A service recommendation method, device, terminal device and storage medium

Abstract

The present application provides a service recommendation method, apparatus, terminal device, and storage medium applicable to the field of data processing technology. The method comprises: obtaining historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information; mapping and enhancing the historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information to generate individual driver behavior information; matching and calculating the individual driver behavior information and driving group type information to obtain matching group type information and matching degree information; and generating service recommendation information based on the individual driver behavior information, matching group type information, matching degree information, a preset service recommendation model, and preset prompt text information. The present application can accurately capture the driver's infotainment service needs and effectively improve the accuracy and reliability of smart cockpit personalized service recommendations.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and in particular relates to a service recommendation method, apparatus, terminal device, and storage medium. Background Art

[0002] Smart cars are becoming the next generation of intelligent devices, following personal computers and smartphones. As the "third space" and the primary interface for human-vehicle interaction, the smart cockpit is the most easily understood, perceived, and accepted area of automotive intelligent technology by users, becoming the key node connecting people, vehicles, and the environment. The interactive functions of the smart cockpit can be divided into three dimensions: driving control, ride comfort, and infotainment. Infotainment, which provides value-added information services and entertainment, is a crucial component of modern cars.

[0003] In existing technologies, smart cockpit infotainment service recommendations are primarily categorized into three main types: collaborative filtering-based recommendations, content-based recommendations, and deep learning-based recommendations. Collaborative filtering-based recommendations aim to identify similar user behavior patterns to predict the likelihood of future interactions, leveraging historical interactions between users and items. Content-based recommendations achieve personalized recommendations through semantic analysis and feature tag matching, addressing the cold start problem. Deep learning-based recommendations analyze user behavior temporal features using models like Transformer and GNN, supporting cross-modal content understanding.

[0004] However, the ability of collaborative filtering-based recommendations to discover long-tail content is limited by data sparsity and bias towards popular content; content-based recommendations run the risk of information narrowing, which may cause users to fall into a fixed content loop; deep learning-based recommendations have poor natural language understanding capabilities and are mostly designed for a single task, resulting in unsatisfactory service recommendation results in complex situations. Summary of the Invention

[0005] In view of this, the embodiments of the present application provide a service recommendation method, apparatus, terminal device and storage medium, aiming to solve the problems existing in the existing intelligent cockpit service recommendation technology, such as the inability to adapt to the complex interaction scenarios with multiple concurrent intentions in the cockpit, and the lack of pertinence and accuracy of the service recommendation results.

[0006] A first aspect of an embodiment of the present application provides a service recommendation method, including:

[0007] Obtain historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information;

[0008] Mapping and enhancing the historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information to generate individual driving behavior information;

[0009] Performing matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information;

[0010] Service recommendation information is generated based on the individual driving behavior information, matching group type information, matching degree information, a preset service recommendation model, and preset prompt text information.

[0011] A second aspect of an embodiment of the present application provides a service recommendation device, including:

[0012] An information acquisition module is used to obtain historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information;

[0013] A driving individual behavior information generation module is used to map and enhance the historical cockpit service usage information, the current cockpit service usage information, and the vehicle driving state information to generate driving individual behavior information;

[0014] a matching calculation module, configured to perform matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information;

[0015] The service recommendation information generation module is used to generate service recommendation information based on the driving individual behavior information, matching group type information, matching degree information, a preset service recommendation model and preset prompt text information.

[0016] A third aspect of an embodiment of the present application provides a terminal device, which includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the service recommendation method described in the first aspect above.

[0017] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, comprising: storing a computer program, which, when executed by a processor, implements the steps of the service recommendation method described in the first aspect above.

[0018] Compared with the prior art, the beneficial effects of the embodiments of the present application are: mapping and enhancing the acquired multi-source information of the smart cockpit, converting it into individual driving behavior information, fully mining the value of data, improving the availability and accuracy of multi-source information, accurately characterizing the driver's behavior characteristics, matching and calculating the individual driving behavior information with the driving group type information, and compensating for the sparsity of individual data by drawing on the rules of group behavior, so that service recommendations can take into account the needs of different types of drivers, enhancing the comprehensiveness of the recommendations, and then combining the service recommendation model and prompt text information to generate service recommendation content, taking into account multiple factors to deeply analyze the driver's potential needs, thereby accurately capturing the driver's infotainment service needs, and effectively improving the accuracy and reliability of personalized service recommendations of the smart cockpit in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 This is a schematic diagram of the implementation flow of the service recommendation method provided in Example 1 of the present application;

[0021] Figure 2 This is a schematic diagram of the implementation flow of the service recommendation method provided in Example 2 of this application;

[0022] Figure 3 This is a schematic diagram of the implementation flow of the service recommendation method provided in Example 3 of this application;

[0023] Figure 4 This is a schematic diagram of the implementation flow of the service recommendation method provided in Example 4 of the present application;

[0024] Figure 5 This is a flowchart illustrating the implementation of the service recommendation method provided in Example 5 of the present application;

[0025] Figure 6 This is a schematic diagram of the implementation flow of the service recommendation method provided in Example 6 of the present application;

[0026] Figure 7 This is a schematic diagram of the structure of the service recommendation device provided in an embodiment of the present application;

[0027] Figure 8 It is a schematic diagram of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0029] In order to illustrate the technical solution described in this application, specific embodiments are provided below.

[0030] Figure 1 The following is a flowchart illustrating an implementation of the service recommendation method provided in Example 1 of the present application, which is described in detail as follows:

[0031] Step S101 , obtaining historical cockpit service usage information, current cockpit service usage information and vehicle driving status information.

[0032] In this embodiment, cabin service usage information may refer to driver usage behavior information for cabin infotainment services (such as music, radio, and navigation) automatically collected by the smart cockpit. Historical cabin service usage information may refer to driver usage behavior information for cabin infotainment services (such as music, radio, and navigation) automatically collected by the smart cockpit during past driving. Current cabin service usage information may refer to driver usage behavior information for cabin infotainment services (such as music, radio, and navigation) automatically collected by the smart cockpit during the current driving process. Vehicle driving status information may refer to information characterizing the vehicle's driving status, such as driving speed and power mode. Cabin service usage information is matched with vehicle driving status information in real time. It is understood that mixed multi-source cabin information may require pre-processing techniques such as rule-based state reduction and behavior noise reduction, and then chronological correlation to form time series data for further computational processing.

[0033] In this embodiment, optionally, historical cockpit service usage information may include historical communication service usage information and historical entertainment service usage information. Historical communication service usage information may refer to information used to characterize the driver's use of phone calls or navigation during the vehicle's past driving process. Historical entertainment service usage information may refer to information used to characterize the driver's use of entertainment behaviors such as listening to music, listening to the radio, or watching videos during the vehicle's past driving process. Therefore, historical entertainment service usage information may specifically include historical music playback information, historical radio playback information, and historical video playback information. Current cockpit service usage information may include current communication service usage information and current entertainment service usage information. Current communication service usage information may refer to information used to characterize the driver's use of phone calls or navigation during the vehicle's current driving process. Current entertainment service usage information may refer to information used to characterize the driver's use of entertainment behaviors such as listening to music, listening to the radio, or watching videos during the vehicle's current driving process. Therefore, current entertainment service usage information may specifically include current music playback information, current radio playback information, and current video playback information.

[0034] Step S102 : Mapping and enhancing the historical cockpit service usage information, the current cockpit service usage information, and the vehicle driving state information to generate individual driving behavior information.

[0035] In this embodiment, historical and current cockpit service usage information can be mapped to corresponding vehicle driving state information. Information enhancement processing and text generation operations are then performed using a preset state-behavior conversion function and a feature text representation function. Based on information attributes and specific rules, historical and current cockpit service usage information, as well as vehicle driving state information, are converted into text information in a uniform format as individual driver behavior information, thereby generating a data set for calculation by a preset service recommendation model. Information attributes can include state attributes, behavior attributes, and cross-attributes; specific rules can include special interaction attribute deletion rules and behavior state matching rules. Individual driver behavior information can be information used to characterize a specific driver's service usage in the smart cockpit.

[0036] In this embodiment, the driving individual behavior information can be represented as the current moment Driving time information , vehicle driving status information , corresponding to the driver's Historical behavior sequence A collection of:

[0037] ;

[0038] Driving time information Including the current moment , boarding time , idling time :

[0039] ;

[0040] Vehicle driving status information Including vehicle operating status , infotainment system status :

[0041] ;

[0042] Among them, the vehicle running status Including vehicle speed , power mode , driving stage :

[0043] ;

[0044] Infotainment system status Including navigation open state , navigation usage time , map zoom status , phone answering status , music playing status , radio station playback status , video playback status :

[0045] ;

[0046] Historical behavior sequence Used to record the specific content of the driver's above five historical interactions.

[0047] Step S103 : performing matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information.

[0048] In this embodiment, driving group behavior information can be used to characterize the service usage of multiple drivers in the smart cockpit. A driving group can include drivers with different service usage habits, such as listening to music during the day, listening to the radio, making phone calls in the afternoon, and playing videos. A driving group can also represent drivers with different driving behavior patterns, such as commuter drivers, professional drivers, and leisure drivers. This can be achieved by calculating the similarity between individual driver behavior information and driving group type information, then filtering the similarity to perform a matching calculation, matching the individual driver within the driving group, and obtaining the driving group type to which the current driver belongs and the degree of match with the driving group type, i.e., matching group type information and matching degree information. The calculated similarity can be cosine similarity.

[0049] Step S104 : generating service recommendation information based on the individual driving behavior information, the matching group type information, the matching degree information, the preset service recommendation model, and the preset prompt text information.

[0050] In this embodiment, the preset service recommendation model can be set based on the LLM model. The preset prompt text information can be manually set to specify the range of the service recommendation model output content to avoid the service recommendation model outputting redundant information. It can be that the driving individual behavior information, matching group type information and matching degree information are used as input data of the service recommendation model, and the prompt text information is set to limit the range of the output data of the service recommendation model, thereby generating service recommendation information for pushing service content to the driver, such as displaying inquiry information to the driver through the display device: "Do you need to turn on navigation?" "Do you need to play music?"

[0051] The service recommendation method provided in the embodiment of the present application maps and enhances the acquired multi-source information of the smart cockpit, converts it into individual driving behavior information, fully taps the value of the data, improves the availability and accuracy of the multi-source information, accurately characterizes the driver's behavior characteristics, and matches and calculates the individual driving behavior information with the driving group type information. By drawing on the group behavior rules to compensate for the sparsity of individual data, the service recommendation can take into account the needs of different types of drivers, thereby enhancing the comprehensiveness of the recommendation. The service recommendation model and prompt text information are then combined to generate service recommendation content, so as to consider multiple factors to deeply analyze the driver's potential needs, thereby accurately capturing the driver's infotainment service needs, and effectively improving the accuracy and reliability of personalized service recommendations in the smart cockpit in complex scenarios.

[0052] Figure 2The following is a flowchart illustrating an implementation of a service recommendation method provided in Example 2 of the present application, which differs from Example 1 above in that:

[0053] The driving group type information is obtained by the following steps:

[0054] Step S201: Acquire multiple driver behavior information and historical vehicle driving status information.

[0055] In this embodiment, the driver behavior information and historical vehicle driving status information can be automatically collected by the smart cockpit. The driver behavior information can be information used to characterize the service usage of multiple drivers in the smart cockpit, and can include the current time Driving time information , vehicle driving status information , corresponding to the driver's Historical behavior sequence . Historical vehicle driving status information can be information representing the status of the vehicle during past driving, and can include vehicle operating status information and infotainment service usage status information during past driving, that is, it can include vehicle driving speed information, power mode information, driving stage information, navigation start status information, navigation usage time information, map zoom status information, call answering status information, music playback status information, radio playback status information and video playback status information during the historical driving process.

[0056] Step S202 : clustering the plurality of driver behavior information and historical vehicle driving state information, and calculating behavior occurrence frequency information and vehicle driving state statistical information.

[0057] In this embodiment, the driver behavior information may be clustered based on a K-means algorithm, and the vehicle driving state information corresponding to the driver behavior information may be extracted from the historical vehicle driving state information. The occurrence frequency information of each driver behavior information in the past driving records and the vehicle driving state statistics may be statistically analyzed. The occurrence frequency information may be an average frequency.

[0058] Step S203 : extracting and integrating the driver behavior information based on the behavior occurrence frequency information and the vehicle driving state statistical information to obtain a plurality of driving group behavior information.

[0059] In this embodiment, based on the behavior frequency information and the vehicle driving status statistical information, features of typical group behavior patterns in the driver behavior information can be extracted, and the extracted features can be integrated into multiple driving behavior information as driving group behavior information.

[0060] Step S204: Obtain driving group type information based on the plurality of driving group behavior information.

[0061] In this embodiment, different driving group behavior information may be calibrated as behavior information of drivers with different types of service usage, thereby generating driving group type information.

[0062] The service recommendation method provided in the embodiment of the present application clusters multiple driver behavior information and historical vehicle driving status information to effectively mine data patterns, extract key features from massive amounts of driver behavior information and historical vehicle driving status information, enhance the availability of multi-source information in the cockpit, extract and integrate the clustered driver behavior information, and generate driving group type information for matching with individual driving information, thereby introducing the driving group type as a supplementary reference for individual driving information in subsequent calculations, and providing reliable data support for accurate cockpit service recommendations.

[0063] Figure 3 The following is a flowchart illustrating an implementation of a service recommendation method provided in Example 3 of the present application, which differs from Example 1 above in that:

[0064] The preset service recommendation model includes a preset residual connection layer, a preset normalization processing layer, and multiple preset service information encoding and decoding layers;

[0065] The step S104 specifically includes:

[0066] Step S301: Generate service recommendation word vector information based on the individual driving behavior information, matching group type information, matching degree information, and preset prompt text information.

[0067] In this embodiment, the driving individual behavior information, matching group type information, matching degree information and preset prompt text information may be subjected to text encoding processing through the encoding layer in the LLM model to generate service recommendation word vector information.

[0068] Step S302: Project the service recommendation word vector information according to a preset spatial weight coefficient to obtain a service recommendation projection vector.

[0069] In this embodiment, the preset spatial weight coefficients may be the query weight coefficient, key weight coefficient, and value weight coefficient obtained after pre-training the multi-head attention mechanism model. The service recommendation word vector may be obtained by multiplying the multiple spatial weight coefficients by the service recommendation word vector, and then scaling, aggregating, and linearly projecting the attention weights based on the multi-head attention mechanism to obtain the service recommendation projection vector.

[0070] Attention weight The calculation method can be expressed as:

[0071] ;

[0072] Among them, the query vector Represents the result of multiplying the query weight coefficient and the service recommendation word vector, key vector Represents the multiplication result of the key weight coefficient and the service recommendation word vector, the value vector The product of the value weight coefficient and the service recommendation word vector, represents the calculation of multi-head attention weights for query vector, key vector, and value vector, represents the normalization to the probability distribution along the key vector dimension, the scaling factor Used to prevent the dot product value from being too large and causing gradient instability.

[0073] The linear projection can be expressed as:

[0074] ;

[0075] in, Used to represent the projection splicing function based on the multi-head attention mechanism, It is used to indicate that the elements are to be spliced together, and h represents the number of heads in the multi-head attention mechanism. It can be the output projection matrix, it can be preset, or it can be obtained by training the multi-head attention mechanism.

[0076] Step S303 : Based on a preset residual connection layer and a preset normalization processing layer, the service recommendation projection vector is calculated to obtain a service recommendation projection variable.

[0077] In this embodiment, the preset residual connection layer can be the residual connection layer included in the LLM model architecture, and the preset normalization processing layer can be the normalization processing layer included in the LLM model architecture. The service recommendation projection vector is first used as input data for the residual connection layer. The intermediate variable calculated by the residual connection layer is then used as input data for the normalization processing layer. The result calculated by the normalization processing layer is used as the service recommendation projection variable.

[0078] Step S304: Generate a service recommendation content vector according to the service recommendation projection variable and a plurality of preset service information encoding and decoding layers.

[0079] In this embodiment, the preset service information encoding and decoding layer may be the service information encoding and decoding layer in the LLM model architecture. The service recommendation projection variable may be used as input data for the service information encoding and decoding layer, and the result calculated by the service information encoding and decoding layer may be used as the service recommendation content vector.

[0080] Step S305: Generate service recommendation information according to the service recommendation content vector.

[0081] In this embodiment, the service recommendation content vector may be format-converted, and the vector form may be converted into a text form for output to obtain the service recommendation information.

[0082] The service recommendation method provided in the embodiment of the present application comprehensively reflects the driver's service demand characteristics by integrating multi-dimensional key information, effectively highlights the key features of the multi-source information in the cabin by transforming the word vector, optimizes the expression of the multi-source information in the cabin, and is used to enhance the targeted service recommendation. The projection variables are calculated through the residual connection layer and the normalization processing layer to ensure the stable transmission of the multi-source information in the cabin and the retention of key features during the calculation process, thereby improving the accuracy and reliability of the cabin service recommendation.

[0083] Figure 4 The flowchart of the service recommendation method provided in the fourth embodiment of the present application is shown. The difference between the fourth embodiment and the third embodiment is that:

[0084] A preset service information encoding and decoding layer includes a plurality of preset service recommendation gating matrices and a plurality of preset service recommendation expert sub-models; the preset service recommendation gating matrices correspond one to one with the preset service recommendation expert sub-models;

[0085] The step S304 specifically includes:

[0086] Step S401: Obtain the initial recommended number of encoding and decoding layers.

[0087] In this embodiment, the recommended calculation encoding and decoding layers may be designed based on the Transformer. The initial recommended calculation encoding and decoding layer number may refer to the number of Transformer layers currently calculated.

[0088] Step S402: Count the number of preset service information encoding and decoding layers to obtain encoding and decoding layer quantity information.

[0089] In this embodiment, the service information encoding and decoding layer can be designed based on the Transformer. The preset number of service information encoding and decoding layers can be the number of Transformer layers in the LLM model. The number of service information encoding and decoding layers can be 32, and the statistically obtained number of encoding and decoding layers is 32.

[0090] Step S403 , determining whether the number of initially recommended calculated encoding and decoding layers is less than the encoding and decoding layer quantity information; if not, proceeding to step S404 ; if so, proceeding to step S405 .

[0091] In this embodiment, it is understood that in the LLM model, multi-layer Transformers are calculated layer by layer. When the number of encoding and decoding layers for the initial recommendation calculation is less than the number of encoding and decoding layers, it means that the calculation has not been completed through all Transformer layers. In this case, the calculation result of the previous Transformer layer needs to be input into the next Transformer layer for calculation until all Transformer layers are calculated. When the number of encoding and decoding layers for the initial recommendation calculation is equal to the number of encoding and decoding layers, it means that all Transformer layers are calculated, and the service recommendation projection variable can be output as the service recommendation content vector.

[0092] Step S404: Using the service recommendation projection variable as a service recommendation content vector.

[0093] In this embodiment, when the number of encoding and decoding layers for the initial recommendation calculation is equal to the number of encoding and decoding layers, it means that all Transformer layers have been calculated, and the service recommendation projection variable can be output as the service recommendation content vector.

[0094] Step S405 : selecting a plurality of preset service recommendation expert sub-models according to a preset service recommendation gating matrix to obtain a plurality of service recommendation calculation sub-models.

[0095] In this embodiment, the preset service recommendation gating matrix can be manually set or calculated by the LLM model during the training process. The sub-model in the LLM model can be selected based on the LoRA mechanism and the MoE architecture, wherein the MoE architecture includes multiple service recommendation expert sub-models, and the service recommendation expert sub-model in the MoE architecture is screened through the LoRA mechanism, and the screened sub-model is the service recommendation calculation sub-model. Specifically, the weights of each service recommendation expert sub-model can be calculated through the service recommendation gating matrix, and then the two service recommendation expert sub-models with the largest weights are selected as the service recommendation calculation sub-models for subsequent calculations.

[0096] Step S406 : generating a service recommendation content intermediate variable according to the service recommendation projection variable, a plurality of service recommendation calculation sub-models, and a plurality of service recommendation gating matrices.

[0097] In this embodiment, the service recommendation expert sub-model can correspond one-to-one with the service recommendation gating matrix. During the calculation process, the service recommendation projection variable is first calculated using the service recommendation gating matrix corresponding to the service recommendation calculation sub-model. The calculation result is then used as the input data of the service recommendation calculation sub-model. The result calculated by the service recommendation calculation sub-model is the intermediate variable of the service recommendation content.

[0098] Step S407: Update the initial recommended number of calculation encoding and decoding layers to obtain the intermediate recommended number of calculation encoding and decoding layers.

[0099] In this embodiment, the current Transformer layer has been calculated, and the current initial recommendation calculation encoding and decoding layer number needs to be incrementally calculated. The calculation result is used as the intermediate recommendation calculation encoding and decoding layer number, so that the service recommendation content intermediate variable is used as the input variable of the next Transformer layer for calculation.

[0100] Step S408: Use the intermediate recommendation calculation encoding and decoding layer number as the initial recommendation calculation encoding and decoding layer number, use the service recommendation content intermediate variable as the service recommendation projection variable, and return to step S403.

[0101] In this embodiment, the current recommendation calculation coding and decoding layer and the calculation intermediate variables are updated to input the calculation results of the previous recommendation calculation coding and decoding layer into the next recommendation calculation coding and decoding layer for calculation until all recommendation calculation coding and decoding layers are calculated.

[0102] In this embodiment, it is optional to recommend that the encoding and decoding layer be calculated as a Transformer layer, and the calculation result of the previous Transformer layer is input into the next Transformer layer for calculation until all Transformer layers are calculated.

[0103] The service recommendation method provided in the embodiment of the present application performs feature extraction on the service recommendation projection variables, comprehensively captures the key features in the multi-source information of the cockpit, improves the accuracy of predicting the driver's potential infotainment intentions, and performs selection processing through multiple preset service recommendation expert sub-models. There is no need to calculate all service recommendation expert sub-models, thereby reducing calculation time and computing overhead, so as to effectively improve the efficiency of service recommendations and reduce the cost of service recommendations while ensuring the accuracy of service recommendation content.

[0104] Figure 5 The flowchart of the service recommendation method provided in the fifth embodiment of the present application is shown. The difference between the fifth embodiment and the fourth embodiment is that the step S405 specifically includes:

[0105] Step S501 : Calculating selection weights of multiple service recommendation sub-models according to the service recommendation projection vector and a preset service recommendation gating matrix.

[0106] In this embodiment, the service recommendation projection vector may be multiplied by a preset service recommendation gating matrix, and the multiplication result is used as the service recommendation sub-model selection weight.

[0107] Step S502: Arrange the plurality of service recommendation sub-model selection weights in descending order to obtain sub-model selection weight sequence information.

[0108] In this embodiment, the service recommendation sub-model selection weights may be arranged in descending order as elements, and the arranged information may be output as a sequence, namely, sub-model selection weight sequence information.

[0109] Step S503 : Based on the preset number of service recommendation expert sub-models, the sub-model selection weight sequence information is selected to obtain a plurality of selected sub-model weight information.

[0110] In this embodiment, the number of preset service recommendation expert sub-models may be manually set, and may be 2. The first two elements of the sub-model selection weight sequence information may be selected as the weight information of the selected sub-model.

[0111] Step S504: determining a plurality of service recommendation expert sub-models corresponding to the weight information of the plurality of selected sub-models, and obtaining a plurality of service recommendation calculation sub-models.

[0112] In this embodiment, the service recommendation expert sub-model corresponding to the selected sub-model weight information is used as the selected sub-model, ie, the service recommendation calculation sub-model, for calculating the service recommendation projection vector.

[0113] In this embodiment, optionally, the input features are compressed into a low-dimensional space through the projection vector through the LoRA mechanism, and then the original dimension is restored through the up-projection vector to form a dynamic weight perturbation. The weight of the service recommendation expert sub-model is dynamically allocated based on the real-time input driving individual behavior characteristics and vehicle driving status characteristics through the MoE mechanism.

[0114] The service recommendation method provided in the embodiment of the present application screens the service recommendation expert sub-model by calculating the service recommendation sub-model selection weight, thereby avoiding the large calculation delay and computing overhead caused by calculating all sub-models, improving the efficiency of smart cockpit service recommendation, and reducing the hardware and software costs of smart cockpit service recommendation.

[0115] Figure 6 The following is a flowchart illustrating an implementation of a service recommendation method provided in Example 6 of the present application, which differs from the above-mentioned Example 4 in that:

[0116] The service recommendation information calculation sub-model includes a plurality of connection weight matrices;

[0117] The step S406 specifically includes:

[0118] Step S601: performing low-rank decomposition processing on the connection weight matrix to obtain a low-rank decomposition matrix.

[0119] In this embodiment, the connection weight matrix may be present in each expert sub-model of the Transformer layer. The connection weight matrix may be subjected to low-rank decomposition using the LoRA mechanism, and the resulting matrix is used as the low-rank decomposition matrix for subsequent calculations.

[0120] Step S602: Calculate a service recommendation variable decomposition matrix according to the low-rank decomposition matrix and the service recommendation projection variables.

[0121] In this embodiment, the service recommendation variable decomposition matrix may be obtained by multiplying the low-rank decomposition matrix and the service recommendation projection variable.

[0122] Step S603 : generating service recommendation content intermediate variables according to the service recommendation projection variables, the connection weight matrix, the service recommendation variable decomposition matrix, and the weight information of the selected sub-model.

[0123] In this embodiment, the service recommendation projection variable may be multiplied layer by layer through the connection weight matrix, the service recommendation variable decomposition matrix and the selected sub-model weight information, and the calculation result is used as the intermediate variable of the service recommendation content.

[0124] In this embodiment, optionally, the low-rank fine-tuning characteristics of LoRA are deeply combined with the dynamic routing mechanism of MoE. The input features are compressed into a low-dimensional space through the projection vector by the LoRA module. The MoE mechanism dynamically allocates the weights of the service recommendation expert sub-model based on the real-time input features.

[0125] The service recommendation method provided in the embodiment of the present application adds a low-rank update matrix to each layer of the service recommendation model through low-rank matrix decomposition, which is used to adjust the parameters of only a few service recommendation expert sub-models in each layer, avoiding adjusting all parameters of all service recommendation expert sub-models, greatly reducing the consumption of computing resources and training time, and dynamically calculating the weight information of the selected sub-model, which can more flexibly select the service recommendation expert sub-model used for calculation, so as to more accurately predict the driver's potential infotainment intentions when facing different driving scenarios and driver needs, thereby improving the accuracy and efficiency of service recommendations.

[0126] Corresponding to the method of the above embodiment, Figure 7 A structural block diagram of a service recommendation device provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown. Figure 7 The exemplary service recommendation device may be the execution subject of the service recommendation method provided in the aforementioned first embodiment.

[0127] Reference Figure 7, the service recommended devices include:

[0128] An information acquisition module 710 is used to acquire historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information;

[0129] The driving individual behavior information generating module 720 is configured to map and enhance the historical cabin service usage information, the current cabin service usage information, and the vehicle driving state information to generate the driving individual behavior information;

[0130] A matching calculation module 730 is configured to perform matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information;

[0131] The service recommendation information generation module 740 is used to generate service recommendation information based on the individual driving behavior information, matching group type information, matching degree information, a preset service recommendation model and preset prompt text information.

[0132] The process of each module in the service recommendation device provided in the embodiment of the present application realizing its own function can be specifically referred to the aforementioned Figure 1 The description of the first embodiment is omitted here.

[0133] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0134] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0135] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0136] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0137] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish descriptions and should not be understood as indicating or implying relative importance. It should also be understood that although the terms "first", "second", etc. are used in the text to describe various elements in some embodiments of the present application, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first table can be named a second table, and similarly, a second table can be named a first table without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0138] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0139] The service recommendation method provided in the embodiments of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). The embodiments of the present application do not impose any restrictions on the specific types of terminal devices.

[0140] For example, the terminal device can be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a TV set-top box (STB), customer premise equipment (CPE) and / or other devices for communicating on a wireless system and a next-generation communication system, such as a mobile terminal in a 5G network or a mobile terminal in a future evolved Public Land Mobile Network (PLMN) network.

[0141] As an example and not a limitation, when the terminal device is a wearable device, the wearable device can also be a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are full-featured, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0142] Figure 8 This is a schematic diagram of the structure of a terminal device provided by an embodiment of the present application. Figure 8 As shown, the terminal device 8 of this embodiment includes: at least one processor 80 ( Figure 8 Only one is shown), a memory 81, wherein the memory 81 stores a computer program 82 that can be run on the processor 80. When the processor 80 executes the computer program 82, the steps in the above-mentioned various service recommendation method embodiments are implemented, such as Figure 1 Alternatively, when the processor 80 executes the computer program 82, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 7Functions of modules 710 to 740 are shown.

[0143] The terminal device 8 can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device can include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will understand that Figure 8 It is only an example of the terminal device 8 and does not constitute a limitation of the terminal device 8. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include an input and sending device, a network access device, a bus, etc.

[0144] The processor 80 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0145] In some embodiments, the memory 81 may be an internal storage unit of the terminal device 8, such as a hard drive or memory of the terminal device 8. The memory 81 may also be an external storage device of the terminal device 8, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the terminal device 8. Furthermore, the memory 81 may include both an internal storage unit of the terminal device 8 and an external storage device. The memory 81 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 81 may also be used to temporarily store data that has been sent or is about to be sent.

[0146] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0147] An embodiment of the present application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, the terminal device implements the steps of any of the above-mentioned method embodiments.

[0148] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0149] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0150] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the above-mentioned embodiment method by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0151] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0152] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0153] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0154] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A service recommendation method, characterized in that: include: Obtain historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information; Mapping and enhancing the historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information to generate individual driving behavior information; Performing matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information; Generate service recommendation information based on the individual driving behavior information, matching group type information, matching degree information, a preset service recommendation model, and preset prompt text information; The preset service recommendation model includes a preset residual connection layer, a preset normalization processing layer, and multiple preset service information encoding and decoding layers; The step of generating service recommendation information based on the individual driving behavior information, the matching group type information, the matching degree information, the preset service recommendation model, and the preset prompt text information specifically includes: Generate service recommendation word vector information based on the individual driving behavior information, matching group type information, matching degree information, and preset prompt text information; Projecting the service recommendation word vector information according to a preset spatial weight coefficient to obtain a service recommendation projection vector; Based on a preset residual connection layer and a preset normalization processing layer, the service recommendation projection vector is calculated to obtain a service recommendation projection variable; Generate a service recommendation content vector according to the service recommendation projection variable and a plurality of preset service information encoding and decoding layers; generating service recommendation information according to the service recommendation content vector; A preset service information encoding and decoding layer includes a plurality of preset service recommendation gating matrices and a plurality of preset service recommendation expert sub-models; the preset service recommendation gating matrices correspond one to one with the preset service recommendation expert sub-models; The step of generating a service recommendation content vector based on the service recommendation projection variable and a plurality of preset service information encoding and decoding layers specifically includes: Get the initial recommended number of encoding and decoding layers; Counting the number of preset service information encoding and decoding layers to obtain encoding and decoding layer quantity information; Determine whether the number of initially recommended calculation encoding and decoding layers is less than the encoding and decoding layer number information; If not, the service recommendation projection variable is used as the service recommendation content vector; If so, multiple preset service recommendation expert sub-models are selected according to the preset service recommendation gating matrix to obtain multiple service recommendation calculation sub-models; generating a service recommendation content intermediate variable according to the service recommendation projection variable, a plurality of service recommendation calculation sub-models, and a plurality of service recommendation gating matrices; Update the initial recommended calculation encoding and decoding layer number to obtain the intermediate recommended calculation encoding and decoding layer number; Using the intermediate recommendation calculation encoding and decoding layer number as the initial recommendation calculation encoding and decoding layer number, using the service recommendation content intermediate variable as the service recommendation projection variable, and returning to the step of determining whether the initial recommendation calculation encoding and decoding layer number is less than the encoding and decoding layer number information; The step of selecting multiple preset service recommendation expert sub-models according to the preset service recommendation gating matrix to obtain multiple service recommendation calculation sub-models specifically includes: Calculating selection weights of multiple service recommendation sub-models according to the service recommendation projection vector and a preset service recommendation gating matrix; Arrange the selection weights of the plurality of service recommendation sub-models in descending order to obtain sub-model selection weight sequence information; Based on the preset number of service recommendation expert sub-models, the sub-model selection weight sequence information is selected to obtain weight information of multiple selected sub-models; Determining multiple service recommendation expert sub-models corresponding to the weight information of the multiple selected sub-models to obtain multiple service recommendation calculation sub-models; The service recommendation information calculation sub-model includes a plurality of connection weight matrices; The step of generating the service recommendation content intermediate variable according to the service recommendation projection variable, the plurality of service recommendation information calculation sub-models, and the plurality of service recommendation gating matrices specifically includes: Performing low-rank decomposition processing on the connection weight matrix to obtain a low-rank decomposition matrix; Calculating a service recommendation variable decomposition matrix according to the low-rank decomposition matrix and the service recommendation projection variable; The service recommendation content intermediate variable is generated according to the service recommendation projection variable, the connection weight matrix, the service recommendation variable decomposition matrix and the selected sub-model weight information.

2. The service recommendation method according to claim 1, wherein: The driving group type information is obtained by the following steps: Obtain multiple driver behavior information and historical vehicle driving status information; Clustering the plurality of driver behavior information and historical vehicle driving state information, and calculating behavior occurrence frequency information and vehicle driving state statistical information; Extracting and integrating the driver behavior information based on the behavior occurrence frequency information and the vehicle driving status statistical information to obtain multiple driving group behavior information; Driving group type information is obtained based on the plurality of driving group behavior information.

3. The service recommendation method according to claim 1, wherein: The historical cockpit service usage information includes historical communication service usage information and historical entertainment service usage information; The historical entertainment service usage information includes historical music playing information, historical radio playing information and historical video playing information; The current cockpit service usage information includes current communication service usage information and current entertainment service usage information; The current entertainment service usage information includes current music playing information, current radio playing information and current video playing information; The vehicle driving state information includes vehicle driving speed information, vehicle power mode information, and vehicle driving stage information.

4. A service recommendation device, characterized in that: include: An information acquisition module is used to obtain historical cockpit service usage information, current cockpit service usage information, and vehicle driving status information; A driving individual behavior information generation module is used to map and enhance the historical cockpit service usage information, the current cockpit service usage information, and the vehicle driving state information to generate driving individual behavior information; a matching calculation module, configured to perform matching calculation on the individual driving behavior information and the driving group type information to obtain matching group type information and matching degree information; A service recommendation information generation module, configured to generate service recommendation information based on the individual driving behavior information, the matching group type information, the matching degree information, a preset service recommendation model, and preset prompt text information; The preset service recommendation model includes a preset residual connection layer, a preset normalization processing layer, and multiple preset service information encoding and decoding layers; The step of generating service recommendation information based on the individual driving behavior information, the matching group type information, the matching degree information, the preset service recommendation model, and the preset prompt text information specifically includes: Generate service recommendation word vector information based on the individual driving behavior information, matching group type information, matching degree information, and preset prompt text information; Projecting the service recommendation word vector information according to a preset spatial weight coefficient to obtain a service recommendation projection vector; Based on a preset residual connection layer and a preset normalization processing layer, the service recommendation projection vector is calculated to obtain a service recommendation projection variable; Generate a service recommendation content vector according to the service recommendation projection variable and a plurality of preset service information encoding and decoding layers; generating service recommendation information according to the service recommendation content vector; A preset service information encoding and decoding layer includes a plurality of preset service recommendation gating matrices and a plurality of preset service recommendation expert sub-models; the preset service recommendation gating matrices correspond one to one with the preset service recommendation expert sub-models; The step of generating a service recommendation content vector based on the service recommendation projection variable and a plurality of preset service information encoding and decoding layers specifically includes: Get the initial recommended number of encoding and decoding layers; Counting the number of preset service information encoding and decoding layers to obtain encoding and decoding layer quantity information; Determine whether the number of initially recommended calculation encoding and decoding layers is less than the encoding and decoding layer number information; If not, the service recommendation projection variable is used as the service recommendation content vector; If so, multiple preset service recommendation expert sub-models are selected according to the preset service recommendation gating matrix to obtain multiple service recommendation calculation sub-models; generating a service recommendation content intermediate variable according to the service recommendation projection variable, a plurality of service recommendation calculation sub-models, and a plurality of service recommendation gating matrices; Update the initial recommended calculation encoding and decoding layer number to obtain the intermediate recommended calculation encoding and decoding layer number; Using the intermediate recommendation calculation encoding and decoding layer number as the initial recommendation calculation encoding and decoding layer number, using the service recommendation content intermediate variable as the service recommendation projection variable, and returning to the step of determining whether the initial recommendation calculation encoding and decoding layer number is less than the encoding and decoding layer number information; The step of selecting multiple preset service recommendation expert sub-models according to the preset service recommendation gating matrix to obtain multiple service recommendation calculation sub-models specifically includes: Calculating selection weights of multiple service recommendation sub-models according to the service recommendation projection vector and a preset service recommendation gating matrix; Arranging the selection weights of the plurality of service recommendation sub-models in descending order to obtain sub-model selection weight sequence information; Based on the preset number of service recommendation expert sub-models, the sub-model selection weight sequence information is selected to obtain weight information of multiple selected sub-models; Determining multiple service recommendation expert sub-models corresponding to the weight information of the multiple selected sub-models to obtain multiple service recommendation calculation sub-models; The service recommendation information calculation sub-model includes a plurality of connection weight matrices; The step of generating the service recommendation content intermediate variable according to the service recommendation projection variable, the plurality of service recommendation information calculation sub-models, and the plurality of service recommendation gating matrices specifically includes: Performing low-rank decomposition processing on the connection weight matrix to obtain a low-rank decomposition matrix; Calculating a service recommendation variable decomposition matrix according to the low-rank decomposition matrix and the service recommendation projection variable; The service recommendation content intermediate variable is generated according to the service recommendation projection variable, the connection weight matrix, the service recommendation variable decomposition matrix and the selected sub-model weight information.

5. A terminal device, characterized in that: The terminal device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images