A Synchronization and Reminder Method Applied in an Intelligent Cockpit

Abstract

The present application provides a method for application synchronization and reminder in an intelligent cockpit, including: monitoring whether a mobile phone application is started; obtaining basic information of the started application, detecting the application category and establishing a category label; judging whether to grant the intelligent cockpit the right to synchronize applications and obtain user data according to whether there are similar applications; obtaining user data and distinguishing public data and private data; calculating an operation recommendation index according to the deficiencies and activity levels of the applications, and determining the type of operation prompt to be recommended, specifically including: detecting the content of the mobile phone application, analyzing the deficiencies of the application, obtaining the application startup data, analyzing the activity level of the application and establishing a label; constructing an operation guidance recommendation model in combination with user characteristic information and the type of operation to be recommended; detecting user behavior, analyzing the behavior of the current user who does not perform operations according to the prompt, and adjusting the operation recommendation model of the current user; analyzing the operation data of all users after the recommendation guidance to generate an improvement plan for the cockpit application.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to a method for application synchronization and reminder in an intelligent cockpit.

Background Art

[0002] Currently, there are a large number of applications and achievable contents in the intelligent cockpit of autonomous driving. However, users often do not know that this function can be used in the cockpit and operate using a mobile phone. As a result, in the cockpit, it is difficult to obtain accurate user operation data and it is also difficult to improve the user's driving. Only when using the APP in the cockpit is more reasonable for users than using it on the mobile phone can we gradually change the user's usage habits. Therefore, how to increase the usage frequency of apps in the intelligent cockpit has become an urgent problem to be solved.

Summary of the Invention

[0003] The present invention provides a method for application synchronization and reminder in an intelligent cockpit, mainly including:

[0004] Monitoring whether the mobile application is started; obtaining the basic information of the started application, detecting the application category and establishing a class label; judging whether to grant the right of the intelligent cockpit application to synchronize and obtain user data according to whether there are similar applications; obtaining user data and distinguishing public data and private data; calculating an operation recommendation index according to the application deficiencies and application activity, and determining the type of operation prompt to be recommended; constructing an operation guidance recommendation model by combining user characteristic information and the type of operation to be recommended; detecting user behavior, analyzing the behavior of the current user not operating according to the prompt, and adjusting the operation guidance recommendation model of the current user; analyzing the operation data of all users after the recommended guidance to generate an improvement plan for the cockpit application;

[0005] Further optionally, the monitoring whether the mobile application is started includes:

[0006] Creating a monitoring process for all mobile application programs that enter the foreground on the user's mobile phone and identifying the application running state; the running state of the application program is divided into a start state and a close state. When the mobile application program is at the top of the task stack, it means the task is in the start state. If the mobile application program is not at the top of the task stack, it means the task is in the close state; when the mobile application program is in the start state, it is marked with start, and when the mobile application program is in the close state, it is marked with close. Query the running state of the mobile application program every 500 milliseconds and record the monitoring result; obtaining the mobile application program corresponding to the top of the task stack that is currently running in the foreground of the mobile phone through the monitoring process, recording the name of the mobile application program, and marking the running state of the mobile application program as start, and the default state of the remaining mobile application programs is close.

[0007] Further optionally, the acquiring basic information of the started application, detecting the application category and establishing the category tag includes:

[0008] Obtain basic information of a mobile application whose current running state is start, including: mobile application name, running state, and application description document; first, segment the application description document into multiple sentences; then, segment the sentences after the mobile application name and document text are segmented, obtain the terms of all text sentences, remove stop words, and mark the part of speech of each word; calculate the TF-IDF value for each term, set a first threshold, and select terms with a TF-IDF value greater than the first threshold as application description document keywords; formulate application classification standards, including the category names of various mobile application programs and the corresponding The mobile application description keywords are denoted as {category name: keywords}; based on the "Synonym Dictionary", the word similarity between the application description document keywords and the mobile application description keywords given by the classification standard is calculated in sequence, denoted as Sim(A,B); wherein A is the keyword extracted from the application description document, and B is the keyword given by the classification standard; all word similarities are averaged, and the mean is taken as the overall similarity, which is denoted as W; w0 is used as the similarity threshold, and matching is performed according to the overall similarity. When the overall similarity is greater than w0, the category name of the current mobile application is obtained, which is denoted as {mobile application name, running status, category}.

[0009] Further optionally, judging whether to grant the smart cockpit application the right to synchronize and obtain user data according to whether similar applications are included includes:

[0010] The basic information of all applications in the smart cockpit is obtained, including: the name of the application in the cockpit and the description document of the application in the cockpit; according to the application category monitoring method, the description document of the application in the cockpit is firstly segmented, the name of the application in the cockpit and the text sentence of the description document of the application in the cockpit are segmented, and the terms of all text sentences after removing stop words are obtained; then the TF-IDF value of each term is calculated, a second threshold is set, and the term with a TF-IDF value greater than the second threshold is selected as the keyword of the description document of the application in the cockpit; the overall similarity between the keywords of the description document of the application in the cockpit and the application classification standard is calculated to classify all applications in the smart cockpit, and corresponding category labels and category lists are established; the mobile application with the current running state of start is taken as the target application, and it is determined whether the category name of the target application is included in the category list of the application in the cockpit; if the category name of the target application is included in the category list of the cockpit application, the smart cockpit application is granted the right to synchronize and obtain user data; if the category name of the target application is not included in the category list of the cockpit application, it means that the target application cannot be synchronized with the application in the cockpit.

[0011] Further optionally, the obtaining of user data and the differentiation between public data and private data include:

[0012] Obtain user data, including user personal information and the operation information of the user in the target application; preprocess the data to construct a user data table, where the user data table includes the following four types of attributes: display identifier, quasi-identifier, sensitive attribute, and non-sensitive attribute; among them, the display identifier is used to indicate the main identity attribute of the user, and the quasi-identifier is used to indicate the secondary identity attribute of the user; the user cannot be located based on the quasi-identifier alone, but a certain user can be potentially identified in combination with other attributes; the sensitive attribute is the user privacy that needs to be protected; the non-sensitive attribute is public data and can be directly published; protect the user privacy data based on the K-anonymity method and perform generalization operations on the user data; first, delete the display identifier, perform iterative searches on the original data table to find a quasi-identifier, and sequentially increase the number of quasi-identifiers to form a larger subset; then, perform the first data generalization on the quasi-identifiers in the original data table, perform K-anonymity detection on the generalized user relationship table. If the anonymized table meets K-anonymity under the quasi-identifier attribute, output the data and synchronize the data to the intelligent cockpit. If it does not meet the requirements, continue to generalize the data until all quasi-identifier subsets meet the K-anonymity standard.

[0013] Further optionally, the calculation of the operation recommendation index according to the application deficiencies and application activity, and the determination of the type of operation prompt to be recommended include:

[0014] Combine the application deficiencies and the application activity Q to calculate the operation recommendation index Z, where the application deficiencies are represented by the operation convenience P, and the operation recommendation index The lower the value of the operation recommendation index Z, the more inconvenient the operation method of the mobile application is in the driving state and it is not suitable for long-term use; given a threshold Z0, when the recommendation index Z of the target application is lower than Z0, output the type of operation guidance with a higher operation convenience P than the current one, and label it with I, II, III; including: detecting the content of the mobile application and analyzing the application deficiencies; obtaining the application startup data, analyzing the application activity and establishing labels.

[0015] The detection of the content of the mobile application and the analysis of the application deficiencies specifically include:

[0016] Monitor the content of mobile applications, analyze the main operation methods of the target application, and calculate the operation convenience of the target application. By analyzing the operation logs of the application, monitor the operation content of users in the application to obtain the main operation methods of the target application. According to the three main types of operation methods of the current application: VR operation, multimedia operation, and touch operation, determine the category to which the operation method of the target application belongs. Among them, multimedia operations include speech, eye contact, facial expressions, lip movements, gestures, and body recognition operations. Denote VR operation as I, multimedia operation as II, and touch operation as III, and assign different scores to operations I to III according to operation convenience. The operation convenience score of the target application is represented by P. The higher P is, the more convenient the operation method is; the lower P is, the more room for improvement the operation method has.

[0017] The obtaining of application startup data, analyzing application activity, and establishing labels specifically includes:

[0018] Read the startup data of the application within one month through the application startup log, and extract the startup time and usage duration of the application. Construct an RFM model to describe the application activity. Among them, recency R represents the number of days passed since the user's last startup of the application until today, frequency F represents the total number of days the user has started the application within one month, and intensity M represents the total usage duration of the application by the user within one month / the total number of days. Standardize the three types of data of progress R, frequency F, and intensity M, classify the application activity based on K-means clustering, and record the class center value of each class. Calculate the distance between the RFM of the target application and the class center value, classify the application activity and establish corresponding labels, and assign different scores according to the application activity. The score of the target application is represented by Q. The higher the Q value, the higher the time and intensity of the user using the target application, and the higher the risk coefficient of using it while driving.

[0019] Further optionally, the constructing of an operation guidance recommendation model by combining user characteristic information and the operation types to be recommended includes:

[0020] Take the category to which the user's target application belongs, the user's main operation method, and the user's RFM model as user characteristic information, and construct a user characteristic matrix; perform normalization processing on the user characteristic data, and then classify all users through k-means clustering into different user groups; then calculate the characteristic similarity between the current user and other users in the same group, and construct a similar user table based on the characteristic similarity; traverse the main operation methods in the similar user table, screen the similar user table according to the operation types to be recommended for the current user, extract the main operation methods included in the screened similar user table, and construct a list of operations to be recommended for the current user, denoted as {user: operation name}; use the characteristic similarity as the preference weight of the user for the operations to be recommended, select the top-n recommended operations from the list of operations to be recommended for the current user and recommend them to the user, and record the operation name and operation type.

[0021] Further optionally, the detecting of user behavior, analyzing the behavior of the current user not operating according to the prompt, and adjusting the operation guidance recommendation model of the current user include:

[0022] After the intelligent cockpit recommends the operation guidance, create a monitoring process to record the operation method selected by the current user; obtain the name, operation type, and operation result of each step of the current user's operation, determine whether the current user's operation is consistent with the operation guidance recommended by the cockpit, and use 1 and 0 as the identifiers of the user operation result; if the user operation is consistent with the operation guidance, the operation result is 1, continue to push the operation guidance until the guidance content ends, and mark it as (operation name, operation type, 1); if the user operation is inconsistent with the operation guidance, the operation result is 0, record the current operation guidance, and mark it as (operation name, operation type, 0); if the user does not operate according to the prompt, adjust the operation guidance recommendation model of the current user; after constructing the list of operations to be recommended for the current user, match the user's most recent operation result according to the operation name, delete the data with an operation result of 0 from the list of operations to be recommended, and then select the top-n from the list of operations to be recommended for the current user and recommend them to the user.

[0023] An application synchronization and reminder method in an intelligent cockpit, characterized in that the system includes:

[0024] Collect the operation data of all users after the intelligent cockpit recommends the guidance, including the target application name, the category to which the target application belongs, the recommended operation name, the operation type, and the operation result; first, perform data preprocessing on the user operation data to delete the data with an empty operation result; then, count the total number of times each operation is recommended and the number of times the operation result is 1 among them. Calculate the average value of the user acceptance probability of each operation with the same operation type; if the user acceptance probability of the target operation is higher than the average value, it means that the target operation conforms to the user's operation habits; if the user acceptance probability of the target operation is lower than the average value of the user acceptance probability of each operation with the same operation type, it means that the usage method of the target operation in the target application does not conform to the user's operation preferences, and the operation method of similar applications in the cockpit needs to be improved.

[0025] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0026] Take the category to which the user's target application belongs, the user's main operation method, and the user RFM model as user feature information, construct a user feature matrix, then construct an operation guidance model, and recommend based on this to the user, and then use the actual feedback of the user after the recommendation as the subsequent improvement direction, based on which the improvement direction is more targeted, ensuring the optimization and improvement efficiency of the intelligent cockpit.

Description of the Drawings

[0027] Figure 1 This is a flowchart of a synchronization and reminder method applied in an intelligent cockpit of the present invention.

[0028] Figure 2 This is a schematic diagram of a recommendation model of a synchronization and reminder method applied in an intelligent cockpit of the present invention.

Detailed implementation manners

[0029] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Figure 1 This is a flowchart of a synchronization and reminder method applied in an intelligent cockpit of the present invention. As Figure 1 shown, a synchronization and reminder method applied in an intelligent cockpit in this embodiment may specifically include:

[0031] Step 101, monitor whether the mobile phone application is started.

[0032] Create a monitoring process for all mobile applications that enter the foreground on the user's mobile phone, and identify the application running status. The running status of the application is divided into a start status and a close status. When the mobile application is at the top of the task stack, it indicates that the task is in the start status. If the mobile application is not at the top of the task stack, it indicates that the task is in the close status. When the mobile application is in the start status, it is marked with start. When the mobile application is in the close status, it is marked with close. Query the running status of the mobile application every 500 milliseconds and record the monitoring result. Obtain the mobile application corresponding to the top of the task stack that is currently running in the foreground of the mobile phone through the monitoring process, record the name of the mobile application, and mark the running status of the mobile application as start. The default status of the remaining mobile applications is close.

[0033] Step 102, obtain the basic information of the started application, detect the application category and establish a category label.

[0034] Obtain the basic information of mobile applications with the current running status of "start", including: mobile application name, running status, and application description document. First, perform sentence segmentation on the application description document, splitting the text of the document into multiple sentences. Then, perform word segmentation on the mobile application name and the sentences after splitting the document text to obtain the terms of all text sentences, remove stop words, and label the part-of-speech of each word. Calculate the TF-IDF value for each term, set a first threshold, and select the terms with TF-IDF values greater than the first threshold as the keywords of the application description document. Develop an application classification standard, which includes the category names of various mobile applications and the corresponding mobile application description keywords, denoted as {category name: keyword}. Calculate the word similarity between the keywords of the application description document and the mobile application description keywords given in the classification standard in sequence based on the "Thesaurus", denoted as Sim(A,B). Where A is the keyword extracted from the application description document, and B is the keyword given in the classification standard. Calculate the average of all word similarities, take the mean as the overall similarity, and denote it as W. Use w0 as the similarity threshold, and perform matching based on the overall similarity. When the overall similarity is greater than w0, obtain the category name of the current mobile application, and denote it as {mobile application name, running status, category}.

[0035] Step 103, determine whether to grant the right to synchronize and obtain user data for the intelligent cockpit application based on whether there are similar applications.

[0036] Obtain the basic information of all applications in the smart cockpit, including: the name of the application in the cockpit and the description document of the application in the cockpit. According to the application category monitoring method, firstly, the description document of the application in the cockpit is segmented, the name of the application in the cockpit and the text sentence of the description document of the application in the cockpit are segmented, and the terms of all text sentences after removing the stop words are obtained. Then, the TF-IDF value of each term is calculated, a second threshold is set, and the term with a TF-IDF value greater than the second threshold is selected as the keyword of the description document of the application in the cockpit. Calculate the overall similarity between the keywords of the description document of the application in the cockpit and the application classification standard to classify all applications in the smart cockpit, and establish corresponding category tags and category lists. Take the mobile application with the current running state of start as the target application, and determine whether the category name of the target application is included in the category list of the application in the cockpit. If the category name of the target application is included in the category list of the cockpit application, the smart cockpit application is granted the right to synchronize and obtain user data. If the category name of the target application is not included in the category list of the cockpit application, it means that the target application cannot be synchronized with the application in the cockpit. There are three mobile applications on the user's phone that enter the foreground, namely WeChat, Baidu Maps and NetEase Cloud Music. Among them, WeChat is the application currently used by the user, so it should be recorded as {WeChat, start}, {Baidu Maps, close}, {NetEase Cloud, close}. If the user switches to NetEase Cloud, it should be recorded as {NetEase Cloud, start}, {Baidu Maps, close}, {WeChat, close}.

[0037] Step 104: Obtain user data and distinguish between public data and private data.

[0038] Obtain user data, including user personal information and user operation information in the target application. Preprocess the data and construct a user data table, which includes the following four types of attributes: display identifier, quasi-identifier, sensitive attribute, and non-sensitive attribute. Among them, the display identifier is used to indicate the user's primary identity attribute, and the quasi-identifier is used to indicate the user's secondary identity attribute. The user cannot be located based on the quasi-identifier, but a user can be potentially identified in combination with other attributes; sensitive attributes are user privacy that needs to be protected; non-sensitive attributes are public data and can be published directly. Protect user privacy data based on the K-anonymity method and generalize user data. First, delete the display identifier, iteratively search the original data table, find a quasi-identifier, and increase the number of quasi-identifiers in turn to form a larger subset. Then, perform the first data generalization on the quasi-identifier in the original data table, and perform K-anonymity detection on the generalized user relationship table. If the anonymous table meets K-anonymity under the quasi-identifier attribute, output the data and synchronize the data to the smart cockpit. If it does not meet the requirements, continue to generalize the data until all quasi-identifier subsets meet the K-anonymity standard. For example, WeChat is an application that is currently in the startup state. After obtaining the basic information of WeChat, the WeChat description document is segmented, the WeChat name and document clauses are segmented, the TF-IDF value of each term is calculated, and 0.7 is used as the threshold to extract {communication, chat, call} as the keyword. The mobile application classification standard defines {communication category: communication, email, chat}. The word similarity is calculated to be 1, 1, 0.0.87, then the overall similarity W = 0.96, and the threshold w0 = 0.9, indicating that WeChat belongs to the communication category, recorded as {WeChat, start, communication category}.

[0039] Step 105 , calculating the operation recommendation index according to the application deficiencies and the application activity, and determining the type of operation prompt to be recommended.

[0040] The operation recommendation index Z is calculated by combining the shortcomings of the application and the application activity Q, where the shortcomings of the application are represented by the operation convenience P, and the operation recommendation index Z = (P + Q) / 2. The lower the value of the operation recommendation index Z, the more inconvenient the operation of the mobile application is when driving, and it is not suitable for long-term use. Given a threshold Z0, when the recommendation index Z of the target application is lower than Z0, an operation guidance type with a higher operation convenience P than the current operation convenience P is output and marked with Ⅰ, Ⅱ, and Ⅲ. For example, applications in the smart cockpit can be divided into network, business, communication, game, multimedia, navigation, communication, system, and news reading. The target application and category are {WeChat, start, communication}, which is included in the category list of the cockpit application, and the smart cockpit application should be granted the right to synchronize and obtain user data.

[0041] Detect the content of mobile applications and analyze their shortcomings.

[0042] Monitor the content of mobile phone applications, analyze the main operation methods of the target application, and calculate the operation convenience of the target application. By analyzing the operation logs of the application, monitor the operation content of users in the application to obtain the main operation methods of the target application. According to the three main types of operation methods of the current application: VR operation, multimedia operation, and touch operation, determine the category to which the operation method of the target application belongs. Among them, multimedia operation includes speech, eye contact, expression, lip movement, gesture, and body recognition operations. Denote VR operation as I, multimedia operation as II, and touch operation as III, and assign different scores to operations I to III according to operation convenience. The operation convenience score of the target application is denoted as P. The higher P is, the more convenient the operation method is, and the lower P is, the more room for improvement the operation method has. Display identifiers such as: user ID card, quasi-identifiers such as: postal code, gender, birthday, sensitive attributes such as: diseases, income, etc., and non-sensitive attributes are public data that can be directly published. For example, obtain the user data relationship table {user mobile phone number, application name, start time, start duration}, and the specific values are as follows: {1360000****, WeChat, 2022-05-01 08:00, 30min}, {1360000****, Baidu Map, 2022-05-01 08:30, 17min},

[0043] {1500000****, Eastmoney, 2022-05-01 08:47, 21min}, {1870000****, Tencent Meeting, 2022-05-01 14:00, 120min}. The table output after processing by the K-anonymity method is as follows: {WeChat, 2022-05-01****, (15min - 30min)}, {Baidu Map, 2022-05-01****, (15min - 30min)}, {Eastmoney, 2022-05-01****, (15min - 30min)}, {Tencent Meeting, 2022-05-01****, (115min - 130min)}.

[0044] Obtain the application start data, analyze the application activity and establish labels.

[0045] By applying the startup log to read the startup data of the application within one month, extract the startup time and usage duration of the application. Construct an RFM model to describe the activity of the application. Among them, the recency R represents the number of days passed since the user's last startup of the application until today, the frequency F represents the total number of days the user has started the application within one month, and the intensity M represents the total usage duration of the application by the user within one month divided by the total number of days. Standardize the three types of data of progress R, frequency F, and intensity M, classify the application activity based on K-means clustering, and record the class center value of each class. Calculate the distance between the RFM of the target application and the class center value, classify the application activity and establish corresponding labels, and assign different scores according to the application activity. The score of the target application is represented by Q. The higher the Q value, the higher the time and intensity of the user using the target application, and the higher the risk coefficient of using it while driving. For example, the convenience score P of WeChat is 0, the application activity score Q is 0, and the calculated Z is 0. If Z0 = 1 is given, then Z < Z0. Therefore, in the driving state, the touch control method of WeChat is not suitable for long-term use in the cockpit. At this time, operation guides with higher convenience should be recommended to the user, such as type I VR operations or type II multimedia operations.

[0046] Step 106: Combine the user characteristic information and the operation type to be recommended to construct an operation guide recommendation model.

[0047] Take the category to which the user's target application belongs, the user's main operation method, and the user's RFM model as user characteristic information, and construct a user characteristic matrix. Normalize the user characteristic data, and then classify all users through k-means clustering into different user groups. Then calculate the feature similarity between the current user and other users in the same group, and construct a similar user table based on the feature similarity. Traverse the main operation methods in the similar user table, screen the similar user table according to the operation type to be recommended for the current user, extract the main operation methods included in the screened similar user table, and construct a list of operations to be recommended for the current user, denoted as {user: operation name}. Use the feature similarity as the preference weight of the user for the operation to be recommended, select the top-n from the list of operations to be recommended for the current user and recommend them to the user, and record the operation name and operation type. For example, if WeChat is the target application and it is identified from the operation log that the main operation method of WeChat is touch control, then its operation method is marked as III. Since the operation convenience is I > II > III, scores of 2, 1, and 0 are assigned to operations I to III respectively, so the score of WeChat P = 0.

[0048] Step 107: Detect the user behavior, analyze the behavior of the current user not operating according to the prompt, and adjust the operation guide recommendation model of the current user.

[0049] After the intelligent cockpit recommendation operation guide, a monitoring process is created to record the operation method selected by the current user. Obtain the name, operation type, and operation result of each step of the current user's operation, determine whether the current user's operation is consistent with the operation guide recommended by the cockpit, and use 1 and 0 as identifiers for the user operation result. If the user operation is consistent with the operation guide, the operation result is 1, and the operation guide is continuously pushed until the guide content ends, and it is marked as (operation name, operation type, 1). If the user operation is inconsistent with the operation guide, the operation result is 0, record the current operation guide, and mark it as (operation name, operation type, 0). If the user does not operate according to the prompt, adjust the operation guide recommendation model for the current user. After constructing the list of operations to be recommended for the current user, match the most recent operation result of the user according to the operation name, delete the data with an operation result of 0 from the list of operations to be recommended, and then select the top-n from the list of operations to be recommended for the current user and recommend them to the user. For example, the target application is WeChat, and RFM = (1, 28, 0.1875). After standardization, we get

[0050] RFM = (0.745136, 1.287966, 1.260569). According to the application activity, the applications are divided into three categories: I. Not used for a long time but not frequently, with a short usage time; II. Recently used but not frequently, with a short usage time; III. Recently used and frequently, with a long usage time. The central values are I(0.4415674, -0.12106, -0.35449), II(-1.4387213, -1.13178, -0.63325), III(0.746029, 1.297896, 1.262598). It is calculated that the RFM of WeChat is the closest to III. Therefore, WeChat belongs to the category of applications that are recently used and frequently, with a long usage time, and the risk coefficient of using it for a long time while driving is relatively high. Since the application activity I < II < III, scores 2, 1, and 0 are assigned to the active types respectively. So the score Q of WeChat is 0.

[0051] Step 108: Analyze the operation data of all users after the recommendation guide to generate an improvement plan for cockpit applications.

[0052] Collect all the operation data of users after the intelligent cockpit recommendation guide, including the target application name, the category to which the target application belongs, the recommended operation name, the operation type, and the operation result. First, perform data preprocessing on the user operation data to delete the data with empty operation results. Then, count the total number of times each operation is recommended and the number of times the operation result is 1. The user acceptance probability = (the number of times the operation result is 1) / (the total number of times recommended). Calculate the average value of the user acceptance probabilities of each operation with the same operation type. If the user acceptance probability of the target operation is higher than the average value, it indicates that the target operation conforms to the user's operation habits. If the user acceptance probability of the target operation is lower than the average value of the user acceptance probabilities of each operation with the same operation type, it indicates that the usage method of the target operation in the target application does not conform to the user's operation preferences, and the operation method of similar applications in the cockpit needs to be improved. For example, the current user is User 1, and the operation types to be recommended are I or II. The list of operations to be recommended obtained after screening is {User 2: Speech recognition, User 3: Gesture recognition, User 4: VR operation}. Among them, speech recognition and gesture recognition belong to type II multimedia technology. The feature similarities between User 1 and Users 2, 3, and 4 are 0.6, 0.9, and 0.3 respectively. Select the top 2 with the highest similarity and recommend them to the user, denoted as {Gesture recognition, II} {Speech recognition, II}.

Claims

1. A method for application synchronization and reminder in an intelligent cockpit, characterized in that, the method includes: Listening for whether a mobile application is started; obtaining the basic information of the started application, detecting the application category and establishing a category label; judging whether to grant the intelligent cockpit the right to synchronize applications and obtain user data according to whether there are similar applications; obtaining user data and distinguishing between public data and private data; calculating an operation recommendation index based on the application deficiencies and application activity, and determining the type of operation prompt to be recommended, specifically including: detecting the content of the mobile application, analyzing the application deficiencies, obtaining the application startup data, analyzing the application activity and establishing a label; combining the user characteristic information and the type of operation to be recommended to construct an operation guide recommendation model; detecting the user behavior, analyzing the behavior of the current user not operating according to the prompt, and adjusting the operation guide recommendation model of the current user; analyzing the operation data of all users after the recommendation guide to generate an improvement plan for the cockpit application.

2. The method according to claim 1, wherein, the listening for whether a mobile application is started includes: Creating a listening process for all mobile applications that enter the foreground on the user's mobile phone and identifying the application running status; the running status of the application is divided into a startup status and a shutdown status. When the mobile application is at the top of the task stack, it means the task is in the startup status. If the mobile application is not at the top of the task stack, it means the task is in the shutdown status; when the mobile application is in the startup status, it is identified by start, and when the mobile application is in the shutdown status, it is identified by close. Query the running status of the mobile application every 500 milliseconds and record the listening result; Obtaining the mobile application corresponding to the top of the task stack of the mobile phone foreground currently running through the listening process, recording the name of the mobile application, and marking the running status of the mobile application as start, and the default status of the remaining mobile applications is close.

3. The method according to claim 2, wherein, the obtaining the basic information of the started application, detecting the application category and establishing a category label includes: Obtain basic information of a mobile application whose current running state is start, including: mobile application name, running state, and application description document; first, segment the application description document into multiple sentences; then, perform word segmentation on the sentences after the mobile application name and the document text are segmented, obtain the terms of all text sentences, remove stop words, and mark the part of speech of each word; calculate the TF-IDF value for each term, set a first threshold, and select terms with a TF-IDF value greater than the first threshold as application description document keywords; formulate application classification standards, including the category names of various mobile application programs and the corresponding The mobile application description keywords are recorded as "category name: keywords"; based on the "Synonym Dictionary", the word similarity between the keywords of the application description document and the mobile application description keywords given by the classification standard is calculated in sequence, which is recorded as Sim(A,B); wherein A is the keyword extracted from the application description document, and B is the keyword given by the classification standard; all word similarities are averaged, and the mean is taken as the overall similarity, which is recorded as W; w0 is used as the similarity threshold, and matching is performed according to the overall similarity. When the overall similarity is greater than w0, the category name of the current mobile application is obtained, and it is recorded as "mobile application name, running status, and category".

4. The method according to claim 3, in, The determination of whether to grant the smart cockpit application the right to synchronize and obtain user data based on whether similar applications are included includes: The basic information of all applications in the smart cockpit is obtained, including: the name of the application in the cockpit and the description document of the application in the cockpit; according to the application category detection method, the description document of the application in the cockpit is firstly segmented, the name of the application in the cockpit and the text sentence of the description document of the application in the cockpit are segmented, and the terms of all text sentences after removing stop words are obtained; then the TF-IDF value of each term is calculated, a second threshold is set, and the term with a TF-IDF value greater than the second threshold is selected as the keyword of the description document of the application in the cockpit; the overall similarity between the keywords of the description document of the application in the cockpit and the application classification standard is calculated to classify all applications in the smart cockpit, and corresponding category labels and category lists are established; the mobile application with the current running state of start is taken as the target application, and it is determined whether the category name of the target application is included in the category list of the application in the cockpit; if the category name of the target application is included in the category list of the cockpit application, the smart cockpit application is granted the right to synchronize and obtain user data; if the category name of the target application is not included in the category list of the cockpit application, it means that the target application cannot be synchronized with the application in the cockpit.

5. The method according to claim 4, in, The obtaining of user data and distinguishing between public data and private data includes: Obtain user data, including user personal information and user operation information in the target application; Preprocess the data to construct a user data table, which includes the following four types of attributes: display identifier, quasi-identifier, sensitive attribute, and non-sensitive attribute; among them, the display identifier is used to indicate the main identity attribute of the user, and the quasi-identifier is used to indicate the secondary identity attribute of the user; the user cannot be located based on the quasi-identifier, but a certain user can be potentially identified in combination with other attributes. The sensitive attribute is the user privacy that needs to be protected; the non-sensitive attribute is public data and can be directly published; the K-anonymity method is used to protect the user privacy data and perform generalization operations on the user data: First, delete the display identifier, iteratively search the original data table to find a quasi-identifier, and sequentially increase the number of quasi-identifiers to form a larger subset; then, perform the first data generalization on the quasi-identifiers in the original data table, and perform K-anonymity detection on the generalized user relationship table. If the anonymized table meets K-anonymity under the quasi-identifier attribute, output the data and synchronize the data to the intelligent cockpit. If it does not meet the requirements, continue to generalize the data until all quasi-identifier subsets meet the K-anonymity standard.

6. The method according to claim 5, wherein, calculating an operation recommendation index according to the application deficiencies and application activity, and determining the type of operation prompt to be recommended, including: Combine the application deficiencies and the application activity Q to calculate the operation recommendation index Z, where the application deficiencies are represented by the operation convenience P, and the operation recommendation index The lower the value of the operation recommendation index Z, the more inconvenient the operation method of the mobile application is in the driving state and it is not suitable for long-term use; given a threshold Z0, when the recommendation index Z of the target application is lower than Z0, output an operation guidance type with a higher operation convenience P than the current one, and label it with Ⅰ, Ⅱ, Ⅲ; detecting the content of the mobile application and analyzing the deficiencies of the application, specifically including: monitoring the content of the mobile application, analyzing the main operation methods of the target application, and calculating the operation convenience of the target application; by analyzing the operation logs of the application, monitoring the operation content of the user in the application, and obtaining the main operation methods of the target application; according to the three main operation methods of the current application: VR operation, multimedia operation, and touch operation, determining the category to which the operation method of the target application belongs; among them, the multimedia operation includes speech, eye contact, expression, lip movement, gesture, and body recognition operations; record VR operation as Ⅰ, multimedia operation as Ⅱ, and touch operation as Ⅲ, and assign different scores to operations Ⅰ to Ⅲ according to the operation convenience. The operation convenience score of the target application is represented by P; the higher the P value, the more convenient the operation method, and the lower the P value, the more room for improvement in the operation method. obtaining the application start data, analyzing the application activity and establishing labels, specifically including: reading the start data of the application within one month through the application start log, and extracting the start time and usage duration of the application; constructing an RFM model to describe the application activity, where the recency R represents the number of days elapsed since the user last started the application until today, the frequency F represents the total number of days the user started the application within one month, and the intensity M represents the total usage duration of the application by the user within one month / the total number of days; standardize the three types of data of recency R, frequency F, and intensity M, classify the application activity based on K-means clustering, and record the class center value of each class; calculate the distance between the RFM of the target application and the class center value of each class, classify the application activity and establish corresponding labels, and assign different scores according to the application activity. The score of the target application is represented by Q; the higher the Q value, the higher the time and intensity of the user using the target application, and the higher the risk coefficient when using it while driving.

7. The method according to claim 6, wherein, the construction of the operation guidance recommendation model by combining user characteristic information and the operation type to be recommended includes: taking the category to which the target application of the user belongs, the main operation mode of the user, and the user RFM model as user characteristic information, constructing a user characteristic matrix; performing normalization processing on the user characteristic data, and then classifying all users through k-means clustering into different user groups; then calculating the characteristic similarity between the current user and other users in the same group, and constructing a similar user table based on the characteristic similarity; traversing the main operation modes in the similar user table, screening the similar user table according to the operation type to be recommended for the current user, extracting the main operation modes included in the screened similar user table, and constructing a list of operations to be recommended for the current user, denoted as "user: operation name"; using the characteristic similarity as the preference weight of the user for the operation to be recommended, selecting the top-n from the list of operations to be recommended for the current user and recommending them to the user, and recording the operation name and operation type.

8. The method according to claim 7, wherein, the detection of the user behavior, the analysis of the behavior that the current user does not operate according to the prompt, and the adjustment of the operation guidance recommendation model of the current user include: after the intelligent cockpit recommends the operation guidance, creating a monitoring process to record the operation mode selected by the current user; obtaining the name, operation type, and operation result of each step of the current user's operation, judging whether the operation of the current user is consistent with the operation guidance recommended by the cockpit, and using 1 and 0 as the identifiers of the user operation result; if the user operation is consistent with the operation guidance, the operation result is 1, and the operation guidance is continuously pushed until the guidance content ends, and it is marked as "operation name, operation type, 1"; if the user operation is inconsistent with the operation guidance, the operation result is 0, recording the current operation guidance, and marking it as "operation name, operation type, 0"; if the user does not operate according to the prompt, adjusting the operation guidance recommendation model of the current user; after constructing the list of operations to be recommended for the current user, matching the most recent operation result of the user according to the operation name, deleting the data with the operation result of 0 from the list of operations to be recommended, and then selecting the top-n from the list of operations to be recommended for the current user and recommending them to the user.

9. The method according to claim 8, wherein, the analysis of the operation data of all users after the recommendation guidance to generate an improvement plan for the cockpit application includes: Collect all the operation data of users after the intelligent cockpit recommendation guidelines, including the target application name, the category to which the target application belongs, the recommended operation name, the operation type, and the operation result. First, perform data preprocessing on the user operation data to delete the data with empty operation results. Then, count the total number of times each operation is recommended and the number of times the operation result is 1, and calculate the average value of the user acceptance probability of each operation with the same operation type. If the user acceptance probability of the target operation is higher than the average value, it indicates that the target operation conforms to the user's operation habits. If the user acceptance probability of the target operation is lower than the average value of the user acceptance probability of each operation with the same operation type, it indicates that the usage method of the target operation in the target application does not conform to the user's operation preferences, and it is necessary to improve the operation method of similar applications in the cockpit.

Images