Charging value image generation method and device, electronic equipment and readable storage medium

Abstract

The application discloses a charging value portrait generation method and device, electronic equipment and a readable storage medium, relates to the technical field of big data mining, and attributes of original charging data are constructed, and classification is performed according to charging behavior fluctuation data, finally, a charging value portrait is constructed according to clustering portrait rules, users are divided into several typical groups with different development values, user behavior characteristics can be accurately mined, portrait dimensions are expanded, and portrait accuracy is improved. The method comprises the following steps: in response to a charging value portrait generation request, obtaining user data of a user to be portraited, processing the user data by using a preset data analysis model to obtain target behavior data; obtaining a threshold classifier and a preset fluctuation threshold, and reading charging behavior fluctuation data in the target behavior data; if the charging behavior fluctuation data is greater than or equal to the preset fluctuation threshold, performing a portrait operation on the target behavior data by using a target clustering classification standard to obtain a charging value portrait.

Description

Technical Field

[0001] The present application relates to the field of big data mining technology, and in particular to a method, device, electronic device and readable storage medium for generating a charging value portrait. Background Art

[0002] With the rapid development of big data mining technology, the scale of new energy vehicles is constantly expanding, leading to an increasing number of electric vehicles (EVs) and charging behaviors. Since the massive number of EVs connected to the power system introduces greater uncertainty, utilizing new energy vehicle big data mining technology to study user charging behavior data and uncover the underlying patterns can help manage and optimize the distribution network. User profiles, derived from studying user behavior, are tagged user models abstracted from data that describe user attributes, preferences, and behavioral habits. Through user profiles, companies can gain a deeper understanding of each user's needs and develop personalized services, such as a charging value profile for car users.

[0003] In the related art, charging value profiles in the new energy vehicle field are generally generated based on big data or questionnaires. Among them, the charging value profile generation method based on big data mainly classifies users according to hard clustering based on partitioning. Usually, multi-category attributes are selected to synthesize multi-dimensional data for clustering, and user clusters and the number of clusters are obtained to generate charging value profiles. However, the applicant has recognized that the charging value profile generation method based on big data is difficult to be accurate to a single user. As a result, the historical behavior data of a single user may appear in multiple user clusters, resulting in low accuracy in mining user behavior characteristics, resulting in insufficient depth and low dimensionality of the generated charging value profile. Summary of the Invention

[0004] In view of this, the present application provides a charging value portrait generation method, device, electronic device and readable storage medium. The main purpose is to solve the problem that the charging value portrait generation method based on big data is difficult to be accurate to a single user, so that the historical behavior data of a single user may appear in multiple user clusters, resulting in low accuracy in mining user behavior characteristics, and causing the generated charging value portrait to be not deep enough and not high in dimensionality.

[0005] According to the first aspect of the present application, a method for generating a charging value profile is provided, the method comprising:

[0006] In response to a request to generate a charging value profile, user data of the user to be profiled is obtained, and the user data is processed using a preset data analysis model to obtain target behavior data;

[0007] Obtaining a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and reading charging behavior volatility data from the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled;

[0008] If the charging behavior volatility data is greater than or equal to the preset volatility threshold, a target clustering classification standard is obtained, and a profiling operation is performed on the target behavior data using the target clustering classification standard to obtain a charging value profile of the user to be profiled.

[0009] Optionally, in response to the charging value profile generation request, obtaining user data of the user to be profiled includes:

[0010] Obtaining the user identifier carried in the charging value profile generation request, taking the user indicated by the user identifier as the user to be profiled, and querying the database for original user data indicated by the user identifier;

[0011] Determining abnormal data, duplicate data, and null value data in the original user data, and deleting the abnormal data, the duplicate data, and the null value data from the original user data, wherein the abnormal data includes hardware abnormal data, software abnormal data, and operation abnormal data;

[0012] Obtaining a preset data format, and adjusting the deleted original user data using the preset data format to obtain first preprocessed data;

[0013] Obtaining a preset time format, and adjusting the first preprocessed data using the preset time format to obtain second preprocessed data;

[0014] Identifying the user name of the user to be profiled in the second pre-processed data, and treating the user information as abnormal user information when the user name is zero or empty;

[0015] Vehicle information is obtained from the second pre-processed data, the abnormal user information is replaced with the vehicle information, and the replaced second pre-processed data is used as the user data.

[0016] Optionally, the processing of the user data using a preset data analysis model to obtain target behavior data includes:

[0017] Obtaining multiple attribute calculation formulas included in the preset data analysis model, the multiple attribute calculation formulas including a charging proximity calculation formula, a total charging frequency calculation formula, a total charging amount calculation formula, a user life cycle calculation formula, and a charging behavior volatility calculation formula;

[0018] Obtain the data statistics cutoff time T of the user to be profiled in the user data now , the last charging time T of the user to be profiled rec , and the data statistics cutoff time T of the user to be profiled using the charging proximity calculation formula now , the last charging time T of the user to be profiled rec Calculate and obtain the charging proximity data R of the user to be profiled, where:

[0019] R=T now -T rec

[0020] Wherein, R represents the charging proximity data;

[0021] Obtain the number of times q the user to be profiled charges on day i from the user data i , and the total charging frequency calculation formula is used to calculate the number of charging times q of the user to be profiled on the i-th day i Calculate and obtain the total charging frequency data F of the user to be profiled, where:

[0022]

[0023] Wherein, F represents the total charging frequency data of the user to be profiled within the data statistical time interval, n is a positive integer, and i is a positive integer;

[0024] Obtain the i-th charging power e of the user to be profiled from the user data i , and the charging amount e of the i-th user to be profiled is calculated using the total charging amount calculation formula i Calculate and obtain the total charging amount data E of the user to be profiled, where:

[0025]

[0026] Wherein, E represents the total charging amount data of the user to be profiled within the data statistical time interval, n is a positive integer, and i is a positive integer;

[0027] Obtain the last charging time T of the user to be profiled from the user data now , the first charging time T of the user to be profiled first , and the last charging time T of the user to be profiled using the user life cycle calculation formula now , the first charging time T of the user to be profiled first Calculate and obtain the user life cycle data L of the user to be profiled, where:

[0028] L=Tnow -T first

[0029] Wherein, L represents the user lifecycle data of the user to be profiled;

[0030] Obtain the time interval t between the nth charging and the n+1th charging of the user to be profiled in the user data gap , and using the charging behavior volatility calculation formula to calculate multiple time intervals T gap Calculation is performed to obtain the charging behavior volatility data V of the user to be profiled, where:

[0031]

[0032] Wherein, V represents the charging behavior volatility data of the user to be profiled, T gap represents the sample set of n time intervals of the user to be profiled within the data statistical time interval, Mean(T gap ) represents the sample set T gap The sample mean of , n is a positive integer, i is a positive integer;

[0033] The charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data corresponding to the user to be profiled are sorted to obtain the target behavior data.

[0034] Optionally, after obtaining a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and reading charging behavior volatility data from the target behavior data, the method further includes:

[0035] If the charging behavior volatility data is less than the preset volatility threshold, then reading the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data from the target behavior data;

[0036] Obtaining a preset classification standard corresponding to the threshold classifier, wherein the preset classification standard includes a plurality of preset classification results;

[0037] Reading a preset proximity range and a preset total frequency range corresponding to each of the preset classification results in the threshold classifier, and determining a target classification result corresponding to the user to be profiled from the multiple preset classification results, wherein the charging proximity data is within the preset proximity range corresponding to the target classification result and the total charging frequency data is within the preset total frequency range corresponding to the target classification result;

[0038] Obtaining the portrait label corresponding to the target classification result, and using the portrait label as the target portrait label corresponding to the user to be profiled;

[0039] The target portrait label, the target classification result, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to generate a charging value portrait of the user to be profiled.

[0040] Optionally, performing a profiling operation on the target behavior data using the target clustering classification standard to obtain a charging value profile includes:

[0041] Obtain multiple portrait categories included in the target cluster classification standard, and read the charging proximity standard value, total charging frequency standard value, total charging amount standard value, user life cycle standard value, and charging behavior volatility standard value corresponding to each portrait category in the target cluster classification standard;

[0042] Obtaining a preset dimensional data point conversion rule, and using the preset dimensional data point conversion rule to process the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category to obtain a standard center point corresponding to each portrait category;

[0043] Reading charging proximity data, total charging frequency data, total charging amount data, and user life cycle data from the target behavior data;

[0044] The charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are processed using the preset dimensional data point conversion rules to obtain a data point to be profiled corresponding to the user to be profiled;

[0045] Calculating the Euclidean distance between the center point of the image to be processed and the standard center point corresponding to each image category to obtain multiple Euclidean distance values, and selecting the Euclidean distance value with the smallest distance value from the multiple Euclidean distance values ​​as the target Euclidean distance value;

[0046] Determining a designated standard center point indicated by the target Euclidean distance value among a plurality of standard center points corresponding to the plurality of portrait categories, and determining the portrait category corresponding to the designated standard center point as the target portrait category of the user to be portraited;

[0047] The target portrait category, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to obtain a charging value portrait of the user to be profiled.

[0048] Optionally, the method further includes:

[0049] Acquire a user data sample, perform normalization calculation on the user data sample to obtain a data sample to be clustered, obtain a preset cluster number range, and extract a plurality of preset cluster numbers within the preset cluster number range;

[0050] Obtaining a silhouette coefficient, and selecting a preset number of clusters from the plurality of preset numbers of clusters using the silhouette coefficient as a target cluster number;

[0051] The target number of clusters is used as the target number of calculations for the target cluster center operation;

[0052] Obtaining a distance probability formula, calculating target cluster centers of the data samples to be clustered using the distance probability formula, and repeatedly calculating the target cluster centers of the data samples to be clustered until the number of calculations reaches the target number of calculations, thereby obtaining a plurality of target cluster centers whose number is equal to the target number of calculations;

[0053] Creating a first blank cluster for each target cluster center to obtain multiple first blank clusters, obtaining multiple data sample points in the data sample to be clustered, and adding the multiple data sample points to the multiple first blank clusters respectively to obtain multiple target clusters;

[0054] For each target cluster, determining the coordinates of a plurality of data sample points included in the target cluster to obtain a plurality of first data point coordinates, averaging the plurality of first data point coordinates to obtain a first average coordinate point, and using the first average coordinate point as a first cluster center corresponding to the target cluster;

[0055] Processing the coordinates of the multiple first data points of each target cluster respectively to obtain a first cluster center corresponding to each target cluster, and obtaining a convergence condition, and processing the first cluster center corresponding to each target cluster using the convergence condition to obtain multiple target clustering results;

[0056] Acquire preset portrait classification rules, use the preset portrait classification rules to process the multiple target clustering results to obtain multiple portrait categories, and use the multiple portrait categories to generate the target clustering classification standard.

[0057] Optionally, the selecting a preset number of clusters from the plurality of preset numbers of clusters using the silhouette coefficient as the target number of clusters includes:

[0058] For each of the plurality of preset cluster numbers, a mean clustering algorithm is obtained, and the mean clustering algorithm is used to calculate the data sample to be clustered and the preset number of clusters to obtain a predicted clustering result;

[0059] Acquire multiple predicted clustering results corresponding to multiple preset cluster numbers, and respectively evaluate and calculate the multiple predicted clustering results using the silhouette coefficient to obtain multiple evaluation coefficients corresponding to the multiple predicted clustering results;

[0060] The multiple evaluation coefficients are arranged in descending order, the evaluation coefficient ranked first is used as the target evaluation coefficient, and the preset cluster number corresponding to the target evaluation coefficient is determined from the multiple preset cluster numbers, and the preset cluster number is used as the target cluster number.

[0061] Optionally, the calculating the target cluster center of the data samples to be clustered by using the distance probability formula includes:

[0062] Selecting any data sample point in the data samples to be clustered as an initial cluster center, and determining other data sample points in the data samples to be clustered except the initial cluster center as a plurality of first sample points;

[0063] Calculating the distance between the initial cluster center and each of the plurality of first sample points to obtain a plurality of sample distance values, and calculating the plurality of sample distance values ​​using the distance probability formula to obtain a plurality of selection probabilities;

[0064] A roulette wheel selection algorithm is obtained, and a selection probability is selected from the multiple selection probabilities using the roulette wheel selection algorithm as a target selection probability, and a first sample point corresponding to the target selection probability is used as a target cluster center.

[0065] Optionally, the adding the plurality of data sample points to the plurality of first blank clusters to obtain a plurality of target clusters includes:

[0066] Calculating the Euclidean distance between each of the data sample points and the multiple target cluster centers respectively to obtain multiple first Euclidean distance values ​​corresponding to each of the data sample points;

[0067] Selecting a first Euclidean distance value with the smallest distance value from the multiple first Euclidean distance values ​​corresponding to each data sample point as a target first Euclidean distance value corresponding to each data sample point;

[0068] For each data sample point, determining a target cluster center corresponding to the target first Euclidean distance value according to the target first Euclidean distance value corresponding to the data sample point, and adding the data sample point to the first blank cluster corresponding to the target cluster center;

[0069] According to the target first Euclidean distance value corresponding to each data sample point, the multiple data sample points are respectively added to the multiple first blank clusters to obtain the multiple target clusters.

[0070] Optionally, the using the convergence condition to process the first cluster center corresponding to each target cluster to obtain multiple target clustering results includes:

[0071] Performing convergence calculation on the first cluster center corresponding to each target cluster to obtain multiple convergence results, and comparing the multiple convergence results with the convergence condition;

[0072] If a convergence result among the multiple convergence results does not meet the convergence condition, obtaining multiple data sample points in the data sample to be clustered, and calculating the Euclidean distance between each data sample point and the multiple first cluster centers respectively, to obtain multiple second Euclidean distance values ​​corresponding to each data sample point;

[0073] Selecting the second Euclidean distance value with the smallest distance value from the multiple second Euclidean distance values ​​corresponding to each data sample point as the target second Euclidean distance value corresponding to each data sample point;

[0074] Creating a second blank cluster for each of the first cluster centers to obtain a plurality of second blank clusters;

[0075] For each data sample point, determining a first cluster center corresponding to the target second Euclidean distance value according to the target second Euclidean distance value corresponding to the data sample point, and adding the data sample point to a second blank cluster corresponding to the first cluster center;

[0076] According to the target second Euclidean distance value corresponding to each data sample point, the multiple data sample points are added to the multiple second blank clusters to obtain multiple designated clusters;

[0077] For each of the designated clusters, read coordinates of multiple second data points in the designated cluster, average the coordinates of the multiple second data points to obtain a second average coordinate point, and use the second average coordinate point as a second cluster center corresponding to the designated cluster;

[0078] Processing the coordinates of the multiple second data points of each designated cluster respectively to obtain a second cluster center corresponding to each designated cluster, generating multiple clustering results using the multiple second cluster centers, and using the multiple clustering results as multiple target clustering results;

[0079] If the multiple convergence results meet the convergence condition, multiple first clustering centers are used to generate multiple target clustering results.

[0080] According to a second aspect of the present application, a device for generating a charging value portrait is provided, the device comprising:

[0081] a processing module, configured to respond to a request for generating a charging value profile, obtain user data of the user to be profiled, and process the user data using a preset data analysis model to obtain target behavior data;

[0082] a reading module, configured to obtain a threshold classifier and a preset fluctuation threshold value corresponding to the threshold classifier, and read charging behavior volatility data from the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled;

[0083] The first portrait module is used to obtain a target cluster classification standard if the charging behavior volatility data is greater than or equal to the preset fluctuation threshold, and use the target cluster classification standard to perform a portrait operation on the target behavior data to obtain a charging value portrait of the user to be portraited.

[0084] Optionally, the processing module is used to obtain the user identifier carried in the charging value portrait generation request, take the user indicated by the user identifier as the user to be profiled, and query the original user data indicated by the user identifier in the database; determine abnormal data, duplicate data, and null value data in the original user data, and delete the abnormal data, the duplicate data, and the null value data in the original user data, and the abnormal data includes hardware abnormal data, software abnormal data, and operation abnormal data; obtain a preset data format, and use the preset data format to adjust the deleted original user data to obtain first preprocessed data; obtain a preset time format, and use the preset time format to adjust the first preprocessed data to obtain second preprocessed data; identify the user name of the user to be profiled in the second preprocessed data, and when the user name is zero or empty, use the user information as abnormal user information; obtain vehicle information in the second preprocessed data, use the vehicle information to replace the abnormal user information, and use the replaced second preprocessed data as the user data.

[0085] Optionally, the processing module is further configured to obtain a plurality of attribute calculation formulas included in the preset data analysis model, wherein the plurality of attribute calculation formulas include a charging proximity calculation formula, a charging total frequency calculation formula, a charging total amount calculation formula, a user life cycle calculation formula, and a charging behavior volatility calculation formula; and obtain the data statistics cutoff time T of the user to be profiled from the user data. now , the last charging time T of the user to be profiled rec , and the data statistics cutoff time T of the user to be profiled using the charging proximity calculation formula now , the last charging time T of the user to be profiled recCalculate and obtain the charging proximity data R of the user to be profiled, where:

[0086] R=T now -T rec

[0087] Wherein, R represents the charging proximity data; the number of charging times q of the user to be profiled on the i-th day is obtained from the user data. i , and the total charging frequency calculation formula is used to calculate the number of charging times q of the user to be profiled on the i-th day i Calculate and obtain the total charging frequency data F of the user to be profiled, where:

[0088]

[0089] Wherein, F represents the total frequency of charging of the user to be profiled within the data statistical time interval, n is a positive integer, i is a positive integer; the charging power e of the user to be profiled for the i-th time is obtained from the user data. i , and the charging amount e of the i-th user to be profiled is calculated using the total charging amount calculation formula i Calculate and obtain the total charging amount data E of the user to be profiled, where:

[0090]

[0091] Wherein, E represents the total amount of charging data of the user to be profiled within the data statistical time interval, n is a positive integer, i is a positive integer; the last charging time T of the user to be profiled is obtained from the user data now , the first charging time T of the user to be profiled first , and the last charging time T of the user to be profiled using the user life cycle calculation formula now , the first charging time T of the user to be profiled first Calculate and obtain the user life cycle data L of the user to be profiled, where:

[0092] L=T now -T first

[0093] Wherein, L represents the user life cycle data of the user to be profiled; the time interval t between the nth charging and the n+1th charging of the user to be profiled is obtained from the user data. gap , and using the charging behavior volatility calculation formula to calculate multiple time intervals t gap Calculation is performed to obtain the charging behavior volatility data V of the user to be profiled, where:

[0094]

[0095] Wherein, V represents the charging behavior volatility data of the user to be profiled, T gap represents the sample set of n time intervals of the user to be profiled within the data statistical time interval, Mean(T gap ) represents the sample set T gap The sample mean of , n is a positive integer, i is a positive integer; the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data corresponding to the user to be profiled are sorted to obtain the target behavior data.

[0096] Optionally, the first portrait module is used to obtain multiple portrait categories included in the target cluster classification standard, and read the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category in the target cluster classification standard; obtain the preset dimension data point conversion rule, and use the preset dimension data point conversion rule to process the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category to obtain the standard center point corresponding to each portrait category; read the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data in the target behavior data; use the preset dimension data point conversion rule to process the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data The frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are processed to obtain the data center point to be profiled corresponding to the user to be profiled; the Euclidean distance between the data center point to be profiled and the standard center point corresponding to each portrait category is calculated to obtain multiple Euclidean distance values, and the Euclidean distance value with the smallest distance value is selected from the multiple Euclidean distance values ​​as the target Euclidean distance value; the designated standard center point indicated by the target Euclidean distance value is determined among the multiple standard center points corresponding to the multiple portrait categories, and the portrait category corresponding to the designated standard center point is determined as the target portrait category of the user to be profiled; the target portrait category, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to obtain a charging value portrait of the user to be profiled.

[0097] Optionally, the device further comprises:

[0098] The second portrait module is used to read the charging proximity data, total charging frequency data, total charging amount data, and user life cycle data in the target behavior data if the charging behavior volatility data is less than the preset fluctuation threshold; obtain the preset classification standard corresponding to the threshold classifier, and the preset classification standard includes multiple preset classification results; read the preset proximity range and preset total frequency range corresponding to each of the preset classification results in the threshold classifier, and determine the target classification result corresponding to the user to be profiled from the multiple preset classification results, wherein the charging proximity data is within the preset proximity range corresponding to the target classification result and the total charging frequency data is within the preset total frequency range corresponding to the target classification result; obtain the portrait label corresponding to the target classification result, and use the portrait label as the target portrait label corresponding to the user to be profiled; organize the target portrait label, the target classification result, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data to generate a charging value portrait of the user to be profiled.

[0099] The generating module is configured to obtain a user data sample, perform normalization calculation on the user data sample to obtain a data sample to be clustered, and obtain a preset cluster number range and extract a plurality of preset cluster numbers in the preset cluster number range; obtain a silhouette coefficient, select a preset cluster number as a target clustering cluster number from the plurality of preset cluster numbers by using the silhouette coefficient; take a value of the target clustering cluster number as a target calculation number of a target clustering center calculation operation; obtain a distance probability formula, calculate a target clustering center of the data sample to be clustered by using the distance probability formula, and repeatedly calculate the target clustering center of the data sample to be clustered until a calculation number reaches the target calculation number, to obtain a plurality of target clustering centers with a number equal to the target calculation number; create a first blank class cluster for each target clustering center to obtain a plurality of first blank class clusters, obtain a plurality of data sample points in the data sample to be clustered, and add the plurality of data sample points to the plurality of first blank class clusters respectively to obtain a plurality of target class clusters; for each target class cluster, determine coordinates of a plurality of data sample points included in the target class cluster to obtain a plurality of first data point coordinates, perform average calculation on the plurality of first data point coordinates to obtain a first average coordinate point, and take the first average coordinate point as a first clustering center corresponding to the target class cluster; process the plurality of first data point coordinates of each target class cluster respectively to obtain a first clustering center corresponding to each target class cluster, obtain a convergence condition, process the first clustering center corresponding to each target class cluster by using the convergence condition, and obtain a plurality of target clustering results; obtain a preset portrait classification rule, process the plurality of target clustering results by using the preset portrait classification rule to obtain a plurality of portrait categories, and generate the target clustering classification standard by using the plurality of portrait categories.

[0100] Optionally, the generating module is configured to, for each preset cluster number in the plurality of preset cluster numbers, obtain a mean clustering algorithm, calculate the data sample to be clustered and the preset cluster number by using the mean clustering algorithm to obtain a predicted clustering result; obtain a plurality of predicted clustering results corresponding to the plurality of preset cluster numbers, and respectively evaluate and calculate the plurality of predicted clustering results by using the silhouette coefficient to obtain a plurality of evaluation coefficients corresponding to the plurality of predicted clustering results; arrange the plurality of evaluation coefficients in descending order, take an evaluation coefficient ranked first as a target evaluation coefficient, and determine a preset cluster number corresponding to the target evaluation coefficient in the plurality of preset cluster numbers, and take the preset cluster number as the target clustering cluster number.

[0101] Optionally, the generation module is used to select any data sample point in the data sample to be clustered as an initial cluster center, determine other data sample points other than the initial cluster center in the data sample to be clustered as multiple first sample points; calculate the distance between the initial cluster center and each first sample point in the multiple first sample points to obtain multiple sample distance values, and use the distance probability formula to calculate the multiple sample distance values ​​to obtain multiple selection probabilities; obtain a roulette wheel selection algorithm, use the roulette wheel selection algorithm to select a selection probability from the multiple selection probabilities as a target selection probability, and use the first sample point corresponding to the target selection probability as a target cluster center.

[0102] Optionally, the generation module is used to calculate the Euclidean distance between each of the data sample points and the multiple target cluster centers respectively to obtain multiple first Euclidean distance values ​​corresponding to each data sample point; select the first Euclidean distance value with the smallest distance value from the multiple first Euclidean distance values ​​corresponding to each data sample point as the target first Euclidean distance value corresponding to each data sample point; for each data sample point, determine the target cluster center corresponding to the target first Euclidean distance value according to the target first Euclidean distance value corresponding to the data sample point, and add the data sample point to the first blank cluster corresponding to the target cluster center; according to the target first Euclidean distance value corresponding to each data sample point, add the multiple data sample points to the multiple first blank clusters respectively to obtain the multiple target clusters.

[0103] Optionally, the generation module is further configured to perform convergence calculation on the first cluster center corresponding to each target cluster to obtain multiple convergence results, and compare the multiple convergence results with the convergence condition; if a convergence result among the multiple convergence results does not meet the convergence condition, obtain multiple data sample points in the data sample to be clustered, calculate the Euclidean distance between each data sample point and the multiple first cluster centers respectively, and obtain multiple second Euclidean distance values ​​corresponding to each data sample point; select the second Euclidean distance value with the smallest distance value from the multiple second Euclidean distance values ​​corresponding to each data sample point as the target second Euclidean distance value corresponding to each data sample point; create a second blank cluster for each first cluster center to obtain multiple second blank clusters; for each data sample point, determine the target second Euclidean distance value pair according to the target second Euclidean distance value corresponding to the data sample point. The method comprises the following steps: selecting the first cluster center corresponding to the first cluster center, adding the data sample points to the second blank cluster corresponding to the first cluster center; adding the multiple data sample points to the multiple second blank clusters according to the target second Euclidean distance value corresponding to each data sample point, and obtaining multiple specified clusters; for each of the specified clusters, reading the coordinates of multiple second data points in the specified cluster, averaging the coordinates of the multiple second data points to obtain a second average coordinate point, and using the second average coordinate point as the second cluster center corresponding to the specified cluster; processing the coordinates of the multiple second data points of each specified cluster respectively to obtain the second cluster center corresponding to each specified cluster, generating multiple clustering results using the multiple second cluster centers, and using the multiple clustering results as multiple target clustering results; wherein, if the multiple convergence results meet the convergence conditions, generating multiple target clustering results using the multiple first cluster centers.

[0104] According to a third aspect of the present application, an electronic device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0105] According to a fourth aspect of the present application, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods in the first aspect are implemented.

[0106] By means of the above technical solution, the present application provides a charging value profile generation method, device, electronic device and readable storage medium. In response to a charging value profile generation request, the present application obtains user data of a user to be profiled, processes the user data using a preset data analysis model to obtain target behavior data, obtains a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, reads charging behavior volatility data from the target behavior data, and the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled. If the charging behavior volatility data is greater than or equal to the preset fluctuation threshold, a target clustering classification standard is obtained, and a profile operation is performed on the target behavior data using the target clustering classification standard to obtain a charging value profile of the user to be profiled. By constructing attributes on the user's original charging data, attribute data of multiple dimensions is obtained, and users are classified according to the attribute data of one dimension, namely the charging behavior volatility data. Finally, a user charging value profile is constructed according to the corresponding clustering profile rules. A large number of users can be divided into several typical groups with different development values. This can not only accurately mine user behavior characteristics and expand the portrait dimension, but also deeply characterize the charging value profile, improve the accuracy of the charging value profile, and accurately describe the user's development value.

[0107] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present application. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0109] Figure 1 A schematic diagram of a method for generating a charging value profile provided in an embodiment of the present application is shown;

[0110] Figure 2A A schematic diagram of a method for generating a charging value profile provided in an embodiment of the present application is shown;

[0111] Figure 2B A data processing diagram provided by an embodiment of the present application is shown;

[0112] Figure 2C A schematic diagram of a mean clustering profiling process provided in an embodiment of the present application is shown;

[0113] Figure 2DA schematic diagram of a charging value profile generation solution provided in an embodiment of the present application is shown;

[0114] Figure 3A A schematic diagram of a charging value profile generation method provided in an embodiment of the present application is shown;

[0115] Figure 3B A schematic diagram of a charging value profile generation method provided in an embodiment of the present application is shown;

[0116] Figure 4 A schematic diagram of the device structure of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0117] The following describes exemplary embodiments of the present application in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0118] The embodiment of the present application provides a method for generating a charging value portrait, such as Figure 1 As shown, the method includes:

[0119] 101. In response to a request to generate a charging value profile, obtain user data of the user to be profiled, and process the user data using a preset data analysis model to obtain target behavior data.

[0120] In the field of new energy vehicles, user charging value portraits are generally constructed based on big data or questionnaires. Since the setting of questionnaire questions in the questionnaire is guiding, the feedback rate is not high, and the authenticity is not strong, it cannot be used as a stable portrait generation method. The existing portrait generation method based on big data uses multiple attributes to synthesize multi-dimensional data for clustering to obtain the user's charging value portrait. However, the hard clustering based on division cannot be accurate to the user and the dimension is not high. As a result, the management side of the electric vehicle field lacks the means to describe the user development value when facing the charging behavior data of large-scale electric vehicles, resulting in a not in-depth portrait of the charging value of automobile users.

[0121] To solve this problem, this application proposes a charging value portrait generation method, which uses data mining technology to construct a user charging value portrait through classification and clustering methods, accurately mines user behavior characteristics, divides a large number of users into several typical groups with different development values, and accurately describes the user development value. The execution subject of this application can be a user portrait construction system, which can be built based on PyCharm (integrated development environment) and Python (computer programming language). The user portrait construction system provides a front-end application for users, that is, a client. Users (such as staff) can build portraits based on front-end application requests, so that the user portrait construction system relies on the computing power of the server to provide users with portrait construction services. The server can be an independent server, or it can provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms and other basic cloud computing servers, so that enterprises can recommend personalized charging plans to users based on different user portraits.

[0122] Because the collected charging record data has problems such as confusing formats, scattered information, and data anomalies, the accuracy of the user's charging value profile is affected when constructing the profile. Therefore, the user profile construction system will first clean the collected raw charging data and then analyze the cleaned data to obtain multi-dimensional user charging behavior data. In the embodiment of the present application, in response to a request to generate a charging value profile, the user profile construction system obtains the user data of the user to be profiled and processes the user data using a preset data analysis model to obtain target behavior data. This can more accurately and deeply explore the behavioral characteristics of charging users and characterize charging users from multiple dimensions.

[0123] 102. Obtain a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and read charging behavior volatility data from the target behavior data.

[0124] In order to more accurately construct a profile of charging users, the user profile construction system first classifies the users, that is, it checks whether the charging user is a frequent charging user or an occasional charging user through the volatility of charging behavior, and then constructs user profiles for each of them. In an embodiment of the present application, the user profile construction system obtains a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and reads the charging behavior volatility data in the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled. In this way, the threshold classifier determines whether the charging record of the user to be profiled is sufficient through volatility, and preliminarily evaluates the value of the user as a development customer.

[0125] 103. If the charging behavior fluctuation data is greater than or equal to the preset fluctuation threshold, a target clustering classification standard is obtained, and the target behavior data is profiled using the target clustering classification standard to obtain the charging value profile of the user to be profiled.

[0126] The charging users can be divided into two categories by the threshold classifier. If the charging behavior fluctuation is higher than the threshold, it means that the user charges frequently. Therefore, for the user with sufficient charging records, the user profile construction system will further use the clustering classification standard constructed by the K-Means++ clustering algorithm to construct the charging value profile. In the embodiment of the present application, if the charging behavior fluctuation data is greater than or equal to the preset fluctuation threshold, it means that the charging record of the user to be profiled is sufficient. The user profile construction system obtains the target clustering classification standard, and uses the target clustering classification standard to profile the target behavior data to obtain the charging value profile. The target clustering classification standard is obtained by processing the user data sample using the K-Means++ clustering algorithm. It can analyze the charging record of the user from multiple dimensions and accurately describe the development value of the user.

[0127] The method provided in the embodiment of the present application responds to the charging value profile generation request, obtains the user data of the user to be profiled, processes the user data using the preset data analysis model to obtain the target behavior data, obtains the threshold classifier and the preset fluctuation threshold corresponding to the threshold classifier, reads the charging behavior fluctuation data in the target behavior data, the charging behavior fluctuation data indicates the fluctuation of the charging behavior of the user to be profiled, if the charging behavior fluctuation data is greater than or equal to the preset fluctuation threshold, the target clustering classification standard is obtained, and the target behavior data is profiled using the target clustering classification standard to obtain the charging value profile of the user to be profiled. By constructing the attributes of the original charging data of the user, attribute data of multiple dimensions is obtained, and the user is classified according to the attribute data of one dimension, that is, the charging behavior fluctuation data. Finally, the charging value profile of the user is constructed according to the corresponding clustering profile rule. A large number of users can be divided into several typical groups with different development values. Not only can the user behavior characteristics be accurately mined to expand the profile dimension, but also the charging value profile can be deeply described to improve the charging value profile accuracy and accurately describe the development value of the user.

[0128] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to completely describe the specific implementation process of the embodiment, the embodiment of the present application provides another charging value profile generation method, as shown in Figure 2A The method comprises:

[0129] 201. In response to a charging value profile generation request, user data of a user to be profiled is obtained.

[0130] The existing charging value portrait generation method is clustered and portrayed in three dimensions of charging start time, charging duration and charging amount, which cannot completely mine the charging behavior of the user, and some only analyze the value of the user without depicting the behavior characteristics of the user. Therefore, the present application proposes a user charging value portrait generation method based on the RFELV model, which can mine and portray the charging behavior characteristics of the user from five dimensions, and analyze the development value at the same time. The generated user portrait is beneficial to the electric power enterprise to realize accurate marketing and personalized service for new energy automobile users. Among them, the RFELV model refers to constructing attributes on the basis of the original charging data of the user to obtain five attribute values of R (Recency, charging proximity), F (Frequency, total charging frequency), E (Energy, total charging amount), L (Length, user life cycle) and V (Volatility, charging behavior volatility).

[0131] In the embodiment of the present application, the user portrait construction system acquires the user identifier carried by the charging value portrait generation request, and takes the user indicated by the user identifier as the user to be portrayed. Then, the user portrait construction system queries the original user data indicated by the user identifier in the database. The original user data can be sourced from the database in the background of the public charging pile data collection system, including Customer_ID (customer ID), Start_Soc (initial remaining power), End_Soc (end remaining power), Charge_kwh (single charging amount), Usetime (charging duration), Chargstarttime (charging start time), Chargendtime (charging end time), Station_ID (charging station ID) and other data fields. The original user data of the embodiment of the present application is described taking the charging record data of the charging piles in Chongqing Banan District power grid within one year as an example.

[0132] In order to ensure the quality and precision of the user charging value portrait result, it is necessary and critical to clean the data before the portrait. Therefore, the user portrait construction system determines the abnormal data, repeated data and null data in the original user data, and deletes the abnormal data, repeated data and null data in the original user data, wherein the abnormal data includes hardware abnormal data, software abnormal data and operation abnormal data.

[0133] Subsequently, the user portrait construction system obtains a preset data format and uses the preset data format to adjust the original user data after deletion, converting the digital string into Int (integer type) format and converting the Dataframe (tabular data structure) format into Array (array) format to obtain the first preprocessed data, which facilitates linear operations on the data. Then, the user portrait construction system obtains a preset time format and uses the preset time format to adjust the first preprocessed data to obtain the second preprocessed data so that the time is converted into a unified format. Then, the user portrait construction system identifies the user name of the user to be profiled in the second preprocessed data, and when the user name is zero or empty, the user information is used as abnormal user information. The original user ID is generally represented by a number, such as 12, 234, etc., while the abnormal user ID is blank or 0, which will have an impact on the subsequent user portrait construction. Therefore, the user portrait construction system will obtain vehicle information in the second preprocessed data, use the vehicle information to replace the abnormal user information, and use the replaced second preprocessed data as user data. For example, when completing abnormal user IDs, you can use the Vehicle_License_No (license plate number) in the adjacent field and the same row. If there is no license plate number information, you can use the Vin (vehicle frame number) in the adjacent field and the same row to fill it in. In other words, replace the blank user ID or the user ID of 0 with the license plate number or frame number. It should be noted that the user profile construction system can use tools such as Pandas (a Python data analysis module) and Numpy (an open source scientific computing library in Python) to complete data processing operations.

[0134] 202. Obtain multiple attribute calculation formulas included in a preset data analysis model.

[0135] In order to further mine the user behavior of the automobile charging user, after processing the original user data, suitable data fields are selected for attribute construction. Therefore, the traditional RFM (a customer relationship management analysis model) model is improved in the application, M (Monetary, consumption amount) is replaced by the total charging amount E to obtain the RFE model, and the RFE model is extended to the RFELV model. Since the consumption amount M and the total charging amount E are proportional, in order to ignore the influence of the time-of-use electricity price, the total charging amount E is used instead of the consumption amount M, which can better reflect the charging demand of the user. The user life cycle L and the charging behavior volatility V are introduced to increase the feature dimension of the model, so that the RFELV attribute value, that is, the R value, the F value, the E value, the L value and the V value, can be constructed. In the embodiment of the application, the user portrait construction system obtains a plurality of attribute calculation formulas included in the preset data analysis model. The preset data analysis model is the RFELV model, the RFELV model involves the calculation of the RFELV attribute value, and the plurality of attribute calculation formulas used include the charging proximity calculation formula, the total charging frequency calculation formula, the total charging amount calculation formula, the user life cycle calculation formula and the charging behavior volatility calculation formula. In this way, the user portrait construction system can construct the R value, the F value, the E value, the L value and the V value through the attribute calculation formula, so as to evaluate the development value of the user based on the five attribute values in the subsequent process, expand the user portrait dimension, and further depict the user portrait.

[0136] 203、In the user data, the data statistical cutoff time of the user to be profiled, the last charging time of the user to be profiled, and the charging proximity data of the user to be profiled are obtained by using the charging proximity calculation formula.

[0137] In the embodiment of the application, the user portrait construction system obtains the data statistical cutoff time T now of the user to be profiled, the last charging time T rec of the user to be profiled, and the charging proximity data R of the user to be profiled by using the charging proximity calculation formula. now rec

[0138] Formula 1: R = T now -T rec ​​

[0139] wherein R represents the charging proximity data, that is, the interval days from the last charging time of the user to the current time (at the data set statistical cutoff time), for example, the last charging time of the user is February 8, 2020, and the data set statistical cutoff time is February 15, 2020, then the number of days obtained by subtracting the two dates is 7 days, which is the R value of the user, and the R value reflects how many days the user has not charged.

[0140] 204、obtain the charging frequency of the i-th day of the user to be profiled in the user data, and calculate the charging frequency of the i-th day of the user to be profiled by using the charging total frequency calculation formula, to obtain the charging total frequency data of the user to be profiled.

[0141] In the embodiment of the present application, the user portrait construction system obtains the charging frequency q i of the i-th day of the user to be profiled in the user data, and calculates the charging frequency q i of the i-th day of the user to be profiled by using the charging total frequency calculation formula, to obtain the charging total frequency data F of the user to be profiled. The charging total frequency calculation formula is formula 2 as follows:

[0142] Formula 2:

[0143] wherein F represents the charging total frequency data of the user to be profiled in the data statistical time interval, q i represents the charging frequency of the i-th day of the user, n is a positive integer, and i is a positive integer. For example, the data set is statistically from February 15, 2019 to February 15, 2020, a total of 365 days, the user portrait construction system obtains the charging frequency of each day of the user, and calculates the total sum of the charging frequency of 365 days.

[0144] 205、obtain the charging capacity of the i-th time of the user to be profiled in the user data, and calculate the charging capacity of the i-th time of the user to be profiled by using the charging total capacity calculation formula, to obtain the charging total capacity data of the user to be profiled.

[0145] In the embodiment of the present application, the user portrait construction system obtains the charging capacity e i of the i-th time of the user to be profiled in the user data, and calculates the charging capacity e i of the i-th time of the user to be profiled by using the charging total capacity calculation formula, to obtain the charging total capacity data E of the user to be profiled. The charging total capacity calculation formula is formula 3 as follows:

[0146] Formula 3:

[0147] wherein E represents the charging total capacity data of the user to be profiled in the data statistical time interval, e iRepresents the amount of electricity charged during the i-th charge, where n is a positive integer and i is a positive integer. For example, if a user charges 300 times during the statistical period of the dataset, the user profile construction system will obtain the amount of electricity charged during each charge and calculate the total amount of electricity charged during the 300 charges.

[0148] 206. Obtain the last charging time and the first charging time of the user to be profiled from the user data, and calculate the last charging time and the first charging time of the user to be profiled using the user life cycle calculation formula to obtain the user life cycle data of the user to be profiled.

[0149] In the embodiment of the present application, the user portrait construction system obtains the last charging time T of the user to be profiled in the user data. now , the first charging time T of the user to be profiled first , and use the user life cycle calculation formula to treat the last charging time T of the profile user now , the first charging time T of the user to be profiled first Calculate and obtain the user life cycle data L of the user to be profiled. The user life cycle calculation formula is the following formula 4:

[0150] Formula 4: L = T now -T first

[0151] Among them, L represents the user life cycle data of the user to be profiled, T now Indicates the last charging time, T first Indicates the first charging time.

[0152] 207. Obtain the time interval between the nth charging and the (n+1th) charging of the user to be profiled from the user data, and calculate multiple time intervals using a charging behavior volatility calculation formula to obtain charging behavior volatility data of the user to be profiled.

[0153] In the embodiment of the present application, the user portrait construction system obtains the time interval t between the nth charging and the n+1th charging of the user to be profiled in the user data. gap , and use the charging behavior volatility calculation formula to calculate multiple time intervals t gap Calculation is performed to obtain the charging behavior volatility data V of the user to be profiled. The charging behavior volatility calculation formula is the following formula 5:

[0154] Formula 5:

[0155] Among them, V represents the charging behavior volatility data of the user to be profiled, which is used to measure the volatility of the time interval of the user's charging behavior and can be defined as the ratio of the standard deviation to the mean.gap Represents the sample set of n time intervals of the user to be profiled within the data statistical time interval, Mean(T gap ) represents the sample set T gap The sample mean of , n is a positive integer, i is a positive integer.

[0156] 208. The charging proximity data, total charging frequency data, total charging amount data, user life cycle data, and charging behavior volatility data corresponding to the profiled user are sorted to obtain target behavior data.

[0157] In the embodiment of the present application, the user profile construction system organizes the charging proximity data, total charging frequency data, total charging amount data, user life cycle data, and charging behavior volatility data corresponding to the profiled user to obtain target behavior data. In this way, the user charging data is calculated through the RFELV model to complete the construction of the RFELV attribute value, so that the user profile construction system can represent the user's charging behavior characteristics through five dimensions: charging proximity, total charging frequency, total charging amount, user life cycle, and charging behavior volatility, and can deeply portray the charging value portrait.

[0158] Based on the above process, a data processing diagram proposed in the embodiment of the present application is as follows:

[0159] like Figure 2B As shown in the figure, the user profile construction system obtains the RFELV attribute value through data cleaning and data construction. Data cleaning removes abnormal data, duplicate data, and null values ​​from the multidimensional raw charging data. Abnormal data includes hardware anomalies, software anomalies, and illegal operation data. Data construction includes four steps: format conversion, time conversion, user ID completion, and attribute construction. Finally, the recency value R, frequency value F, total power consumption E, charging time L, and coefficient of variation V are obtained.

[0160] 209. Obtain a threshold classifier and a preset fluctuation threshold value corresponding to the threshold classifier, read the charging behavior volatility data in the target behavior data, and compare the charging behavior volatility data with the preset fluctuation threshold value; if the charging behavior volatility data is greater than or equal to the preset fluctuation threshold value, execute the following step 210; if the charging behavior volatility data is less than the preset fluctuation threshold value, execute the following step 211.

[0161] Existing profiling methods have a large granularity when classifying electric vehicle users, making it impossible to achieve fine-grained classification of electric vehicle users and provide more accurate personalized services for different users. Furthermore, existing studies have only analyzed the characteristics of user charging behavior without evaluating the value of each type of electric vehicle user. Therefore, this application constructs an RF threshold classifier. Since the RFELV model introduces charging behavior volatility to characterize user charging behavior, and the V value for users with a lack of charging records will fail, these users can be screened out through the RF threshold classifier as user data with a lack of charging records, while the original data with a valid V value is used as user data with sufficient charging records.

[0162] In an embodiment of the present application, the user portrait construction system obtains a threshold classifier and a preset fluctuation threshold value corresponding to the threshold classifier. The preset fluctuation threshold value can be used to classify users with obvious characteristics, determine which of the two categories the target behavior data belongs to, namely, user data lacking charging records and user data with sufficient charging records, and then perform user portraits for them, which can improve the accuracy of the charging value portrait. Next, the user portrait construction system reads the charging behavior volatility data in the target behavior data. The charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled. The charging behavior volatility data can be used to see whether the user charges frequently. Subsequently, the user portrait construction system compares the charging behavior volatility data with the preset fluctuation threshold value; if the charging behavior volatility data is greater than or equal to the preset fluctuation threshold value, it indicates that the user charges frequently and belongs to user data with sufficient charging user records, and then the following step 210 is executed; if the charging behavior volatility data is less than the preset fluctuation threshold value, it indicates that the user does not charge frequently and belongs to user data lacking charging records, and then the following step 211 is executed.

[0163] 210. If the charging behavior volatility data is greater than or equal to the preset volatility threshold, the target clustering classification standard is obtained, and the target behavior data is profiled using the target clustering classification standard to obtain a charging value profile of the user to be profiled.

[0164] For users with sufficient charging records, the user profile building system analyzes the RFELV attribute value based on the cluster classification criteria constructed by the K-Means++ clustering algorithm and outputs a charging value profile. In this embodiment of the present application, if the charging behavior volatility data is greater than or equal to the preset volatility threshold, it indicates that the user charges frequently and has sufficient charging user records. The user profile building system then obtains the target cluster classification criteria and obtains multiple profile categories included in the target cluster classification criteria, for example, the target cluster classification criteria has four categories. Next, the user profile construction system reads the corresponding standard values ​​for charging recency, total charging frequency, total charging volume, user lifecycle, and charging behavior volatility for each profile category from the target cluster classification criteria. For example, Category 1 has an R value of 213.10 days, an F value of 16.23 times, an E value of 404.41 kilowatts, an L value of 59.30 days, a V value of 1.13, and a profile label of churn and value+. Category 2 has an R value of 34.83 days, an F value of 124.38 times, and an E value of 3255. 52 kilowatts, L value is 230.79 days, V value is 1.98, portrait label is potential type, value ++; Category three: R value is 51.04 days, F value is 15.06 times, E value is 362.20 kilowatts, L value is 105.73 days, V value is 1.25, portrait label is retention type, value +++; Category four: R value is 40.14 days, F value is 33.70 times, E value is 832.14 kilowatts, L value is 226.74 days, V value is 2.05, portrait label is high value, value ++++.

[0165] Subsequently, the user portrait construction system obtains the preset dimensional data point conversion rules, and uses the preset dimensional data point conversion rules to process the charging proximity standard value, total charging frequency standard value, total charging amount standard value, user life cycle standard value, and charging behavior volatility standard value corresponding to each portrait category to obtain the standard center point corresponding to each portrait category. The user portrait construction system then reads the charging proximity data, total charging frequency data, total charging amount data, and user life cycle data in the target behavior data, and uses the preset dimensional data point conversion rules to process the charging proximity data, total charging frequency data, total charging amount data, user life cycle data, and charging behavior volatility data to obtain the center point of the user to be profiled. Then, the user portrait construction system calculates the Euclidean distance between the center point of the user to be profiled and the standard center point corresponding to each portrait category, obtains multiple Euclidean distance values, and selects the Euclidean distance value with the smallest distance value among the multiple Euclidean distance values ​​as the target Euclidean distance value. A five-dimensional center point can be obtained through the R standard value, F standard value, E standard value, L standard value, and V standard value corresponding to each category. The division of users is determined based on the distance between the user's R value, F value, E value, L value, and V value (the user's five-dimensional center point) and the five-dimensional points of these four categories (the five-dimensional center points corresponding to categories one, two, three, and four). The user is divided into the category to which the five-dimensional center point has the smallest Euclidean distance. For example, if a user's RFELV attribute value is closest to the five-dimensional center point corresponding to the churn type, then the user belongs to the churn type (low-value) user. In other words, the R value, F value, E value, L value, and V value of each category of users will be close to the R standard value, F standard value, E standard value, L standard value, and V standard value of the corresponding category, and the difference will not be too large. Therefore, every user can be divided into categories according to this logic. It should be noted that the target clustering classification standard is obtained by clustering the data sample set using the K-Means++ clustering algorithm. Since the clustering center results after clustering different data sample sets are different, there is no absolute clustering center, and the target clustering classification standard may also be different. This application does not specifically limit the generation method of the classification standard.

[0166] Next, the user profile construction system determines the designated standard center point indicated by the target Euclidean distance value from the multiple standard center points corresponding to the multiple profile categories, and determines the profile category corresponding to the designated standard center point as the target profile category of the user to be profiled. Finally, the user profile construction system organizes the target profile category, charging proximity data, total charging frequency data, total charging volume data, user lifecycle data, and charging behavior volatility data to obtain a charging value profile for the user to be profiled.

[0167] The user profile construction system can categorize users with sufficient charging records into four types: churners, potential users, retention users, and high-value users. The clustering results for each category are analyzed and assigned semantic labels. The value of each category is then described with a "+" to form a user profile. Category 1: Churners, or low-value users. Their total charging frequency (F), total charging volume (E), and user lifetime (L) are all low. While the charging behavior volatility (V) indicates regular charging, the charging proximity (R) indicates that these users have not visited the local area for a long time and are likely to leave. The probability of return is very low, resulting in a "+" value. Category 2: Potential users, or potential users. Their R values ​​are small, their E and F values ​​are large, and their V values ​​are irregular, but their lifetime is long. These users are more likely to return for charging and have considerable growth potential. However, while these users have extensive charging records, their behavior is irregular, making it difficult to establish a stable charging relationship through customized charging solutions. Therefore, their value is "++." Category 3: Retention users. The R value for this type of user indicates potential churn, while the V value indicates that their charging intervals fluctuate little, resulting in a long lifespan. If discounts are offered regularly to retain these users, they may develop a long-term, regular charging behavior and potentially become high-value users, thus assigning this user a value of "+++." Category 4: High Value. The R, F, and E values ​​indicate that this type of user is a loyal local customer with a high demand for electricity. They have recently maintained their charging behavior and have the longest lifespan of all user categories. Because they have numerous charging records to support and maintain their charging behavior, this user behavior profile has the highest value, "++++."

[0168] In this way, the user portrait construction system can comprehensively evaluate the user's development value based on the attributes of R value, F value, E value, L value, and V value. For example, users with small R value and small F value are not stable, and the possibility of such users going to local charging stations is not high, so such users are given low-value labels in the final value evaluation. In addition, each user ID can be directly represented by the portrait label, that is, the user can be found through the portrait label to which the portrait belongs, which is conducive to power companies to achieve precise marketing and personalized services for new energy vehicle users. It should be noted that as the various attribute values ​​of RFELV change, the user's value changes accordingly, and users can be divided into several typical groups with different development values. It can not only accurately mine user behavior characteristics, but also deeply portray user portraits and accurately describe user development value.

[0169] 211. If the charging behavior volatility data is less than the preset volatility threshold, the threshold classifier is used to perform a profiling operation on the target behavior data to obtain a charging value profile of the user to be profiled.

[0170] User data lacking charging records is data filtered out by the RF threshold classifier. The RF threshold classifier serves as a reference standard for data segmentation. Since there is an RF threshold range representation, the user profile construction system can directly perform the profile operation. In an embodiment of the present application, if the charging behavior volatility data is less than the preset volatility threshold, it means that the user does not charge frequently and is user data lacking charging records. Then, the user profile construction system reads the charging proximity data, total charging frequency data, total charging amount data, and user life cycle data from the target behavior data. Next, the user profile construction system obtains the preset classification standard corresponding to the threshold classifier and reads the preset proximity range and preset total frequency range corresponding to each preset classification result in the threshold classifier. The preset classification standard includes multiple preset classification results, the preset proximity range is also the R value range, and the preset total frequency range is also the F value range. For example, the preset classification standard includes classification result one as (R>30) & (2≤F≤5), classification result two as (R≤30) & (2≤F≤5), and classification result three as F=1. Subsequently, the user portrait construction system determines the target classification result corresponding to the user to be profiled from multiple preset classification results, wherein the charging proximity data is within the preset proximity range corresponding to the target classification result and the total charging frequency data is within the preset total frequency range corresponding to the target classification result. For example, if the R value corresponding to the user is 27 days and the F value is 4 times, then the user belongs to classification result two. Then, the user portrait construction system obtains the portrait label corresponding to the target classification result and uses the portrait label as the target portrait label corresponding to the user to be profiled. For example, the portrait label corresponding to classification result one is EX tourist type, the portrait label corresponding to classification result two is tourist type, and the portrait label corresponding to classification result three is individual tourist type, then the user belongs to tourist type user. Finally, the user portrait construction system organizes the target portrait label, target classification result, charging proximity data, total charging frequency data, total charging amount data, user life cycle data, and charging behavior volatility data to generate a charging value portrait of the user to be profiled.

[0171] In this way, the user profile construction system can use a threshold classifier to categorize users with no charging history into three types: EX-Tourists, Tourists, and Individuals. Tourists can be defined as users whose activities are centered around the charging station and who do not stray too far from the area. EX-Tourists represent former visitors. These users have an R value greater than 30 days and an F value between 2 and 5 times, indicating that they have visited the local charging station before and have returned to charge, but only rarely. These users may be staying in the area for a short period of time, perhaps for tourism, short business trips, or returning from work in another area. This indicates that they are unlikely to return to charge in the near future. Tourists have an R value less than 30 days and an F value between 2 and 5 times. This indicates that these users may have recently arrived in the area. Their R value indicates that they are more likely to return to charge than EX-Tourists, but it is unclear whether they will engage in long-term charging behavior in the future. Individual users have an F value of 1, indicating they charge only once. They are likely passing through the area and are not active in the area, so their return visit is highly unlikely, if ever. Of course, this group may include newcomers who may become permanent residents, but this cannot be identified based solely on historical charging data. In short, the commercial value of these users is not worth considering.

[0172] In an optional implementation scheme, the user portrait process is divided into clustering and threshold classification processes. The results of clustering and threshold classification are used to match each user record with a corresponding portrait label. The portrait label is a behavioral semantic summary based on the analysis of clustering and classification results. The user portrait construction system obtains user data samples, performs normalization calculations on the user data samples, and obtains data samples to be clustered. The user portrait construction system obtains a preset cluster number range and extracts multiple preset cluster numbers within the preset cluster number range. For example, if the cluster number range is (2,12), the multiple cluster numbers extracted are 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 and 12, a total of 11 cluster numbers. Next, the user portrait construction system sets up a for loop and uses the silhouette coefficient to evaluate the K-Means++ clustering results corresponding to each cluster number. Finally, the cluster number corresponding to the clustering result with the largest silhouette coefficient is selected as the optimal cluster number. Specifically, for each of the multiple preset cluster numbers, the user portrait construction system obtains a mean clustering algorithm, that is, a K-Means++ clustering algorithm, and uses the mean clustering algorithm to calculate the clustered data sample and the preset cluster number to obtain a predicted clustering result. Then, the user portrait construction system obtains multiple predicted clustering results corresponding to the multiple preset cluster numbers, and uses the silhouette coefficient to evaluate and calculate the multiple predicted clustering results respectively, to obtain multiple evaluation coefficients corresponding to the multiple predicted clustering results. Subsequently, the user portrait construction system arranges the multiple evaluation coefficients in descending order, takes the evaluation coefficient ranked first as the target evaluation coefficient, and determines the preset cluster number corresponding to the target evaluation coefficient among the multiple preset cluster numbers, and takes the preset cluster number as the target cluster number. For example, the silhouette coefficient evaluation shows that the number of clusters 5 is the optimal number of clusters.

[0173] Subsequently, the user portrait construction system uses the numerical value of the target number of clusters as the target number of calculations for calculating the target cluster center, that is, 5 target cluster centers are obtained through 5 calculations. Specifically, the calculation process for one of the target cluster centers is as follows: the user portrait construction system selects any data sample point in the data sample to be clustered as the initial cluster center, and determines other data sample points other than the initial cluster center in the data sample to be clustered as multiple first sample points. Then, the user portrait construction system calculates the distance between the initial cluster center and each first sample point in the multiple first sample points to obtain multiple sample distance values. The sample distance value refers to the shortest distance in space between the first sample point and the current initial cluster center. Subsequently, the user portrait construction system uses the distance probability formula to calculate the multiple sample distance values ​​to obtain multiple selection probabilities. The distance probability formula is the following formula 6:

[0174] Formula 6:

[0175] Where D(x) represents the shortest distance in space between the first sample point and the current initial cluster center, and X is a positive integer. The distance probability formula can be used to calculate the probability of each first sample point being selected as the next cluster center. The user profile construction system then obtains a roulette wheel selection algorithm, uses the roulette wheel selection algorithm to select a selected probability from multiple selection probabilities as the target selected probability, and uses the first sample point corresponding to the target selected probability as the target cluster center. In this way, the user profile construction system repeatedly calculates the target cluster center of the data samples to be clustered until the number of calculations reaches the target number, resulting in a number of target cluster centers equal to the target number of calculations.

[0176] Then, the user profile construction system creates a first blank cluster for each target cluster center, obtaining multiple first blank clusters. Then, the user profile construction system obtains multiple data sample points in the data sample to be clustered, and adds the multiple data sample points to multiple first blank clusters respectively, obtaining multiple target clusters. Specifically, the user profile construction system calculates the Euclidean distance between each data sample point and the multiple target cluster centers respectively, obtaining multiple first Euclidean distance values ​​corresponding to each data sample point. The calculation formula used to calculate the Euclidean distance is the following formula 7:

[0177] Formula 7:

[0178] Among them, d (x，y) Represents the distance between the data sample point and the target cluster center in the mathematical space, and also represents the degree of proximity between the two points. i is a positive integer, and n is a positive integer. Next, the user portrait construction system selects the first Euclidean distance value with the smallest distance value from the multiple first Euclidean distance values ​​corresponding to each data sample point as the target first Euclidean distance value corresponding to each data sample point. For each data sample point, based on the target first Euclidean distance value corresponding to the data sample point, the target cluster center corresponding to the target first Euclidean distance value is determined, and the data sample point is added to the first blank cluster corresponding to the target cluster center. In this way, based on the target first Euclidean distance value corresponding to each data sample point, the user portrait construction system can add multiple data sample points to multiple first blank clusters respectively to obtain multiple target clusters.

[0179] Furthermore, to improve clustering accuracy, the user profile construction system updates the cluster center of each cluster, using the average coordinate value of all data points in each cluster as the new cluster center. Specifically, for each target cluster, the user profile construction system determines the coordinates of multiple data sample points included in the target cluster, obtains multiple first data point coordinates, averages the multiple first data point coordinates to obtain a first average coordinate point, and uses the first average coordinate point as the first cluster center corresponding to the target cluster. In this way, the user profile construction system processes the multiple first data point coordinates of each target cluster separately to obtain the first cluster center corresponding to each target cluster.

[0180] Subsequently, it is judged whether the new clustering center converges, and if the clustering center converges, the clustering result is output; if the clustering center does not converge, the Euclidean distance between each data sample point in the sample set and the clustering center is recalculated to form a new class cluster. Specifically, the user portrait construction system performs convergence calculation on each first clustering center corresponding to each target class cluster to obtain a plurality of convergence results, and compares the plurality of convergence results with the convergence condition, in order to determine whether the new clustering center changes compared with the previous clustering center, or whether the change is large. It should be noted that whether the clustering center converges is judged together for all class clusters, for example, if four clustering centers are calculated, the four clustering centers are judged at the same time. If the plurality of convergence results meet the convergence condition, it means that there is no change or the change is within a very small range, and the clustering center is convergent, then a plurality of target clustering results are generated using the plurality of first clustering centers; if one of the plurality of convergence results does not meet the convergence condition, it means that there is a change or the change is large, then a plurality of data sample points in the data sample to be clustered are obtained, and the Euclidean distance between each data sample point and the plurality of first clustering centers is calculated to obtain a plurality of second Euclidean distance values corresponding to each data sample point. Next, the user portrait construction system selects the smallest second Euclidean distance value as the target second Euclidean distance value corresponding to each data sample point from the plurality of second Euclidean distance values corresponding to each data sample point. Subsequently, the user portrait construction system creates a second blank class cluster for each first clustering center to obtain a plurality of second blank class clusters. For each data sample point, the first clustering center corresponding to the target second Euclidean distance value is determined according to the target second Euclidean distance value of the data sample point, and the data sample point is added to the second blank class cluster corresponding to the first clustering center. Then, the user portrait construction system adds the plurality of data sample points to the plurality of second blank class clusters according to the target second Euclidean distance value corresponding to each data sample point to obtain a plurality of specified class clusters. For each specified class cluster, a plurality of second data point coordinates are read in the specified class cluster, and the plurality of second data point coordinates are averaged to obtain a second average coordinate point, and the second average coordinate point is taken as a second clustering center corresponding to the specified class cluster. Finally, the user portrait construction system processes the plurality of second data point coordinates of each specified class cluster to obtain a second clustering center corresponding to each specified class cluster, generates a plurality of clustering results using the plurality of second clustering centers, and takes the plurality of clustering results as a plurality of target clustering results.

[0181] Finally, the user portrait construction system obtains preset portrait classification rules, uses the preset portrait classification rules to process multiple target clustering results, obtains multiple portrait categories, and uses multiple portrait categories to generate target clustering classification standards. For example, in category one, the R value is 213.10 days, the F value is 16.23 times, the E value is 404.41 kilowatts, the L value is 59.30 days, the V value is 1.13, the portrait label is the churn type, and the value is +; in category two, the R value is 34.83 days, the F value is 124.38 times, the E value is 3255.52 kilowatts, the L value is 230.79 days, the V value is 1.98, the portrait label is the potential type, and the value is ++; in category three, the R value is 51.04 days, the F value is 15.06 times, the E value is 362.20 kilowatts, the L value is 105.73 days, the V value is 1.25, the portrait label is the retention type, and the value is +++; in category four, the R value is 40.14 days, the F value is 33.70 times, the E value is 832.14 kilowatts, the L value is 226.74 days, the V value is 2.05, the portrait label is the high value, and the value is ++++. In this way, a multi-dimensional, high-standard target clustering classification standard can be constructed, so that the user portrait construction system can conduct in-depth and accurate portraits based on the mined user behavior characteristics.

[0182] Based on the above process, a schematic diagram of a mean clustering profiling process proposed in the embodiment of the present application is as follows:

[0183] like Figure 2C As shown, the normalized data sample is input into the K-means() function. The optimal cluster value, denoted as K, is obtained based on the silhouette coefficient method. Next, a point is randomly selected from the data sample as the initial cluster center, and the shortest distance D(X) between each sample and the existing cluster center in space is calculated. Subsequently, the probability of each sample being selected as the next cluster center is calculated according to the formula, and the next cluster center is selected using the roulette wheel method based on this probability. A check is performed to determine whether K cluster centers have been selected. If not, the process jumps to the step of randomly selecting a point from the data sample as the initial cluster center and continues until K cluster centers are selected. If K cluster centers have been selected, the Euclidean distance between each data point and the K cluster centers is calculated, and each point is assigned to the cluster center closest to it, forming a cluster. The cluster center of each cluster is then updated, with the average of the coordinates of all data points in each cluster serving as the new cluster center. Determine whether the cluster centers have converged. If so, output the clustering results. If not, recalculate the Euclidean distance between each data point and the K cluster centers, assign each point to the cluster center closest to it, and form a new cluster. Update the cluster center of each new cluster and re-determine whether the cluster centers have converged. Finally, analyze the behavioral semantics based on the clustering results of each class to obtain a semantic label.

[0184] Therefore, the schematic diagram of the solution flow for generating a charging value portrait proposed in the embodiment of the present application is as follows:

[0185] like Figure 2D As shown in the figure, in order to ensure the quality and accuracy of the charging value portrait results, the user portrait construction system will perform data preprocessing operations on the original user data. First, the data is cleaned based on the original charging data, and then the data of suitable fields are filtered from the cleaned data for attribute construction to construct the RFELV attribute value. Then, the RF threshold classifier is used to classify users with obvious characteristics, and the data is divided into two categories, namely user data with a lack of charging records and user data with sufficient charging records, and then user portraits are generated for each of them. For user data with a lack of charging records, since the user behavior characteristics can be directly summarized by the RF threshold classifier, the user portrait can be directly obtained through portrait analysis. For user data with sufficient charging records, the K-Means++ clustering algorithm is used for clustering based on the RFELV attribute value, the performance of each clustering result is evaluated and the best result is selected for output, and finally, a portrait analysis is performed to obtain the user portrait.

[0186] The method provided in the embodiment of the present application, in response to a request for generating a charging value portrait, obtains user data of the user to be profiled, processes the user data using a preset data analysis model to obtain target behavior data, obtains a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, reads charging behavior volatility data in the target behavior data, the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled, if the charging behavior volatility data is greater than or equal to the preset fluctuation threshold, obtains a target clustering classification standard, performs a profile operation on the target behavior data using the target clustering classification standard, obtains a charging value portrait of the user to be profiled, constructs attributes of the user's original charging data to obtain attribute data of multiple dimensions, and classifies the user according to the attribute data of one of the dimensions, that is, the charging behavior volatility data, and finally constructs a user charging value portrait according to the corresponding clustering portrait rule, which can divide a large number of users into several typical groups with different development values, can not only accurately mine user behavior characteristics and expand the portrait dimension, but also can deeply portray the charging value portrait, improve the accuracy of the charging value portrait, and accurately describe the user's development value.

[0187] Further, as Figure 1 In the specific implementation of the method, the embodiment of the present application provides a charging value portrait generation device, such as Figure 3A As shown, the device includes: a processing module 301, a reading module 302 and a first portrait module 303.

[0188] The processing module 301 is configured to respond to a request for generating a charging value profile, obtain user data of the user to be profiled, and process the user data using a preset data analysis model to obtain target behavior data;

[0189] A reading module 302 is configured to obtain a threshold classifier and a preset fluctuation threshold value corresponding to the threshold classifier, and read charging behavior volatility data from the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled;

[0190] The first portrait module 303 is used to obtain a target cluster classification standard if the charging behavior volatility data is greater than or equal to the preset volatility threshold, and perform a portrait operation on the target behavior data using the target cluster classification standard to obtain a charging value portrait of the user to be portraited.

[0191] In a specific application scenario, the processing module 301 is used to obtain the user identifier carried in the charging value portrait generation request, take the user indicated by the user identifier as the user to be profiled, and query the original user data indicated by the user identifier in the database; determine abnormal data, duplicate data, and null value data in the original user data, and delete the abnormal data, the duplicate data, and the null value data in the original user data, and the abnormal data includes hardware abnormal data, software abnormal data, and operation abnormal data; obtain a preset data format, and use the preset data format to adjust the deleted original user data to obtain first preprocessed data; obtain a preset time format, and use the preset time format to adjust the first preprocessed data to obtain second preprocessed data; identify the user name of the user to be profiled in the second preprocessed data, and when the user name is zero or empty, use the user information as abnormal user information; obtain vehicle information in the second preprocessed data, use the vehicle information to replace the abnormal user information, and use the replaced second preprocessed data as the user data.

[0192] In a specific application scenario, the processing module 301 is further used to obtain multiple attribute calculation formulas included in the preset data analysis model, the multiple attribute calculation formulas including the charging proximity calculation formula, the total charging frequency calculation formula, the total charging amount calculation formula, the user life cycle calculation formula, and the charging behavior volatility calculation formula; obtain the data statistics cutoff time T of the user to be profiled in the user data now , the last charging time T of the user to be profiled rec , and the data statistics cutoff time T of the user to be profiled using the charging proximity calculation formula now , the last charging time T of the user to be profiledrec Calculate and obtain the charging proximity data R of the user to be profiled, where:

[0193] R=T now -T rec

[0194] Wherein, R represents the charging proximity data; the number of charging times q of the user to be profiled on the i-th day is obtained from the user data. i , and the total charging frequency calculation formula is used to calculate the number of charging times q of the user to be profiled on the i-th day i Calculate and obtain the total charging frequency data F of the user to be profiled, where:

[0195]

[0196] Wherein, F represents the total frequency of charging of the user to be profiled within the data statistical time interval, n is a positive integer, i is a positive integer; the charging power e of the user to be profiled for the i-th time is obtained from the user data. i , and the charging amount e of the i-th user to be profiled is calculated using the total charging amount calculation formula i Calculate and obtain the total charging amount data E of the user to be profiled, where:

[0197]

[0198] Wherein, E represents the total amount of charging data of the user to be profiled within the data statistical time interval, n is a positive integer, i is a positive integer; the last charging time T of the user to be profiled is obtained from the user data now , the first charging time T of the user to be profiled first , and the last charging time T of the user to be profiled using the user life cycle calculation formula now , the first charging time T of the user to be profiled first Calculate and obtain the user life cycle data L of the user to be profiled, where:

[0199] L=T now -T first

[0200] Wherein, L represents the user life cycle data of the user to be profiled; the time interval t between the nth charging and the n+1th charging of the user to be profiled is obtained from the user data. gap , and using the charging behavior volatility calculation formula to calculate multiple time intervals t gap Calculation is performed to obtain the charging behavior volatility data V of the user to be profiled, where:

[0201]

[0202] Wherein, V represents the charging behavior volatility data of the user to be profiled, T gap represents the sample set of n time intervals of the user to be profiled within the data statistical time interval, Mean(T gap ) represents the sample set T gap The sample mean of , n is a positive integer, i is a positive integer; the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data corresponding to the user to be profiled are sorted to obtain the target behavior data.

[0203] In a specific application scenario, the first portrait module 303 is used to obtain multiple portrait categories included in the target cluster classification standard, and read the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category in the target cluster classification standard; obtain the preset dimensional data point conversion rule, and use the preset dimensional data point conversion rule to process the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category to obtain the standard center point corresponding to each portrait category; read the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data in the target behavior data; use the preset dimensional data point conversion rule to process the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data The total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are processed to obtain the data center point to be profiled corresponding to the user to be profiled; the Euclidean distance between the data center point to be profiled and the standard center point corresponding to each portrait category is calculated to obtain multiple Euclidean distance values, and the Euclidean distance value with the smallest distance value is selected from the multiple Euclidean distance values ​​as the target Euclidean distance value; the designated standard center point indicated by the target Euclidean distance value is determined among the multiple standard center points corresponding to the multiple portrait categories, and the portrait category corresponding to the designated standard center point is determined as the target portrait category of the user to be profiled; the target portrait category, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to obtain a charging value portrait of the user to be profiled.

[0204] In specific application scenarios, such as Figure 3B As shown, the device also includes: a second portrait module 304 and a generation module 305.

[0205] The second portrait module 304 is used to read the charging proximity data, total charging frequency data, total charging amount data, and user life cycle data in the target behavior data if the charging behavior volatility data is less than the preset volatility threshold; obtain the preset classification standard corresponding to the threshold classifier, and the preset classification standard includes multiple preset classification results; read the preset proximity range and preset total frequency range corresponding to each of the preset classification results in the threshold classifier, and determine the target classification result corresponding to the user to be profiled in the multiple preset classification results, wherein the charging proximity data is within the preset proximity range corresponding to the target classification result and the total charging frequency data is within the preset total frequency range corresponding to the target classification result; obtain the portrait label corresponding to the target classification result, and use the portrait label as the target portrait label corresponding to the user to be profiled; organize the target portrait label, the target classification result, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data to generate a charging value portrait of the user to be profiled.

[0206] The generation module 305 is used to obtain user data samples, perform normalization calculation on the user data samples to obtain data samples to be clustered, and obtain a preset cluster number range, extract multiple preset cluster numbers within the preset cluster number range; obtain a silhouette coefficient, use the silhouette coefficient to select a preset cluster number as the target cluster number among the multiple preset cluster numbers; use the value of the target cluster number as the target number of calculations for calculating the target cluster center; obtain a distance probability formula, use the distance probability formula to calculate the target cluster center of the data samples to be clustered, and repeatedly calculate the target cluster center of the data samples to be clustered until the number of calculations reaches the target number of calculations, and obtain multiple target cluster centers whose number is equal to the target number of calculations; create a first blank cluster for each target cluster center to obtain multiple first blank clusters, obtain multiple data sample points in the data samples to be clustered, and The multiple data sample points are respectively added to the multiple first blank clusters to obtain multiple target clusters; for each target cluster, the coordinates of the multiple data sample points included in the target cluster are determined to obtain multiple first data point coordinates, the multiple first data point coordinates are averaged to obtain a first average coordinate point, and the first average coordinate point is used as the first cluster center corresponding to the target cluster; the multiple first data point coordinates of each target cluster are respectively processed to obtain the first cluster center corresponding to each target cluster, and the convergence condition is obtained, and the first cluster center corresponding to each target cluster is processed using the convergence condition to obtain multiple target clustering results; the preset portrait classification rule is obtained, the preset portrait classification rule is used to process the multiple target clustering results to obtain multiple portrait categories, and the target cluster classification standard is generated using the multiple portrait categories.

[0207] In a specific application scenario, the generation module 305 is used to obtain a mean clustering algorithm for each of the multiple preset cluster numbers, and use the mean clustering algorithm to calculate the data sample to be clustered and the preset cluster number to obtain a predicted clustering result; obtain multiple predicted clustering results corresponding to the multiple preset cluster numbers, and use the silhouette coefficient to evaluate and calculate the multiple predicted clustering results respectively to obtain multiple evaluation coefficients corresponding to the multiple predicted clustering results; arrange the multiple evaluation coefficients in descending order, and use the evaluation coefficient ranked first as the target evaluation coefficient, and determine the preset cluster number corresponding to the target evaluation coefficient among the multiple preset cluster numbers, and use the preset cluster number as the target clustering cluster number.

[0208] In a specific application scenario, the generation module 305 is used to select any data sample point in the data sample to be clustered as the initial cluster center, and determine other data sample points other than the initial cluster center in the data sample to be clustered as multiple first sample points; calculate the distance between the initial cluster center and each first sample point in the multiple first sample points to obtain multiple sample distance values, and use the distance probability formula to calculate the multiple sample distance values ​​to obtain multiple selection probabilities; obtain a roulette wheel selection algorithm, use the roulette wheel selection algorithm to select a selection probability from the multiple selection probabilities as the target selection probability, and use the first sample point corresponding to the target selection probability as the target cluster center.

[0209] In a specific application scenario, the generation module 305 is used to respectively calculate the Euclidean distance between each of the data sample points and the multiple target cluster centers to obtain multiple first Euclidean distance values ​​corresponding to each data sample point; select the first Euclidean distance value with the smallest distance value from the multiple first Euclidean distance values ​​corresponding to each data sample point as the target first Euclidean distance value corresponding to each data sample point; for each data sample point, determine the target cluster center corresponding to the target first Euclidean distance value according to the target first Euclidean distance value corresponding to the data sample point, and add the data sample point to the first blank cluster corresponding to the target cluster center; according to the target first Euclidean distance value corresponding to each data sample point, add the multiple data sample points to the multiple first blank clusters respectively to obtain the multiple target clusters.

[0210] In a specific application scenario, the generation module 305 is also used to perform convergence calculation on the first cluster center corresponding to each target cluster to obtain multiple convergence results, and compare the multiple convergence results with the convergence condition; if a convergence result among the multiple convergence results does not meet the convergence condition, multiple data sample points in the data sample to be clustered are obtained, and the Euclidean distance between each data sample point and the multiple first cluster centers is calculated respectively to obtain multiple second Euclidean distance values ​​corresponding to each data sample point; the second Euclidean distance value with the smallest distance value is selected from the multiple second Euclidean distance values ​​corresponding to each data sample point as the target second Euclidean distance value corresponding to each data sample point; a second blank cluster is created for each first cluster center to obtain multiple second blank clusters; for each data sample point, the target second Euclidean distance is determined according to the target second Euclidean distance value corresponding to the data sample point. The method comprises the following steps: obtaining a first cluster center corresponding to a distance value, adding the data sample point to the second blank cluster corresponding to the first cluster center; adding the multiple data sample points to the multiple second blank clusters according to the target second Euclidean distance value corresponding to each data sample point, and obtaining multiple specified clusters; for each of the specified clusters, reading multiple second data point coordinates in the specified cluster, averaging the multiple second data point coordinates to obtain a second average coordinate point, and using the second average coordinate point as the second cluster center corresponding to the specified cluster; processing the multiple second data point coordinates of each specified cluster respectively to obtain the second cluster center corresponding to each specified cluster, generating multiple clustering results using the multiple second cluster centers, and using the multiple clustering results as multiple target clustering results; wherein, if the multiple convergence results meet the convergence conditions, generating multiple target clustering results using the multiple first cluster centers.

[0211] The device provided in the embodiment of the present application obtains user data of the user to be profiled in response to a request for generating a charging value portrait, processes the user data using a preset data analysis model to obtain target behavior data, obtains a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, reads charging behavior volatility data in the target behavior data, the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled, and if the charging behavior volatility data is greater than or equal to the preset fluctuation threshold, obtains a target clustering classification standard, performs a profile operation on the target behavior data using the target clustering classification standard, and obtains a charging value portrait of the user to be profiled. By constructing attributes of the user's original charging data, attribute data of multiple dimensions is obtained, and the user is classified according to the attribute data of one of the dimensions, that is, the charging behavior volatility data. Finally, a user charging value portrait is constructed according to the corresponding clustering portrait rule, which can divide a large number of users into several typical groups with different development values. It can not only accurately mine user behavior characteristics and expand the portrait dimension, but also deeply portray the charging value portrait, improve the accuracy of the charging value portrait, and accurately describe the user's development value.

[0212] It should be noted that for other corresponding descriptions of the functional units involved in the charging value portrait generation device provided in the embodiment of the present application, please refer to Figure 1 and Figures 2A to 2D The corresponding description in will not be repeated here.

[0213] In an exemplary embodiment, see Figure 4 , also provides an electronic device comprising a bus, a processor, a memory, and a communication interface, and may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor is configured to execute the program stored in the memory and perform the charging value profile generation method described in the above embodiment.

[0214] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the charging value portrait generation method.

[0215] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented through hardware or by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0216] Those skilled in the art will understand that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application.

[0217] Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the implementation scenario description, or can be modified accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module or further split into multiple submodules.

[0218] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0219] The above disclosure only describes several specific implementation scenarios of the present application. However, the present application is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the scope of protection of the present application.

Claims

1. A method for generating a charging value portrait, characterized in that: include: In response to a request to generate a charging value profile, user data of the user to be profiled is obtained, and the user data is processed using a preset data analysis model to obtain target behavior data; Obtaining a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and reading charging behavior volatility data from the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled; If the charging behavior volatility data is greater than or equal to the preset fluctuation threshold, the target clustering classification standard is obtained, and the target behavior data is subjected to a profiling operation using the target clustering classification standard to obtain a charging value portrait of the user to be profiled, including: obtaining multiple portrait categories included in the target clustering classification standard, reading the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category in the target clustering classification standard, obtaining a preset dimensional data point conversion rule, and processing the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category using the preset dimensional data point conversion rule to obtain the standard center point corresponding to each portrait category, and reading the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle standard value, and the charging behavior volatility standard value in the target behavior data. periodic data, use the preset dimensional data point conversion rule to process the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data to obtain the data center point to be profiled corresponding to the user to be profiled, calculate the Euclidean distance between the data center point to be profiled and the standard center point corresponding to each portrait category, and obtain multiple Euclidean distance values, select the Euclidean distance value with the smallest distance value from the multiple Euclidean distance values ​​as the target Euclidean distance value, determine the designated standard center point indicated by the target Euclidean distance value from the multiple standard center points corresponding to the multiple portrait categories, and determine the portrait category corresponding to the designated standard center point as the target portrait category of the user to be profiled, organize the target portrait category, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data to obtain a charging value portrait of the user to be profiled.

2. The method according to claim 1, characterized in that The step of obtaining user data of the user to be profiled in response to the charging value profile generation request includes: Obtaining the user identifier carried in the charging value profile generation request, taking the user indicated by the user identifier as the user to be profiled, and querying the database for original user data indicated by the user identifier; Determining abnormal data, duplicate data, and null value data in the original user data, and deleting the abnormal data, the duplicate data, and the null value data from the original user data, wherein the abnormal data includes hardware abnormal data, software abnormal data, and operation abnormal data; Obtaining a preset data format, and adjusting the deleted original user data using the preset data format to obtain first preprocessed data; Obtaining a preset time format, and adjusting the first preprocessed data using the preset time format to obtain second preprocessed data; Identifying the user name of the user to be profiled in the second pre-processed data, and treating the user information as abnormal user information when the user name is zero or empty; Vehicle information is obtained from the second pre-processed data, the abnormal user information is replaced with the vehicle information, and the replaced second pre-processed data is used as the user data.

3. The method according to claim 1, characterized in that The method of processing the user data using a preset data analysis model to obtain target behavior data includes: Obtaining multiple attribute calculation formulas included in the preset data analysis model, the multiple attribute calculation formulas including a charging proximity calculation formula, a total charging frequency calculation formula, a total charging amount calculation formula, a user life cycle calculation formula, and a charging behavior volatility calculation formula; Obtain the data statistics cutoff time T of the user to be profiled in the user data now , the last charging time T of the user to be profiled rec , and the data statistics cutoff time T of the user to be profiled using the charging proximity calculation formula now , the last charging time T of the user to be profiled rec Calculate and obtain the charging proximity data R of the user to be profiled, where: R=T now -T rec Wherein, R represents the charging proximity data; Obtain the number of times q the user to be profiled charges on day i from the user data i , and the total charging frequency calculation formula is used to calculate the number of charging times q of the user to be profiled on the i-th day i Calculate and obtain the total charging frequency data F of the user to be profiled, where: Wherein, F represents the total charging frequency data of the user to be profiled within the data statistical time interval, n is a positive integer, and i is a positive integer; Obtain the i-th charging power e of the user to be profiled from the user data i , and the charging amount e of the i-th user to be profiled is calculated using the total charging amount calculation formula i Calculate and obtain the total charging amount data E of the user to be profiled, where: Wherein, E represents the total charging amount data of the user to be profiled within the data statistical time interval, n is a positive integer, and i is a positive integer; Obtain the last charging time T of the user to be profiled from the user data now , the first charging time T of the user to be profiled first , and the last charging time T of the user to be profiled using the user life cycle calculation formula now , the first charging time T of the user to be profiled first Calculate and obtain the user life cycle data L of the user to be profiled, where: L=T now -T first Wherein, L represents the user lifecycle data of the user to be profiled; Obtain the time interval t between the nth charging and the n+1th charging of the user to be profiled in the user data gap , and using the charging behavior volatility calculation formula to calculate multiple time intervals t gap Calculation is performed to obtain the charging behavior volatility data V of the user to be profiled, where: Wherein, V represents the charging behavior volatility data of the user to be profiled, t gap represents the sample set of n time intervals of the user to be profiled within the data statistical time interval, Mean(T gap ) represents the sample set T gap The sample mean of , n is a positive integer, i is a positive integer; The charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data corresponding to the user to be profiled are sorted to obtain the target behavior data.

4. The method according to claim 1, wherein After obtaining a threshold classifier and a preset fluctuation threshold corresponding to the threshold classifier, and reading charging behavior volatility data from the target behavior data, the method further includes: If the charging behavior volatility data is less than the preset volatility threshold, then reading the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data from the target behavior data; Obtaining a preset classification standard corresponding to the threshold classifier, wherein the preset classification standard includes a plurality of preset classification results; Reading a preset proximity range and a preset total frequency range corresponding to each preset classification result in the threshold classifier, and determining a target classification result corresponding to the user to be profiled from the multiple preset classification results, wherein the charging proximity data is within the preset proximity range corresponding to the target classification result and the total charging frequency data is within the preset total frequency range corresponding to the target classification result; Obtaining the portrait label corresponding to the target classification result, and using the portrait label as the target portrait label corresponding to the user to be profiled; The target portrait label, the target classification result, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to generate a charging value portrait of the user to be profiled.

5. The method according to claim 1, wherein The method further comprises: Acquire a user data sample, perform normalization calculation on the user data sample to obtain a data sample to be clustered, obtain a preset cluster number range, and extract a plurality of preset cluster numbers within the preset cluster number range; Obtaining a silhouette coefficient, and selecting a preset number of clusters from the plurality of preset numbers of clusters using the silhouette coefficient as a target cluster number; The target number of clusters is used as the target number of calculations for the target cluster center operation; Obtaining a distance probability formula, calculating target cluster centers of the data samples to be clustered using the distance probability formula, and repeatedly calculating the target cluster centers of the data samples to be clustered until the number of calculations reaches the target number of calculations, thereby obtaining a plurality of target cluster centers whose number is equal to the target number of calculations; Creating a first blank cluster for each target cluster center to obtain multiple first blank clusters, obtaining multiple data sample points in the data sample to be clustered, and adding the multiple data sample points to the multiple first blank clusters respectively to obtain multiple target clusters; For each target cluster, determining the coordinates of a plurality of data sample points included in the target cluster to obtain a plurality of first data point coordinates, averaging the plurality of first data point coordinates to obtain a first average coordinate point, and using the first average coordinate point as a first cluster center corresponding to the target cluster; Processing the coordinates of the multiple first data points of each target cluster respectively to obtain a first cluster center corresponding to each target cluster, and obtaining a convergence condition, and processing the first cluster center corresponding to each target cluster using the convergence condition to obtain multiple target clustering results; Acquire preset portrait classification rules, use the preset portrait classification rules to process the multiple target clustering results to obtain multiple portrait categories, and use the multiple portrait categories to generate the target clustering classification standard.

6. The method according to claim 5, characterized in that The selecting a preset number of clusters from the plurality of preset numbers of clusters using the silhouette coefficient as the target number of clusters comprises: For each of the plurality of preset cluster numbers, a mean clustering algorithm is obtained, and the mean clustering algorithm is used to calculate the data sample to be clustered and the preset number of clusters to obtain a predicted clustering result; Acquire multiple predicted clustering results corresponding to multiple preset cluster numbers, and respectively evaluate and calculate the multiple predicted clustering results using the silhouette coefficient to obtain multiple evaluation coefficients corresponding to the multiple predicted clustering results; The multiple evaluation coefficients are arranged in descending order, the evaluation coefficient ranked first is used as the target evaluation coefficient, and the preset cluster number corresponding to the target evaluation coefficient is determined from the multiple preset cluster numbers, and the preset cluster number is used as the target cluster number.

7. The method according to claim 5, characterized in that The calculating the target cluster center of the data sample to be clustered by using the distance probability formula includes: Selecting any data sample point in the data samples to be clustered as an initial cluster center, and determining other data sample points in the data samples to be clustered except the initial cluster center as a plurality of first sample points; Calculating the distance between the initial cluster center and each of the plurality of first sample points to obtain a plurality of sample distance values, and calculating the plurality of sample distance values ​​using the distance probability formula to obtain a plurality of selection probabilities; A roulette wheel selection algorithm is obtained, and a selection probability is selected from the multiple selection probabilities using the roulette wheel selection algorithm as a target selection probability, and a first sample point corresponding to the target selection probability is used as a target cluster center.

8. The method according to claim 5, characterized in that The step of adding the plurality of data sample points to the plurality of first blank clusters to obtain a plurality of target clusters includes: Calculating the Euclidean distance between each of the data sample points and the multiple target cluster centers respectively to obtain multiple first Euclidean distance values ​​corresponding to each of the data sample points; Selecting a first Euclidean distance value with the smallest distance value from the multiple first Euclidean distance values ​​corresponding to each data sample point as a target first Euclidean distance value corresponding to each data sample point; For each data sample point, determining a target cluster center corresponding to the target first Euclidean distance value according to the target first Euclidean distance value corresponding to the data sample point, and adding the data sample point to the first blank cluster corresponding to the target cluster center; According to the target first Euclidean distance value corresponding to each data sample point, the multiple data sample points are respectively added to the multiple first blank clusters to obtain the multiple target clusters.

9. The method according to claim 5, characterized in that The first cluster center corresponding to each target cluster is processed using the convergence condition to obtain multiple target clustering results, including: Performing convergence calculation on the first cluster center corresponding to each target cluster to obtain multiple convergence results, and comparing the multiple convergence results with the convergence condition; If a convergence result among the multiple convergence results does not meet the convergence condition, obtaining multiple data sample points in the data sample to be clustered, and calculating the Euclidean distance between each data sample point and the multiple first cluster centers respectively, to obtain multiple second Euclidean distance values ​​corresponding to each data sample point; Selecting the second Euclidean distance value with the smallest distance value from the multiple second Euclidean distance values ​​corresponding to each data sample point as the target second Euclidean distance value corresponding to each data sample point; Creating a second blank cluster for each of the first cluster centers to obtain a plurality of second blank clusters; For each data sample point, determining a first cluster center corresponding to the target second Euclidean distance value according to the target second Euclidean distance value corresponding to the data sample point, and adding the data sample point to a second blank cluster corresponding to the first cluster center; According to the target second Euclidean distance value corresponding to each data sample point, the multiple data sample points are added to the multiple second blank clusters to obtain multiple designated clusters; For each of the designated clusters, read coordinates of multiple second data points in the designated cluster, average the coordinates of the multiple second data points to obtain a second average coordinate point, and use the second average coordinate point as a second cluster center corresponding to the designated cluster; Processing the coordinates of the multiple second data points of each designated cluster respectively to obtain a second cluster center corresponding to each designated cluster, generating multiple clustering results using the multiple second cluster centers, and using the multiple clustering results as multiple target clustering results; If the multiple convergence results meet the convergence condition, multiple first clustering centers are used to generate multiple target clustering results.

10. A charging value portrait generation device, characterized in that: include: a processing module, configured to respond to a request for generating a charging value profile, obtain user data of the user to be profiled, and process the user data using a preset data analysis model to obtain target behavior data; a reading module, configured to obtain a threshold classifier and a preset fluctuation threshold value corresponding to the threshold classifier, and read charging behavior volatility data from the target behavior data, wherein the charging behavior volatility data indicates the volatility of the charging behavior of the user to be profiled; A first profiling module is configured to obtain a target clustering classification standard if the charging behavior volatility data is greater than or equal to the preset volatility threshold, and perform a profiling operation on the target behavior data using the target clustering classification standard to obtain a charging value profile of the user to be profiled; The first portrait module is also used to obtain multiple portrait categories included in the target cluster classification standard, read the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category in the target cluster classification standard, obtain the preset dimension data point conversion rule, use the preset dimension data point conversion rule to process the charging proximity standard value, the total charging frequency standard value, the total charging amount standard value, the user life cycle standard value, and the charging behavior volatility standard value corresponding to each portrait category, obtain the standard center point corresponding to each portrait category, read the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data in the target behavior data, and use the preset dimension data point conversion rule to process the charging proximity data, the total charging frequency data, the total charging amount data, and the user life cycle data. The data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are processed to obtain the data center point to be profiled corresponding to the user to be profiled, the Euclidean distance between the data center point to be profiled and the standard center point corresponding to each portrait category is calculated to obtain multiple Euclidean distance values, the Euclidean distance value with the smallest distance value is selected from the multiple Euclidean distance values ​​as the target Euclidean distance value, the designated standard center point indicated by the target Euclidean distance value is determined from the multiple standard center points corresponding to the multiple portrait categories, and the portrait category corresponding to the designated standard center point is determined as the target portrait category of the user to be profiled, the target portrait category, the charging proximity data, the total charging frequency data, the total charging amount data, the user life cycle data, and the charging behavior volatility data are sorted to obtain a charging value portrait of the user to be profiled.

11. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A readable storage medium having a computer program stored thereon, 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 9 are implemented.

Images