customized services for specific owners, improving the owner's satisfaction and usage experience.

[0079] The above-mentioned scenario tag can refer to a tag that describes the usage environment, time, location, and other scenario conditions where the vehicle is located. The scenario tag can include, but is not limited to, time, space, road conditions, weather conditions, etc. The specific scenario tag needs to be determined according to the actual situation of different scenarios and is not limited here. The scenario tag can be used to guide the system to identify the current or expected vehicle usage situation and provide relevant services in the corresponding scenario.

[0080] The above-mentioned service tag can refer to a classification identifier for the service operations that the system can provide. The service tag can include, but is not limited to, air conditioning control, music playback, navigation service, and seat adjustment, etc. The specific service tag needs to be determined according to the functions and services of the vehicle system and is not limited here. The service tag can be used to help the system match the service with the owner's needs and scenarios, ensuring that the provided service is accurate and meets the owner's immediate needs.

[0081] The above-mentioned filtered feature data can refer to a data set obtained by filtering the original feature data through the owner tag and the scenario tag, which is associated with a specific owner and scenario. The filtered feature data can be used to enhance the relevance of the data, making the subsequent action mining more accurate.

[0082] The above-mentioned initial execution action data may refer to a set of frequently executed service operations directly identified from the target feature data based on service tags in the initial stage of action mining. The initial execution action data can be used to preliminarily reveal the behavior patterns of vehicle owners under specific service tags.

[0083] In an alternative embodiment, the action mining of the target feature data in combination with the target tags includes the following two modes: In the first mode, the target feature data is preliminarily screened based on the vehicle owner tag and the scenario tag. This screening step ensures that the system only focuses on the data closely related to a specific vehicle owner and scenario, removing irrelevant or inapplicable information, and improving the efficiency and pertinence of data processing. Subsequently, the system conducts in-depth analysis on the screened data based on the service tag to mine the execution action data of the vehicle owner when using the intelligent cockpit service. This mode ensures the in-depth customization of personalized services and can provide services that are accurately matched to the specific behaviors and preferences of the vehicle owner. The second mode is to first perform initial action mining based on the service tag to obtain a series of possible execution action data, and then screen these initial actions in combination with the vehicle owner tag and the scenario tag. The advantage of this mode is that it can quickly generate a candidate set of service operations and then make fine-tuning through specific conditions of the vehicle owner and the scenario to ensure the accuracy and personalization of the service. The combined use of the two mining modes not only significantly improves the driving experience of vehicle owners, reduces the frequency of manual operations, enhances driving safety, but also promotes the in-depth interaction between vehicle owners and the system, and improves the trust and satisfaction of vehicle owners.

[0084] Optionally, screening the target feature data based on the vehicle owner tag and the scenario tag to obtain the screened feature data includes: determining the correlation degree between the vehicle owner tag and the scenario tag; performing action mining on the target feature data based on the correlation degree to obtain the screened feature data.

[0085] The above-mentioned correlation degree may refer to a quantitative relationship between the vehicle owner tag and the scenario tag. The correlation degree can be obtained through association rule learning by the Apriori algorithm (abbreviated as the Apriori algorithm). The correlation degree can be used to provide the system with a means to quantify the connection between the vehicle owner's behavior and the scenario characteristics, enabling the system to perform more accurate action mining based on this connection.

[0086] In an alternative embodiment, first, the association degree between the vehicle owner label and the scenario label is calculated through the Apriori algorithm. The calculation of the association degree helps the system identify the deep connection between the vehicle owner and the scenario, ensuring that the service can truly meet the personalized needs of the vehicle owner. Secondly, action mining is performed on the target feature data based on the association degree, which can quickly screen out the behavioral patterns highly relevant to specific vehicle owners and scenarios from the massive data, reducing the complexity and time consumption of data processing and improving the efficiency of providing personalized services. This data screening method combining the association degree and action mining enables the system to achieve highly personalized service push, not only significantly improving the driving experience of vehicle owners, enhancing their trust and dependence on the system, but also promoting the deep interaction between the system and vehicle owners and improving the market competitiveness of the system.

[0087] For example, Figure 2 is a schematic diagram of a framework mining process according to an embodiment of the present application. As Figure 2 shown, Figure 2 is the process of using the Apriori algorithm to perform framework mining on the actions of vehicle owner A. First, when the number of actions is 1, the association degree of each action is determined. For example, the association degree of listening to the news is 0.001, the association degree of turning on the air conditioner is 0.5, and the association degree of listening to the radio is 0.5. The action with the lowest association degree is removed, that is, pruning the action with the lowest association degree, namely removing the action of listening to the news with the lowest association degree. Then, the two actions with high association degrees are combined, and the association degree when the number of actions is 2 is determined. For example, the association degree of turning on the air conditioner and listening to the radio is 0.4, and the association degree of turning on the air conditioner and turning on the navigation is 0.0043. The action with the lowest association degree is removed, that is, pruning the actions of turning on the air conditioner and turning on the navigation with low association degrees, namely removing the actions of turning on the air conditioner and turning on the navigation with the lowest association degree. And so on, the item set that meets the minimum association degree, that is, the frequent item set, is determined.

[0088] Optionally, action mining is performed on the screened feature data based on the service label to obtain at least one execution action data, including: determining the feature data corresponding to different execution action data in the screened feature data based on the service label; determining the action execution frequency of different execution action data based on the feature data corresponding to different execution action data; and determining at least one execution action data from different execution action data based on the action execution frequency.

[0089] The above action execution frequency may refer to the number or frequency of a certain execution action data appearing in the filtered feature dataset under specific scenarios and service label conditions. The action execution frequency is a key indicator for measuring the universality and habituality of the action being performed by the vehicle owner in that scenario. The types of action execution frequency may include, but are not limited to, short-term frequency, long-term frequency, scenario frequency, and service frequency, etc. The specific action execution frequency needs to be determined according to actual requirements and is not limited here. The action execution frequency can be used to reflect the behavior habits of vehicle owners in specific services and scenarios, helping the system identify which service operations are the most commonly used or preferred by vehicle owners, providing a data basis for the system to provide personalized services.

[0090] In an alternative embodiment, first, by analyzing the filtered feature data and service labels, the feature data corresponding to different execution action data is determined; secondly, based on the feature data corresponding to different execution action data, the action execution frequency is calculated, and by quantifying the operation habits of vehicle owners under specific services, the system can identify the actions most frequently performed by vehicle owners; finally, according to the action execution frequency, at least one execution action data is determined from different execution action data. This frequency-based parameter mining method can more accurately reflect the preference settings and operation habits of vehicle owners in specific scenarios, ensuring that the system can automatically provide the services that best meet the needs of vehicle owners based on the high-frequency operation habits of vehicle owners, thereby not only reducing the operation burden of vehicle owners, but also significantly enhancing the personalization and comfort of the driving experience, and enhancing the dependence and satisfaction of vehicle owners on the system.

[0091] Optionally, parameter mining is performed on the target feature data based on the target label and at least one execution action data to obtain configuration parameters, including: based on the target label, determining the action feature data corresponding to at least one execution action data in the target feature data; performing parameter mining on the action feature data to obtain configuration parameters.

[0092] The above action feature data may refer to specific configuration or behavior parameters related to a specific execution action. The types of action feature data may include, but are not limited to, air-conditioning control parameters, music playback parameters, navigation setting parameters, and seat adjustment parameters, etc. The specific action feature data needs to be determined according to the execution action data and the target label and is not limited here. The action feature data can be used to describe the preference settings or operation details of vehicle owners when performing a certain action, refining the action requirements of vehicle owners to improve the service experience.

[0093] In an alternative embodiment, first, action feature data corresponding to execution action data is determined from target feature data based on a target label. The determination of the action feature data enables the system to grasp the fine-tuning preferences of the vehicle owner when performing a specific service. Second, parameter mining is performed on the action feature data to obtain refined configuration parameters. Through the above process of in-depth analysis of the target label and the execution action data, the system can accurately mine the action feature data related to the execution action from the target feature data and further refine the personalized configuration parameters of the vehicle owner, which not only deepens the personalization degree of the system service but also significantly improves the practicability of the service and user satisfaction.

[0094] For example, Figure 3 is a schematic diagram of a parameter mining process according to an embodiment of the present application. As Figure 3 shown, Figure 3 is a process of parameter mining for the actions of vehicle owner A using the Apriori algorithm. First, it is determined that the main actions of vehicle owner A include listening to the radio, turning on the air conditioner, and turning on the fragrance. Among them, the parameters included in listening to the radio are columns (4 columns), the parameters included in turning on the air conditioner are temperature (7 temperatures) and wind speed magnitude (4 magnitudes), and the fragrance modes (4 modes) included in turning on the fragrance. Then, the correlation degrees of different parameters corresponding to the above three actions are determined. For example, for the parameter column 1 included in the action of listening to the radio, the corresponding correlation degree is 0.1, column 2 is 0.2, column 3 is 0.05, and column 4 is 0.65. Further, the parameter with a higher correlation degree is determined from the above four parameters as the high-frequency parameter, that is, radio column 4 with a correlation degree of 0.65. Based on the above parameter mining process, similarly, the high-frequency parameters for the action of turning on the air conditioner are air conditioner temperature 25 degrees with a correlation degree of 0.75 and air conditioner wind speed with a correlation degree of 0.85; the high-frequency parameter corresponding to the action of turning on the fragrance is fragrance mode 1 with a correlation degree of 0.7. Through the above process, on the basis of framework mining, parameter mining is carried out. By continuously refining the execution action data of the user and determining the high-frequency parameters based on the correlation degree, the actual needs of the user are clarified, thereby providing a more customized and refined process for constructing a scenario mode. Furthermore, the needs of different user groups can be met based on the above scenario mode construction process.

[0095] Based on an embodiment of the present application, a method for constructing a scenario mode first requires framework mining and then parameter mining. The corresponding relationship between framework mining and parameter mining in different scenarios is shown in Table 2:

[0096] Table 2

[0097]

[0098] As shown in Table 2, for the music recommendation scenario, the car owners in this scenario include car owner A, car owner B, car owner C, and car owner D. During driving, the above car owners only performed the action of playing music. Therefore, there is no need to perform frame mining, and only parameter mining is required. Specifically, after performing parameter mining on car owner A, it is determined that songs by singer A need to be played. After performing parameter mining on car owner B, it is determined that songs by singer B need to be played. After performing parameter mining on car owner C, it is determined that songs by singer B and singer C need to be played. After performing parameter mining on car owner D, it is determined that no music is needed.

[0099] For the homecoming suggestion scenario, the car owners in this scenario include car owner E, car owner F, and car owner G. During the journey home, the above car owners performed multiple actions such as music selection, air conditioning control, and seat adjustment. Therefore, frame mining needs to be performed on them, and then parameter mining. Specifically, after frame mining, the actions of car owner E are determined to be listening to the radio and turning on the air conditioning. On this basis, parameter mining is performed to determine that listening to the radio specifically means listening to program A, turning on the air conditioning specifically means setting the temperature to 22 degrees, and the wind speed is medium; after frame mining, the actions of car owner F are determined to be listening to the news and turning on the air conditioning. On this basis, parameter mining is performed to determine that listening to the news specifically means listening to channel A, turning on the air conditioning specifically means setting the temperature to 22 degrees, and the wind speed is high; after frame mining, the action of car owner G is determined to be turning on the air conditioning. On this basis, parameter mining is performed to determine that turning on the air conditioning specifically means setting the temperature to 26 degrees and the wind speed is low.

[0100] In the above mining process, first, through frame mining, the service actions that car owners tend to perform in a specific scenario are quickly identified, thus quickly locking in the basic scope of the recommended services, avoiding the undifferentiated analysis of all possible services. On this basis, parameter mining is performed to further refine the configuration parameters of the recommended services, thereby improving the accuracy of service recommendations and the user experience.

[0101] Optionally, parameter mining is performed on the action feature data to obtain configuration parameters, including: based on the action feature data, determining the parameter setting frequencies of different parameters corresponding to at least one execution action data; determining the configuration parameters from different parameters based on the parameter setting frequencies.

[0102] The above parameter setting frequency may refer to the number or proportion of various different parameter configurations that appear in the dataset when performing a specific action. The types of parameter setting frequencies may include but are not limited to temperature setting frequencies, music type frequencies, and light brightness frequencies, etc. The specific parameter setting frequencies need to be determined according to the parameter type and the car owner's habits, and are not limited here. The parameter setting frequency can be used to reflect the specific preference habits of car owners for parameters when using specific services and is a key data indicator for personalized service customization.

[0103] In an alternative embodiment, first, the system determines the parameter setting frequencies of different parameters corresponding to the execution action data based on the action feature data. Then, based on the calculated parameter setting frequencies, the system determines the configuration parameters that best suit the owner's needs from different parameters. By calculating the parameter setting frequencies, the configuration parameters of the execution action can be further refined, taking into account the diversity of service operations and the personalized needs of the owner in specific service operations, and more specific configuration parameters that conform to the owner's preferences can be discovered. Through this process of automatically identifying and applying the parameter configuration that the owner is most accustomed to, the vehicle system can provide a more accurate and smooth service experience, reduce unnecessary operations, and enhance driving safety.

[0104] Determining the target label of the target feature data based on at least one cluster center in step S120 includes: outputting at least one cluster center; receiving feedback information on the at least one cluster center; and naming the target feature data based on the feedback information to obtain the target label.

[0105] The above feedback information may refer to the direct or indirect feedback of the user on the cluster center or the services, operations, or configurations recommended by the system. The types of feedback information may include, but are not limited to, information such as user evaluations, confirmations, adjustments, or rejections. The specific feedback information needs to be determined according to the actual situation and is not limited here. The feedback information can be used to help the system verify whether the cluster center accurately reflects the usage habits or preferences of the owner, and can also guide the system to make targeted adjustments to be closer to the actual needs of the owner.

[0106] In an alternative embodiment, first, the system outputs the cluster centers formed based on the user behavior data. This step is the result of clustering analysis of the user data based on unsupervised learning algorithms, aiming to discover user groups with similar behavior patterns. Subsequently, the system receives the feedback information of the user on these cluster centers. This step not only collects the user's recognition or corrective opinions on the system analysis results, but also provides a direct basis for the system to correct the clustering model and improve service recommendations. Finally, the system names the target feature data based on the collected feedback information to obtain the target label. This process transforms the abstract cluster centers into concrete labels, not only enhancing the personalized experience of the user for the cockpit service, but also promoting the system's accurate understanding of the user's behavior patterns. The above process not only enhances the personalization and intelligence of the service through the user feedback mechanism, but also continuously improves the accuracy of the service and the user experience through continuous interactive learning. At the same time, the transparency and interpretability of the system are enhanced through the feedback information, enabling the user to understand the data logic behind the system recommendations and enhancing the trust in the personalized service.

[0107] Step S110 includes: clustering a plurality of first feature data to obtain a first clustering result; performing a preset operation on at least one feature data among the plurality of first feature data to obtain a plurality of second feature data; clustering the plurality of second feature data to obtain a second clustering result; and determining target feature data from the plurality of first feature data based on the first clustering result and the second clustering result.

[0108] The above-mentioned first clustering result may refer to the result of grouping or classifying data using a clustering algorithm based on the original set of first feature data. Each group or class represents a group of users or scenarios with similar characteristics. The types of the first clustering result may include, but are not limited to, clustering results based on scenarios, clustering results based on user behavior, etc. The specific first clustering result needs to be determined according to the actual situation and is not limited here. The first clustering result can be used to provide a preliminary classification of user groups or scenarios, helping the system understand the macro distribution of user behavior patterns.

[0109] The above-mentioned at least one feature data may refer to one feature data determined manually or randomly from the plurality of first feature data.

[0110] The above-mentioned preset operation may refer to a specific processing or transformation operation performed on at least one feature data in the first feature data set. The types of the preset operation may include, but are not limited to, feature selection, replacement operation, deletion operation, and feature normalization, etc. The specific preset operation needs to be determined according to the actual requirements and is not limited here. The preset operation can be used to improve the clustering effect. By adjusting certain features of the first feature data, the interpretability of clustering can be enhanced, the accuracy of classification can be improved, or more refined behavior patterns can be discovered.

[0111] The above-mentioned first feature data may refer to the originally collected data set. The first feature data may include, but are not limited to, various types of data such as user's basic information, driving habits, music preferences, temperature control preferences, geographical location information, time information, etc. The specific first feature data needs to be determined according to the actual data acquisition situation and is not limited here. The first feature data can be used as the basis for clustering analysis to provide the original behavior pattern of the vehicle owner.

[0112] The above-mentioned second feature data may refer to the set of first feature data after being processed by the preset operation, and its features may be enhanced, weakened, or modified. The second feature data can be used to provide the required post-data pattern by adjusting the preset operation on the first feature data, thus better supporting the development of personalized services.

[0113] The above-mentioned second clustering result may refer to the clustering analysis result based on the second feature data set processed by a preset operation. The types of the second clustering result may include, but are not limited to, the clustering result based on scenarios, the clustering result based on user behavior, etc. The specific second clustering result needs to be determined according to the actual situation and is not limited here. The second clustering result can be used to enable the system to identify the improvements brought by the preset operation by comparing with the first clustering result, further adjust the clustering strategy, and finally determine a more accurate and personalized user group division or scenario pattern recognition.

[0114] In an alternative embodiment, first, perform preliminary clustering on the first feature data to obtain a first clustering result, enabling the system to identify the macroscopic distribution of users or scenario patterns and providing a direction for the preliminary customization of services. Then, through the preset operation, perform a deletion operation or a replacement operation on the first feature data to remove redundant data and obtain second feature data. Next, perform clustering on the second feature data to obtain a second clustering result and deepen the clustering analysis. Finally, based on the first clustering result and the second clustering result, determine target feature data from multiple first feature data, that is, those feature data that can significantly distinguish different user groups or scenario patterns. This process screens out the target feature data through two rounds of clustering and preset operations, ensuring that the data used to construct the recommendation pattern is of high quality and closely related to the target scenario, while improving the data screening efficiency and accuracy and reducing the interference of redundant data.

[0115] Optionally, determining target feature data from multiple first feature data based on the first clustering result and the second clustering result includes: determining the influence level of at least one feature data based on the first clustering result and the second clustering result, where the influence level is used to represent the influence degree of at least one feature data on the clustering; and determining the target feature data in at least one feature data based on the influence level.

[0116] The above-mentioned influence level may refer to a measure used to quantify the influence degree of feature data on the clustering analysis process and results. The influence level describes the contribution degree of specific feature data to the user group division or scenario pattern recognition during clustering. The types of the influence level may include, but are not limited to, the stability influence level, the quality influence level, and the interpretability influence level, etc. The specific influence level needs to be determined according to the actual needs and is not limited here. The influence level can be used to screen and improve the feature set, perform personalized service customization, and dynamically adjust service strategies, etc.

[0117] In an alternative embodiment, first, the system obtains a preliminary classification result of user behavior or scenario patterns through clustering of the first feature data. Subsequently, by performing a preset operation on these feature data and clustering again, a second clustering result is obtained. Then, based on the first clustering result and the second clustering result, the influence level of the feature data is determined. The determination of the influence level enables the system to screen out key features from multiple feature data, improving the clustering analysis result, thereby ensuring that the constructed scenario pattern focuses more on the key parameters affecting user behavior and scenario features, and enhancing the practicality and effectiveness of the pattern.

[0118] Optionally, a plurality of first feature data of the target vehicle are obtained, including: collecting multi-dimensional data of the target vehicle; performing feature extraction on the multi-dimensional data to obtain a plurality of first feature data.

[0119] Obtained through a variety of sensors equipped in the vehicle system, such as the Global Positioning System (GPS for short), ambient light sensor, temperature sensor, humidity sensor, pressure sensor, sound sensor, etc., which are used to monitor multi-dimensional data such as the changes in the internal and external environment of the vehicle, user behavior habits, and vehicle operating status in real time. These sensors can be in-vehicle cameras, microphones, infrared sensors, etc., which are used to capture information such as the user's facial expressions, voice commands, and body postures.

[0120] By accessing the cloud database, information from different data sources is integrated, such as social media data, online shopping records, weather forecasts, traffic conditions, etc., to construct the user's behavior patterns and scenario predictions and obtain multi-dimensional data.

[0121] Through Bluetooth or other wireless communication technologies, the system can synchronize data with devices such as the user's mobile phone and smartwatch, collect information such as the user's schedule, health data, music playlist, etc., to more comprehensively understand the user's personal preferences and real-time needs.

[0122] The above multi-dimensional information acquisition methods are only examples, and the specific acquisition method should be determined according to the owner's habits and actual situations, and is not limited here.

[0123] The above multi-dimensional data may refer to a series of data collected by the system covering multiple aspects such as user behavior, preferences, environmental conditions, and time information. The multi-dimensional data may include but are not limited to user basic information, driving habit data, music preference data, temperature control preference data, and environmental condition data, etc. The specific multi-dimensional data needs to be determined according to the vehicle information acquisition device, and is not limited here. The multi-dimensional data can be used to reflect the user's specific behavior at different times and scenarios, provide a long-term record of the user's basic information and usage habits, and provide comprehensive data support for the personalized customization of intelligent cockpit services.

[0124] In an alternative embodiment, first, the system comprehensively collects multi-dimensional data of the target vehicle through multi-type sensors and data acquisition technologies, which covers a series of data points from user basic information, driving habits to environmental conditions, vehicle status, etc., providing a rich and comprehensive information source for in-depth analysis of the system; subsequently, the system extracts features from the multi-dimensional data, screening and extracting core feature data that can characterize user behavior and preferences from the original data, namely the first feature data. This process usually involves complex operations such as data cleaning, preprocessing, feature selection, and feature construction, aiming to transform the original and redundant multi-dimensional data into highly abstract and structured feature data, providing accurate data input for subsequent clustering analysis and personalized service customization. The above process, by collecting multi-dimensional data and performing feature extraction, provides a comprehensive and accurate data basis for subsequent clustering and mining, ensuring that the constructed scenario patterns are based on sufficient and diverse data and can comprehensively reflect user needs and behavior patterns.

[0125] Embodiment 2

[0126] The embodiment of the present application also provides a construction device 40 for a scenario pattern. Figure 4 It is a structural diagram of a construction device for a scenario pattern according to the embodiment of the present application. Please refer to Figure 4 , including: an acquisition module 410, configured to execute the step of acquiring target feature data of the target vehicle, where the target feature data is obtained by screening multiple first feature data according to the importance degree of the multiple first feature data of the target vehicle for label learning; a determination module 420, configured to execute the step of clustering the target feature data to obtain at least one clustering center, and determining a target label of the target feature data based on the at least one clustering center; a mining module 430, configured to execute the step of mining the target feature data based on the target label to obtain at least one execution action data and configuration parameters corresponding to the at least one execution action data; a construction module 440, configured to execute the step of constructing a scenario pattern corresponding to the target label based on the at least one execution action data and the configuration parameters.

[0127] Optionally, the mining module is configured to perform action mining on the target feature data based on the target label to obtain at least one execution action data; and perform parameter mining on the target feature data based on the target label and the at least one execution action data to obtain configuration parameters.

[0128] Optionally, the target label includes an owner label, a scenario label, and a service label. The owner label is used to represent the owner information of the owner corresponding to the target vehicle. The scenario label is used to represent the scenario information of the scenario where the target vehicle is located. The service label is used to represent the service information of the service operation performed by the target vehicle. The mining module is further configured to perform one of the following: screen the target feature data based on the owner label and the scenario label to obtain the screened feature data, and perform action mining on the screened feature data based on the service label to obtain at least one execution action data; perform action mining on the target feature data based on the service label to obtain multiple initial execution action data, and screen the multiple initial execution action data based on the owner label and the scenario label to obtain at least one execution action data.

[0129] Optionally, the mining module is further configured to determine the correlation degree between the owner label and the scenario label; perform action mining on the target feature data based on the correlation degree to obtain the screened feature data.

[0130] Optionally, the mining module is further configured to, based on the service label, determine the feature data corresponding to different execution action data in the screened feature data; determine the action execution frequency of different execution action data based on the feature data corresponding to different execution action data; and determine at least one execution action data from different execution action data based on the action execution frequency.

[0131] Optionally, the mining module is further configured to, based on the target label, determine the action feature data corresponding to at least one execution action data in the target feature data; perform parameter mining on the action feature data to obtain configuration parameters.

[0132] Optionally, the mining module is further configured to, based on the action feature data, determine the parameter setting frequency of different parameters corresponding to at least one execution action data; and determine the configuration parameters from different parameters based on the parameter setting frequency.

[0133] Optionally, the determination module is configured to output at least one clustering center; receive feedback information on the at least one clustering center; and name the target feature data based on the feedback information to obtain the target label.

[0134] Optionally, the acquisition module is configured to cluster multiple first feature data to obtain a first clustering result; perform a preset operation on at least one feature data among the multiple first feature data to obtain multiple second feature data; cluster the multiple second feature data to obtain a second clustering result; and determine the target feature data from the multiple first feature data based on the first clustering result and the second clustering result.

[0135] Optionally, the obtaining module is further configured to determine the influence level of at least one piece of feature data based on the first clustering result and the second clustering result, where the influence level is used to represent the influence degree of at least one piece of feature data on clustering; and determine target feature data from at least one piece of feature data based on the influence level.

[0136] Optionally, the obtaining module is further configured to collect multi-dimensional data of the target vehicle; and perform feature extraction on the multi-dimensional data to obtain a plurality of first feature data.

[0137] Embodiment III

[0138] The embodiment of the present application further provides an electronic device 50, Figure 5 which is a structural diagram of the electronic device provided according to the embodiment of the present application. Please refer to Figure 5 and includes a processor 510 and a memory 520. Among them, the memory 510 is used to store a computer program; the processor 520 is used to execute the program stored on the memory 510 to implement the method for constructing a scene pattern introduced in any embodiment of the present application.

[0139] The embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the method for constructing a scene pattern introduced in any embodiment of the present application is implemented.

[0140] In the present application, "a plurality of" refers to two or more.

[0141] In the present application, unless otherwise clearly defined, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a structural connection or an electrical connection; it may be a direct connection or an indirect connection through an intermediate medium, and it may be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations.

[0142] The terms "first", "second", "third", "fourth", etc. (if any) in the present application are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0143] The term "and / or" in the present application is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects.

[0144] Unless otherwise specified, all steps of this application can be carried out sequentially or randomly. For example, the method includes steps A and B, which means that the method can include steps A and B carried out sequentially, or steps B and A carried out sequentially. For example, it is mentioned that the method may further include step C, which means that step C can be added to the method in any order. For example, the method can include steps A, B, and C, or steps A, C, and B, or steps C, A, and B, etc.

[0145] The foregoing are only preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A method for constructing a scene mode, characterized in that, Including: Obtaining target feature data of a target vehicle, where the target feature data is used to screen the multiple first feature data according to the importance of the multiple first feature data of the target vehicle for label learning; Clustering the target feature data to obtain at least one clustering center, and determining a target label of the target feature data based on the at least one clustering center; Mining the target feature data based on the target label to obtain at least one execution action data and configuration parameters corresponding to the at least one execution action data; Constructing a scenario pattern corresponding to the target label based on the at least one execution action data and the configuration parameters; 2. The method according to claim 1, wherein Mining the target feature data based on the target label to obtain at least one execution action data and configuration parameters corresponding to the at least one execution action data, including: Performing action mining on the target feature data based on the target label to obtain the at least one execution action data; Performing parameter mining on the target feature data based on the target label and the at least one execution action data to obtain the configuration parameters.

3. The method according to claim 2, wherein The target label includes an owner label, a scenario label, and a service label. The owner label is used to represent the owner information of the owner corresponding to the target vehicle, the scenario label is used to represent the scenario information of the scenario where the target vehicle is located, and the service label is used to represent the service information of the service operation executed by the target vehicle; Performing action mining on the target feature data based on the target label to obtain the at least one execution action data, including one of the following: Screening the target feature data based on the owner label and the scenario label to obtain screened feature data, and performing action mining on the screened feature data based on the service label to obtain the at least one execution action data; Performing action mining on the target feature data based on the service label to obtain multiple initial execution action data, and screening the multiple initial execution action data based on the owner label and the scenario label to obtain the at least one execution action data.

4. The method according to claim 3, characterized in that, Screening the target feature data based on the owner label and the scenario label to obtain screened feature data, including: Determining the correlation degree between the owner label and the scenario label; Performing action mining on the target feature data based on the correlation degree to obtain the screened feature data.

5. The method according to claim 3, wherein Performing action mining on the screened feature data based on the service label to obtain the at least one execution action data, including: Based on the service label, determining the feature data corresponding to different execution action data in the screened feature data; Based on the feature data corresponding to the different execution action data, determining the action execution frequency of the different execution action data; Determining the at least one execution action data from the different execution action data based on the action execution frequency.

6. The method according to claim 2, wherein Performing parameter mining on the target feature data based on the target label and the at least one execution action data to obtain the configuration parameters, including: Based on the target label, determine the action feature data in the target feature data corresponding to the at least one execution action data; Perform parameter mining on the action feature data to obtain the configuration parameters.

7. The method according to claim 6, characterized in that, Performing parameter mining on the action feature data to obtain the configuration parameters includes: Based on the action feature data, determine the parameter setting frequencies of different parameters corresponding to the at least one execution action data; Determine the configuration parameters from the different parameters based on the parameter setting frequencies.

8. The method according to claim 1, characterized in that Determine the target label of the target feature data based on the at least one cluster center, including: Output the at least one cluster center; Receive feedback information on the at least one cluster center; Name the target feature data based on the feedback information to obtain the target label.

9. The method according to claim 1, wherein Obtain the target feature data of the target vehicle, including: Cluster the multiple first feature data to obtain a first clustering result; Perform a preset operation on at least one of the multiple first feature data to obtain multiple second feature data; Cluster the multiple second feature data to obtain a second clustering result; Determine the target feature data from the multiple first feature data based on the first clustering result and the second clustering result.

10. The method according to claim 9, wherein Determining the target feature data from the multiple first feature data based on the first clustering result and the second clustering result includes: Based on the first clustering result and the second clustering result, determine the influence level of the at least one feature data, where the influence level is used to represent the influence degree of the at least one feature data on the clustering; Determine the target feature data among the at least one feature data based on the influence level.

11. The method according to claim 9, characterized in that, Obtain multiple first feature data of the target vehicle, including: Collect multi-dimensional data of the target vehicle; Perform feature extraction on the multi-dimensional data to obtain the multiple first feature data.

12. An electronic device, characterized in that, Includes a processor and a memory, where The memory is used to store computer programs; The processor is used to execute the programs stored on the memory to implement the method according to any one of claims 1-11.

Images