A cross-platform car machine function classification and adaptation optimization method and system

Abstract

The application relates to the technical field of vehicles and discloses a cross-platform vehicle machine function classification and adaptive optimization method and system, which comprises the following steps: acquiring screen hardware parameters of different vehicle models, mapping the screen hardware parameters to a same vector space for clustering, and dividing the vehicle machine into n2 discrete screen hardware levels; for each screen hardware level, a corresponding vehicle machine function classification scheme is designed, user preference portraits are generated by collecting use behavior data of the vehicle machine functions of different users, and then the individualized function recommendation list is generated in combination with the vehicle machine function classification scheme; for a target vehicle machine to be adapted, the screen hardware level of the target vehicle machine is determined, the corresponding vehicle machine function classification scheme and the individualized function recommendation list are called, and a vehicle machine function list conforming to the characteristics of the corresponding vehicle model of the target vehicle machine is adaptively generated; the application can automatically adapt and generate an individualized vehicle machine function list according to the hardware capacity of the vehicle machine and the user preference, and the practicability and user experience of the vehicle machine function are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicles, more particularly, the present application relates to a cross-platform car machine function classification and adaptive optimization method and system. BACKGROUND

[0002] With the rapid development of intelligent and networked vehicles, the functions of in-vehicle infotainment systems are becoming increasingly rich. However, due to the large differences in hardware configurations and screen parameters of different vehicle models, the function adaptation and individualization of car machine systems are insufficient, making it difficult to meet the diverse needs of users. There have been some related researches on function optimization and self-adaptation of in-vehicle systems in the prior art, but there are still some deficiencies.

[0003] Patent application No. CN110888578A discloses a vehicle function optimization method, device and vehicle with the same. The method collects the target operation action of the driver and passenger, identifies the corresponding vehicle function, and recommends the vehicle function to the driver and passenger according to the operation frequency of the target operation action. Although this method can recommend functions according to the user's operation habits, it does not consider the differences in hardware configurations of different vehicle models, and cannot achieve cross-platform function adaptation. In addition, this method only recommends based on operation frequency, without deeply mining user preference characteristics, and the individualization degree is insufficient.

[0004] Patent application No. CN118760376A discloses a self-adaptive display method and device for car machine interface, vehicle and storage medium. The method obtains the use time and operation mode of vehicle-mounted software under different driving states, constructs a software frequently used item list for different scenes, and automatically switches the interface according to the vehicle driving state. This method can dynamically adjust the car machine interface according to the driving scene, improving the user experience. However, this method does not consider the differences in car machine hardware platforms, and the interface arrangement and interaction mode cannot be uniformly optimized for screens of different sizes and resolutions. At the same time, this method mainly focuses on the scene adaptation of information display, and does not involve the classification and personalized recommendation of car machine functions.

[0005] In summary, the existing car machine function optimization and self-adaptive methods generally have the problems of poor cross-platform uniformity and insufficient individualization. Due to the large differences in screen size and resolution of different vehicle models, the car machine interface layout and interaction mode are difficult to unify, and the individualized needs of different users are difficult to meet. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the present application provides a cross-platform car machine function classification and adaptive optimization method and system, which realizes personalized function recommendation based on the unified classification of car machine functions, significantly improving the user experience of car machine human-computer interaction.

[0007] To achieve the above object, the present application provides the following technical solutions:

[0008] A cross-platform car machine function classification and adaptive optimization method, comprising:

[0009] Obtaining screen hardware parameters of different vehicle models, mapping the screen hardware parameters to the same vector space for clustering, and dividing the car machine into n2 discrete screen hardware levels;The screen hardware parameters include car machine screen parameters and hardware configuration parameters;

[0010] For each screen hardware level, a corresponding car machine function classification scheme is designed;Collecting different user behavior data of car machine functions, generating user preference portraits;Combining the user preference portraits and the car machine function classification scheme, generating a personalized function recommendation list;

[0011] For the target car machine to be adapted, the car machine screen parameters and hardware configuration parameters of the target car machine are obtained, the screen hardware level of the target car machine is determined, the corresponding car machine function classification scheme and personalized function recommendation list are retrieved according to the screen hardware level of the target car machine, and the car machine function list conforming to the characteristics of the corresponding vehicle model of the target car machine is adaptively generated.

[0012] Further, the mapping of the screen hardware parameters to the same vector space for clustering comprises:

[0013] Pretreatment and feature selection are performed on the screen hardware parameters to obtain clustering features;

[0014] The clustering features are used as the coordinate axes of a multi-dimensional vector space, and the screen hardware parameters of each car machine are represented as an n1-dimensional screen hardware parameter vector in the multi-dimensional vector space;

[0015] The screen hardware parameter vector of the car machine is used as the input, and the K-means clustering algorithm is used to divide the car machine system of different vehicle models into n2 screen hardware levels.

[0016] Further, the pretreatment and feature selection of the screen hardware parameters to obtain clustering features comprise:

[0017] The screen hardware parameters are cleaned to remove outliers and missing values;

[0018] The normalized screen hardware parameters are normalized;

[0019] The normalized screen hardware parameters are removed;According to the normalized screen hardware parameters, the parameter pairs are constructed, the Pearson correlation coefficient between the parameter pairs is calculated, and for the parameter pairs with an absolute value of the correlation coefficient greater than a predetermined correlation threshold, one parameter in the parameter pair is removed and the other parameter is used as a reserved parameter;

[0020] Perform principal component analysis on the screen hardware parameters after removing redundant parameters, calculate the variance contribution rate of each principal component, and take the principal components with a variance contribution rate exceeding a preset contribution rate threshold as clustering features.

[0021] Further, the corresponding car machine function classification scheme for each screen hardware level includes:

[0022] According to the hardware configuration parameters of the car machine, obtain the computing performance index of the car machine;

[0023] According to the car machine screen parameters of the car machine, obtain the display effect index of the car machine;

[0024] Calculate the function complexity index and resource occupation index of the car machine;

[0025] According to the computing performance index, display effect index, function complexity index and resource occupation index of the car machine, a support vector machine (SVM) algorithm is used to construct a function classification model, and the car machine functions are divided into three levels: basic functions, extended functions and advanced functions.

[0026] Further, the use behavior data includes demographic characteristics and function use records of the user;

[0027] The generation of the user preference profile includes:

[0028] The demographic characteristics and function use records of each user are constructed into a user behavior feature vector, the user behavior feature vector is clustered, and the users are divided into n3 user groups;

[0029] The average use frequency and duration of each user group on different functions are counted, a user-function preference matrix is constructed, and a function preference profile of each user group is generated; the rows of the user-function preference matrix represent different user groups, the columns represent each car machine function, and the matrix elements represent the preference degree of a certain user group to a certain car machine function.

[0030] Further, the generation of the personalized function recommendation list includes:

[0031] Based on the user preference profile, obtain a preferred function list of different user groups;

[0032] For each screen hardware level, high-frequency preferred functions are screened from the corresponding car machine function classification scheme to obtain a candidate function set;

[0033] Generate a personalized function recommendation list from the candidate function set.

[0034] Further, the obtaining of the preferred function list of different user groups includes:

[0035] sequencing each row of the user-function preference matrix in descending order to obtain a function ranking list of each user group in descending order of preference degree;

[0036] extracting N preferred functions with the highest preference degree from each function ranking list to generate a preferred function list of each user group;

[0037] counting the occurrence frequency of each preferred function in the preferred function list, marking the preferred function with an occurrence frequency exceeding a preset frequency threshold as a high-frequency preferred function, and generating a high-frequency preferred function set.

[0038] Further, the filtering of the high-frequency preferred function from the corresponding car machine function classification scheme to obtain the candidate function set comprises:

[0039] including all basic functions and extended functions in the high-frequency preferred function set into the candidate function set;

[0040] calculating the resource occupation level SE of each high-level function in the high-frequency preferred function set;

[0041] including the high-level function in the high-frequency preferred function set with a resource occupation level SE lower than a preset resource occupation level threshold into the candidate function set.

[0042] Further, the generation of the personalized function recommendation list from the candidate function set comprises:

[0043] obtaining the demographic characteristics and historical function use records of the user to be recommended, constructing a user behavior feature vector of the user to be recommended, calculating the similarity between the user behavior feature vector of the user to be recommended and the user behavior feature vector of each user group, and finding the user group with the highest similarity as the user group to which the user to be recommended belongs;

[0044] obtaining the preferred function list of the user group to which the user to be recommended belongs, and finding the intersection of the preferred function list and the candidate function set to obtain a candidate recommended function set;

[0045] obtaining the historical function preference of the user to be recommended based on the historical function use records of the user to be recommended, and selecting M functions from the candidate recommended function set to generate a personalized function recommendation list according to the historical function preference of the user to be recommended.

[0046] A cross-platform car machine function classification and adaptive optimization system for implementing the above-mentioned cross-platform car machine function classification and adaptive optimization method, the system comprising:

[0047] a car machine grading module: obtaining screen hardware parameters of different car models, mapping the screen hardware parameters to the same vector space for clustering, and dividing the car machine into n2 discrete screen hardware levels; the screen hardware parameters include car machine screen parameters and hardware configuration parameters;

[0048] The personalized recommendation module: for each screen hardware level, a corresponding car machine function classification scheme is designed; the use behavior data of different users on the car machine function is collected, and a user preference portrait is generated; the user preference portrait and the car machine function classification scheme are combined to generate a personalized function recommendation list;

[0049] The adaptation module: for the target car machine to be adapted, the car machine screen parameters and hardware configuration parameters of the target car machine are obtained, the screen hardware level of the target car machine is determined, the corresponding car machine function classification scheme and personalized function recommendation list are retrieved according to the screen hardware level of the target car machine, and a car machine function list conforming to the characteristics of the corresponding car model of the target car machine is adaptively generated.

[0050] An electronic device comprising a memory, a central processing unit, and a computer program stored on the memory and executable on the central processing unit, wherein the central processing unit executes the computer program to implement the cross-platform car machine function classification and adaptation optimization method described above.

[0051] A computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed to implement the cross-platform car machine function classification and adaptation optimization method described above.

[0052] Compared with the prior art, the beneficial effects of the present application are:

[0053] The present application can map different car models to a unified screen hardware level system through feature extraction and cluster analysis of car machine screen hardware parameters, providing a basis for subsequent function classification and optimization. This hardware parameter mapping method can effectively overcome the adaptation difficulties caused by car machine hardware fragmentation.

[0054] The present application designs a function classification scheme matched with the hardware capability for each screen hardware level, and divides the car machine function into three levels of basic, extended, and advanced. This hierarchical adaptation strategy can maximize the use of car machine hardware performance, reasonably configure functions at each level, and avoid resource waste or performance deficiency.

[0055] The present application generates a user function preference portrait by collecting user car machine use behavior data and using clustering and collaborative filtering algorithms. Based on the user portrait, a personalized recommendation function list is generated, which can accurately match user needs, provide more considerate car machine function services, and bring a smoother human-computer interaction experience.

[0056] The cross-platform car machine function classification and adaptive optimization method has strong universality and expansibility. For the access of a new vehicle model, only the screen hardware parameters thereof need to be obtained, and the corresponding screen hardware level, function classification scheme and recommendation list can be automatically matched to realize one-stop function adaptive optimization, and the manual development cost is greatly reduced.

[0057] The cross-platform car machine function classification and adaptive optimization method fully taps the value of user usage data, realizes the trinity of hardware perception, user perception and function perception, and builds an end-to-end, full-process car machine personalized function customization optimization platform. This provides a personalized, fine function management and operation means for intelligent networked vehicles. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0059] Figure 1 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application;

[0060] Figure 2 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which maps the screen hardware parameters to the same vector space for clustering;

[0061] Figure 3 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which obtains the clustering features;

[0062] Figure 4 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which designs the car machine function classification scheme;

[0063] Figure 5 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which generates the user preference portrait;

[0064] Figure 6 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which obtains the preferred function list of different user groups;

[0065] Figure 7 The method flow chart of the cross-platform car machine function classification and adaptive optimization method in the present application, which obtains the candidate function set;

[0066] Figure 8 FIG. 1 is a functional module diagram of a cross-platform car machine function classification and adaptive optimization system according to an embodiment of the present application. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of the present application.

[0068] Embodiment 1

[0069] Referring to FIG. 1, the embodiment provides a cross-platform car machine function classification and adaptive optimization method, which comprises the following steps: Figure 1 In step S1000, screen hardware parameters of different vehicle models are acquired, the screen hardware parameters are mapped to the same vector space for clustering, and the car machine is divided into n2 discrete screen hardware levels; the screen hardware parameters comprise car machine screen parameters and hardware configuration parameters.

[0070] Further, step S1000 comprises the following steps:

[0071] In step S1100, car machine screen parameters of different vehicle models are acquired; the car machine screen parameters comprise car machine screen size parameters, car machine screen resolution parameters, car machine screen pixel density parameters, and car machine screen gamut parameters.

[0072] Specifically, the car machine screen size parameters comprise screen diagonal length, aspect ratio, etc.; the car machine screen resolution parameters comprise horizontal pixel number and vertical pixel number; the car machine screen pixel density parameters are calculated through the screen diagonal length and the resolution; and the car machine screen gamut parameters comprise coverage rates of different gamut standards such as Adobe RGB (Adobe Red Green Blue color space), sRGB (standard Red Green Blue color space), NTSC (American National Television System Committee color television broadcast standard), etc. In step S1100, car machine screen parameters of mainstream vehicle models on the market are collected in a large range, and the heterogeneity of different car machine screens is comprehensively understood.

[0073]

[0074] ​The car machine screen size parameter reflects the physical size of the car machine screen, and is a key factor for determining the UI layout. By measuring the diagonal length of the screen, the approximate size of the screen can be obtained; and the screen aspect ratio describes the length-width ratio of the screen, which affects the display method of the UI elements, and common aspect ratios include 4:3, 16:9, 16:10, etc. The screen resolution represents the image fineness of the screen, and the higher the resolution, the richer the image details that the screen can display. By collecting the number of horizontal and vertical pixels of the screen, the numerical value of the resolution can be accurately described. For example, 1280x720 represents 1280 horizontal pixels and 720 vertical pixels. The pixel density reflects the sharpness of the screen, expressed as PPI (Pixels Per Inch). PPI can be calculated from the screen size and resolution, and the higher the PPI, the more delicate the display effect. The color gamut parameter represents the color performance of the screen, and different color gamut standards have different coverage ranges of colors. The Adobe RGB color gamut covers a wider color space, especially in green and cyan, and is suitable for high-end display requirements such as printing; the sRGB color gamut is the most widely used standard, covering most of the colors perceivable by the human eye; the NTSC color gamut is derived from the television broadcast standard, emphasizing the performance of red and yellow. By collecting the coverage rate data of the car machine screen under different color gamut standards, the color display capability of the car machine can be quantitatively evaluated.

[0075] The car machine screen parameters depict the characteristics of the car screen from multiple dimensions such as screen physical size, display precision, and color performance, both quantitative parameter values and qualitative level descriptions, providing solid data support for subsequent screen classification. The acquisition of car machine screen parameters requires a lot of manpower and resources, which is a long and meticulous process through on-site measurement, consulting vehicle manuals, and comparing screen specifications. However, complete and accurate screen parameter data is the basis for cross-platform interface adaptation, so step S1100 is crucial. Collecting comprehensive car machine screen parameters unifies the reference standard for subsequent adaptation and optimization, avoids analysis blind spots caused by data missing, and is a necessary condition to ensure adaptation effect.

[0076] In step S1200, the hardware configuration parameters of each car model car machine system are obtained; the hardware configuration parameters include CPU model, memory capacity parameter, GPU model, and storage device parameter;

[0077] Specifically, the CPU model includes the number of CPU cores, the main frequency, the cache capacity, and the like; the memory capacity parameters include the specifications and models of the running memory and the storage memory; the GPU model includes the number of GPU cores, the frequency, the video memory capacity, and the like; the storage device parameters include the storage capacity, the storage medium type (such as eMMC, UFS), the read-write speed, and the like; and the step S1200 evaluates the computing performance differences of different car machine systems by collecting the hardware configurations of the car machine systems of various vehicle models.

[0078] The CPU, as the central processor of the car machine, directly determines the operation speed and concurrent processing capability of the car machine through its performance parameters. By comparing the architectures (such as ARM, X86), the process technology, the number of cores, the main frequency, and the like of different CPUs, the computing power of the car machine can be quantitatively analyzed. For example, a car machine equipped with Qualcomm Snapdragon 865 has more advanced 5nm process and more CPU cores than a car machine equipped with Snapdragon 835, and the operation speed can be increased by more than 30%. The memory capacity is another factor that affects the smoothness of the car machine operation, and the specifications of the running memory and the storage memory can be collected to determine the data processing capability of the car machine. An 8GB LPDDR5 (Low Power Double Data Rate 5th Generation Memory) running memory can simultaneously drive multiple large software, while a 256GB UFS3.1 (Universal Flash Storage 3.1 Standard) mobile phone has larger storage space and faster data reading speed. The GPU is responsible for the graphics rendering task of the car machine, and the model and frequency parameters of the GPU can be collected to evaluate the graphics processing performance of the car machine. For example, a car machine equipped with Mali-G78 has more GPU cores and higher frequency than a car machine equipped with Mali-G72, and the graphics rendering speed can be increased by about 50%. The type and capacity of the storage device also affect the user experience of the car machine, and the collection of the storage capacity and medium type can analyze the data storage capability of the car machine. A 256GB UFS storage has more space to store map data and music files, while the read-write of eMMC storage is relatively slow, which may cause problems such as lag.

[0079] The hardware configuration parameters evaluate the performance level of the car machine from the perspectives of computing, graphics, storage, and the like, covering the key hardware factors that affect the user experience of the car machine. The collection of hardware parameters requires referring to the product manual of the car machine, disassembling the device for testing, comparing the score data of similar products, and the like, and the process is relatively complex and professional. However, reliable hardware parameter evaluation data is the basis for ensuring the software adaptation effect. Ignoring the hardware performance differences and developing software in a one-size-fits-all manner will inevitably cause user experience problems such as lag and crash. Collecting comprehensive hardware configuration parameters and clarifying the hardware endowments of different car machines are important bases for judging the adaptation and optimization effect. Only within the scope allowed by the hardware conditions, can the optimization space of the software algorithm be realized, and the best user experience can be achieved.

[0080] Step S1300, mapping the screen hardware parameters into the same vector space for clustering, and dividing the car machine into n2 discrete screen hardware levels.

[0081] Further, as shown in Figure 2 Step S1300 includes:

[0082] Step S1310, preprocessing and feature selection of the screen hardware parameters to obtain clustering features;

[0083] Further, as shown in Figure 3 Step S1310 includes:

[0084] Step S1311, cleaning the screen hardware parameters to remove obvious outliers and missing values;

[0085] Step S1312, normalizing the cleaned screen hardware parameters;

[0086] Step S1313, removing redundant parameters from the normalized screen hardware parameters; constructing parameter pairs according to the normalized screen hardware parameters, calculating the Pearson correlation coefficient between the parameter pairs, and removing one parameter in the parameter pair and taking the other parameter as the reserved parameter for the parameter pair whose absolute value of the correlation coefficient is greater than a preset correlation threshold;

[0087] Step S1314, performing principal component analysis on the screen hardware parameters after removing the redundant parameters, and calculating the variance contribution rate of each principal component, and taking the principal component whose variance contribution rate exceeds a preset contribution rate threshold as the clustering feature.

[0088] Specifically, outliers are usually caused by measurement errors, data entry errors and other factors, far away from normal observation values, which will interfere with subsequent statistical analysis. For example, the general car machine has an internal existence of 1GB-12GB, and if the memory of an individual car machine is as high as 128GB, it is likely to be an abnormal value caused by misrecording, which needs to be removed. Similarly, missing values of individual parameters will also affect the overall data quality and need to be estimated and completed or directly removed. After completing the processing of outliers and missing values, the effectiveness and cleanliness of the data can be improved. The screen hardware parameters each have different dimensions, and the numerical values differ greatly. For example, the screen size is measured in inches, generally between 5-15 inches, while the resolution value can reach tens of millions. If the original values are directly used for clustering, the influence of the resolution will be much greater than that of the screen size. Normalization processing unifies different parameters to a similar numerical interval and gives them similar importance. Common normalization methods include Min-Max normalization, Z-score normalization, etc. The normalized parameters are easier to compare comprehensively.

[0089] Step S1313 removes redundant parameters through correlation analysis. There can be significant linear correlation between some parameters, carrying similar information. For example, resolution is highly correlated with pixel density, and including both in clustering will put more weight on display accuracy. Redundant parameters will interfere with clustering results and prolong computation time, and need to be optimized and removed. Here, Pearson correlation coefficient is used to measure linear correlation between any two parameters, and the closer the absolute value of the coefficient is to 1, the stronger the positive or negative correlation. Parameter pairs with a correlation coefficient exceeding a preset correlation threshold carry similar information and one of them can be removed. Step S1314 further extracts clustering features using principal component analysis. Principal component analysis is a commonly used dimensionality reduction method that can condense multiple parameters into a few principal components while preserving data features. Each principal component is a linear combination of original parameters and carries most of the information of the original data. The variance contribution rate of a principal component measures its ability to explain the original data, and the higher the contribution rate, the more complete the information preservation. By analyzing the variance contribution rate of each principal component, a few principal components with high contribution rates are selected as clustering features, which can reduce computational complexity while maximizing information preservation.

[0090] For example, assume that 5 key parameters of 100 car machines are collected: screen size, resolution, pixel density, CPU frequency, and memory capacity. Some data is shown in Table 1:

[0091] Table 1: Partial data list of 5 key parameters of 100 car machines

[0092]

[0093] The steps S1311-S1314 are performed on this set of data in turn:

[0094] Step S1311: Clean up outliers and missing values. The screen size of 230 in row 4 is obviously abnormal, and the row is deleted.

[0095] Step S1312: Min-Max normalization, mapping each column of original data to the interval [0, 1].

[0096] Step S1313: Remove redundant parameters. Calculate the Pearson correlation coefficient of all parameter pairs, and the results are shown in Table 2:

[0097] Table 2: Pearson correlation coefficient result data table

[0098]

[0099] The absolute values of the correlation coefficients exceeding 0.9 are: screen size and resolution, screen size and CPU frequency. Remove resolution and CPU frequency, and keep screen size, pixel density, and memory capacity.

[0100] Step S1314: Principal component analysis. The normalized and redundant parameter-removed data matrix is subjected to principal component analysis, and the variance contribution rates of the principal components are 80%, 15%, and 5%, respectively. The cumulative contribution rate of the first two principal components is 95%, which is sufficient to represent most of the information of the original data. The two principal components are selected as the clustering features.

[0101] Step S1320: The clustering features obtained in step S1310 are taken as the coordinate axes of a multi-dimensional vector space, and the screen hardware parameters of each car machine are represented as an n1-dimensional screen hardware parameter vector in the multi-dimensional vector space.

[0102] For example, assuming that the screen hardware parameters of the car machine still retain 6 key parameters after dimension reduction, they can be represented as a 6-dimensional screen hardware parameter vector:

[0103] ;

[0104] wherein pixel density represents the screen pixel density, color gamut represents the color gamut coverage, cpu_freq represents the CPU frequency, gpu_freq represents the GPU frequency, ram_size represents the memory capacity, storage_size represents the storage capacity, and vehicle_vector represents a 6-dimensional screen hardware parameter vector.

[0105] In the multi-dimensional vector space, the screen hardware parameters of each vehicle model correspond to a position coordinate. The specific numerical value of the vehicle model in each parameter dimension determines its position in the vector space. By calculating the vector distance between the vehicle models, the similarity of the car machine screen hardware parameters can be quantitatively evaluated, providing a mathematical basis for subsequent car machine hardware classification and clustering.

[0106] Specifically, by modeling the car machine screen hardware parameters as a vector space, the multi-dimensional heterogeneous parameters of different vehicle models can be mapped to a unified feature space for comparison. Each coordinate axis in the feature space represents a key parameter, and the scale of the coordinate axis reflects the numerical size of the parameter. The specific value of each vehicle model screen hardware parameter determines its position coordinates in the feature space, and the position coordinates reflect the comprehensive performance of the vehicle model in each key parameter. In the feature space, the Euclidean distance between vehicle models reflects their similarity in hardware performance. The closer the distance between the vehicle models, the closer the hardware parameters, and the more similar the adaptation optimization strategy. Therefore, the mapping of car machine screen hardware parameters to vector space provides a direct geometric representation of the differences in screen hardware of different vehicle models, and lays a model foundation for using clustering algorithms for hardware classification. By unifying multiple dimensions of parameters into the same space, the overall hardware performance of the vehicle model can be evaluated without the need for independent threshold judgment and weighted summation of each single parameter, avoiding the influence of subjective factors and improving the objectivity and interpretability of hardware classification.

[0107] Step S1330, using the K-means clustering algorithm to divide the car machine systems of different vehicle models into n2 screen hardware levels with the screen hardware parameter vector of the car machine as input.

[0108] Specifically, K-means clustering divides similar car machines into the same screen hardware level, and there are obvious differences in hardware performance between car machines in different levels. The division of screen hardware levels takes into account key parameters such as screen display, computing performance, and storage capacity, reflecting the overall difference in hardware performance of the vehicle model. High-level car machines have better screen display effect, stronger computing power, and larger storage space, and can adapt to more sophisticated UI interfaces and run more complex functions; while low-level car machines are limited by hardware and need to adapt to simplified UI and functions to ensure basic experience. K-means clustering makes full use of the distribution characteristics of multiple vehicle models in hardware parameters, and through iterative search for the optimal clustering center, it converts the hardware heterogeneity of vehicle models into several discrete levels, which is convenient for subsequent interface and function classification adaptation. Compared with the hardware classification method of single threshold division, K-means clustering can adaptively divide the levels according to the distribution of the data itself, and has stronger objectivity and flexibility. In addition, K-means clustering also has the advantages of high computational efficiency and good interpretability, and the screen hardware level of each vehicle model is clear, the hardware difference is intuitive and comparable, which is more conducive to the development of classification adaptation strategy.

[0109] For example, assume that the car machine clustering is divided into 3 screen hardware levels: low, medium and high. The clustering result can be as follows:

[0110] Low-end car machine: screen pixel density below 200ppi, color gamut coverage below 90%sRGB, equipped with low-end CPU / GPU, main frequency below 1.5GHz, memory below 4GB, storage capacity below 32GB. Representative models include some entry-level SUVs, economy cars, etc.

[0111] Medium car machine: screen pixel density between 200-300ppi, color gamut coverage between 90%-95%sRGB, equipped with mid-end CPU / GPU, main frequency between 1.5-2.5GHz, memory between 4-8GB, storage capacity between 32-128GB. Representative models include some mid-end SUVs, compact cars, etc.

[0112] High-end car machine: screen pixel density above 300ppi, color gamut coverage above 95%sRGB, equipped with flagship CPU / GPU, main frequency above 2.5GHz, memory above 8GB, storage capacity above 128GB. Representative models include some luxury SUVs, flagship cars, etc.

[0113] For low-end car machines, the UI interface should adopt a more simple flat design style, reduce visual effects such as gradients, shadows, and transparency, and reduce the complexity of interface rendering; at the same time, simplify the car machine function, only keep navigation, music and other core applications, and do not provide extended functions such as video playback, voice assistant, etc.

[0114] For medium car machines, the UI interface can adopt flat design, moderately use visual elements such as gradients and images, but avoid overly complex multi-layer superposition; integrate navigation, music, radio and other commonly used applications, provide video playback, voice assistant and other extended functions, but should limit video bitrate and voice interaction complexity.

[0115] For high-end car machines, the UI interface can adopt a realistic design, using a large number of visual elements such as gradients, shadows, and motion effects to create a refined interface experience; in addition to covering all applications of low and medium car machines, it can also integrate innovative applications exclusive to large-screen car machines, such as KTV, gaming and other entertainment functions, making full use of high-configuration screens, CPUs and memories to provide an extreme car experience.

[0116] By K-means clustering to classify car machine screen hardware, combined with grade division to develop targeted UI design and function implementation strategies, while ensuring user experience, it avoids excessive requirements for low-end car machine hardware, thereby reducing the development cost of adaptation and optimization, and improving the delivery efficiency of car machine system. In addition, hardware grade division is beneficial for car companies to divide configuration requirements of different models according to positioning and cost, and hardware procurement can also adapt to business strategies.

[0117] Step S2000, for each screen hardware level, design the corresponding car machine function classification scheme; collect different user behaviors of using car machine functions, generate user preference portrait; combine the user preference portrait and the car machine function classification scheme to generate a personalized function recommendation list;

[0118] Further, step S2000 includes:

[0119] Step S2100, for each screen hardware level, design the corresponding car machine function classification scheme; the car machine function classification scheme includes three levels of basic function, extended function and advanced function;

[0120] Further, as Figure 4 shown, step S2100 includes:

[0121] Step S2110, according to the hardware configuration parameters of different screen hardware level car machines, obtain the computing performance index of different screen hardware level car machines;

[0122] ;

[0123] Among them:

[0124] , the computing performance index of the car machine under the screen hardware level L.

[0125] , the reference adjustment coefficient of the screen hardware level L;

[0126] Effect: used to balance the hardware configuration difference of different screen hardware levels. It takes different values between different screen hardware levels, and the value range is adjusted according to the overall performance of the car machine screen hardware level. The preferred value range is generally between 1.0 and 2.5, the higher the level, the greater. Ensure that the computing performance index of different screen hardware levels will not appear too large fluctuation, and appropriately enlarge the performance of high level car machine, highlight the level difference.

[0127] Obtaining method: the optimal value can be obtained through regression analysis or machine learning model optimization of historical performance data.

[0128] , the CPU performance average of the screen hardware level L;

[0129] Effect: represents the average CPU performance of the car machine under the screen hardware level L, which is used to evaluate the computing processing ability of the car machine of this level. CPU performance is the core of car machine operation, which affects the computing speed and response time. The high and low of The core computing capabilities in the system.

[0130] Acquisition method: Obtain the CPU performance data of all vehicle infotainment systems within screen hardware level L, and take the average value.

[0131] CPU base coefficient;

[0132] effect: Controlling average CPU performance Logarithmic magnification effect, balance Numerical pairs The impact of computation. Balancing the performance of different CPU levels to prevent low-performance CPUs from causing problems. The value is too low. The preferred value range is generally between 0.1 and 0.8, and the specific value should be adjusted according to the CPU performance range of the screen hardware level L.

[0133] Acquisition method: Analyze the distribution of CPU performance data for each hardware level to ensure low... Time The calculations maintain a moderate impact.

[0134] CPU index adjustment factor;

[0135] effect: It is an exponential adjustment factor that affects CPU performance. It amplifies with changes in CPU performance exponentially. Controlling CPU performance has... The increase or decrease in the value ensures that the CPU's high and low performance are balanced. The impact is more reasonable, distinguishing different levels. The preferred value range is generally between 0.8 and 1.5, and the value depends on the CPU's impact on the performance index;

[0136] Acquisition method: It can be obtained through regression fitting of CPU performance data at different levels, making... It better reflects the actual performance of the vehicle's infotainment system.

[0137] Memory capacity impact coefficient;

[0138] effect: It is memory capacity The influence coefficient makes The changes can be more accurately reflected in Adjusting memory capacity is important. The contribution of [the system / mechanism] enables devices with larger memory to [perform certain functions / functions]. This demonstrates superior performance. The recommended optimal value range is between 0.2 and 0.6, with slightly larger values ​​for higher levels.

[0139] Acquisition method: according to the influence of memory on multitasking performance, historical data regression fitting is determined.

[0140] , the average memory capacity of the screen hardware level L;

[0141] Effect: Reflects the ability of the car machine in multitasking, the larger the memory represents the more applications and tasks the car machine can support, and the greater the performance improvement;

[0142] Acquisition method: take the average of the memory data of all car machines in the screen hardware level L.

[0143] , the benchmark denominator coefficient;

[0144] Effect: Adjust the influence of GPU performance and storage speed on , balance the influence of GPU and storage speed, ensure the moderate influence of both in . The preferred value range is generally between 0.5 and 1.2;

[0145] Acquisition method: observe the distribution of GPU and storage performance data of different levels, and fit the optimal value.

[0146] , storage adjustment coefficient;

[0147] Effect: Used to adjust the influence of storage speed in , ensure that low-speed storage devices do not significantly reduce , so that calculation remains stable in low-level devices. The preferred value range is between 0.1 and 1.0, and the higher the level, the higher the value can be taken;

[0148] Acquisition method: fitting or empirical judgment through historical storage data.

[0149] , the average GPU performance of the car machine under the screen hardware level L;

[0150] Effect: It is the average value of the GPU performance of the car machine under the screen hardware level L, used to evaluate the image processing and rendering capability. Higher GPU performance reflects better graphics and video processing capability, improving the overall performance index;

[0151] Acquisition method: take the average of the GPU performance data of all car machines in the screen hardware level L.

[0152] , the average storage speed of the car machine under the screen hardware level L;

[0153] Effect: Storage speed affects application startup time and data processing speed. The higher the storage performance, the more significant the improvement.

[0154] Obtaining method: Average the vehicle storage speed data in the screen hardware level L.

[0155] , resolution influence coefficient;

[0156] Effect: Used to adjust the screen resolution mean effect on performance, avoiding the negative impact of high-resolution screens on performance. The preferred numerical range is generally between 0.1 and 0.6, and the specific value is related to the screen level and resolution requirements;

[0157] Obtaining method: Analyze the data of the impact of resolution on performance.

[0158] , screen resolution mean;

[0159] Effect: Reflects the visual display quality of the vehicle, the higher the resolution, the better the visual experience, but also requires higher performance;

[0160] Obtaining method: Average the vehicle resolution data in the screen hardware level L.

[0161] This formula maps the hardware configuration parameters of different screen hardware levels L to , so that the calculation performance index of each level can reasonably reflect its comprehensive hardware performance. Through the adjustment of , and other parameters, the formula can adapt to the differences between different levels, ensuring that the amplitude is obvious in higher levels, while maintaining a relatively stable in low-level devices.

[0162] Step S2120, according to the vehicle screen parameters of different screen hardware level vehicles, obtain the display effect index of different screen hardware level vehicles;

[0163] ;

[0164] Where:

[0165] represents the number of samples under the current screen hardware level;

[0166] represents the​ One vehicle infotainment system sample;

[0167] Indicates the first The weighting factor for each vehicle infotainment system sample is used to control the impact of different screen hardware levels;

[0168] Indicates the first Pixel density parameters of individual vehicle infotainment system samples;

[0169] Indicates the first Screen resolution parameters for a single in-vehicle infotainment system sample;

[0170] Indicates the first Screen size parameters for a sample vehicle infotainment system;

[0171] Indicates the first Screen color gamut coverage parameters for individual vehicle infotainment system samples;

[0172] This is a parameter for adjusting pixel density; it reflects the weight of pixel density in the display performance index.

[0173] These are parameters for adjusting resolution; they reflect the weight of resolution in the display quality index.

[0174] These are parameters for adjusting the screen's color gamut coverage; they reflect the weight of color gamut in the display performance index.

[0175] It is the adjustment angle of color gamut weight, used to control the relative contribution to the display color quality under different screen hardware levels;

[0176] A very small positive number (e.g.) ), used to avoid cases where the denominator is zero.

[0177] Weighting factor The index can be set based on the representativeness of each vehicle infotainment system sample within the current level. This is mainly determined by the sales share or application breadth of the current vehicle infotainment system within a certain level, ensuring that representative data in the sample has a greater impact on the display effect index.

[0178] Pixel density The clarity of the screen display can be calculated from the screen's resolution and size. This parameter affects the visual clarity component of the display quality index.

[0179] Screen resolution , which is used to evaluate the maximum resolution that the screen can display, usually represented by the physical resolution of the screen (e.g., 1920x1080). This parameter determines the level of detail in the displayed image and affects the clarity of the display.

[0180] Screen size represents the physical size of the screen, usually measured in inches. Screen size affects the user's visual experience, as a larger screen at the same resolution will reduce the display density, so it is used as the denominator in the formula to balance the display effect.

[0181] Color gamut coverage is used to evaluate the color performance of the screen, usually expressed as a percentage of the color gamut range covered by the screen (e.g., percentage coverage of sRGB, Adobe RGB). Color gamut coverage determines the richness and authenticity of the screen colors, so the function is introduced in the formula to simulate the sensitivity of the human eye to color display quality at different levels.

[0182] Adjustment parameters These parameters are used to adjust the weight of different display factors on the display effect index to adapt to the differences between different levels of screen hardware. By adjusting the values of these parameters, personalized weight configurations can be provided for different levels of screen hardware.

[0183] When the pixel density increases, the display effect index will rise accordingly, but its effect will gradually tend to be flat, which is consistent with the visual saturation effect of the human eye on high-pixel-density screens.

[0184] When the resolution increases, the display effect index will increase linearly, indicating that at the same screen size, the higher the resolution, the better the display effect.

[0185] When the screen size increases,

[0186] the display effect index will decrease slightly, as a larger screen will dilute the display density and affect the display clarity. When the color gamut coverage

[0187] increases, the display effect index has a periodic trend in its increase, simulating the advantages of color quality at high color gamut coverage.

[0188] Through the non-linear adjustment of this formula, the display effect index can comprehensively reflect the differences between different levels of screens and the visual performance of specific screen parameters, thus more accurately evaluating the display effect of screen hardware.The formula integrates multiple parameters such as resolution, pixel density, screen size and color gamut coverage, and realizes clustering in the same vector space in a weighted and mapped manner, so as to calculate the corresponding display effect index for different levels of screen hardware. By setting different weighting factors for each level and adjusting the values of the parameters , the formula can be flexibly applied in different levels and reflect the differences between different levels. At the same time, through the nonlinear processing of the dependent variables (such as resolution, size, color gamut), the display effect index is further enhanced in expressiveness, so that it takes a sample average in the same level, and maintains sufficient discrimination in different levels.

[0189] Step S2130, for each screen hardware level, according to the computing performance index and the display effect index of the vehicle machine, and combining the functional complexity and the resource occupation level, the vehicle machine function is divided into three levels of basic function, extended function and high-level function.

[0190] Further, step S2130 includes:

[0191] Step S2131, for each screen hardware level, calculating the functional complexity index of the vehicle machine;

[0192] Step S2132, for each screen hardware level, calculating the resource occupation index of the vehicle machine;

[0193] Step S2133, for each screen hardware level, according to the computing performance index, the display effect index, the functional complexity index and the resource occupation index of the vehicle machine, using the support vector machine (SVM) algorithm, a function classification model is constructed, and the vehicle machine function is divided into three levels of basic function, extended function and high-level function.

[0194] Specifically, in step S2131, the complexity index represents the implementation difficulty and logical complexity of the vehicle machine function; the complexity index is obtained by analyzing the code amount, the number of called APIs, the types of involved third-party libraries and other indicators, and the higher the index, the more complex the function implementation. The code amount contained in the function is an intuitive basis for complexity evaluation, and the larger the code amount, the more business logic the function involves, and the more complex the interaction between modules. In addition, the number of system APIs and third-party libraries called by the function also reflects its complexity from the side, and the more APIs and libraries called, the more system resources the function may need to coordinate, and the more complex the logical flow. In the actual evaluation process, code static analysis tools can be used to count the number of code lines, API calls and referenced libraries of each function module, and experience thresholds can be set to quantitatively score the complexity index.

[0195] In step S2132, the resource occupation index represents the consumption level of hardware resources when the car machine function is running; the resource occupation index is obtained by statistically averaging the CPU occupation rate, memory occupation, storage space occupation, network traffic consumption and other indicators when the function is running, and the higher the index, the more resource consumption. The CPU occupation rate reflects the occupation level of processor resources when the function is running, and the higher the occupation rate, the more likely it means that the function contains a large number of computing-intensive tasks. The memory occupation represents the consumption of memory resources when the function is running, and the larger the occupation, the more data needs to be processed. The storage space occupation measures the degree of dependence of the function on local storage resources, and the larger the occupation, the more user data or cache data needs to be saved locally. Network traffic consumption reflects the occupation intensity of data transmission resources, and the higher the traffic, the more likely it means that the function needs to frequently interact with network data. In actual evaluation, performance analysis tools can be used to collect resource consumption indicators in real time when the function is running, and the weighted average can be calculated according to the experience weight to finally obtain the resource occupation index.

[0196] Step S2133 uses the support vector machine (SVM) algorithm to train a function classification model, which realizes automatic layering of car machine functions. The calculation performance index, display effect index, complexity index, and resource occupation index are used as input features of the model, which can comprehensively evaluate the hardware adaptation difficulty of different functions from the perspectives of hardware adaptation and software implementation. The function level is used as the output label of the model, which can clearly define the target adaptation level of different functions, and learn the corresponding relationship between different indexes and function levels through a training sample set. Basic functions usually include basic operations of the car machine, such as turning on and off, brightness adjustment, time display, etc., which have low requirements for hardware adaptation; extended functions usually include car navigation, multimedia playback, Bluetooth phone, etc., which have medium requirements for hardware adaptation; advanced functions may include voice assistants, mobile phone projection, driving assistance, etc., which have high requirements for hardware adaptation. Building a suitable training sample set is crucial for training a high-quality classification model. The training samples need to cover different levels of car machine hardware and various types of functions, and the function level of each sample needs to be labeled by professionals. The SVM model learns the mapping relationship between features and labels in the sample set to build an optimal classification plane, which can correctly classify different functions into target levels.

[0197] One advantage of the function classification model is its objectivity and efficiency. Traditional function classification mainly relies on expert experience, which is subjective and time-consuming. However, a trained classification model can automatically infer the target level of adaptation based on the hardware resource consumption of a function, greatly improving the classification efficiency. The classification result can directly guide the subsequent interface and interaction optimization, and advanced functions can use more refined UI design and interaction methods, while low-level functions should simplify UI elements and interaction steps as much as possible.

[0198] For example, assume that a car machine belongs to a high-end level, and the model input features of its multimedia playback function are as follows:

[0199] Computing performance index: 0.8 (can smoothly run high-definition video decoding);

[0200] Display effect index: 0.9 (has a display screen with high contrast and wide color gamut);

[0201] Complexity index: 0.6 (the logic of audio and video decoding, subtitle rendering, etc. is relatively complex);

[0202] Resource occupation index: 0.7 (video decoding and audio playback occupy a large amount of CPU and memory resources);

[0203] Inputting the above indexes into the trained function classification model, the function level output by the model is an extended function. This indicates that multimedia playback is a medium-difficulty function for a high-end car machine, and UI elements can be appropriately enriched in interface design to provide cover browsing, lyric synchronization, and other extended functions. Rich interaction methods such as voice control and gesture operation can also be introduced in interaction design to improve user experience.

[0204] For example, assume that a car machine belongs to a low-end level, and the model input features of its navigation function are as follows:

[0205] Computing performance index: 0.4 (limited processor performance, difficult to smoothly run complex navigation algorithms);

[0206] Display effect index: 0.5 (low display screen size and resolution);

[0207] Complexity index: 0.8 (path planning, map rendering, and other algorithm logic are very complex);

[0208] Resource occupation index: 0.9 (real-time navigation will occupy a large amount of CPU, memory, and storage resources for a long time);

[0209] The output level of the navigation function for a low-end car machine is a high-level function, which means that the navigation function may exceed the hardware capability range of the car machine and needs to be significantly simplified. In UI design, vector elements of the map should be removed as much as possible, and only necessary road lines should be retained to avoid using a large number of 3D building models. In interaction, the operation process should also be simplified to provide one-key navigation and other fast functions to reduce the user's operation burden. At the same time, the navigation algorithm should be optimized to provide basic path guidance function under the premise of reducing resource consumption as much as possible.

[0210] Another big advantage of the function grading model is its interpretability. Each grading decision can be mapped to a specific hardware adaptation index, and engineers can analyze the index composition to quickly locate possible hardware bottlenecks and carry out targeted optimization. The importance of different hardware indicators to the grading decision can also be evaluated during the model training process, further guiding hardware selection and algorithm optimization. In the long run, the grading model can also be continuously optimized and updated by introducing more actual function adaptation cases for continuous learning, constantly iterating to improve grading accuracy.

[0211] In summary, steps S2131 to S2133 quantitatively evaluate the adaptation difficulty of different car machine functions through complexity index and resource occupation index, etc. hardware indicators, and use SVM algorithm to build a grading model to automatically divide the functions into three levels: basic, extended and advanced, greatly improving the pertinence and efficiency of cross-platform adaptation. This quantitative grading method overcomes the subjectivity of traditional empirical grading, with clear and transparent decision-making basis and clear and traceable optimization guidance. The grading results can directly guide the layered optimization of interface, interaction and algorithm, and through the differentiation of adaptation strategies, the hardware potential of different levels of car machines is fully utilized, avoiding one-size-fits-all overdesign or simple design. It can be predicted that this method will significantly improve the adaptation efficiency and quality of car machine software, providing solid technical support for car machine cross-platform development.

[0212] Step S2100 designs a differentiated function grading scheme according to the screen hardware level of the car machine, divides the car machine functions into three levels: basic function, extended function and advanced function, and realizes the fine management and resource allocation optimization of car machine functions. The computing performance index comprehensively considers CPU frequency, core number, GPU frequency, memory capacity, storage capacity and other hardware configuration parameters, quantitatively evaluates the computing power of different levels of car machines, and is an important decision basis for function grading. The display effect index comprehensively measures screen size, resolution, pixel density, color gamut and other display-related parameters, and directly reflects the differences in visual experience of different levels of car machines, affecting the design decisions of UI layout, motion effect, color matching and other aspects.

[0213] For car machines of different hardware levels, according to the high and low of their computing performance index and display effect index, and considering the complexity and resource occupation of the function itself, the car machine functions are divided into three levels: basic function, extended function and advanced function. Specifically, it includes:

[0214] Basic function: is the core function to ensure the basic use of car machine, with low resource occupation and low computing complexity, simple interface, such as radio, car recorder, etc. All levels of car machines must have basic functions.

[0215] Extended function: is the extended function of improving the use experience of the car machine, the resource occupation and the calculation complexity are moderate, the interface is beautiful, such as voice assistant, online music, etc. The medium and high level car machine needs to have the extended function.

[0216] Advanced function: is the cutting-edge function of showing the top performance of the car machine, the resource occupation is large, the calculation complexity is high, the interface is cool, such as AR real scene navigation, gesture control, etc. Only high level car machine needs to adapt to advanced function.

[0217] According to the hardware level and resource endowment of the car machine, different levels of functions are reasonably arranged, which meets the basic needs of users while avoiding resource waste and smoothness problems. Through the differentiated function management mechanism, the high configuration car machine and the low configuration car machine respectively play the hardware advantages, provide the function selection conforming to the positioning, and the user can obtain the use experience matching the hardware strength. In addition, the function grading facilitates the developers and designers to formulate targeted adaptation strategy, the high configuration car machine focuses on experience optimization, the low configuration car machine focuses on performance optimization, which greatly improves the cross-platform adaptation efficiency. The function grading reasonably divides the basic, extended and advanced three levels through the combination of qualitative and quantitative methods, not simply piles up the number of functions, but bases on the hardware performance and user experience, considers not only the complexity of the function itself, but also the carrying capacity of the hardware, balances the adaptation effect, research and development efficiency and cost investment, is a practical grading adaptation scheme.

[0218] Step S2200, collecting the use behavior data of different users on the car machine function, generating user preference portrait;

[0219] Further, as shown in Figure 5 , step S2200 includes:

[0220] Step S2210, collecting the use behavior data of different users on the car machine function; the use behavior data includes the demographic characteristics of the user and the function use record;

[0221] Specifically, the demographic characteristics of the user include gender, age, occupation, income level, driving age, etc.; the function use record includes the frequency and duration of the user using each function on the car machine. Step S2210 records the interactive behavior of the user and the car machine function through user behavior collection software development kit (SDK), questionnaire survey, car machine log and other channels. Demographic information helps to divide users into different groups, and use records reflect the differences in function preferences of different groups.

[0222] Illustratively, part of the user behavior data collected by a car machine is shown in Table 3:

[0223] Table 3 User behavior data table

[0224] User ID Gender Age Group Navigation Usage Music Play Duration Bluetooth Phone Duration U1001 Male 35-44 years old 120 times / month 10 hours / month 6 hours / month U1002 Female 25-34 years old 90 times / month 30 hours / month 1 hour / month U1003 Male 45-54 years old 60 times / month 5 hours / month 18 hours / month

[0225] According to the demographic characteristics, it can be preliminarily judged that U1001 and U1003 are middle-aged male business people, and navigation and Bluetooth phone are the most needed; U1002 is a young woman, and pays more attention to entertainment functions such as music playing.

[0226] In step S2220, the demographic characteristics and function use records of each user are constructed into a user behavior feature vector, the user behavior feature vectors are clustered, and the users are divided into n3 user groups;

[0227] Specifically, the demographic similarity and use behavior similarity are comprehensively considered during clustering, and similar users are divided into the same group. The finally obtained n3 groups have significant differences in demographic characteristics and use preferences.

[0228] For example, the user behavior feature vectors are subjected to K-means clustering, and the clustering effect is good when the number of clusters is 3, and the users are divided into 3 user groups:

[0229] User group 1: mainly male users, mostly over 40 years old, mostly private business owners and enterprise managers, and prefers navigation and Bluetooth phone functions. It is speculated that this type of user has frequent travel and call needs.

[0230] User group 2: mainly female users, mostly under 40 years old, mostly white-collar workers and students, and prefers music, video and other entertainment functions. It is speculated that this type of user has stronger entertainment needs.

[0231] User group 3: mainly taxi drivers, professional vehicles, more male, long use time of car machine functions, and the use scene is mainly productivity functions such as window message and order management.

[0232] In step S2230, the average use frequency and duration of each user group on different functions are counted, a user-function preference matrix is constructed, and a function preference portrait of each user group is generated; the rows of the user-function preference matrix represent different user groups, the columns represent each car machine function, and the matrix elements represent the preference degree of a user group to a car machine function.

[0233] Specifically, the preference degree is measured according to the average use frequency and duration of a group on a function and other indicators. The row vector represents the overall preference of a user group to each function, which is called the function preference portrait of the group.

[0234] For example, according to the clustering result of step S2220, the user-function preference matrix is shown in Table 4:

[0235] Table 4 User-function preference matrix

[0236] Navigation Music Video Bluetooth Phone Voice Assistant User Group 1 0.8 0.3 0.2 0.6 0.5 User Group 2 0.4 0.9 0.8 0.2 0.7 User Group 3 0.7 0.2 0.1 0.8 0.6

[0237] The matrix element takes a value between 0 and 1, and the greater the value, the stronger the preference.

[0238] From the matrix, it can be seen that user group 1 prefers navigation and Bluetooth phone, which is a business-oriented user; user group 2 prefers music and video, which is an entertainment-oriented user; and user group 3 prefers navigation and Bluetooth phone, and also considers voice assistants, which is a productivity-oriented user. This is consistent with the previous speculation of the characteristics of the groups. The row vector of each group is the functional preference profile of the group, which directly reflects the differences in functional needs of different groups.

[0239] The collection of user behavior data for car machine functions is the data basis for functional hierarchical adaptation. Only by deeply obtaining the behavior characteristics and use demands of different user groups can it be determined which functions are more important to a particular group and should be prioritized for adaptation and optimization. User behavior collection needs to comprehensively and meticulously record the real interaction between the user and the car machine, involving multiple aspects such as burying points, data cleaning, privacy protection, and requires high data engineering capabilities. Once the user behavior big data accumulates to a certain extent, it will form a key decision support for functional optimization and adaptation.

[0240] The clustering method used is the classic K-means algorithm, which has fast operation speed and good interpretability, and can quickly converge to the optimal clustering in a million-level user group.

[0241] On the basis of obtaining user clustering, the user-function preference matrix is further introduced to quantitatively represent the functional preferences of the groups. The preference matrix directly presents the preference differences of different groups in different functions, and vividly depicts the overall preference profile of the groups. For example, it can be clearly seen from the matrix that user group 1 and user group 3 have strong preferences for business functions such as navigation and Bluetooth phone, while user group 2 prefers entertainment functions such as music and video. These preference profiles will provide direct optimization ideas for subsequent functional hierarchical adaptation and guide individualized function recommendation for different groups.

[0242] Step S2200 proposes an automated and quantitative method for analyzing user function preferences. This method uses demographic information and usage records as its data foundation, and employs unsupervised clustering and matrix analysis as its main tools. It can automatically extract user groups and function preference profiles from large-scale, fine-grained user behavior data, inferring the functional needs and tendencies of different user groups. Compared with traditional subjective grouping and qualitative descriptions, this method boasts higher intelligence and computational efficiency, yielding more accurate and objective group profiles, significantly improving the targeting and effectiveness of in-vehicle system function adaptation.

[0243] In summary, step S2200 collected comprehensive user behavior data and automatically generated user group segments and functional preference profiles through clustering and matrix analysis. This data and model support is crucial for the hierarchical adaptation of in-vehicle infotainment system functions. In subsequent function weight calculations and hierarchical adaptation, the user preference information obtained here will be fully utilized to determine the importance of different functions from a human perspective, and to formulate personalized adaptation strategies for different user groups, maximizing the human-computer interaction experience of the in-vehicle infotainment system.

[0244] Step S2300: For each screen hardware level, a personalized function recommendation list is generated by combining the user preference profile and the vehicle system function classification scheme.

[0245] Further, step S2300 includes:

[0246] Step S2310: Based on user preference profiles, obtain a list of preferred functions for different user groups;

[0247] Furthermore, such as Figure 6 As shown, step S2310 includes:

[0248] Step S2311: Sort each row of the user-feature preference matrix in descending order to obtain a list of features ranked from high to low in terms of user group preference.

[0249] Step S2312: Extract the N most preferred functions from each function sorting list and generate a list of preferred functions for each user group;

[0250] Step S2313: Count the frequency of each preference function in the preference function list, mark the preference functions whose frequency exceeds the preset frequency threshold as high-frequency preference functions, and generate a set of high-frequency preference functions.

[0251] Specifically, step S2311 can obtain a function ranking list of user group preference degrees from high to low by descendingly sorting the row vectors of the user-function preference matrix. For example, for the jth user group, if the preference degree of the car machine navigation is 0.8, the preference degree of the car machine music is 0.6, and the preference degree of the car machine telephone is 0.4, the preference function ranking is: navigation > music > telephone. Step S2312 further intercepts the top N positions of the ranking list to obtain the N functions with the highest preference of each group, and N can be flexibly set according to actual needs, usually 5-10. The intercepted result is called a preference function list, which reflects the core function preference of the user. Step S2313 further counts the occurrence frequency of each function in the different group preference function lists. The preference function with an occurrence frequency higher than a preset frequency threshold can be regarded as a high-frequency preference function at the screen hardware level. These functions are highly consistent with the actual needs of most users, and are worthy of special adaptation and optimization. Through the preference function analysis, the actual preference of different user groups for the car machine functions is clarified, which is beneficial to matching user demands in personalized recommendation and improving the accuracy of recommendation. The frequency statistics result also reveals the common preference of the mainstream user group. These preference functions should be given priority consideration in interface layout and resource configuration to maximize user demand and optimize user experience.

[0252] Step S2320, for each screen hardware level, the high-frequency preference functions are screened from the corresponding car machine function classification scheme to obtain a candidate function set;

[0253] Further, as shown in Figure 7 Step S2320 includes:

[0254] Step S2321, the basic functions and the extended functions in the high-frequency preference function set are all included in the candidate function set;

[0255] Step S2322, the resource occupation level SE of each high-level function in the high-frequency preference function set is calculated;

[0256] Step S2323, the high-level function in the high-frequency preference function set with a resource occupation level SE lower than a preset resource occupation level threshold is included in the candidate function set.

[0257] Specifically, step S2320 screens the high-frequency preference functions from the corresponding car machine function classification scheme to form a candidate function set for different screen hardware levels. The so-called high-frequency preference function refers to a car machine function with a higher frequency of use and a higher degree of popularity in a specific user group. Different hardware levels adapt different high-frequency preference functions to achieve differentiated function recommendation.

[0258] The generation of the candidate function set follows two principles: first, the basic functions and extended functions of the car machine are unconditionally included in the candidate set. The basic functions are the core functions that guarantee the normal use of the car machine, such as navigation, Bluetooth phone, etc., which belong to the demand type requirement and the adaptation surface should be wide enough. The extended functions are the characteristic functions of the car machine manufacturer, such as voice assistant, mobile phone interconnection, etc., which belong to the differentiation selling point and also need to be included in the candidate set. Secondly, the high-level functions need to consider the constraints of hardware resources when included in the candidate set. High-level functions usually have high requirements for computing, storage, display and other hardware resources, and can only run smoothly when the resource occupation level is lower than the hardware carrying capacity.

[0259] The screening process of the high-frequency preference function into the candidate function set is as follows: step S2321, the basic functions and extended functions are directly included in the candidate set; step S2322, the resource occupation level SE of each high-level function is calculated. SE can be evaluated by function call times, memory usage, frame rate, etc.; step S2323, the high-level functions with SE lower than the resource occupation level threshold are included in the candidate set. The setting of the threshold needs to consider the function quality and hardware level, which guarantees the experience of the recommended function and prevents hardware overload.

[0260] For example, assuming that the high-frequency preference functions of a certain mid-end car machine include: navigation (basic), Bluetooth phone (basic), intelligent voice (extended), video player (high-level), 3D game (high-level). Navigation, Bluetooth phone, intelligent voice directly enter the candidate set. The SE evaluation of the video player is medium to low, and the SE evaluation of the 3D game is medium to high. If the resource occupation threshold of the mid-end car machine is medium, the video player can be included in the candidate set, and the 3D game is excluded due to SE exceeding the standard.

[0261] The generation of the candidate function set combines the use preferences of different user groups and the hardware adaptation ability, which has both general functions for all users and exclusive functions for segmented users, and also eliminates the necessary hardware bottleneck, preliminarily realizing the hierarchical adaptation of functions. Compared with treating all high-frequency functions equally, different hardware levels of car machines can obtain more targeted function combinations under the candidate set method, improving the individuality of car machine functions.

[0262] Step S2330, generating a personalized function recommendation list from the candidate function set.

[0263] Further, step S2330 includes:

[0264] Step S2331, obtaining the demographic characteristics and historical function use records of the user to be recommended, constructing the user behavior feature vector of the user to be recommended; calculating the similarity between the user behavior feature vector of the user to be recommended and the user behavior feature vector of each user group, and finding the user group with the highest similarity as the user group to which the user to be recommended belongs;

[0265] Step S2332, obtaining the preference function list of the user group to which the user to be recommended belongs, intersecting with the candidate function set to obtain the candidate recommended function set;

[0266] Step S2333, obtaining the historical function preference of the user to be recommended based on the historical function use record of the user to be recommended, and selecting M functions from the candidate recommended function set to generate a personalized function recommendation list according to the historical function preference of the user to be recommended.

[0267] Specifically, step S2330 generates a personalized function recommendation list from the candidate function set, which realizes the effect of further matching the hierarchical adaptive candidate function to the individual user. Personalized recommendation is based on user portrait and user historical behavior, and excavates the group characteristics and individual preferences of the user to find out the head function that best fits the current user in the candidate function set.

[0268] The process of generating a personalized function recommendation list is as follows: step S2331, obtaining the demographic characteristics (such as gender, age, occupation) and historical function use record (such as use time, use frequency) of the user to be recommended, and constructing a feature vector reflecting the behavior characteristics of the user. Based on the similarity matching of the feature vector and each user group, the group to which the current user is most likely to belong is found; step S2332, extracting the preference function list of the group to which the current user belongs, intersecting with the candidate function set to obtain the candidate recommended set considering the group preference and hardware adaptation; step S2333, on the basis of the candidate recommended set, combining the historical preference of the current user, screening out the M head functions with the highest degree of fit to generate a personalized recommendation list.

[0269] The user behavior feature vector encodes the static attributes and dynamic behavior of the user into a unified mathematical representation, which is convenient for calculating the similarity between users. Common algorithms for similarity matching between users and groups include cosine similarity, Euclidean distance, Pearson correlation coefficient, etc. The intersection operation further highlights the hardware adaptability of the candidate function on the basis of respecting the group preference. The integration of historical preferences not only learns from the user's own past choices, but also can excavate the implicit preferences of similar users through collaborative filtering algorithms. The length M of the final recommendation list needs to balance the accuracy and diversity of the recommendation, and generally considers dynamic adjustment according to the business scenario.

[0270] For example, a certain SUV model is equipped with the aforementioned mid-end car machine, and its high-frequency preference function is filtered through step S2320 to obtain the candidate function set {navigation, Bluetooth phone, intelligent voice, video player}. A young male user uses the car machine for the first time, and the system obtains his characteristics: male, 26 years old, programmer, no historical use record, and accordingly matches to the "young male technical population". The preferred function of this population is {navigation, intelligent voice, video player, 3D game}, and the intersection of the candidate set is obtained as the candidate recommendation set {navigation, intelligent voice, video player}. Finally, according to the selection of the similar population, a recommendation list of 2 functions {navigation, intelligent voice} is generated for him.

[0271] Personalized function recommendation further connects group preferences and individual characteristics on the basis of hierarchical adaptation, providing a tailor-made function combination. Compared with mass recommendation, personalized recommendation allows each user to obtain exclusive functions that better meet their needs, without having to passively accept redundant functions. The utilization efficiency of car machine resources and user experience are improved, and the matching degree and stickiness of functions are further enhanced.

[0272] In summary, filtering high-frequency preference functions into a candidate function set and generating a personalized function recommendation list based on user characteristics is an effective method to achieve hierarchical adaptation of car machine functions and improve the personalized experience of car machines. This method fully explores the differences between different user groups and different car machine hardware, realizes the leap from unified adaptation to hierarchical adaptation and then to individual adaptation, and makes the car machine more intelligently match user needs. Of course, this method is not a one-time success, and it also needs to accumulate user feedback in practical application, dynamically optimize the adaptation strategy and recommendation algorithm, in order to find the best balance point between hardware utilization and user experience.

[0273] Step S3000, for the target car machine to be adapted, obtaining the car machine screen parameters and hardware configuration parameters of the target car machine, determining the screen hardware level of the target car machine, according to the screen hardware level of the target car machine, calling the corresponding car machine function classification scheme and personalized function recommendation list, and adaptively generating a car machine function list that meets the characteristics of the corresponding car model of the target car machine.

[0274] Specifically, the car machine screen parameters of the target car machine to be adapted are acquired, including screen size, resolution, pixel density, color gamut, etc. These parameters can be obtained by consulting the car machine hardware manual, disassembling the device to measure, or cooperating with the car company, etc. The hardware configuration parameters of the target car machine are acquired, including CPU model, main frequency, memory capacity, GPU model, storage capacity, etc. These parameters reflect the computing and storage performance of the car machine. They can be collected through car machine system information, running score software, etc. The screen and hardware parameters of the target car machine are preprocessed according to the method of step S1300, such as feature extraction and normalization, to obtain a multi-dimensional screen hardware parameter vector of the target car machine. The preprocessing procedure here is completely consistent with steps S1310-S1320, which is to ensure that the target car machine parameters and the vehicle clustering samples are in the same vector space. The Euclidean distance between the screen hardware parameter vector of the target car machine and each cluster center is calculated, and it is divided into the nearest cluster level. The cluster center and the number of clusters come from the clustering result of step S1330.

[0275] For example, assume that the car machine is divided into three screen hardware levels: low configuration, medium configuration, and high configuration through step S1330, and their cluster centers are respectively:

[0276] Low configuration: ;The cluster center of the screen hardware level of low configuration is:

[0277] Medium configuration: ;The cluster center of the screen hardware level of medium configuration is:

[0278] High configuration: ;The cluster center of the screen hardware level of high configuration is:

[0279] The screen hardware parameter vector of a newly launched SUV car machine is: ,The distances between the three cluster centers are calculated:

[0280] The smallest distance is

[0281] , So the target SUV car machine is divided into the medium configuration level.

[0282] According to the screen hardware level to which the target car machine belongs, a corresponding function classification scheme and a personalized function recommendation list are called. The function classification scheme comes from step S2000, and different functions are classified for different screen hardware levels. The personalized function recommendation list is generated by considering hardware level, target vehicle positioning, user preferences, etc. According to the function classification scheme called, all functions supported by the vehicle model are sorted to form an initial function list. The functions that do not match the positioning of the vehicle model and have low user acceptance are removed to obtain a simplified candidate car machine function list. From the candidate car machine function list, functions suitable for the screen hardware level of the car machine are selected. For different levels of car machines, the form and complexity of the functions will be different:

[0283] Low-end car machine: select basic functions, simple interface layout, and mainly use keys and touch for interaction, and focus on function availability.

[0284] Medium car machine: select advanced functions, fine interface design, and increase voice, gesture, etc. for interaction, and balance functionality and aesthetics.

[0285] High-end car machine: select flagship functions, rich interface effects, advanced interaction, integrate AR, intelligent recommendation, etc. and emphasize luxury and technology.

[0286] The items in the personalized function recommendation list are supplemented to the filtered function list. The personalized function recommendation list considers the user portrait of the target vehicle model and other factors, which can improve the use experience and loyalty of the vehicle model users. The final function list is sorted, the key functions are highlighted, and the layout is optimized to generate a function scheme that matches the screen hardware of the target car machine. The scheme fully considers the characteristics of the car machine hardware in terms of function content and display form.

[0287] Step S3000 filters and supplements car machine functions at different granularities, seeks matching in multiple dimensions such as hardware level, vehicle positioning, user preferences, etc., maximizes the use of hardware performance, and takes into account user experience and product characteristics. Compared with simply stacking functions, hierarchical adaptation can achieve reasonable configuration of functions in limited hardware resources. The overall optimization and arrangement of the function list highlights the display of key functions and also takes into account the aesthetic coordination of the layout, which can improve user experience in both function and visual aspects.

[0288] For example, a certain mid-range sedan is equipped with a car machine screen hardware parameter: screen size 8 inches, resolution 1280*720, pixel density 210ppi, color gamut coverage 95%sRGB, CPU 1.8GHz, memory 4GB, storage 32GB. According to the above steps:

[0289] 1. The car machine screen hardware parameter is pre-processed to obtain a 7-dimensional screen hardware parameter vector (210, 95%, 1.8, 600, 4, 32, 8). Among them, 600 is the estimated GPU frequency.

[0290] 2. Calculate the distance between the vector and the cluster center. Assuming that the nearest cluster center is the medium matching cluster center, it is divided into the medium matching hardware level.

[0291] 3. Retrieve the function classification scheme designed for medium hardware, and the personalized function recommendation list of the car model user.

[0292] 4. Screen the supported navigation, music, Bluetooth phone functions, etc. Eliminate high-end functions such as mobile phone projection, AR navigation, etc. Keep the advanced version of the function with beautiful interface and friendly interaction.

[0293] 5. Supplement the car networking service, driving behavior analysis and other characteristic functions from the personalized function recommendation list.

[0294] 6. Sort and layout the personalized function recommendation list, highlight the main functions such as music and navigation, and arrange other functions according to the use frequency and logical relationship.

[0295] Finally, output a set of car machine function list that fits the hardware performance of the car machine, reflects the characteristics of the car model, and takes into account the preferences of the user. The list is adapted and differentiated by presetting, reasonably configuring limited hardware resources to each function, and smoothly realizing the function requirements, bringing a high-quality human-vehicle interaction experience.

[0296] The scheme of the embodiment can tailor a matching function list for the hardware heterogeneity of different car models, adapt to hardware, users and products at multiple levels, and has strong practical value. This hierarchical adaptation method can be promoted to the development of more car-mounted systems.

[0297] Embodiment 2

[0298] This embodiment provides a cross-platform car machine function classification and adaptation optimization system based on embodiment 1, as shown in Figure 8 , comprising:

[0299] Car machine grading module: obtain the screen hardware parameters of different car models, map the screen hardware parameters to the same vector space for clustering, and divide the car machine into n2 discrete screen hardware levels; the screen hardware parameters include car machine screen parameters and hardware configuration parameters;

[0300] The personalized recommendation module: for each screen hardware level, a corresponding car machine function classification scheme is designed; the use behavior data of different users on the car machine function is collected, and a user preference portrait is generated; the user preference portrait and the car machine function classification scheme are combined to generate a personalized function recommendation list;

[0301] The adaptation module: for the target car machine to be adapted, the car machine screen parameters and hardware configuration parameters of the target car machine are obtained, the screen hardware level of the target car machine is determined, the corresponding car machine function classification scheme and personalized function recommendation list are retrieved according to the screen hardware level of the target car machine, and a car machine function list conforming to the characteristics of the corresponding car model of the target car machine is adaptively generated.

[0302] In the car machine grading module, the screen hardware parameters of different car models include:

[0303] Step S1100, obtaining car machine screen parameters of different car models; the car machine screen parameters include car machine screen size parameters, car machine screen resolution parameters, car machine screen pixel density parameters, and car machine screen gamut parameters;

[0304] Step S1200, obtaining hardware configuration parameters of car machine systems of each car model; the hardware configuration parameters include CPU model, memory capacity parameters, GPU model, and storage device parameters;

[0305] In the car machine grading module, the screen hardware parameters are mapped to the same vector space for clustering, including:

[0306] Step S1310, preprocessing and feature selection are performed on the screen hardware parameters to obtain clustering features;

[0307] Step S1320, taking the clustering features obtained in step S1310 as the coordinate axes of a multi-dimensional vector space, and representing the screen hardware parameters of each car machine as an n1-dimensional screen hardware parameter vector in the multi-dimensional vector space;

[0308] Step S1330, taking the screen hardware parameter vector of the car machine as input, and using the K-means clustering algorithm to divide the car machine systems of different car models into n2 screen hardware levels.

[0309] The step S1310 includes:

[0310] Step S1311, cleaning the screen hardware parameters to remove obvious outliers and missing values;

[0311] Step S1312, normalizing the cleaned screen hardware parameters;

[0312] Step S1313, redundant parameter elimination is performed on the normalized screen hardware parameters; parameter pairs are constructed according to the normalized screen hardware parameters, Pearson correlation coefficients between the parameter pairs are calculated, and for the parameter pairs with an absolute value of the correlation coefficient greater than a preset correlation threshold, one parameter in the parameter pair is eliminated, and the other parameter is taken as a reserved parameter.

[0313] Step S1314, principal component analysis is performed on the screen hardware parameters after the redundant parameters are eliminated, and a variance contribution rate of each principal component is calculated, and a principal component with a variance contribution rate exceeding a preset contribution rate threshold is taken as a clustering feature.

[0314] In the personalized recommendation module, the corresponding car machine function classification scheme for each screen hardware level includes:

[0315] Step S2100, for each screen hardware level, a corresponding car machine function classification scheme is designed; the car machine function classification scheme includes three levels of basic function, extended function and advanced function;

[0316] Step S2200, the use behavior data of the car machine function of different users is collected, and a user preference portrait is generated;

[0317] Step S2300, for each screen hardware level, the user preference portrait and the car machine function classification scheme are combined to generate a personalized function recommendation list.

[0318] The step S2100 includes:

[0319] Step S2110, according to the hardware configuration parameters of the car machine of different screen hardware levels, the computing performance index of the car machine of different screen hardware levels is obtained;

[0320] Step S2120, according to the car machine screen parameters of the car machine of different screen hardware levels, the display effect index of the car machine of different screen hardware levels is obtained;

[0321] Step S2130, for each screen hardware level, according to the computing performance index and the display effect index of the car machine, and in combination with the function complexity and the resource occupation level, the car machine function is divided into three levels of basic function, extended function and advanced function.

[0322] The step S2130 includes:

[0323] Step S2131, for each screen hardware level, the function complexity index of the car machine is calculated;

[0324] Step S2132, for each screen hardware level, the resource occupation index of the car machine is calculated;

[0325] Step S2133, for each screen hardware level, a function classification model is constructed by using a support vector machine (SVM) algorithm according to the computing performance index, the display effect index, the function complexity index and the resource occupation index of the vehicle machine, so as to divide the vehicle machine functions into three levels of basic functions, extended functions and advanced functions.

[0326] The step S2200 includes:

[0327] Step S2210, collecting usage behavior data of different users on vehicle machine functions; the usage behavior data includes demographic characteristics and function usage records of the users.

[0328] Step S2220, constructing demographic characteristics and function usage records of each user into a user behavior feature vector, clustering the user behavior feature vector, and dividing the users into n3 user groups.

[0329] Step S2230, statistically analyzing the average usage frequency and duration of each user group on different functions, constructing a user-function preference matrix, and generating a function preference portrait of each user group; the rows of the user-function preference matrix represent different user groups, the columns represent vehicle machine functions, and the matrix elements represent the preference degree of a certain user group to a certain vehicle machine function.

[0330] The step S2300 includes:

[0331] Step S2310, obtaining a preferred function list of different user groups based on the user preference portrait.

[0332] Step S2320, for each screen hardware level, high-frequency preferred functions are screened from the corresponding vehicle machine function classification scheme to obtain a candidate function set.

[0333] Step S2330, generating a personalized function recommendation list from the candidate function set.

[0334] The step S2310 includes:

[0335] Step S2311, descendingly sorting each row of the user-function preference matrix to obtain a function sorting list of each user group in descending order of preference degree.

[0336] Step S2312, intercepting N preferred functions with the highest preference degree from each function sorting list to generate a preferred function list of each user group.

[0337] Step S2313, statistically analyzing the occurrence frequency of each preferred function in the preferred function list, marking the preferred functions with an occurrence frequency higher than a preset frequency threshold as high-frequency preferred functions, and generating a high-frequency preferred function set.

[0338] The step S2320 comprises:

[0339] Step S2321, including all the basic functions and extended functions in the high-frequency preference function set into the candidate function set;

[0340] Step S2322, calculating the resource occupation level SE of each high-level function in the high-frequency preference function set;

[0341] Step S2323, including the high-level function in the high-frequency preference function set whose resource occupation level SE is lower than the preset resource occupation level threshold into the candidate function set.

[0342] The step S2330 comprises:

[0343] Step S2331, obtaining the demographic characteristics and historical function use records of the user to be recommended, constructing the user behavior feature vector of the user to be recommended; calculating the similarity between the user behavior feature vector of the user to be recommended and the user behavior feature vector of each user group, and finding the user group with the highest similarity as the user group to which the user to be recommended belongs;

[0344] Step S2332, obtaining the preference function list of the user group to which the user to be recommended belongs, and finding the intersection of the candidate function set to obtain the candidate recommended function set;

[0345] Step S2333, obtaining the historical function preference of the user to be recommended based on the historical function use records of the user to be recommended, and selecting M functions from the candidate recommended function set to generate a personalized function recommendation list according to the historical function preference of the user to be recommended.

[0346] Embodiment 3

[0347] The embodiment discloses an electronic device, which can include one or more processors and one or more memories. Among them, the memory stores computer readable code, and the computer readable code can execute a cross-platform car machine function classification and adaptation optimization method as described above when executed by one or more processors.

[0348] The method or system according to the embodiments of the present application can also be implemented by means of an architecture of an electronic device. The electronic device can include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. The storage device in the electronic device, such as the ROM or the hard disk, can store a cross-platform car machine function classification and adaptation optimization method provided by the present application. The cross-platform car machine function classification and adaptation optimization method can for example include: obtaining screen hardware parameters of different vehicle models, mapping the screen hardware parameters to the same vector space for clustering, and dividing the car machine into n2 discrete screen hardware levels; the screen hardware parameters include car machine screen parameters and hardware configuration parameters; for each screen hardware level, a corresponding car machine function classification scheme is designed; user behavior data of the use of the car machine function is collected, and a user preference portrait is generated; in combination with the user preference portrait and the car machine function classification scheme, a personalized function recommendation list is generated; for a target car machine to be adapted, the car machine screen parameters and the hardware configuration parameters of the target car machine are obtained, the screen hardware level of the target car machine is determined, and according to the screen hardware level of the target car machine, the corresponding car machine function classification scheme and the personalized function recommendation list are retrieved, and a car machine function list conforming to the characteristics of the corresponding vehicle model of the target car machine is adaptively generated.

[0349] Further, the electronic device can also include a user interface. Of course, the architecture disclosed in the present application is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present application can be omitted according to actual needs.

[0350] Embodiment 4

[0351] The present embodiment discloses a computer-readable storage medium having computer-readable instructions stored thereon. When the computer-readable instructions are executed by a processor, a cross-platform car machine function classification and adaptation optimization method according to the embodiments of the present application can be executed. The storage medium includes but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory can include, for example, random access memory (RAM) and cache memory, etc. The non-volatile memory can include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0352] In addition, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions executable by a processor to perform instructions corresponding to the method steps provided by the present application, for example: obtaining screen hardware parameters of different vehicle models, mapping the screen hardware parameters to the same vector space for clustering, and dividing the vehicle machine into n2 discrete screen hardware levels; the screen hardware parameters include vehicle machine screen parameters and hardware configuration parameters; for each screen hardware level, a corresponding vehicle machine function classification scheme is designed; user preference portraits are generated by collecting different user behaviors of using the vehicle machine functions; an individualized function recommendation list is generated in combination with the user preference portraits and the vehicle machine function classification scheme; for a target vehicle machine to be adapted, the vehicle machine screen parameters and hardware configuration parameters of the target vehicle machine are obtained, the screen hardware level of the target vehicle machine is determined, and the corresponding vehicle machine function classification scheme and individualized function recommendation list are retrieved according to the screen hardware level of the target vehicle machine, to adaptively generate a vehicle machine function list that meets the characteristics of the corresponding vehicle model of the target vehicle machine. When this computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0353] The methods and systems, devices of the present application can be implemented in many ways. For example, the methods and systems, devices of the present application can be implemented through software, hardware, firmware, or any combination of software, hardware, and firmware. The above described order of steps for the method is merely for illustration, and the steps of the method of the present application are not limited to the above specifically described order, unless otherwise specifically stated. In addition, in some embodiments, the present application can also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Thus, the present application also covers the recording medium storing the programs for executing the method according to the present application.

[0354] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive repetition.

[0355] The specific embodiments described above further illustrate the objects, technical solutions, and advantages of the present application. It should be understood that the above description is merely a specific embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A cross-platform vehicle infotainment system function classification and adaptation optimization method, characterized in that, The method includes: The screen hardware parameters of different vehicle models are obtained, and the screen hardware parameters are mapped to the same vector space for clustering, dividing the vehicle system into n2 discrete screen hardware levels; the screen hardware parameters include vehicle system screen parameters and hardware configuration parameters. For each screen hardware level, a corresponding vehicle infotainment system (VMS) function classification scheme is designed, which divides VMS functions into three levels: basic functions, extended functions, and advanced functions. User behavior data on VMS functions is collected from different users. Based on this data, users are clustered into n³ user groups, and a user-function preference matrix is ​​constructed to generate a function preference profile for each user group, i.e., a user preference profile. Based on these profiles, a list of preferred functions for each user group is obtained. The frequency of each preferred function in the list is counted, and functions with a frequency exceeding a preset threshold are marked as high-frequency preferred functions, generating a high-frequency preferred function set. All basic and extended functions in the high-frequency preferred function set are included in a candidate function set. The resource consumption level (SE) of each advanced function in the high-frequency preferred function set is calculated. Advanced functions in the high-frequency preferred function set with an SE below a preset threshold are included in the candidate function set. A personalized function recommendation list is generated from the candidate function set. The process of generating a personalized feature recommendation list from the candidate feature set includes: obtaining the demographic characteristics and historical feature usage records of the user to be recommended, and constructing a user behavior feature vector of the user to be recommended; calculating the similarity between the user behavior feature vector of the user to be recommended and the user behavior feature vectors of each user group, and finding the user group with the highest similarity as the user group to which the user to be recommended belongs; obtaining the preferred feature list of the user group to which the user to be recommended belongs, and finding the intersection with the candidate feature set to obtain the candidate recommendation feature set; obtaining the historical feature preferences of the user to be recommended based on the historical feature usage records of the user to be recommended, and selecting M features from the candidate recommendation feature set to generate a personalized feature recommendation list based on the historical feature preferences of the user to be recommended; For the target vehicle infotainment system to be adapted, obtain the screen parameters and hardware configuration parameters of the target vehicle infotainment system, determine the screen hardware level of the target vehicle infotainment system, and based on the screen hardware level of the target vehicle infotainment system, retrieve the corresponding vehicle infotainment system function classification scheme and personalized function recommendation list, and adaptively generate a vehicle infotainment system function list that conforms to the characteristics of the corresponding vehicle model.

2. The method for cross-platform vehicle infotainment system function classification and adaptation optimization according to claim 1, characterized in that, The step of mapping screen hardware parameters to the same vector space for clustering includes: Preprocessing and feature selection of screen hardware parameters yields clustering features; Clustering features are used as coordinate axes in a multidimensional vector space, and the screen hardware parameters of each vehicle system are represented as an n1-dimensional screen hardware parameter vector in the multidimensional vector space. Using the screen hardware parameter vector of the vehicle infotainment system as input, the K-means clustering algorithm is used to divide the vehicle infotainment system of different models into n2 screen hardware levels.

3. The method for cross-platform vehicle infotainment function classification and adaptation optimization according to claim 2, characterized in that, The preprocessing and feature selection of screen hardware parameters to obtain clustering features include: Clean the screen hardware parameters to remove outliers and missing values; The hardware parameters of the cleaned screen are normalized. Redundant parameters are removed from the normalized screen hardware parameters; parameter pairs are constructed based on the normalized screen hardware parameters, and the Pearson correlation coefficient between the parameter pairs is calculated. For parameter pairs whose absolute value of the correlation coefficient is greater than the preset correlation threshold, one parameter in the parameter pair is removed, and the other parameter is retained. Principal component analysis is performed on the screen hardware parameters after removing redundant parameters. The variance contribution rate of each principal component is calculated, and the principal components whose variance contribution rate exceeds the preset contribution rate threshold are used as clustering features.

4. The method for cross-platform vehicle infotainment function classification and adaptation optimization according to claim 1, characterized in that, The corresponding vehicle infotainment function classification scheme designed for each screen hardware level includes: Based on the vehicle's hardware configuration parameters, the vehicle's computing performance index is obtained; Based on the vehicle's infotainment system screen parameters, obtain the vehicle's display effect index; Calculate the functional complexity index and resource consumption index of the vehicle's infotainment system; Based on the vehicle's computing performance index, display effect index, functional complexity index, and resource consumption index, a functional hierarchy model is constructed using the Support Vector Machine (SVM) algorithm, dividing the vehicle's functions into three levels: basic functions, extended functions, and advanced functions.

5. The method for cross-platform vehicle infotainment function classification and adaptation optimization according to claim 1, characterized in that, The usage behavior data includes users' demographic characteristics and function usage records; The method for generating the user preference profile includes: Each user's demographic characteristics and functional usage records are used to construct a user behavior feature vector. The user behavior feature vectors are then clustered to divide the users into n3 user groups. The average usage frequency and duration of different functions for each user group are statistically analyzed to construct a user-function preference matrix and generate a function preference profile for each user group. The rows of the user-function preference matrix represent different user groups, the columns represent various in-vehicle functions, and the matrix elements represent the degree of preference of a certain user group for a certain in-vehicle function.

6. The cross-platform vehicle infotainment system function classification and adaptation optimization method according to claim 5, characterized in that, The list of functions for obtaining preferences from different user groups includes: Sort each row of the user-feature preference matrix in descending order to obtain a list of features ranked from highest to lowest preference level for each user group. Extract the N most preferred features from each feature ranking list and generate a list of preferred features for each user group.

7. A cross-platform vehicle infotainment system for function classification and adaptation optimization, used to implement the cross-platform vehicle infotainment system for function classification and adaptation optimization method according to any one of claims 1-6, characterized in that, The system includes: Vehicle infotainment system classification module: acquires screen hardware parameters of different vehicle models, maps the screen hardware parameters to the same vector space for clustering, and divides the vehicle infotainment system into n2 discrete screen hardware levels; the screen hardware parameters include vehicle infotainment system screen parameters and hardware configuration parameters; Personalized recommendation module: Design corresponding vehicle infotainment function classification schemes for each screen hardware level; collect user behavior data of different users on vehicle infotainment functions to generate user preference profiles; combine the user preference profiles and vehicle infotainment function classification schemes to generate a personalized function recommendation list; Adaptation module: For the target vehicle infotainment system to be adapted, obtain the screen parameters and hardware configuration parameters of the target vehicle infotainment system, determine the screen hardware level of the target vehicle infotainment system, and based on the screen hardware level of the target vehicle infotainment system, retrieve the corresponding vehicle infotainment system function classification scheme and personalized function recommendation list, and adaptively generate a vehicle infotainment system function list that conforms to the characteristics of the corresponding vehicle model.

Images