Electricity payment collection strategy design method, system and storage medium

Abstract

The present invention belongs to the field of electric power technology, and is a method, system and storage medium for designing electric power payment collection strategies. The method includes the following steps: obtaining and processing payment collection SMS sending records and user payment records, and performing correlation analysis between the two through time to determine the relationship between SMS and payment; using a clustering algorithm to calculate the user's payment preference day; constructing characteristic indicators and generating an indicator system, aggregating by user number, extracting SMS sending date and payment collection amount from the payment collection SMS, and constructing a reference indicator for user SMS effectiveness analysis; calculating the characteristic variable IV value of the reference indicator; selecting a reference indicator with strong predictive ability based on the IV value, and mining the concentration and discrete trends of the data to design a user SMS effectiveness label; and designing a corresponding payment collection strategy based on the design results of the user SMS effectiveness label. The present invention constructs a SMS effectiveness analysis model to refine features and solidify them into rule labels, which helps in the formulation of payment collection strategies and achieves the accurate sending of effective payment collection SMS.

Description

Technical Field

[0001] The present invention belongs to the field of electric power technology, and specifically relates to an electric power bill collection strategy design method, system and storage medium. Background Art

[0002] At present, the power grid system mainly uses various text messages to urge electricity bills and payment reminders. The text messages with the reminder function include: payment reminder, reminder notice, electricity bill issuance, power outage for arrears, smart payment warning, smart payment before power outage, smart payment after power outage text messages, payment reminders, etc. Among them, payment reminders and reminder notices are manually triggered to urge, which is the current mainstream SMS reminder method; other types are automatically triggered after the user meets certain conditions, including reminders after the user pays, and reminders after the user's current electricity bill drops to a certain threshold. Relatively speaking, they are slightly more functional and slightly less reminder.

[0003] Currently, the collection staff mainly collects the payment through SMS messages sent to all users in arrears at fixed time intervals. In this mode, many users do not respond to the collection messages, resulting in meaningless sending of a large number of SMS messages, and it is impossible to carry out differentiated SMS collection based on the individual characteristics of users. Summary of the invention

[0004] In order to solve the problems existing in the existing payment collection business model, the present invention provides a method, system and storage medium for designing an electricity payment collection strategy. By establishing a text message effectiveness mechanism based on the payment behavior of electricity users, the historical payment collection situation of electricity users is converted into numerical indicators, and a text message effectiveness analysis model is constructed to refine features and solidify them into rule labels, which helps in the formulation of payment collection strategies to efficiently and quickly distinguish user types, and provides effective technical support for accurately sending effective payment collection text messages, saving useless text messages and reducing costs.

[0005] On the one hand, the technical solution adopted by the present invention is: a method for designing a strategy for collecting electricity bills, the method comprising the following steps:

[0006] S1. Obtain the sending record of reminder SMS and the user payment record;

[0007] S2. Process the user's payment records and the reminder SMS sending records, and perform correlation analysis on the user's payment records and the reminder SMS sending records by time to determine the relationship between the SMS and the payment;

[0008] S3. Use clustering algorithm to calculate the user's preferred payment date. The cluster is regarded as a dense object area separated by a low-density area in the data space, and the high-density area is regarded as the user's concentrated payment area. The user's historical payment date data is clustered to calculate whether the user tends to pay on a certain day and the corresponding preferred date;

[0009] S4. Construct characteristic indicators based on the data processed in step S2 and step S3, and generate an indicator system, aggregate the data by user number, count the payment information and arrears related information of the user within a period of time, and extract the SMS sending date and the amount of the reminder from the reminder SMS to construct a reference indicator for analyzing the effectiveness of the user SMS;

[0010] S5, calculating the IV value of the characteristic variable of the reference indicator to evaluate the characteristic prediction ability of the reference indicator;

[0011] S6. According to the IV value of the characteristic variable of the reference indicator, select the reference indicator with strong prediction ability, and mine the concentration and dispersion trends of the data to design the user SMS effectiveness label;

[0012] S7. Design a corresponding payment collection strategy based on the result of designing the user SMS effectiveness label in S6.

[0013] On the other hand, the present invention also provides a power bill collection strategy design system, which includes the following modules:

[0014] The data acquisition module is used to obtain the sending records of reminder SMS messages and user payment records;

[0015] The data association analysis processing module is used to process the user payment records and the collection SMS sending records, and to analyze the user payment records and the collection SMS sending records by time to determine the relationship between the SMS and the payment;

[0016] The payment preference day calculation module uses a clustering algorithm to calculate the user's payment preference day. The cluster is regarded as a dense object area separated by a low-density area in the data space, and the high-density area is regarded as the user's concentrated payment area. The user's historical payment day data is clustered to calculate whether the user tends to pay on a certain day and the corresponding preferred date;

[0017] The indicator construction module is used to construct characteristic indicators based on the data processed by the data association analysis processing module and the payment preference day calculation module, and generate an indicator system, aggregate by user number, count the user's payment information and arrears related information within a period of time, and extract the SMS sending date and reminder amount from the reminder SMS to construct reference indicators for user SMS effectiveness analysis;

[0018] A characteristic variable IV value calculation module is used to calculate the characteristic variable IV value of the reference indicator to evaluate the characteristic prediction ability of the reference indicator;

[0019] The user SMS effectiveness label design module is used to select reference indicators with strong predictive ability according to the characteristic variable IV value of the reference indicator, and to mine the concentration and dispersion trends of the data to design the user SMS effectiveness label;

[0020] The payment collection strategy design module designs corresponding payment collection strategies based on the results of the user SMS effectiveness label design;

[0021] The payment preference day calculation module uses the DBSCAN clustering algorithm to cluster the user's payment days as samples. If the user's payment days are closely clustered within the range of a certain day, that day is taken as the user's payment preference day.

[0022] The characteristic variable IV value calculation module uses the reference indicator data and divides the reminder SMS into positive and negative samples based on whether the payment is paid within a few days after the reminder SMS is sent processed by the data association analysis processing module. The data of users who have paid within a few days after the reminder SMS is sent is set as a positive sample, and the data of users who have not paid within a few days is set as a negative sample, and constructs relevant features to convert the problem into a binary classification problem of machine learning; the characteristic variable IV value of the reference indicator is calculated for encoding the input variable and evaluating the predictive ability. The size of the characteristic variable IV value of the reference indicator indicates the strength of the variable's predictive ability; and then the characteristic indicators with strong predictive ability are selected according to the size of the characteristic variable IV value.

[0023] On the other hand, the present invention also provides a storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the various steps of the electricity bill collection strategy design method of the present invention are implemented.

[0024] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0025] The present invention establishes a text message effectiveness mechanism based on the payment behavior of electricity users, converts the historical payment collection information of electricity users into numerical indicators, constructs a text message effectiveness analysis model through a machine learning algorithm, refines features and solidifies them into rule labels, and helps business personnel to formulate collection strategies based on the analysis of user text message effectiveness, so as to efficiently and quickly distinguish user types, and provide effective technical support for accurately sending effective collection text messages, saving useless text messages and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram of a flow chart of a method for designing a power bill collection strategy in an embodiment of the present invention;

[0027] Figure 2 A heat map showing the cross analysis of the user's preferred payment day and the SMS sending day in an embodiment of the present invention;

[0028] Figure 3 The structure block diagram of the power bill collection strategy design system in the embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be further described in detail below in conjunction with embodiments and drawings, but the implementation manner of the present invention is not limited thereto.

[0030] Example 1

[0031] like Figure 1 As shown, this embodiment is a method for designing a power bill collection strategy. The method is based on the analysis of the effectiveness of user text messages. The technical means adopted include data acquisition, data analysis and processing, using the DBSCAN clustering algorithm (Density-Based Spatial Clustering of Applications with Noise) to calculate the preferred payment date, construct a feature width table, calculate the feature IV value, design a user text message effectiveness label, and design a bill collection strategy. The method specifically includes the following steps:

[0032] Step 1, data acquisition: obtain data tables such as the collection of payment reminder SMS sending records and user payment records, and filter key information such as SMS categories, subcategories, SMS sending time, SMS content, SMS receiving number, user payment time, user payment method, etc. from the collection of payment reminder SMS sending records and user payment records. This key information will be used for subsequent user SMS effectiveness analysis.

[0033] Step 2: Data analysis and processing: Process the user's payment records and the collection SMS sending records, and conduct correlation analysis on the user's payment records and the collection SMS sending records by time to determine the relationship between the SMS and the payment.

[0034] This embodiment associates the time between the reminder SMS sending record and the user payment record to calculate the payment frequency after the reminder SMS is sent. After analysis, the user payment frequency is extremely high within 2 days after the reminder SMS is sent, and the payment frequency tends to be stable from 3 days after the reminder SMS is sent. Based on the above phenomenon, it can be considered that the user payment behavior within 2 days after the reminder SMS is sent is affected by the SMS, and the SMS that prompts the user to pay within 2 days is a valid SMS.

[0035] Step 3: Use the DBSCAN clustering algorithm to calculate the user's preferred payment date. This algorithm is a density-based clustering algorithm that considers clusters as dense object areas separated by low-density areas in the data space, while high-density areas can be considered as user concentrated payment areas. The user's historical payment date data is clustered to calculate whether the user tends to pay on a certain day and the corresponding preferred date.

[0036] Due to the different situations of each user, the payment habits of different users are quite different. The most important one is the difference in the payment time of users. It is particularly important to understand the payment time habits of users. However, the number of electricity users is too large to be completed by manpower. Therefore, the DBSCAN clustering algorithm is used to analyze the payment date samples of users to determine the user's preferred payment date.

[0037] DBSCAN is a density-based clustering algorithm. This type of density clustering algorithm generally assumes that the category can be determined by the density of sample distribution. Samples of the same category are closely connected, that is, there must be samples of the same category not far away from any sample of the category. In addition, this type of density clustering algorithm can find outliers while clustering and is not sensitive to outliers in the data set. According to the principle of DBSCAN, it can be clearly found that if the user's payment day is clustered as a sample, if the user's payment day is closely clustered within the range of a certain day, the day obviously has a special meaning to the user, making the user tend to pay on that day, and the day is regarded as the user's payment preference day.

[0038] The calculation process of the user's payment preference day specifically includes the following steps:

[0039] Step 31, obtain a sample X of the historical payment date data of a certain user, and determine whether the number of user samples is greater than or equal to 12, that is, the number of samples for at least 1 year. If the sample is insufficient, return a null value. If the sample is sufficient, according to the definition of the user's payment preference day, the user should pay around a certain date, so the neighborhood radius ε is set to 1.

[0040] Let any point be p, then the corresponding ε neighborhood is:

[0041] N ε ={q∈D|dist(p,q)≤ε}

[0042] The corresponding density is:

[0043] ρ(x)=|N ε (x)|

[0044] The upper quartile of the total number of samples is used to calculate the minimum point minPts of the user, where N X The number of samples for this user is:

[0045] minPts=N X *0.75

[0046] Among them, the upper quartile refers to the degree of dispersion of skewed data when the data is described by the quartile statistical descriptive analysis method, that is, when all the data are arranged from small to large, the number that is exactly arranged in the upper 1 / 4 position is called the upper quartile; that is, in terms of percentage, the upper quartile refers to the number at the 75% position.

[0047] Step 32: Perform clustering according to the radius ε set in step 31 and the minimum point minPts obtained to obtain the core point of the payment day density corresponding to the user. The core point x is calculated as follows:

[0048] ρ(x)≥minPts

[0049] Step 33: Finally, all the core points x of the user are averaged to calculate the user's preferred day c, and all user sample data are iterated to calculate the user's payment preference day in turn.

[0050] Step 4, construct characteristic indicators: construct characteristic indicators based on the data processed in steps 2 and 3, and generate an indicator system, aggregate by user number, and count the payment information and arrears related information of users in a period of time (for example, in the past year, that is, in the past 12 months), including payment reminder collection rate, collection notice collection rate, intelligent payment warning collection rate, power outage warning number, arrears power outage number, overdue number, average payment time, payment preference day, etc.; and extract information such as SMS sending date and collection amount from the collection SMS to construct the reference indicators shown in Table 1 below for analysis of the effectiveness of user SMS:

[0051] Table 1 Main reference indicators for analysis of user SMS effectiveness

[0052] type index describe 1 Payment reminder rate in the past 12 months The situation of each user being urged to pay during the payment period during the probation period 2 Collection characteristics and collection rates in the last 12 months During the probation period, each user was urged to pay after exceeding the payment period 3 Number of power outage warnings in the past 12 months The number of times each user was notified of power outages due to unpaid bills during the inspection period 4 Smart payment warning and payment reminder rate in the past 12 months The payment status of each user after reaching the reminder threshold during the probation period 5 Average payment duration in the past 12 months Average time taken by each user to pay after receiving SMS during the inspection period 6 SMS reminder reliance rate in the past 12 months Examine whether each user relies on SMS reminders to pay 7 Do you have a preferred payment date? Check whether each user is used to paying on a certain day 8 The number of days between the preferred day and the day the SMS is sent Evaluate the time interval between the date the reminder SMS is sent and the date the user prefers 9 Preferred payment method Examine each user's payment habits 10 Number of overdue payments in the past 12 months Examine each user's payment habits 11 Is the SMS reminder successful? Whether the payment is successfully collected within 2 days after this message is sent

[0053] Step 5: Calculate the IV value (Information Value) of the characteristic variable of the reference indicator to evaluate the characteristic prediction ability of the reference indicator. The specific steps include the following:

[0054] Step 51: Calculate the characteristic variable IV value of the reference indicator.

[0055] Since the impact of payment reminder text messages on electricity users is difficult to evaluate and it is difficult to grasp the overall rules through business experience, the reference indicator data is used to divide the payment reminder text messages into positive and negative samples based on whether the payment is paid within a certain number of days (for example, within 2 days) after the payment reminder text message is sent based on the association analysis in step 2. The data of users who have paid within 2 days after the text message is sent is set as a positive sample, and the data of users who have not paid within 2 days is set as a negative sample. Relevant features are constructed, and the problem is converted into a binary classification problem of machine learning. Big data analysis is used to understand the overall characteristics.

[0056] In the binary classification problem of machine learning, the IV value of the characteristic variable of the reference indicator is calculated, which is mainly used to encode the input variable and evaluate its predictive ability. The size of the IV value of the characteristic variable of the reference indicator indicates the strength of the variable's predictive ability.

[0057] Step 52: feature screening, screening feature indicators with strong predictive capabilities according to the size of the feature variable IV value.

[0058] Take the text messages of power users in a certain period of time as samples, use the indicators constructed in step 4 as features, calculate the IV values ​​of all feature variables, and select the features with strong influence as the reference for judgment. Among them, when the IV value of the feature variable is less than 0.03, it means no predictive ability; the IV value between 0.03 and 0.09 indicates weak predictive ability; the IV value between 0.1 and 0.29 indicates medium predictive ability; the IV value between 0.3 and 0.49 indicates strong predictive ability; and the IV value greater than 0.5 indicates that the feature has extremely strong predictive ability.

[0059] After selecting the electricity user's bill collection SMS data for a certain month in this step, based on the calculation in step 51, the main characteristic variable IV value is shown in Table 2 below:

[0060] Table 2 IV values ​​of main reference indicators

[0061] Serial number feature IV Value 1 Payment reminder rate in the past 12 months 0.531806084 2 The number of days between the payment preference date and the SMS sending date 0.31677815 3 Average payment duration in the past 12 months 0.251459993 4 Number of overdue payments in the last 12 months 0.162446425 5 The collection rate of payment reminders in the past 12 months 0.152550962 6 Number of power outage warnings in the past 12 months 0.095186498 7 SMS reminder reliance rate in the past 12 months 0.065895712 8 Smart payment warning and reminder rate in the past 12 months 0.058285148 9 Do you have a preferred payment date? 0.058157343 10 Preferred payment method 0.058134454

[0062] Step 6: Design user SMS effectiveness labels: According to the IV value of the characteristic variable of the reference indicator, select reference indicators with strong predictive ability, and explore the concentration and dispersion trends of the data to design SMS effectiveness labels.

[0063] This step constructs a user SMS effectiveness evaluation model, which is used to design user SMS effectiveness labels. The constructed evaluation model uses xgboost for supervised learning, and uses the feature index obtained by screening according to the feature variable IV value in step 5 as the feature input, and whether the user pays within 2 days after receiving the SMS as the label input. The user SMS effectiveness evaluation model is trained based on the above feature input and label input. Take the user-related data before the evaluation month, calculate the corresponding feature index, and input it into the user SMS effectiveness evaluation model. The model output is the probability that the user pays after receiving the SMS in the evaluation month (also known as the payment reminder rate). The higher the output result of the probability of paying after receiving the SMS, the greater the possibility that the user pays after receiving the SMS in the evaluation month.

[0064] From the features selected based on IV values, we can see that the impact on the effectiveness of user reminder SMS messages is basically due to the user's payment reminder rate (i.e. historical payment status) and the time difference between the SMS sending date and the user's payment preference date. Therefore, based on these two types of features, user effectiveness labels and time effectiveness labels are set respectively.

[0065] User effectiveness label: The characteristics of the user's payment reminder rate in the past 12 months have a strong predictive ability. Combined with the sample data distribution, users whose payment reminder rate in the past 12 months (i.e., model evaluation result, or model output probability) is above 0.75 are evaluated as high SMS effectiveness users, users whose payment reminder rate in the past 12 months is between 0.2 and 0.75 are evaluated as medium SMS effectiveness users, and users whose payment reminder rate in the past 12 months is lower than 0.2 are evaluated as low SMS effectiveness users.

[0066] Time zone effectiveness label: Based on the user payment preference day obtained by the above steps, combined with the SMS sending day, a heat map of the payment preference day and the SMS sending day is created (e.g. Figure 2 ) analysis, it can be seen that the combination of the two has a great impact on the effectiveness of the collection SMS. Figure 2 The heat map shown is divided into three SMS time effectiveness intervals based on the interval between the SMS sending time and the user's payment preference day. When the SMS sending day is earlier than the user's payment preference day and the SMS sending day is close to the user's payment preference day, for example, the user's payment preference day is within 2 days after the SMS sending day, the collection success rate is the highest, which is the high SMS effectiveness time zone; when the SMS sending day is earlier than the user's payment preference day but the SMS sending day is not close to the user's payment preference day, the collection success rate is the lowest, which is the low SMS effectiveness time zone; when the SMS sending day is later than the user's payment preference day, there is still a certain collection effect, which is the medium SMS effectiveness time zone. For example, suppose a user prefers to pay on the 15th; around the 15th, assuming that the user forgets to pay because he is busy or has other things to do, if a text message reminder is sent after the time when the user is most accustomed to paying (for example, on the 18th), the user will most likely take into account factors such as the payment penalty and respond positively to the reminder text message; however, if a reminder text message is sent to the user much earlier than the 15th, for example on the 10th, because the user is accustomed to paying on the 15th, he will most likely ignore the reminder text message.

[0067] Based on the above steps, the user SMS effectiveness labels are sorted as shown in Table 3:

[0068] Table 3 User SMS effectiveness label rules table

[0069] Tag Type Label Business significance User effectiveness label SMS service users Model output probability>=0.75 User effectiveness label High SMS efficiency users 0.2<=Model output probability <0.75 User effectiveness label Low SMS efficiency users Model output probability < 0.2 Time Zone Effect Tags High SMS Effectiveness Time Zone The SMS sending date is close to the preferred date Time Zone Effect Tags Time zone for SMS The SMS is sent on a date later than the preferred date Time Zone Effect Tags Low SMS effectiveness time zone The SMS is sent on a date earlier than the preferred date

[0070] Step 7: Based on the result of designing the user SMS effectiveness label in step 6, design a corresponding collection strategy. The specific example is shown in Table 4 below:

[0071] Table 4 Strategy design example table

[0072]

[0073] That is to say, the collection strategy designed in this step is: for users with high effectiveness of SMS collection, users with payment preference days will receive collection SMS in the high effectiveness time zone, while users without payment preference days will remain unchanged; for users with medium effectiveness of SMS collection, users with payment preference days will receive collection SMS in the high effectiveness time zone, while users without payment preference days will receive collection SMS on the 10th by default; for users with low effectiveness of SMS collection, users with payment preference days will receive collection SMS in the high effectiveness time zone, while users without payment preference days will receive collection SMS on the 10th by default.

[0074] Example 2

[0075] Based on the same concept, the present invention also provides a power bill collection strategy design system based on user SMS effectiveness analysis, such as Figure 3 As shown, it includes:

[0076] A data acquisition module, used to implement step 1 of embodiment 1, to acquire the sending record of the reminder SMS and the user payment record;

[0077] A data analysis and processing module, used to implement step 2 of embodiment 1, process user payment records and reminder SMS sending records, associate user payment records and reminder SMS sending records with time, and determine the relationship between the SMS and the payment;

[0078] The payment preference day calculation module is used to implement step 3 of embodiment 1, calculate the user's payment preference day using a clustering algorithm, regard the cluster as a dense object area separated by a low-density area in the data space, and the high-density area is considered to be a user's concentrated payment area, cluster the user's historical payment day data, and calculate whether the user tends to pay on a certain day and the corresponding preferred date;

[0079] A characteristic index construction module is used to implement step 4 of Example 1, construct characteristic indexes based on the data processed by the data association analysis processing module and the payment preference day calculation module, and generate an index system, aggregate by user number, count the payment information and arrears related information of the user within a period of time, and extract the SMS sending date and the amount of the reminder from the reminder SMS to construct a reference index for the effectiveness analysis of the user SMS;

[0080] A characteristic variable IV value calculation module, used to implement step 5 of embodiment 1, calculate the characteristic variable IV value of the reference index to evaluate the characteristic prediction ability of the reference index;

[0081] A user SMS effectiveness label design module is used to implement step 6 of Example 1, select reference indicators with strong predictive ability according to the characteristic variable IV value of the reference indicator, and mine the concentration and dispersion trends of the data to design the user SMS effectiveness label;

[0082] The payment collection strategy design module is used to implement step 7 of Example 1, and designs a corresponding payment collection strategy based on the result of designing the user's SMS effectiveness label.

[0083] Among them, the payment preference day calculation module adopts the DBSCAN clustering algorithm to cluster the user's payment day as a sample. If the user's payment day is closely clustered within the range of a certain day, that day will be used as the user's payment preference day.

[0084] The characteristic variable IV value calculation module uses the reference indicator data and divides the reminder SMS into positive and negative samples based on whether the payment is paid within a few days after the reminder SMS is sent processed by the data association analysis processing module. The data of users who have paid within a few days after the reminder SMS is sent is set as a positive sample, and the data of users who have not paid within a few days is set as a negative sample, and constructs relevant features to convert the problem into a binary classification problem of machine learning; the characteristic variable IV value of the reference indicator is calculated for encoding the input variable and evaluating the predictive ability. The size of the characteristic variable IV value of the reference indicator indicates the strength of the variable's predictive ability; and then the characteristic indicators with strong predictive ability are selected according to the size of the characteristic variable IV value.

[0085] For the system described in this embodiment, since the implementation of the technical solution corresponds to the method of Example 1, the description of this embodiment is relatively brief. For the corresponding technical features, please refer to the description of each step in Example 1, which will not be repeated here.

[0086] Example 3

[0087] Based on the same inventive concept, this embodiment provides a storage medium on which computer instructions are stored; when the above-mentioned computer instructions are executed by the processor, steps 1 to 7 of the electricity bill collection strategy design method of embodiment 1 are implemented to complete the design of the electricity bill collection strategy.

[0088] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a CD, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0089] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for designing a strategy for collecting electricity bills, characterized in that: The following steps are involved: S1. Obtain the sending record of reminder SMS and the user payment record; S2. Process the user's payment records and the reminder SMS sending records, and perform correlation analysis on the user's payment records and the reminder SMS sending records by time to determine the relationship between the SMS and the payment; S3. Use clustering algorithm to calculate the user's preferred payment date. The cluster is regarded as a dense object area separated by a low-density area in the data space, and the high-density area is regarded as the user's concentrated payment area. The user's historical payment date data is clustered to calculate whether the user tends to pay on a certain day and the corresponding preferred date; S4. Construct characteristic indicators based on the data processed in step S2 and step S3, and generate an indicator system, aggregate the data by user number, count the payment information and arrears related information of the user within a period of time, and extract the SMS sending date and the amount of the reminder from the reminder SMS to construct a reference indicator for analyzing the effectiveness of the user SMS; S5, calculating the IV value of the characteristic variable of the reference indicator to evaluate the characteristic prediction ability of the reference indicator; S6. According to the IV value of the characteristic variable of the reference indicator, select the reference indicator with strong prediction ability, and mine the concentration and dispersion trends of the data to design the user SMS effectiveness label; S7. Design a corresponding payment collection strategy based on the result of designing the user SMS effectiveness label in S6; Step S5 includes: S51, using the reference indicator data, based on the correlation analysis in step S2, whether the payment reminder SMS is paid within a few days after it is sent as a basis for dividing the payment reminder SMS into positive and negative samples, setting the data of the user paying within a few days after the payment reminder SMS is sent as a positive sample, and setting the data of the user not paying within a few days as a negative sample, and constructing relevant features to convert the problem into a binary classification problem of machine learning; calculating the characteristic variable IV value of the reference indicator for encoding the input variable and evaluating the prediction ability, and the size of the characteristic variable IV value of the reference indicator indicates the strength of the variable's prediction ability; S52. Select characteristic indicators with strong predictive ability according to the size of the characteristic variable IV value.

2. The method for designing a power bill collection strategy according to claim 1, characterized in that: Step S3 uses the DBSCAN clustering algorithm to cluster the user's payment days as samples. If the user's payment days are closely clustered within a certain day, the day is used as the user's preferred payment day.

3. The method for designing a power bill collection strategy according to claim 2, characterized in that: Step S3 includes the following steps: S31. Obtain a sample X of the historical payment date data of a certain user, and determine whether the number of samples is greater than or equal to 12. If the number of samples is insufficient, return a null value; if the number of samples is sufficient, according to the definition of the user's payment preference date, the user should pay the fee before or after a certain date, so the neighborhood radius is set. is 1; Let any point be p, then the corresponding The neighborhood is: ； The corresponding density is: ； The upper quartile of the total number of samples is used to calculate the minimum point minPts of the user, where N x The number of samples for this user is: ； S32, according to the set radius The obtained minimum point minPts is clustered to obtain the core point of the payment day density corresponding to the user. The core point x is calculated as follows: ； S33, averaging all the core points x of the user to calculate the user's preferred day c; and iterating all user sample data to calculate the user's payment preference day in turn.

4. The method for designing a power bill collection strategy according to claim 1, characterized in that: The payment information and arrears-related information in step S4 include payment reminder collection rate, collection notice collection rate, intelligent payment warning collection rate, number of power outage warnings, number of arrears power outages, number of overdue payments, average payment time and payment preference day.

5. The method for designing a power bill collection strategy according to claim 1, characterized in that: When step S52 screens characteristic indicators, when the IV value of the characteristic variable is less than 0.03, it indicates no predictive ability; when the IV value is between 0.03 and 0.09, it indicates weak predictive ability; when the IV value is between 0.1 and 0.29, it indicates medium predictive ability; when the IV value is between 0.3 and 0.49, it indicates strong predictive ability; when the IV value is greater than 0.5, it indicates extremely strong predictive ability.

6. The method for designing a power bill collection strategy according to claim 1, characterized in that: Step S6 designs a user SMS effectiveness label by constructing a user SMS effectiveness evaluation model; the constructed evaluation model uses xgboost for supervised learning, and uses the feature index obtained by screening according to the feature variable IV value in step S5 as the feature input, and whether the user pays the bill within a few days after receiving the SMS as the label input, and obtains the user SMS effectiveness evaluation model based on the above feature input and label input training; obtains user-related data before the evaluation month, calculates the corresponding feature index, and inputs it into the user SMS effectiveness evaluation model, then the model output is the probability that the user pays the bill after receiving the SMS in the evaluation month.

7. The method for designing a power bill collection strategy according to claim 6, characterized in that: The user SMS effectiveness tag designed in step S6 includes a user effectiveness tag and a time effectiveness tag; For user effectiveness labels, users with a model output probability of more than 0.75 are evaluated as high SMS effectiveness users, users with a model output probability between 0.2 and 0.75 are evaluated as medium SMS effectiveness users, and users with a model output probability below 0.2 are evaluated as low SMS effectiveness users; For the time effectiveness label, based on the user's payment preference day and the SMS sending day, a heat map analysis of the payment preference day and the SMS sending day is conducted; according to the heat map, the interval between the SMS sending time and the user's payment preference day is divided into three SMS time effectiveness intervals: when the SMS sending day is earlier than the user's payment preference day and the SMS sending day is close to the user's payment preference day, the collection success rate is the highest, which is the high SMS effectiveness time zone; when the SMS sending day is earlier than the user's payment preference day but the SMS sending day is not close to the user's payment preference day, the collection success rate is the lowest, which is the low SMS effectiveness time zone; when the SMS sending day is later than the user's payment preference day, it is the medium SMS effectiveness time zone; The collection strategy designed in step S7 is: for users with high effectiveness of SMS collection, users with a payment preference day will receive collection SMS in a high effectiveness time zone, while users without a payment preference day will remain unchanged; for users with medium effectiveness of SMS collection, users with a payment preference day will receive collection SMS in a high effectiveness time zone, while users without a payment preference day will receive collection SMS on the 10th by default; for users with low effectiveness of SMS collection, users with a payment preference day will receive collection SMS in a high effectiveness time zone, while users without a payment preference day will receive collection SMS on the 10th by default.

8. The electricity bill collection strategy design system is characterized by: Includes the following modules: The data acquisition module is used to obtain the sending records of reminder SMS messages and user payment records; The data association analysis processing module is used to process the user payment records and the collection SMS sending records, and to analyze the user payment records and the collection SMS sending records by time to determine the relationship between the SMS and the payment; The payment preference day calculation module uses a clustering algorithm to calculate the user's payment preference day. The cluster is regarded as a dense object area separated by a low-density area in the data space, and the high-density area is regarded as the user's concentrated payment area. The user's historical payment day data is clustered to calculate whether the user tends to pay on a certain day and the corresponding preferred date; The indicator construction module is used to construct characteristic indicators based on the data processed by the data association analysis processing module and the payment preference day calculation module, and generate an indicator system, aggregate by user number, count the user's payment information and arrears related information within a period of time, and extract the SMS sending date and reminder amount from the reminder SMS to construct reference indicators for user SMS effectiveness analysis; A characteristic variable IV value calculation module is used to calculate the characteristic variable IV value of the reference indicator to evaluate the characteristic prediction ability of the reference indicator; The user SMS effectiveness label design module is used to select reference indicators with strong predictive ability according to the characteristic variable IV value of the reference indicator, and to mine the concentration and dispersion trends of the data to design the user SMS effectiveness label; The payment collection strategy design module designs corresponding payment collection strategies based on the results of the user SMS effectiveness label design; The payment preference day calculation module uses the DBSCAN clustering algorithm to cluster the user's payment days as samples. If the user's payment days are closely clustered within the range of a certain day, that day is taken as the user's payment preference day. The characteristic variable IV value calculation module uses the reference indicator data and divides the reminder SMS into positive and negative samples based on whether the payment is paid within a few days after the reminder SMS is sent processed by the data association analysis processing module. The data of users who have paid within a few days after the reminder SMS is sent is set as a positive sample, and the data of users who have not paid within a few days is set as a negative sample, and constructs relevant features to convert the problem into a binary classification problem of machine learning; the characteristic variable IV value of the reference indicator is calculated for encoding the input variable and evaluating the predictive ability. The size of the characteristic variable IV value of the reference indicator indicates the strength of the variable's predictive ability; and then the characteristic indicators with strong predictive ability are selected according to the size of the characteristic variable IV value.

9. A storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by the processor, each step of the method for designing an electricity bill collection strategy described in any one of claims 1 to 7 is implemented.

Images