CN116522227A - Charging user image drawing method and device based on multi-label data, medium and equipment - Google Patents

Abstract

The present application discloses a charging user portrait method and device, medium and equipment based on multi-label data. The method includes: obtaining the original charging data of multiple charging users, and preprocessing the original charging data to obtain data fields corresponding to multiple preset attributes; respectively performing clustering operations on the data fields corresponding to each preset attribute to obtain The cluster center corresponding to the preset attribute and the portrait label corresponding to each cluster center; obtain the charging data of the user to be analyzed, and determine the target label of the user to be analyzed in the portrait label according to the charging data to be analyzed, according to the target label Generate user portraits of users to be analyzed. The method of the present application solves the problem that the authenticity of the survey questionnaire in the existing method is not strong, and the hard clustering cannot be accurate to the user, which leads to insufficient understanding of the user and the inability to fully understand the charging behavior of the user.

Description

Charging user portrait method and device, medium, and equipment based on multi-label data

technical field

The present application relates to the field of electric vehicles, in particular to a charging user portrait method and device, medium and equipment based on multi-label data.

Background technique

With the continuous expansion of new energy vehicles, a large number of electric vehicles (EV: Electric Vehicle) will be connected to the power system in the near future, which will bring more uncertainties to the power system. In this regard, studying the laws behind user charging behavior data mining will be beneficial to the optimization of distribution network management. In the field of user behavior research, user portraits, as a labeled user model abstracted from data to describe user attributes, preferences, and behavior habits, have been widely used in the field of e-commerce. Enterprises can use user portraits to gain a deeper understanding The needs of each user, so as to develop personalized services.

In the field of new energy vehicles, user portraits are generally based on big data or questionnaire surveys. Due to the instructive setting of questionnaire questions, low feedback rate, and weak authenticity, it cannot be used as a stable portrait method. The portrait method of big data mainly completes user classification through hard clustering based on division. Usually, multi-category attributes are used to synthesize multi-dimensional data for clustering. The final result is the user cluster and the number of clusters, but it is not accurate to the user. , the historical data of a single user may appear in multiple clusters.

Contents of the invention

In view of this, this application provides a charging user portrait method and device, medium and equipment based on multi-label data, which solves the problem that the authenticity of the questionnaire in the existing method is not strong, and the hard clustering cannot be accurate to the user, which leads to the lack of confidence in the user. The understanding is not deep enough to fully understand the problem of user charging behavior.

According to one aspect of the present application, a charging user portrait method based on multi-label data is provided, including:

Acquiring original charging data of multiple charging users, and performing preprocessing on the original charging data to obtain data fields corresponding to multiple preset attributes;

performing a clustering operation on the data fields corresponding to each of the preset attributes to obtain the cluster centers corresponding to the preset attributes and the portrait labels corresponding to each of the cluster centers;

Acquiring charging data to be analyzed of the user to be analyzed, determining a target tag of the user to be analyzed in the portrait tag according to the charging data to be analyzed, and generating a user portrait of the user to be analyzed according to the target tag.

Optionally, determining the target tag of the user to be analyzed in the portrait tag according to the charging data to be analyzed includes:

Extracting a field group to be analyzed corresponding to each of the preset attributes from the charging data to be analyzed, wherein the field group to be analyzed includes at least one field to be analyzed and an action occurrence corresponding to each field to be analyzed time;

determining the time decay coefficient of each field to be analyzed according to the moment when the action occurs;

For each of the preset attributes, calculate the matching degree between the field group to be analyzed and the portrait label of each of the preset attributes based on the time decay coefficient, and determine that the portrait label with the highest matching degree is the The target tags of the users to be analyzed under the preset attributes are described.

Optionally, the preprocessing of the raw charging data includes:

Data cleaning is performed on the original charging data to obtain data after cleaning, wherein the data cleaning includes at least one of the following: abnormal data cleaning, repeated data cleaning, and null data cleaning;

Data fields corresponding to each of the preset attributes are respectively extracted from the cleaned data, wherein the preset attributes include duration attributes, power anxiety attributes, and charging power attributes.

Optionally, the data fields corresponding to each of the preset attributes are respectively extracted from the cleaned data, including:

Extracting the value corresponding to the charging duration field from the cleaned data as the data field corresponding to the duration attribute;

Extracting the value corresponding to the initial remaining power field and the value corresponding to the end remaining power field from the cleaned data, as the data field corresponding to the power anxiety attribute;

According to the value corresponding to the field of single charging amount and the value corresponding to the field of single charging duration in the data after cleaning, calculate the average charging power of single pile, and use the average charging power of single pile as the corresponding value of the charging power attribute data field.

Optionally, before extracting data fields corresponding to each of the preset attributes from the cleaned data, the method further includes:

converting each of the cleaned data into a preset data format;

extracting time data from the cleaned data, and converting the time data into a preset time format;

It is judged whether the user identification is missing in the cleaned data, and if it is missing, the user identification is supplemented by verification data, wherein the verification data is data provided by the charging service system.

Optionally, the clustering operation is performed on the data fields corresponding to each of the preset attributes to obtain the cluster centers corresponding to the preset attributes and the portrait labels corresponding to each of the cluster centers, including:

Each of the preset attributes is used as the attribute to be clustered, the data field corresponding to the attribute to be clustered is used as the field to be clustered, and each of the fields to be clustered is mapped to a mathematical space to be clustered data point;

Set the number K of clusters corresponding to the attribute to be clustered, and select K data points as cluster centers in the data points to be clustered, and use each cluster center as a cluster;

Calculate the distance between each of the data points to be clustered and each of the cluster centers, and divide the data points to be clustered into clusters corresponding to the nearest cluster centers;

Update the clustering centers of the clusters according to the coordinates of each data point in each cluster, and return to the step of calculating the distance between each of the data points to be clustered and each of the clustering centers, until satisfying Default stop condition;

According to the data points in the cluster where each cluster center is located, the portrait label corresponding to the cluster center is determined.

Optionally, the setting the cluster number K corresponding to the attribute to be clustered includes:

preset a plurality of cluster values, and respectively calculate the silhouette coefficient corresponding to the data points to be clustered under each of the cluster values;

Determine the cluster value with the largest silhouette coefficient as the number of clusters.

According to another aspect of the present application, a charging user portrait device based on multi-label data is provided, the device comprising:

A preprocessing module, configured to obtain the original charging data of multiple charging users, and preprocess the original charging data to obtain data fields corresponding to multiple preset attributes;

A clustering module, configured to perform clustering operations on the data fields corresponding to each of the preset attributes, to obtain the cluster centers corresponding to the preset attributes and the portrait labels corresponding to each of the cluster centers;

The portrait module is used to obtain the charging data to be analyzed of the user to be analyzed, determine the target tag of the user to be analyzed in the portrait tag according to the charging data to be analyzed, and generate the user to be analyzed according to the target tag user portrait.

Optionally, the portrait module is used for:

Extracting a field group to be analyzed corresponding to each of the preset attributes from the charging data to be analyzed, wherein the field group to be analyzed includes at least one field to be analyzed and an action occurrence corresponding to each field to be analyzed time;

determining the time decay coefficient of each field to be analyzed according to the moment when the action occurs;

For each of the preset attributes, calculate the matching degree between the field group to be analyzed and the portrait label of each of the preset attributes based on the time decay coefficient, and determine that the portrait label with the highest matching degree is the The target tags of the users to be analyzed under the preset attributes are described.

Optionally, the preprocessing module is used for:

Data cleaning is performed on the original charging data to obtain data after cleaning, wherein the data cleaning includes at least one of the following: abnormal data cleaning, repeated data cleaning, and null data cleaning;

Data fields corresponding to each of the preset attributes are respectively extracted from the cleaned data, wherein the preset attributes include duration attributes, power anxiety attributes, and charging power attributes.

Optionally, the preprocessing module is used for:

Extracting the value corresponding to the charging duration field from the cleaned data as the data field corresponding to the duration attribute;

Extracting the value corresponding to the initial remaining power field and the value corresponding to the end remaining power field from the cleaned data, as the data field corresponding to the power anxiety attribute;

According to the value corresponding to the field of single charging amount and the value corresponding to the field of single charging duration in the data after cleaning, calculate the average charging power of single pile, and use the average charging power of single pile as the corresponding value of the charging power attribute data field.

Optionally, the preprocessing module is used for:

converting each of the cleaned data into a preset data format;

extracting time data from the cleaned data, and converting the time data into a preset time format;

It is judged whether the user identification is missing in the cleaned data, and if it is missing, the user identification is supplemented by verification data, wherein the verification data is data provided by the charging service system.

Optionally, the clustering module is used for:

Each of the preset attributes is used as the attribute to be clustered, the data field corresponding to the attribute to be clustered is used as the field to be clustered, and each of the fields to be clustered is mapped to a mathematical space to be clustered data point;

Set the number K of clusters corresponding to the attribute to be clustered, and select K data points as cluster centers in the data points to be clustered, and use each cluster center as a cluster;

Calculate the distance between each of the data points to be clustered and each of the cluster centers, and divide the data points to be clustered into clusters corresponding to the nearest cluster centers;

Update the clustering centers of the clusters according to the coordinates of each data point in each cluster, and return to the step of calculating the distance between each of the data points to be clustered and each of the clustering centers, until satisfying Default stop condition;

According to the data points in the cluster where each cluster center is located, the portrait label corresponding to the cluster center is determined.

Optionally, the clustering module is used for:

preset a plurality of cluster values, and respectively calculate the silhouette coefficient corresponding to the data points to be clustered under each of the cluster values;

Determine the cluster value with the largest silhouette coefficient as the number of clusters.

According to another aspect of the present application, a medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the above method for charging user portrait based on multi-label data is implemented.

According to another aspect of the present application, a device is provided, including a storage medium and a processor, the storage medium stores a computer program, and when the processor executes the computer program, the above-mentioned charging user portrait method based on multi-label data is implemented.

With the help of the above technical solution, this application utilizes multi-dimensional clustering method to mine data, conduct portrait analysis and build user portraits. The constructed portraits are beneficial to power companies to optimize management and formulate better charging schemes. Compared with the traditional questionnaire survey method, the original charging data of this method comes from the public charging pile data collection system, which has higher authenticity and a larger amount of data, so the portrait is more accurate. Compared with the traditional method of classifying users based on partition-based hard clustering, it is possible to profile each user specifically and accurately to each attribute of a specific user, so the profile results are more refined and more practical.

The above description is only an overview of the technical solution of the present application. In order to better understand the technical means of the present application, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and understandable , the following specifically cites the specific implementation manner of the present application.

Description of drawings

The drawings described here are used to provide a further understanding of the application and constitute a part of the application. The schematic embodiments and descriptions of the application are used to explain the application and do not constitute an improper limitation to the application. In the attached picture:

Fig. 1 shows a schematic flow diagram of a charging user portrait method based on multi-label data provided by an embodiment of the present application;

Fig. 2 shows a schematic flow diagram of another charging user portrait method based on multi-label data provided by the embodiment of the present application;

Fig. 3 shows a schematic flow diagram of a preprocessing scheme of another charging user portrait method based on multi-label data provided by an embodiment of the present application;

Fig. 4 shows a schematic flowchart of another clustering scheme based on multi-label data charging user portrait method provided by the embodiment of the present application;

Fig. 5 shows a structural block diagram of a charging user portrait device based on multi-label data provided by an embodiment of the present application.

Detailed ways

Hereinafter, the present application will be described in detail with reference to the drawings and embodiments. It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

In this embodiment, a charging user portrait method based on multi-label data is provided, as shown in Figure 1, the method includes:

Step 101, acquiring the original charging data of multiple charging users, and preprocessing the original charging data to obtain data fields corresponding to multiple preset attributes;

Step 102, performing a clustering operation on the data fields corresponding to each preset attribute to obtain the cluster centers corresponding to the preset attributes and the portrait labels corresponding to each cluster center;

Step 103: Obtain the charging data to be analyzed of the user to be analyzed, determine the target tag of the user to be analyzed in the portrait tag according to the charging data to be analyzed, and generate a user portrait of the user to be analyzed according to the target tag.

The charging user portrait method based on multi-label data provided by the embodiment of this application is used to mine user charging behavior data, perform portrait analysis, and construct user portraits. The constructed portraits are beneficial to power companies to optimize management and formulate better charging schemes .

Specifically, first obtain the original charging data of the charging user, wherein the original data comes from the database in the background of the public charging pile data acquisition system, for example, customer_id (customer ID), start_soc (starting remaining power), end_soc can be obtained in the database (End remaining power), chargkwh (single charge), usetime (charging duration), chargestarttime (charging start time), chargeendtime (charging end time), station_id (charging station ID) and other fields. It is understandable that the larger the amount of raw data, the more accurate the portrait effect.

Preprocess the obtained raw charging data to obtain the fields corresponding to each preset attribute, and then cluster each attribute to achieve multi-dimensional clustering. The clustering method can use any general method, which is not described here. Do limited. Multiple cluster centers are obtained for each attribute, and its corresponding portrait label is obtained according to the specificity of each cluster center. The portrait label is used to describe a certain aspect of the user's characteristics, such as preferring to charge during peak hours, etc. Among them, the normalization operation can be performed on the field before clustering to unify its dimensions, and the denormalization operation can be performed after the clustering operation to restore it.

Preferably, the field corresponding to each preset attribute is a numerical field, that is, the field value is a numerical value instead of a text. For example, if each address in the field corresponding to station_id (charging station ID) is represented by text, such as: Dajiang Station, then this field is considered to be a field with a label rather than a numerical field.

Based on this, when analyzing the profile of the user to be analyzed, the target tag corresponding to each preset attribute of the user to be analyzed can be determined according to the charging data to be analyzed, and multiple target tags are comprehensively analyzed to describe the characteristics of the user from multiple dimensions , to obtain the user profile of the user to be analyzed.

This embodiment utilizes the method of multi-dimensional clustering to mine data, conduct portrait analysis and build user portraits. The constructed portraits are beneficial to power companies to optimize management and formulate better charging schemes. Compared with the traditional questionnaire survey method, the original charging data in this embodiment comes from the public charging pile data collection system, which has higher authenticity and a larger amount of data, so the portrait is more accurate. Compared with the traditional method of classifying users based on partition-based hard clustering, it is possible to profile each user specifically and accurately to each attribute of a specific user, so the profile results are more refined and more practical.

Figure 2 shows a flow chart of the technical solution of the method for charging user portraits based on multi-label data according to an embodiment of the present application. Unlabeled fields, where the preprocessing includes two steps of data cleaning and attribute construction. After that, normalization operation, k-means clustering operation, and denormalization operation are performed on the output fields in turn to obtain the cluster centers of the four attributes and the portrait labels corresponding to each cluster center. Furthermore, in the specific analysis process, the data to be analyzed corresponding to the users to be analyzed can be analyzed from the four directions of charging period, charging time, charging rate and power anxiety, and the target labels corresponding to the four directions can be obtained respectively, and then the data can be obtained from the four directions. Depict user portraits to obtain user characteristics of users to be analyzed.

Furthermore, as a refinement and extension of the specific implementation of the above-mentioned embodiment, in order to fully describe the specific implementation process of this embodiment, another charging user portrait method is provided. In this method, according to the charging data to be analyzed in the portrait label Determine the target label of the user to be analyzed in , including the following steps:

Step 201, respectively extract the field group to be analyzed corresponding to each preset attribute from the charging data to be analyzed, wherein the field group to be analyzed includes at least one field to be analyzed and the action occurrence time corresponding to each field to be analyzed;

Step 202, determine the time decay coefficient of each field to be analyzed according to the moment when the action occurs;

Step 203, for each preset attribute, calculate the matching degree between the field group to be analyzed and the portrait label of each preset attribute based on the time decay coefficient, and determine the portrait label with the highest matching degree as the user to be analyzed under the preset attribute target label for .

In steps 301-303, the user to be analyzed is analyzed from multiple preset attributes, that is, multiple dimensions. Firstly, the field group to be analyzed corresponding to the preset attribute is extracted from the charging data to be analyzed corresponding to the user to be analyzed. The specific extraction method is similar to the aforementioned attribute construction method and will not be repeated here. In a specific application scenario, each preset attribute can be analyzed separately in turn. Since each preset attribute corresponds to multiple cluster centers, and each cluster center corresponds to a portrait label, for each preset attribute, Which portrait tag matches the field group to be analyzed can be analyzed, and then the matched portrait tag is used as the target tag. After the matching is completed, multiple target tags can be obtained, and the target tags correspond to the preset attributes one by one, and then the portrait of the user to be analyzed can be generated according to the multiple target tags.

Wherein, as time goes by, the user's preference may change, for example, last month, the user prefers to charge at night, but this month, the user prefers to charge during the day. Considering the influence of time, in the specific matching process, the time decay coefficient can be introduced, and the target label can be determined by the time decay method. Specifically, each field to be analyzed corresponds to an action occurrence time, which is used to indicate the time when the corresponding action occurs, and the time decay coefficient is set for the field to be analyzed according to the time when the action occurs. It can be understood that the earlier the action occurs, the More decay, and relatively less recent action decay. In the process of matching target tags, the time decay degree can be used as a weight. For example, the charging record of the user last month is displayed as charging at night, while the charging record of this month is displayed as charging during the day. It is considered that the weight of charging time in the previous month is relatively small, and the weight of charging time in this month is relatively large. Therefore, the portrait label of charging during the day is selected as the target label corresponding to the charging period attribute.

Optionally, preprocessing the raw charging data includes the following steps:

Step 301, performing data cleaning on the original charging data to obtain the cleaned data, wherein the data cleaning includes at least one of the following: abnormal data cleaning, repeated data cleaning and null data cleaning;

Step 302, respectively extract the data field corresponding to each preset attribute from the cleaned data, wherein the preset attribute includes the duration attribute, battery anxiety attribute and charging power attribute, extract the data field corresponding to each preset attribute from the cleaned data respectively A data field, where the preset attributes include duration attributes, battery anxiety attributes, and charging power attributes.

In steps 301-302, preprocessing operations include data cleaning and attribute construction. Specifically, during the data cleaning process, data that has a negative impact on the profiling method, such as abnormal data, duplicate data, and null data, can be eliminated. The abnormal data includes hardware abnormal data, software abnormal data, illegal operation data, and the like. Through the data cleaning operation, the quality and accuracy of the portrait results can be improved.

After the data cleaning operation is completed, the data field corresponding to each preset attribute can be extracted from the cleaned data, and then when the clustering operation is performed on each preset attribute, the corresponding data field of the preset attribute can be analyzed in a targeted manner. The data fields of similar users are divided into one category to further improve the effect of multi-dimensional clustering.

Optionally, the data fields corresponding to each preset attribute are respectively extracted from the cleaned data, including the following steps:

Step 401, extracting the value corresponding to the charging duration field from the cleaned data as the data field corresponding to the duration attribute;

Step 402, extracting the value corresponding to the initial remaining power field and the value corresponding to the end remaining power field from the cleaned data as the data field corresponding to the power anxiety attribute;

Step 403 , according to the value corresponding to the single charge field and the value corresponding to the single charge duration field in the cleaned data, calculate the average charging power of the single pile, and use the average charging power of the single pile as the data field corresponding to the charging power attribute.

In steps 401-403, usetime (charging duration) can be extracted as the data field corresponding to the duration attribute, and start_soc (starting remaining power) and end_soc (starting remaining power) form a two-dimensional data field corresponding to the power anxiety attribute; The average charging power of a single pile obtained by dividing chargkwh (single charging amount) and usetime (charging duration) is used as the data field corresponding to the charging power attribute.

Among them, the calculation formula corresponding to the charging power attribute is as follows:

Wherein, E _n represents the charging capacity of the nth charging, T _n represents the charging time of the nth charging, and P _n represents the average power within the nth charging time.

Optionally, before extracting the data fields corresponding to each preset attribute from the cleaned data, the following steps are further included:

Step 501, converting each cleaned data into a preset data format;

Step 502, extracting the time data in the cleaned data, and converting the time data into a preset time format;

In step 503, it is judged whether the user identification is missing in the cleaned data, and if it is missing, the user identification is supplemented by verification data, wherein the verification data is data provided by the charging service system.

In steps 501-503, the format of the cleaned data is unified and the user ID is completed. Specifically, since the acquired data may have errors such as full-width input, half-width input, space symbols, and wrong field formats, the format of the cleaned data is converted into a unified data format for subsequent operations. After the format conversion, the time conversion can be performed on the time data alone to ensure that the format of all time data is unified, for example, both 2020.04.20 and 2020/04/20 are uniformly converted to 20200420. Finally, the user ID completion operation is performed on the data. It is understandable that for missing data, corresponding measures can be taken according to the importance of the data. For example, the method of direct deletion can be used for missing data with low importance, while the important High missing data can be completed from other channels, calculated using other fields, or filled in based on historical experience. Since the user ID is relatively important, it is possible to read data from the charging business system and perform Completion method to improve data integrity.

Figure 3 shows a schematic diagram of the preprocessing flow of the charging user portrait method based on multi-label data according to an embodiment of the present application. As shown in the figure, the preprocessing includes data cleaning and attribute screening/construction steps, and the multidimensional raw data is processed in the data cleaning step Perform abnormal data cleaning, repeated data cleaning, and null data cleaning operations, and then enter the attribute screening/construction step, first perform format conversion, time conversion, and user ID completion operations, and then perform attribute construction on the obtained data to obtain three different types. The attribute data is the data field corresponding to the instant length attribute, the data field corresponding to the two-dimensional battery anxiety attribute, and the data field corresponding to the charging power attribute.

Optionally, clustering operations are performed on the data fields corresponding to each preset attribute to obtain the cluster centers corresponding to the preset attributes and the portrait labels corresponding to each cluster center, including the following steps:

Step 601, taking each preset attribute as the attribute to be clustered, the data field corresponding to the attribute to be clustered as the field to be clustered, and mapping each field to be clustered to a data point to be clustered in the mathematical space;

Step 602, setting the number K of clusters corresponding to the attribute to be clustered, and selecting K data points as cluster centers among the data points to be clustered, and using each cluster center as a cluster;

Step 603, respectively calculating the distance between each data point to be clustered and each cluster center, and dividing the data points to be clustered into clusters corresponding to the nearest cluster center;

Step 604, update the cluster center of the cluster according to the coordinates of each data point in each cluster, and return to the step of calculating the distance between each data point to be clustered and each cluster center, until the preset stop is satisfied condition;

Step 605, according to the data points in the cluster where each cluster center is located, determine the portrait label corresponding to the cluster center.

In steps 601-605, each preset attribute is used as an attribute to be clustered, and k-means is used to perform a clustering operation on each attribute to be clustered. K-means is a kind of clustering algorithm, the principle is simple, easy to understand, and the operation speed is fast. In a specific clustering operation process, firstly, each field to be clustered is mapped to a data point to be clustered in a mathematical space, wherein each data point to be clustered can be expressed in the form of a vector. Then determine the specific value of the cluster number K, wherein the K value is used to identify K clusters obtained through clustering, that is, K groups. After determining the K value, select K data points in the set composed of all data points to be clustered as the cluster center, and for each data point in the set, calculate its distance from the cluster center, which cluster center is the closest It depends on which clustering center it belongs to. When all data points are determined to belong to the clustering center, the first round of clustering operation is completed. At this time, K clusters are formed, and each cluster includes a clustering center and Several other data points.

After the first round of clustering operation, within each cluster, a new cluster center is calculated according to the coordinates of all data points in the cluster, and the next round of clustering operation is performed again, and the operation steps are the same as those of the first round of clustering The operation is similar and will not be repeated here. The above operation is performed repeatedly until the preset stop condition is met, then the clustering operation is considered to be completed. After the clustering operation is completed, K clusters are formed, and each cluster includes a cluster center and several other data points. At this time, the portrait label of the cluster can be generated according to the characteristics of the data points in each cluster, that is, The portrait label corresponding to the cluster center.

Among them, the preset stop condition can be that the number of iterations reaches the maximum number of iterations, or that the cluster center converges, etc., for example, in several iterations, if the distance between the new cluster center and the old cluster center is equal to If it is less than the preset threshold, the cluster center is considered to be convergent.

By clustering each preset attribute, multiple cluster centers corresponding to each preset attribute can be obtained, and then the portrait analysis is performed from the perspective corresponding to each preset attribute, and the portrait label and label explanation are obtained.

Optionally, setting the cluster number K corresponding to the attribute to be clustered includes the following steps:

Step 701, preset a plurality of cluster values, and respectively calculate the silhouette coefficient corresponding to the data points to be clustered under each cluster value;

Step 702, determine the cluster value with the largest silhouette coefficient as the number of clusters.

In steps 701-702, the number of clusters K is determined using the silhouette coefficient method. It can be understood that the silhouette coefficient is a coefficient used to describe the similarity between the target cluster and other clusters. The larger the value, the higher the matching relationship between the target and its own cluster, and the higher the matching relationship with other clusters. The lower can be used to describe the compactness of a class. In the specific application process, first read the data points to be clustered, and then calculate the silhouette coefficients corresponding to the data points to be clustered under each preset K value, and determine the K value with the largest silhouette coefficient as the best K value, Use this as the final K value.

This embodiment uses the silhouette coefficient method to determine the optimal K value, compared with the method of randomly determining the K value and performing clustering, the clustering effect of this embodiment is better.

Fig. 4 shows a flow chart of a clustering scheme of a charging user portrait method based on multi-label data according to an embodiment of the present application. As shown in the figure, the clustering scheme includes the following steps:

Step 1: Input the normalized n-dimensional data samples to be clustered.

Step 2: Get the optimal cluster value based on the silhouette coefficient method, denoted as K, which is the optimal cluster number, randomly select K points in the mathematical space as the initial clustering center of the sample set, and initialize.

Step 3: For each data point in the sample set, calculate the Euclidean distance between them and the cluster center, and assign them to the cluster center with the shortest distance to form a cluster.

Step 4: Update the cluster center of each cluster, and use the coordinate average of all data points in each cluster as the new cluster center.

Step 5: Determine whether the new cluster center is convergent, and if convergent, output the clustering result; if not, skip to step 3 and continue to execute.

The Euclidean distance calculation formula used in this method is as follows:

Among them, d represents the distance in the mathematical space between two n-dimensional data points x and y, and also represents the degree of adjacency of these two points.

In a specific application scenario, the one-year charging record data of a grid charging pile in a certain area is collected. For customer_id (customer ID), start_soc (starting remaining power), end_soc (end remaining ), usetime (charging duration), chargestarttime (charging start time), chargeendtime (charging end time), station_id (charging station ID) and other data fields for mining. Based on the integrated development environment of PyCharm, the construction of the portrait device is completed by using the Python language. Use pandas and numpy tools to complete data cleaning, data screening/construction and data classification. Clustering operations are done using sklearn tools.

In the data cleaning step, the cleaning of abnormal data, repeated data, and null value data is completed. The abnormal data includes: hardware abnormal data, software abnormal data, and illegal operation data.

Use the K-means clustering method to cluster the SOC attribute, that is, the battery anxiety attribute (stat_SOC, end_SOC), the charging duration attribute, and the charging power attribute, and use the silhouette coefficient method to output the optimal cluster number. Clustering is performed several times and the clustering results are analyzed. Finally, the SOC data clustering analysis outputs 5 types of behavior portrait labels; the charging time cluster analysis outputs 3 types of portrait labels; the charging power cluster analysis outputs 4 types of portrait labels, after clustering portraits The results are shown in Table 1 below.

Table 1 Results of charging behavior clustering portrait

After that, all the portrait labels obtained by the clustering method and the direct classification method are used for final portrait label matching by the time decay method. Among them, the clustering method can use the aforementioned k-means method, or other clustering methods can be used for direct classification The method is to classify and obtain portrait labels based on historical experience, and select the final portrait label that best represents the user's recent behavior from each type of behavioral portrait label.

The image analysis is as follows:

(1) The power attribute data clustering results are divided into four labels according to the relative size of the cluster centers: extremely low, medium speed, fast, and super fast. Among them, the proportion of extremely fast users is extremely low, the number of medium-speed users is the largest, and the number of low-speed and fast users is equal. To a certain extent, this portrait can reflect the matching between the user and the charging pile, as well as the relative speed of charging.

(2) The user's charging time can be divided into three categories, namely the first category - short time, the second category - long time, and the third category - general duration. Such classification labels are determined according to the relative size of the user's usage and charging time and the number in the cluster.

(3) SOC attributes can be divided into four categories:

Depletion-type large-amount charging type: This type of user chooses to recharge when the power is close to zero, and when the charging ends, it is close to full power, so this type of user consumes a relatively large amount of power during charging, which is defined as depletion-type large-amount charging type. Among all users, this type of users has the largest number and proportion, and belongs to the conventional type.

Economical: This type of user is mainly characterized by low-level charging when the power is close to zero. This type of user may clearly know the mileage required for their trip, and they will arrange to charge to the appropriate amount of power to save charging time. This type of user is defined as economical.

Severe mileage anxiety: This type of users has obvious behavioral characteristics. The cluster center shows that they usually choose to recharge when the SOC is not lower than 50%. Such behavior shows that they are very likely to be users who want to travel long distances. Defined as severe range anxiety.

Mild mileage anxiety: Such users choose to recharge when the battery is about 30% left. They may be worried about the mileage of the remaining power, miss the opportunity to charge this time, and the remaining power may be difficult to support to find the next charging station, so it is defined as mild mileage anxiety.

After that, for any user to be analyzed, each attribute can be analyzed according to the data to be analyzed, and it can be judged that the user is more suitable for the portrait label in each attribute, and then a user portrait portrayed from multiple angles can be obtained.

It should be understood that the sequence numbers of the steps in the above embodiments do not mean the order of execution, and the execution order of each process should be determined by its functions and internal logic, and should not constitute any limitation to the implementation process of the embodiment of the present invention.

Further, as a specific implementation of the charging user portrait method based on multi-label data, an embodiment of the present application provides a charging user portrait device based on multi-label data, as shown in FIG. 5 , the device includes: a preprocessing module, Clustering module and portrait module.

A preprocessing module, configured to obtain the original charging data of multiple charging users, and preprocess the original charging data to obtain data fields corresponding to multiple preset attributes;

The clustering module is used to perform a clustering operation on the data field corresponding to each preset attribute, and obtain the cluster center corresponding to the preset attribute and the portrait label corresponding to each cluster center;

The portrait module is used to obtain the charging data to be analyzed of the user to be analyzed, determine the target label of the user to be analyzed in the portrait label according to the charging data to be analyzed, and generate the user portrait of the user to be analyzed according to the target label.

In a specific application scenario, optionally, the portrait module is used to:

Extracting a field group to be analyzed corresponding to each preset attribute from the charging data to be analyzed, wherein the field group to be analyzed includes at least one field to be analyzed and an action occurrence time corresponding to each field to be analyzed;

Determine the time decay coefficient of each field to be analyzed according to the moment when the action occurs;

For each preset attribute, calculate the matching degree between the field group to be analyzed and the portrait label of each preset attribute based on the time decay coefficient, and determine the portrait label with the highest matching degree as the target label of the user to be analyzed under the preset attribute .

In a specific application scenario, optionally, the preprocessing module is used to:

Perform data cleaning on the original charging data to obtain the cleaned data, wherein the data cleaning includes at least one of the following: abnormal data cleaning, repeated data cleaning, and null data cleaning;

The data fields corresponding to each preset attribute are respectively extracted from the cleaned data, wherein the preset attribute includes the duration attribute, battery anxiety attribute, and charging power attribute.

In a specific application scenario, optionally, the preprocessing module is used to:

Extract the value corresponding to the charging duration field from the cleaned data as the data field corresponding to the duration attribute;

Extract the value corresponding to the initial remaining power field and the value corresponding to the end remaining power field from the cleaned data as the data field corresponding to the power anxiety attribute;

According to the value corresponding to the single charge field and the value corresponding to the single charge duration field in the cleaned data, calculate the average charging power of the single pile, and use the average charging power of the single pile as the data field corresponding to the charging power attribute.

In a specific application scenario, optionally, the preprocessing module is used to:

Convert each cleaned data into a preset data format respectively;

Extract the time data from the cleaned data and convert the time data to a preset time format;

It is judged whether the user identification is missing in the cleaned data, and if it is missing, the user identification is completed by using the verification data, wherein the verification data is the data provided by the charging service system.

In a specific application scenario, optionally, the clustering module is used to:

Each preset attribute is used as the attribute to be clustered, the data field corresponding to the attribute to be clustered is used as the field to be clustered, and each field to be clustered is mapped to a data point to be clustered in the mathematical space;

Set the cluster number K corresponding to the attribute to be clustered, and select K data points as cluster centers among the data points to be clustered, and use each cluster center as a cluster;

Calculate the distance between each data point to be clustered and each cluster center, and divide the data points to be clustered into the cluster corresponding to the nearest cluster center;

Update the cluster center of the cluster according to the coordinates of each data point in each cluster, and return to the step of calculating the distance between each data point to be clustered and each cluster center, until the preset stop condition is met;

According to the data points in the cluster where each cluster center is located, the portrait label corresponding to the cluster center is determined.

In a specific application scenario, optionally, the clustering module is used to:

Preset multiple cluster values, and calculate the silhouette coefficient corresponding to the data points to be clustered under each cluster value;

Determine the cluster value with the largest silhouette coefficient as the number of clusters.

It should be noted that for other corresponding descriptions of the functional modules involved in the multi-tag data-based charging user portrait device provided in the embodiment of the present application, reference may be made to the corresponding descriptions in the above methods, and details are not repeated here.

Based on the above method, correspondingly, an embodiment of the present application also provides a storage medium on which a computer program is stored, and when the program is executed by a processor, the above method for charging user portraits based on multi-label data is implemented.

Based on this understanding, the technical solution of the present application can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.), including several The instructions are used to make an electronic device (which may be a personal computer, a server, or a network device, etc.) execute the methods described in various implementation scenarios of this application.

Based on the above methods shown in Figures 1 to 4, and the virtual device embodiment shown in Figure 5, in order to achieve the above purpose, the embodiment of the present application also provides a device, which can be a personal computer, a server, or a network device. Etc., the electronic device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to realize the above-mentioned charging user portrait method based on multi-label data as shown in Figures 1 to 4 .

Optionally, the electronic device may further include a user interface, a network interface, a camera, a radio frequency (Radio Frequency, RF) circuit, a sensor, an audio circuit, a WI-FI module, and the like. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the like, and optional user interfaces may also include a USB interface, a card reader interface, and the like. Optionally, the network interface may include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface) and the like.

Those skilled in the art can understand that the structure of an electronic device provided in this embodiment does not constitute a limitation to the electronic device, and may include more or less components, or combine some components, or arrange different components.

The storage medium may also include an operating system and a network communication module. An operating system is a program that manages and preserves hardware and software resources of an electronic device, and supports the operation of information processing programs and other software and/or programs. The network communication module is used to realize the communication between various controls inside the storage medium, and communicate with other hardware and software in the physical device.

Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software plus a necessary general-purpose hardware platform, or by hardware.

Those skilled in the art can understand that the accompanying drawing is only a schematic diagram of a preferred implementation scenario, and the units or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art can understand that the units in the devices in the implementation scenario can be distributed among the devices in the implementation scenario according to the description of the implementation scenario, or can be located in one or more devices different from the implementation scenario according to corresponding changes. The units of the above implementation scenarios can be combined into one unit, or can be further split into multiple sub-units.

The serial numbers of the above application are for description only, and do not represent the pros and cons of the implementation scenarios. The above disclosures are only several specific implementation scenarios of the present application, but the present application is not limited thereto, and any changes conceivable by those skilled in the art shall fall within the protection scope of the present application.

Claims (10)

1. A method for charging user image based on multi-tag data, the method comprising:

acquiring original charging data of a plurality of charging users, and preprocessing the original charging data to obtain data fields corresponding to a plurality of preset attributes;

clustering operation is carried out on the data fields corresponding to each preset attribute respectively, so that a clustering center corresponding to the preset attribute and portrait tags corresponding to each clustering center are obtained;

and acquiring charging data to be analyzed of the user to be analyzed, determining a target tag of the user to be analyzed in the portrait tag according to the charging data to be analyzed, and generating a user portrait of the user to be analyzed according to the target tag.

2. The method according to claim 1, wherein the determining the target tag of the user to be analyzed from the portrait tags according to the charging data to be analyzed includes:

Extracting a field group to be analyzed corresponding to each preset attribute from the charging data to be analyzed, wherein the field group to be analyzed comprises at least one field to be analyzed and action occurrence time corresponding to each field to be analyzed;

determining a time attenuation coefficient of each field to be analyzed according to the action occurrence time;

and calculating the matching degree of the field group to be analyzed and the portrait tag of each preset attribute according to each preset attribute based on the time attenuation coefficient, and determining the portrait tag with the highest matching degree as the target tag of the user to be analyzed under the preset attribute.

3. The method of claim 1, wherein the preprocessing the raw charging data comprises:

performing data cleaning on the original charging data to obtain cleaned data, wherein the data cleaning comprises at least one of the following steps: abnormal data cleaning, repeated data cleaning and null data cleaning;

and respectively extracting data fields corresponding to each preset attribute from the cleaned data, wherein the preset attributes comprise a duration attribute, an electric quantity anxiety attribute and a charging power attribute.

4. A method according to claim 3, wherein extracting the data field corresponding to each preset attribute from the cleaned data respectively includes:

extracting a numerical value corresponding to a charging time length field from the cleaned data to serve as a data field corresponding to the time length attribute;

extracting a value corresponding to the initial residual electric quantity field and a value corresponding to the end residual electric quantity from the cleaned data to serve as a data field corresponding to the electric quantity anxiety attribute;

and calculating the single pile charging average power according to the value corresponding to the single charge amount field and the value corresponding to the single charge duration field in the cleaned data, and taking the single pile charging average power as the data field corresponding to the charging power attribute.

5. A method according to claim 3, wherein before extracting the data field corresponding to each preset attribute from the cleaned data, the method further comprises:

respectively converting each cleaned data into a preset data format;

extracting time data in the cleaned data, and converting the time data into a preset time format;

Judging whether the user identification is missing in the cleaned data, if so, complementing the user identification by using check data, wherein the check data is data provided by a charging service system.

6. The method according to claim 1, wherein the clustering operation is performed for the data field corresponding to each preset attribute to obtain a cluster center corresponding to the preset attribute and a portrait tag corresponding to each cluster center, and the clustering method comprises:

respectively taking each preset attribute as an attribute to be clustered, taking a data field corresponding to the attribute to be clustered as a field to be clustered, and mapping each field to be clustered into a data point to be clustered in a mathematical space;

setting a clustering cluster number K corresponding to the attribute to be clustered, selecting K data points from the data points to be clustered as clustering centers, and taking each clustering center as a cluster;

respectively calculating the distance between each data point to be clustered and each clustering center, and dividing the data points to be clustered into clusters corresponding to the closest clustering centers;

updating the clustering center of each cluster according to the coordinates of each data point in each cluster, and returning to the step of respectively calculating the distance between each data point to be clustered and each clustering center until a preset stopping condition is met;

And determining the portrait label corresponding to each cluster center according to the data point in the cluster where each cluster center is located.

7. The method of claim 6, wherein the setting the number K of clusters corresponding to the attribute to be clustered includes:

presetting a plurality of cluster values, and respectively calculating contour coefficients corresponding to the data points to be clustered under each cluster value;

and determining the cluster value with the largest profile coefficient as the cluster number.

8. A charged user imaging device based on multi-tag data, the device comprising:

the preprocessing module is used for acquiring original charging data of a plurality of charging users, and preprocessing the original charging data to obtain data fields corresponding to a plurality of preset attributes;

the clustering module is used for carrying out clustering operation on the data fields corresponding to each preset attribute respectively to obtain a clustering center corresponding to the preset attribute and portrait tags corresponding to each clustering center;

and the portrait module is used for acquiring charging data to be analyzed of a user to be analyzed, determining a target label of the user to be analyzed in the portrait label according to the charging data to be analyzed, and generating a user portrait of the user to be analyzed according to the target label.

9. A storage medium having stored thereon a program or instructions which, when executed by a processor, implement the method of any of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, characterized in that the processor implements the method of any one of claims 1 to 7 when executing the program.