Electricity overdue risk prediction method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a method and device for predicting the risk of overdue electricity bills, electronic equipment and storage medium. The method comprises: obtaining electricity consumption data and payment data of a target user; determining target electricity consumption behavior characteristics of the target user based on the user type of the target user, according to the electricity consumption data and the payment data; determining the user value level of the target user based on the target electricity consumption behavior characteristics by using a pre-trained risk prediction model, and determining the overdue risk level of the target user according to the user value level. In this way, by dual dimension division of user value level and overdue risk level, the user value level and the overdue risk level can be bound, thereby improving the prediction accuracy of the risk prediction model for the overdue risk.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, and storage medium for predicting the risk of overdue electricity bills. Background Technology

[0002] With the diversification of users' electricity needs, electricity bill collection efficiency has become one of the core indicators of power supply companies' operation and management. Current methods for predicting overdue electricity bill risks largely rely on manual experience and simple rule settings, which cannot comprehensively and accurately identify the risk of users defaulting on their electricity bills.

[0003] In view of this, how to comprehensively and accurately identify the risk of users' electricity bill overdue has become an urgent technical problem to be solved. Summary of the Invention

[0004] In view of this, the purpose of this disclosure is to provide a method, apparatus, electronic device and storage medium for predicting the risk of overdue electricity bills in order to solve or partially solve the above-mentioned technical problems.

[0005] To achieve the above objectives, a first aspect of this disclosure proposes a method for predicting the risk of overdue electricity bills, the method comprising:

[0006] Obtain the target users' electricity consumption and payment data; Based on the user type of the target user, the target electricity consumption behavior characteristics of the target user are determined according to the electricity consumption data and the payment data. Using a pre-trained risk prediction model, the user value level of the target user is determined based on the target electricity consumption behavior characteristics, and the overdue electricity bill risk level of the target user is determined based on the user value level.

[0007] Based on the same inventive concept, a second aspect of this disclosure proposes a device for predicting the risk of overdue electricity bills, comprising: The acquisition module is configured to acquire the target user's electricity consumption data and payment data. The feature determination module is configured to determine the target electricity consumption behavior features of the target user based on the user type of the target user, according to the electricity consumption data and the payment data. The risk prediction module is configured to use a pre-trained risk prediction model to determine the user value level of the target user based on the target electricity consumption behavior characteristics, and to determine the overdue electricity bill risk level of the target user based on the user value level.

[0008] Based on the same inventive concept, a third aspect of this disclosure proposes an electronic device including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described above when executing the computer program.

[0009] Based on the same inventive concept, a fourth aspect of this disclosure provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to perform the methods described above.

[0010] As described above, this disclosure provides a method, apparatus, electronic device, and storage medium for predicting electricity bill delinquency risk. It involves acquiring the target user's electricity consumption and payment data. Based on the target user's user type, the target user's target electricity consumption behavior characteristics are determined according to the electricity consumption and payment data. These characteristics comprehensively reflect the target user's electricity consumption and payment behavior. A pre-trained risk prediction model is used to determine the target user's user value level based on the target electricity consumption behavior characteristics, and then the target user's electricity bill delinquency risk level is determined based on the user value level. Thus, by using a dual-dimensional division of user value level and electricity bill delinquency risk level, the user value level and the electricity bill delinquency risk level can be linked, thereby improving the accuracy of the risk prediction model in predicting electricity bill delinquency risk and enabling comprehensive and accurate identification of the target user's electricity bill delinquency risk level. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating the method for predicting the risk of overdue electricity bills according to an embodiment of this disclosure; Figure 2 A flowchart illustrating a method for predicting the risk of overdue electricity bills for individual users, as described in this embodiment of the disclosure. Figure 3 A flowchart of a method for predicting the risk of overdue electricity bills for merchant users according to an embodiment of this disclosure; Figure 4 This is a schematic diagram of the structure of the electricity bill overdue risk prediction device according to an embodiment of the present disclosure; Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0014] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0015] Based on the background description, electricity bill collection is a crucial task in the operation of power companies, directly affecting their economic benefits and cash flow. Accurately identifying the risk of overdue electricity bills helps power companies take corresponding measures in advance, such as implementing differentiated collection strategies for users with different risk levels, ensuring timely and full collection of electricity bills.

[0016] Current methods for identifying overdue electricity bill risks primarily rely on human experience and simple rules, depending solely on past payment records and frequency. This approach is highly subjective and fails to comprehensively capture the dynamic changes in user electricity consumption behavior and potential risk factors, resulting in low efficiency when dealing with large amounts of user data. Early data processing focused on basic statistical analysis, but electricity user consumption and payment behavior are influenced by various factors such as seasons and economic conditions, exhibiting complex nonlinear and dynamic characteristics. Basic statistics struggle to uncover patterns and regularities, leading to insufficient accuracy in identifying overdue electricity bill risks. Single machine learning models, such as logistic regression and decision trees, have also been studied. However, logistic regression assumes linear relationships in data features, making it poorly adaptable to the complex nonlinearities of electricity data. Decision trees are prone to overfitting and have weak generalization ability. Furthermore, single models utilize limited data feature dimensions, failing to comprehensively consider multiple aspects of user behavior, resulting in insufficient accuracy and reliability in risk identification. While user electricity consumption and payment behavior exhibit clear time-series characteristics, previous time-series analyses have largely been limited to simple trend and periodic analyses, making it difficult to uncover long-term dependencies and complex dynamic patterns. In the field of feature extraction, existing methods focus on statistical and simple time-series feature extraction, which struggles to handle long-sequence dependency information in electricity data. Related machine learning models, when used alone for identifying electricity bill delinquency risk, suffer from high computational costs, sensitivity to noise, limited time-series feature processing capabilities, and insufficient generalization performance, and cannot fully integrate multiple feature information. Related regression models are insufficient in characterizing complex nonlinear relationships and multi-factor interactions in electricity data, making it difficult to accurately predict electricity bill delinquency risk. Therefore, an innovative method is urgently needed to deeply mine user dynamic behavior patterns and accurately identify electricity bill delinquency risk.

[0017] Electricity marketing management in commercial complexes presents unique challenges for electricity bill collection: businesses operate in diverse sectors (restaurants, retail, entertainment, etc.), exhibiting complex and varied electricity consumption patterns; their business performance is susceptible to fluctuations in economic conditions, customer traffic, and seasonality, leading to dynamic changes in their ability to repay electricity bills. The limitations of existing risk identification methods become apparent in this scenario.

[0018] Current methods for predicting electricity bill delinquency risk rely on manual experience and simple rule settings. Judgment conditions are manually set based on a user's historical electricity consumption and payment history, including electricity bills, electricity usage, electricity safety, outstanding payments, late fees, and power outages. For example, the number of power outages due to unpaid bills in the past year and the number of late fees incurred in the past six months are used to filter customers into lists with different levels of delinquency risk. This method is highly subjective, lacks quantitative methods, and makes it difficult to comprehensively and accurately identify users' electricity bill delinquency risk.

[0019] Machine learning methods (e.g., logistic regression models, decision trees, etc.) are used to predict user debt risk. Logistic regression constructs a linear regression equation, using various user features as independent variables and whether a user is in arrears as the dependent variable, to find the linear relationship between features and the risk of arrears. However, the relationships between features in electricity data are complex and not simply linear, making logistic regression models less adaptable. Decision tree models, on the other hand, make decisions based on a tree structure, classifying samples according to different features. Each internal node is a test of an attribute, the branches are the test outputs, and the leaf nodes are the categories. However, decision trees are prone to overfitting due to noise and complex structures in the data, resulting in poor generalization ability. Furthermore, they have limited dimensions for utilizing data features, making it difficult to comprehensively consider multiple aspects of user behavior, leading to low accuracy and reliability in risk identification.

[0020] Treating electricity bill delinquency risk as a binary classification (high risk / low risk) problem. Features such as user electricity consumption, bill amount, number of delinquencies, and account age are extracted from the marketing system, and classifiers are built using algorithms such as logistic regression, support vector machine (SVM), or random forest. The model aims to learn the mapping relationship between features in historical data and whether or not a payment is ultimately delinquent, and to predict new users. This type of method achieves preliminary automation, but feature engineering often remains at the level of superficial statistics, and the models are typically black boxes with poor business interpretability.

[0021] By analyzing time series trends, we can observe the overall upward or downward trend of electricity consumption or payment amounts over time, and by performing periodic analysis, we can identify periodic payment patterns such as monthly and quarterly. However, this analytical method struggles to uncover long-term dependencies and complex dynamic patterns in time series data, and cannot effectively capture the dynamic adjustment patterns of payment behavior.

[0022] The relevant methods for predicting the risk of overdue electricity bills lack dynamic behavior modeling capabilities, have insufficient time-series feature mining, and suffer from model simplification and poor integration, making it difficult for the identification of overdue electricity bill risks to adapt to the complexity and dynamism of the electricity market.

[0023] As mentioned above, how to comprehensively and accurately identify the risk of users' electricity bill overdue payments has become an important research question.

[0024] Based on the above description, such as Figure 1 As shown in this embodiment, the method for predicting the risk of overdue electricity bills includes: Step 101: Obtain the target user's electricity consumption data and payment data.

[0025] In practice, electricity consumption data refers to relevant data on the target user's electricity usage, including at least one of the following: electricity consumption, load curve, and electricity consumption time distribution. Payment data refers to relevant data on the target user's electricity bill payments, including at least one of the following: payment time, payment amount, number of days of delay, and historical arrears records.

[0026] After obtaining the target user's electricity consumption data and payment data, the process also includes: preprocessing the electricity consumption data and payment data to obtain preprocessed electricity consumption behavior data. The preprocessing includes at least one of the following: time alignment, outlier handling, and missing value handling.

[0027] Specifically, time alignment is performed on electricity consumption data and payment data to obtain matched electricity consumption behavior data. Abnormal electricity consumption data and abnormal payment data are removed from the matched electricity consumption behavior data to obtain normal electricity consumption behavior data. Missing electricity consumption data and missing payment data in the normal electricity consumption behavior data are imputed using the median algorithm to obtain complete electricity consumption behavior data, which is then used as the preprocessed electricity consumption behavior data.

[0028] Step 102: Based on the user type of the target user, determine the target electricity consumption behavior characteristics of the target user according to the electricity consumption data and the payment data.

[0029] In practice, the user type of the target user is determined. When the target user is an individual user, their electricity consumption behavior characteristics are determined based on electricity usage and payment data, and these characteristics are used as the target user's electricity consumption behavior characteristics. When the target user is a business user, their electricity consumption behavior characteristics are determined based on electricity usage and payment data, and these characteristics are used as the target user's electricity consumption behavior characteristics.

[0030] Step 103: Using a pre-trained risk prediction model, determine the user value level of the target user based on the target electricity consumption behavior characteristics, and determine the overdue electricity bill risk level of the target user based on the user value level.

[0031] In practice, the target electricity consumption behavior characteristics are divided into target electricity consumption characteristics and target payment characteristics. Using a pre-trained risk prediction model, cluster value clusters are determined based on the target electricity consumption characteristics, and the user value level of the target user is determined from these clusters. The payment characteristics corresponding to the target value clusters are determined from the target payment characteristics, and the overdue electricity bill risk level of the target user is determined based on these payment characteristics.

[0032] This disclosure integrates technologies and methods from multiple fields, including electricity marketing, data mining, machine learning, behavioral pattern mining, and risk identification and early warning. It aims to use advanced data processing and analysis methods to solve the practical problem of identifying overdue electricity bill risks in electricity marketing, thereby improving the risk management level and operational efficiency of power companies.

[0033] The embodiments disclosed herein can utilize advanced data processing and model building methods to improve the accuracy, timeliness, and automation of identifying electricity bill overdue risk, specifically achieving the following technical effects: (1) Solve the problem that traditional methods rely solely on static features and are difficult to capture the temporal evolution of user behavior. By constructing a multi-dimensional dynamic feature system (including features such as temporal changes, cross-correlation, and pattern classification), it accurately depicts the dynamic patterns of electricity consumption fluctuations, sudden changes in payment habits, and cross-cycle behavior dependence, which is more in line with the evolutionary logic of electricity bill overdue risk.

[0034] (2) Solve the problem of dependence on manually set rules and screening features in traditional methods, realize data-driven automated risk identification, reduce the bias introduced by subjective judgment, and improve the efficiency and reliability of electricity bill overdue risk identification.

[0035] Through the above embodiments, electricity consumption data and payment data of target users are obtained. Based on the user type of the target user, the target electricity consumption behavior characteristics of the target user are determined according to the electricity consumption data and payment data. The target electricity consumption behavior characteristics can comprehensively reflect the target user's electricity consumption behavior and payment behavior. Using a pre-trained risk prediction model, the user value level of the target user is determined based on the target electricity consumption behavior characteristics, and the overdue electricity bill risk level of the target user is determined according to the user value level. In this way, by dividing the user value level and the overdue electricity bill risk level into two dimensions, the user value level and the overdue electricity bill risk level can be linked, thereby improving the accuracy of the risk prediction model in predicting the overdue electricity bill risk and enabling comprehensive and accurate identification of the target user's overdue electricity bill risk level.

[0036] In some embodiments, step 102 includes: Step 1021: Determine the user type of the target user.

[0037] In practice, the historical electricity consumption and meter type of the target user are obtained. The user type of the target user is determined based on the historical electricity consumption and meter type. If it is determined that the target user's historical electricity consumption is less than a first electricity consumption threshold and the meter type is a single-phase meter, then the target user's user type is determined to be an individual user. If it is determined that the target user's historical electricity consumption is greater than or equal to a second electricity consumption threshold and the meter type is a three-phase meter, then the target user's user type is determined to be a business user.

[0038] For example, the first electricity consumption threshold is 1000 kWh, and the second electricity consumption threshold is 3000 kWh. If the target user's historical electricity consumption is less than 1000 kWh and the meter type is a single-phase meter, then the target user's user type is determined to be an individual user. If the target user's historical electricity consumption is greater than or equal to 3000 kWh and the meter type is a three-phase meter, then the target user's user type is determined to be a business user.

[0039] Step 1022: In response to determining that the user type of the target user is an individual user, determine the individual electricity consumption behavior characteristics based on the electricity consumption data and the payment data, and use the individual electricity consumption behavior characteristics as the target electricity consumption behavior characteristics of the target user.

[0040] In practice, when the target user is an individual user, the individual's electricity consumption behavior characteristics are determined based on electricity consumption data and payment data, and these characteristics are used as the target user's electricity consumption behavior characteristics. These individual electricity consumption behavior characteristics include: basic electricity consumption characteristics, basic payment characteristics, fluctuation characteristics, behavioral correlation characteristics, electricity consumption trend characteristics, and electricity consumption pattern sequence characteristics.

[0041] Alternatively, in step 1023, in response to determining that the user type of the target user is a merchant user, the merchant's electricity consumption behavior characteristics are determined based on the electricity consumption data and the payment data, and the merchant's electricity consumption behavior characteristics are used as the target electricity consumption behavior characteristics of the target user.

[0042] In practice, when the target user is a merchant, the merchant's electricity consumption behavior characteristics are determined based on electricity usage data and payment data, and these characteristics are used as the target user's electricity consumption behavior characteristics. These characteristics include: business electricity consumption density, basic electricity consumption stability characteristics, business electricity consumption stability characteristics, electricity consumption activity characteristics, energy consumption characteristics, energy efficiency synergy characteristics, peak-valley electricity consumption variation rate, electricity price matching degree, payment method distribution entropy, and payment delay duration.

[0043] The above scheme determines the user type of the target user. If the target user is an individual, their electricity consumption behavior characteristics are determined based on electricity usage and payment data, and these characteristics are then used as the target user's target electricity consumption behavior characteristics. If the target user is a business user, their electricity consumption behavior characteristics are determined based on electricity usage and payment data, and these characteristics are then used as the target user's target electricity consumption behavior characteristics. This method extracts corresponding target electricity consumption behavior characteristics based on the target user's user type, making the characteristics more accurate and aligned with the user type. Therefore, based on these target electricity consumption behavior characteristics, the risk level of overdue electricity bills for the target user can be accurately identified.

[0044] In some embodiments, step 1022 includes: Step 1022A: Determine basic electricity consumption characteristics from the electricity consumption data; wherein, the basic electricity consumption characteristics include: total electricity consumption, average electricity consumption duration, average electricity consumption amount, and electricity consumption behavior attributes.

[0045] In practice, total electricity consumption refers to the electricity consumption of an individual user within a preset time period. Total electricity consumption includes at least one of the following: daily total electricity consumption, weekly total electricity consumption, monthly total electricity consumption, and annual total electricity consumption.

[0046] Average electricity usage duration is the duration of electricity consumption by an individual user within a preset time period. Average electricity usage duration includes at least one of the following: daily average electricity usage duration, weekly average electricity usage duration, monthly average electricity usage duration, and annual average electricity usage duration.

[0047] Average electricity consumption refers to the electricity consumption of an individual user within a preset time period. Average electricity consumption includes at least one of the following: daily average electricity consumption, weekly average electricity consumption, monthly average electricity consumption, and annual average electricity consumption.

[0048] Electricity consumption behavior attributes are the attribute characteristics related to the electricity consumption behavior of individual users. Among them, electricity consumption behavior attributes include at least one of the following: whether there is any illegal electricity consumption behavior, whether there are any complaints, whether there is any transfer of ownership, and whether there is any change in the applicable electricity price.

[0049] Step 1022B: Determine basic payment characteristics from the payment data; wherein, the basic payment characteristics include: amount in arrears, number of arrears, electricity bill notification method, payment method, number of times payment method has been changed, payment frequency, payment duration, and fluctuation characteristics of payment amount.

[0050] In practice, the outstanding amount refers to the individual user's unpaid electricity bill. The number of times a payment is overdue can be monthly or annual. The electricity bill notification method is a notification system that informs the individual user about their electricity usage. The payment method can be the payment method used by the individual user to pay their electricity bill. The number of times the payment method has changed can be a coefficient representing the number of times the individual user has changed their payment method. The payment frequency can be the monthly electricity bill payment frequency, determined by the number of payments made each month. The payment duration can be the time elapsed since the last payment. The fluctuation characteristics of the payment amount can be determined based on the amount of each payment.

[0051] Step 1022C: Determine the electricity consumption volatility and peak-valley difference based on the electricity consumption data; determine the standard deviation of payment delay and electricity load parameters based on the payment data; and use the electricity consumption volatility, peak-valley difference, standard deviation of payment delay, and electricity load parameters as fluctuation characteristics.

[0052] In practice, electricity consumption volatility can be the electricity consumption volatility of an individual user within a preset time period. This volatility reflects the stability of an individual user's electricity consumption. The preset time period can be adjusted according to business needs, and different preset time periods may reflect different electricity consumption volatility rates. The preset time period can be the most recent month, the most recent week, or the most recent 90 days. Specifically, the daily electricity consumption of an individual user within the preset time period is determined based on electricity consumption data, and the electricity consumption volatility of the individual user within the preset time period is determined based on the daily electricity consumption. ,in, For individual users, within a preset time period Daily electricity consumption The number of days in the preset time period.

[0053] The standard deviation of payment delay is the moving standard deviation of the number of days of payment delay. It reflects the degree of fluctuation in an individual user's payment behavior. Specifically, based on payment data, the actual number of days of payment delay and the average number of days of payment delay for each individual user in each payment cycle are determined. The standard deviation of the individual user's payment delay is then determined based on the actual number of days of payment delay and the average number of days of payment delay. ,in, For historical payment periods, For the current payment cycle, For individual users in the first The actual number of days of payment delay within each payment cycle, For individual users in the first The average number of days of payment delay within each payment cycle.

[0054] Peak-valley electricity consumption difference is the difference between the peak and valley values ​​in an individual user's electricity load curve. It is used to quantify an individual user's electricity consumption patterns. Specifically, a payment period is selected as the statistical period. Based on electricity consumption data, the maximum and minimum daily electricity consumption of each individual user within the statistical period are determined. The peak-valley electricity consumption difference is obtained by subtracting the maximum and minimum daily electricity consumption. ,in, To account for the peak and valley differences in electricity consumption, This represents the maximum daily power consumption. This represents the minimum daily power consumption.

[0055] Electricity load parameters can be the electricity load parameters of an individual user within a payment cycle. These parameters include average load and load factor. The process of determining the average load includes: selecting a payment cycle as the statistical period; determining the load value of the individual user for each time period within the statistical period based on electricity consumption data; and determining the average load based on the load values ​​of the individual user for multiple time periods within the statistical period. ,in, For average load, For the first Load values ​​for each time period This refers to the number of time periods within the statistical period. The process of determining the load factor includes: selecting a payment period as the statistical period; determining the load value of individual users for each time period within the statistical period based on electricity consumption data; determining the maximum daily power consumption of individual users within the statistical period based on electricity consumption data; and determining the load factor based on the load value for each time period and the maximum daily power consumption. ,in, For load factor, For the first Load values ​​for each time period This refers to the number of time periods within the statistical period. This represents the maximum daily power consumption.

[0056] Step 1022D: The ratio of the outstanding amount to the total electricity consumption is used as the payment ability feature. The month-on-month change in electricity consumption is determined based on the electricity consumption of two adjacent periods. The payment ability feature and the month-on-month change in electricity consumption are used as behavioral association features.

[0057] In practice, the system obtains the outstanding amount and total electricity consumption for the current billing cycle, and uses the ratio of the outstanding amount to the total electricity consumption as a payment ability characteristic. This payment ability characteristic measures an individual user's ability to pay for each unit of electricity consumed. A higher ratio indicates a weaker payment ability, meaning a higher risk of overdue electricity bills.

[0058] The month-on-month change in electricity consumption measures the fluctuation range of electricity consumption between two adjacent periods. Specifically, it obtains the electricity consumption of the current payment period and the electricity consumption of the previous payment period, and determines the month-on-month change in electricity consumption based on the current period's electricity consumption and the previous period's electricity consumption. .

[0059] The number of days of payment delay over multiple consecutive payment periods is statistically analyzed, and the correlation coefficient between the number of days of payment delay and the month-on-month change in electricity consumption is calculated. This correlation coefficient is used as a behavioral association feature. Specifically, the behavioral association feature is used to identify the association between payment ability and electricity consumption behavior.

[0060] Step 1022E: Convert the electricity consumption data into an electricity consumption sequence, and determine the electricity consumption trend characteristics based on the electricity consumption sequence.

[0061] In practice, electricity consumption data will be converted into an electricity consumption sequence. The electricity consumption series was analyzed using Seasonal and Trend Decomposition using Loess (STL decomposition), which is based on locally weighted regression. Decomposed into trend term, seasonal term, and residual term. ,in, This is a trend term, reflecting the long-term trend of electricity consumption. This is a seasonal item, reflecting the periodic and seasonal fluctuations in electricity consumption; This is the residual term, reflecting the random fluctuations in electricity consumption. Based on the electricity consumption sequence. Extracting trend items slope characteristics Average fluctuation characteristics of seasonal terms and the standard deviation of the residuals , trend item slope characteristics Average fluctuation characteristics of seasonal terms and the standard deviation of the residuals Identify the risk of overdue electricity bills based on electricity consumption trend characteristics.

[0062] Step 1022F: Analyze and process the electricity consumption data through a sliding window to obtain electricity consumption pattern sequence features.

[0063] In practice, sliding window and K-means clustering are used to classify individual users’ electricity consumption patterns into stable, fluctuating and sudden drop types, and category features are generated based on individual users’ electricity consumption patterns.

[0064] Specifically, 1) Extract hourly electricity consumption data for individual users from the electricity consumption data. Determine the size and step size of the sliding window. The sliding window size can be set to 15 days (adjustable according to actual conditions), and the step size can be set to 1 day, i.e., moving one time unit at a time to analyze the electricity consumption data of individual users segment by segment. 2) For the electricity consumption data within each sliding window, calculate the statistical characteristics of the electricity consumption data. For example, statistical characteristics can be mean, standard deviation, maximum value, minimum value, skewness, kurtosis, etc. The statistical characteristics can reflect the basic situation of user electricity consumption within the sliding window. 3) Use principal component analysis to reduce the dimensionality of the statistical characteristics. 4) Use the dimensionality-reduced statistical characteristics within each sliding window as the input vector, and use the K-means algorithm to cluster (using the silhouette coefficient to determine the optimal number of clusters K) to divide the electricity consumption patterns of individual users into K categories. 5) For each sliding window, assign corresponding category labels to the cluster category to which the electricity consumption pattern belongs. For example, the category labels can be stable, fluctuating, or sudden drop, etc., and perform one-hot encoding on the category labels. 6) Integrate the category labels within each sliding window into a sequence of individual user electricity consumption patterns in chronological order. Use the frequency of different categories and the number of transitions between categories as features for risk identification.

[0065] Step 1022G: The basic electricity consumption characteristics, the basic payment characteristics, the fluctuation characteristics, the behavioral correlation characteristics, the electricity consumption trend characteristics, and the electricity consumption pattern sequence characteristics are used as personal electricity consumption behavior characteristics.

[0066] In practice, the characteristics of personal electricity consumption behavior include: basic electricity consumption characteristics, basic payment characteristics, fluctuation characteristics, behavioral correlation characteristics, electricity consumption trend characteristics, and electricity consumption pattern sequence characteristics.

[0067] The above scheme determines basic electricity consumption characteristics, basic payment characteristics, fluctuation characteristics, behavioral correlation characteristics, electricity consumption trend characteristics, and electricity consumption pattern sequence characteristics based on electricity consumption data and payment data. These characteristics are then used as individual electricity consumption behavior characteristics. In this way, when the target user is an individual user, individual electricity consumption behavior characteristics can be comprehensively and accurately extracted based on electricity consumption data and payment data.

[0068] In some embodiments, step 1023 includes: Step 1023A: Determine the electricity consumption of merchants during business hours and the total electricity consumption for the whole day from the electricity consumption data, and use the ratio of the electricity consumption during business hours to the total electricity consumption for the whole day as the electricity consumption density.

[0069] In practice, the operational electricity consumption density refers to the electricity consumption density of a business user during its business hours. For example, for a restaurant, electricity consumption data is extracted from the business hours of 10:00-14:00 and 17:00-22:00 to determine the average operational electricity consumption during these hours and the average daily electricity consumption. The ratio of the average operational electricity consumption to the average daily electricity consumption is used as the operational electricity consumption density. A low operational electricity consumption density indicates that the business user's electricity consumption is low during the business hours, which may suggest poor business performance or non-operation, and the risk of unpaid electricity bills.

[0070] Step 1023B: Determine the first variation parameter of the average daily electricity consumption of merchant users within a preset time period based on the electricity consumption data, and determine the basic electricity consumption stability characteristics based on the first variation parameter.

[0071] In practice, the basic electricity consumption stability characteristic is the coefficient of variation of the average daily electricity consumption of merchants over the past month. A low value for the basic electricity consumption stability characteristic indicates that the merchant's operations are basically stable.

[0072] Step 1023C: Determine a second variation parameter of the average electricity consumption of merchants during business hours based on the electricity consumption of the business, and determine the stability characteristics of the electricity consumption of the business based on the second variation parameter.

[0073] In practice, the stability characteristic of business electricity consumption is the coefficient of variation of the average electricity consumption of merchants over the past three or six months of business operations. A low value for the stability characteristic of business electricity consumption indicates that the main business of merchants is stable.

[0074] Step 1023D: Determine the merchant's non-business hours, and based on the electricity consumption data, determine the active duration during which the electricity consumption exceeds a preset electricity threshold within the non-business hours. Use the ratio of the active duration to the non-business hours as the electricity consumption activity feature.

[0075] In practice, the electricity activity characteristic can be the electricity activity during non-business hours. Specifically, the preset electricity threshold can be a preset multiple of the basic electricity consumption, for example, 1.5 times the basic electricity consumption. The percentage of time during non-business hours when electricity consumption exceeds 1.5 times the basic electricity consumption is counted, and this percentage is used as the electricity activity characteristic. A high value for the electricity activity characteristic indicates excessive electricity consumption during non-business hours, which indirectly reflects chaotic management of business users and the risk of unpaid electricity bills.

[0076] Step 1023E: Determine the monthly electricity consumption of the merchant user based on the electricity consumption data, and use the ratio of the monthly electricity consumption to the merchant's business area as an energy consumption characteristic.

[0077] In practice, the energy consumption characteristic can be the energy consumption index per unit area for businesses. Specifically, the monthly electricity consumption of businesses is determined based on electricity usage data, and the ratio of monthly electricity consumption to the business's operating area is used as the energy consumption characteristic. This energy consumption characteristic is then compared with energy consumption characteristics of similar types and within the same area range to obtain the energy consumption index. A low energy consumption index indicates poor business performance or problems with electricity meter readings.

[0078] Step 1023F: Determine the daily average load rate and average power based on the electricity consumption data, and determine the energy efficiency synergy characteristics based on the daily average load rate and average power.

[0079] In practice, the daily average load factor and average power consumption of business users are determined based on electricity consumption data. Energy efficiency synergy characteristics are then determined based on these factors. The calculation formula is as follows: Energy Efficiency Synergy Characteristics = (Daily Average Load Factor - 0.5) × Average Power. A lower energy efficiency synergy characteristic indicates poor energy efficiency and high electricity costs for business users, indirectly reflecting their operational pressure.

[0080] Step 1023G: Determine the peak and off-peak electricity consumption of the merchant user during the peak and off-peak hours of the current month and the peak and off-peak electricity consumption of the previous month based on the electricity consumption data, and determine the peak and off-peak electricity consumption change rate based on the peak and off-peak electricity consumption of the current month and the peak and off-peak electricity consumption of the previous month.

[0081] In practice, the peak-valley electricity consumption variation rate can be the peak-valley electricity price response coefficient for business users. Specifically, the off-peak electricity consumption of the current month and the previous month is determined based on electricity consumption data, and the peak-valley electricity consumption variation is determined based on the current month's off-peak electricity consumption and the previous month's off-peak electricity consumption. When the peak-valley electricity consumption variation rate drops significantly, it indicates that the business user is not performing well.

[0082] Step 1023H: Determine the electricity consumption and electricity cost within a preset time period based on the electricity consumption data, and determine the electricity cost matching degree based on the electricity consumption and electricity cost.

[0083] In practice, the electricity consumption and cost within a preset time period are determined based on electricity usage data, and the electricity cost matching degree is determined based on the electricity consumption and cost. Electricity cost matching degree = Electricity cost / (Electricity consumption × Average electricity price). When the electricity cost matching degree is significantly lower than 1, it indicates that the merchant / user may have outstanding payments or have not paid the full amount in this transaction. The number of consecutive periods with an electricity cost matching degree below 0.9 is used as the electricity cost matching degree characteristic.

[0084] Step 1023I: Determine the payment method distribution entropy and payment delay duration based on the payment data.

[0085] In practice, the distribution of payment methods used by merchants and users over the past year is statistically analyzed based on payment data, and the entropy of the payment method distribution is determined accordingly. A lower entropy value indicates that the payment methods are stable.

[0086] The payment delay duration can be measured in days. It is determined from payment data by counting the number of days between the initial payment reminder and the actual payment, and this interval is used as the payment delay duration. A longer payment delay duration indicates a more delayed response from merchants to payment reminders and a weaker willingness to pay.

[0087] Based on electricity consumption data, determine the month-on-month change in electricity consumption and calculate the ratio of the month-on-month change in electricity consumption to the number of days of delayed payment. Calculate the month-on-month growth rate of electricity consumption and the month-on-month growth rate of the number of days of delayed payment. When both electricity consumption growth and the number of days of delayed payment are positive and relatively small, it indicates that the business user may be expanding their business but their cash flow cannot keep up, and the business user faces the risk of electricity bill arrears.

[0088] Based on electricity consumption data, determine the baseline electricity consumption during non-business hours and the week-on-week decrease rate of the baseline electricity consumption percentage. Also based on electricity consumption data, determine the average electricity consumption during business hours and the average daily electricity consumption for businesses. Use the ratio of average business hours electricity consumption to average daily electricity consumption as the week-on-week decrease rate of electricity density. Determine if there are any impact load events by calculating the volatility of electricity consumption at the minute level. When the volatility is extremely low, there may be no impact load, indicating that core electrical equipment does not need to be started, causing the impact load to disappear. This suggests a sharp decline in revenue and a higher risk of unpaid electricity bills for businesses.

[0089] If the week-on-week decrease rate of basic electricity consumption ratio exceeds 30%, and the week-on-week decrease rate of electricity density exceeds 20%, and there is no impact load, then the operating risk coefficient is set to 1; otherwise, it is set to 0.

[0090] Step 1023J: The business electricity consumption density, the basic electricity consumption stability characteristics, the business electricity consumption stability characteristics, the electricity consumption activity characteristics, the energy consumption characteristics, the energy efficiency synergy characteristics, the peak-valley electricity consumption change rate, the electricity bill matching degree, the payment method distribution entropy, and the payment delay duration are taken as the characteristics of the merchant's electricity consumption behavior.

[0091] In specific implementation, the characteristics of merchants' electricity consumption behavior include: business electricity consumption density, basic electricity consumption stability characteristics, business electricity consumption stability characteristics, electricity consumption activity characteristics, energy consumption characteristics, energy efficiency synergy characteristics, peak-valley electricity consumption change rate, electricity bill matching degree, payment method distribution entropy, and payment delay duration.

[0092] The above scheme determines the following characteristics based on electricity consumption and payment data: business electricity consumption density, basic electricity consumption stability characteristics, business electricity consumption stability characteristics, electricity consumption activity characteristics, energy consumption characteristics, energy efficiency synergy characteristics, peak-valley electricity consumption variation rate, electricity price matching degree, payment method distribution entropy, and payment delay duration. These characteristics are then used as the characteristics of merchant electricity consumption behavior. Thus, when the target user type is a merchant user, the merchant's electricity consumption behavior characteristics can be comprehensively and accurately extracted based on electricity consumption and payment data.

[0093] In some embodiments, step 103 includes: Step 1031: Divide the target electricity consumption behavior characteristics into target electricity consumption characteristics and target payment characteristics.

[0094] Step 1032: Using a pre-trained risk prediction model, determine cluster value clusters based on the target electricity consumption characteristics, and determine the user value level of the target user from the cluster value clusters.

[0095] Step 1033: Determine the payment characteristics corresponding to the target value cluster from the target payment characteristics, and determine the overdue electricity bill risk level of the target user based on the payment characteristics corresponding to the target value cluster.

[0096] In practice, the target electricity consumption behavior characteristics are divided into target electricity consumption characteristics and target payment characteristics. Using a pre-trained risk prediction model, the target electricity consumption characteristics are clustered to obtain cluster value clusters. The target value clusters are then determined from these cluster value clusters using the Bayesian information criterion, and these target value clusters are used as the user value levels for target users. The cluster value clusters include: high-value clusters, medium-value clusters, and low-value clusters.

[0097] Based on the cluster value clusters, the target users' target electricity consumption characteristics are divided into three subsets, corresponding to the high-value cluster (subset A), the medium-value cluster (subset B), and the low-value cluster (subset C), respectively. The supplementary payment features for subset A are the number of late payment fees, the stability of payment channels, and the fluctuation of monthly payment amounts. The supplementary payment features for subset B are the payment capacity coefficient, the behavioral correlation coefficient, and the number of times payment channels are changed. The supplementary payment features for subset C are the amount of arrears, the average payment duration, and the default / violation mark. Cluster identifier features (A=0, B=1, C=2) are added to each subset feature matrix. Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is performed on the feature matrices of the three subsets to determine the overdue electricity bill risk level (-1 represents an abnormal risk cluster, 0 represents a low-risk cluster, 1 represents a medium-risk cluster, and 2 represents a high-risk cluster). The silhouette coefficient of each value risk combination is determined. When the silhouette coefficient is greater than 0.5, it indicates that the clustering effect is good. Otherwise, the parameters are adjusted and the clustering is repeated to ultimately form a variety of value risk combinations such as high-value low-risk and high-value high-risk.

[0098] The above scheme divides target electricity consumption behavior characteristics into target electricity consumption characteristics and target payment characteristics. Using a pre-trained risk prediction model, cluster value clusters are determined based on the target electricity consumption characteristics, and the user value level of the target user is determined from these clusters. The payment characteristics corresponding to the target value clusters are determined from the target payment characteristics, and the electricity bill delinquency risk level of the target user is determined based on these payment characteristics. In this way, first determining the user value level of the target user through clustering, and then determining the electricity bill delinquency risk level based on the user value level, can solve the interpretability problem of single clustering and improve the accuracy and interpretability of the electricity bill delinquency risk level.

[0099] In some embodiments, step 1032 includes: Step 1032A: Normalize the continuous features in the target electricity consumption characteristics to obtain the first input feature, and encode the discrete features in the target electricity consumption characteristics to obtain the second input feature.

[0100] In practice, the target electricity consumption characteristics are divided into continuous characteristics and discrete characteristics. The continuous characteristics are normalized by mean to obtain the first input feature, which is the input feature obtained after normalizing the continuous characteristics in the target electricity consumption characteristics. The discrete characteristics are encoded using one-hot encoding to obtain the second input feature, which is the input feature obtained after encoding the discrete characteristics in the target electricity consumption characteristics.

[0101] Step 1032B: Using a pre-trained risk prediction model, cluster the first input feature and the second input feature to obtain cluster value clusters.

[0102] In practice, the first and second input features are integrated into a multidimensional sequence X. The multidimensional sequence X is then input into a pre-trained risk prediction model, which performs clustering on the multidimensional sequence X to obtain cluster value clusters.

[0103] The clusters are categorized into three value clusters: high-value, medium-value, and low-value. When the target users are individual users, high-value clusters represent individual users with large electricity consumption, stable load, and a growing trend. Medium-value clusters represent individual users with moderate electricity consumption and stability. Low-value clusters represent individual users with small electricity consumption, large load fluctuations, and a declining trend. When the target users are business users, high-value clusters represent high-quality business users with high energy efficiency per unit area, high electricity consumption stability, and stable operations. Medium-value clusters represent business users with relatively high stability-related indicators and some operational volatility. Low-value clusters represent business users with low energy efficiency per unit area, extremely unstable basic electricity consumption, and chaotic and difficult business models.

[0104] Step 1032C: Determine the target value cluster from the clustered value clusters using Bayesian information criteria, and use the target value cluster as the user value level of the target user.

[0105] In practice, the optimal number of components K of the Gaussian Mixture Model (GMM) is determined by the Bayesian Information Criterion (BIC). The K value corresponding to the minimum of the Bayesian Information Criterion (BIC) is selected as the target value cluster, and the target value cluster is used as the user value level of the target user.

[0106] The above scheme normalizes the continuous features in the target electricity consumption characteristics to obtain the first input feature, and encodes the discrete features to obtain the second input feature. Using a pre-trained risk prediction model, the first and second input features are clustered to obtain cluster value clusters. The target value cluster is determined from these cluster value clusters using the Bayesian information criterion, and this target value cluster is used as the user value level of the target user. In this way, the user value level of the target user can be quickly and accurately determined using the Bayesian information criterion.

[0107] In some embodiments, the pre-training process of the risk prediction model includes: Step 103A: Obtain sample data of user electricity consumption, and divide the sample data into a first sample sequence, a second sample sequence, and a third sample sequence according to time sequence.

[0108] Step 103B: Train the neural network using the first sample sequence and the second sample sequence to obtain a first initial prediction model, and verify the first initial prediction model using the third sample sequence to obtain a first target prediction model.

[0109] Step 103C: Train the neural network using the first sample sequence and the third sample sequence to obtain a second initial prediction model, and verify the second initial prediction model using the second sample sequence to obtain a second target prediction model.

[0110] Step 103D: Train the neural network using the second sample sequence and the third sample sequence to obtain a third initial prediction model, and verify the third initial prediction model using the first sample sequence to obtain a third target prediction model.

[0111] Step 103E: Determine the risk prediction model based on the first target prediction model, the second target prediction model, and the third target prediction model.

[0112] In practice, sample data of user electricity consumption is acquired and divided into a first sample sequence, a second sample sequence, and a third sample sequence according to time sequence. The first and second sample sequences are used to train a neural network to obtain a first initial prediction model, and the third sample sequence is used to validate the first initial prediction model to obtain a first target prediction model. The first and third sample sequences are then used to train the neural network to obtain a second initial prediction model, which is then validated using the second sample sequence to obtain a second target prediction model. Finally, the second and third sample sequences are used to train the neural network to obtain a third initial prediction model, which is then validated using the first sample sequence to obtain a third target prediction model. A risk prediction model is then determined based on these three target prediction models. In this way, the first, second, and third sample sequences enable time-series cross-validation of the risk prediction model.

[0113] In addition, the sample data can be divided into 5 sample sequences according to time order. The first to fourth sample sequences are used as the training set and the fifth sample sequence is used as the test set. Then, the first to third and fifth sample sequences are used as the training set and the fourth sample sequence is used as the test set. The risk identification accuracy and abnormal risk recall rate of the risk prediction model can be verified in turn.

[0114] The above scheme obtains sample data on user electricity consumption, which is then divided into a first sample sequence, a second sample sequence, and a third sample sequence according to time sequence. The first and second sample sequences are used to train a neural network to obtain a first initial prediction model, and the third sample sequence is used to validate the first initial prediction model to obtain a first target prediction model. The first and third sample sequences are then used to train the neural network to obtain a second initial prediction model, which is then validated using the second sample sequence to obtain a second target prediction model. The second and third sample sequences are then used to train the neural network to obtain a third initial prediction model, which is then validated using the first sample sequence to obtain a third target prediction model. A risk prediction model is then determined based on the first, second, and third target prediction models. In this way, the first, second, and third sample sequences enable time-series cross-validation of the risk prediction model.

[0115] Through the above embodiments, electricity consumption data and payment data of target users are obtained. Based on the user type of the target user, the target electricity consumption behavior characteristics of the target user are determined according to the electricity consumption data and payment data. The target electricity consumption behavior characteristics can comprehensively reflect the target user's electricity consumption behavior and payment behavior. Using a pre-trained risk prediction model, the user value level of the target user is determined based on the target electricity consumption behavior characteristics, and the overdue electricity bill risk level of the target user is determined according to the user value level. In this way, by dividing the user value level and the overdue electricity bill risk level into two dimensions, the user value level and the overdue electricity bill risk level can be linked, thereby improving the accuracy of the risk prediction model in predicting the overdue electricity bill risk and enabling comprehensive and accurate identification of the target user's overdue electricity bill risk level.

[0116] It should be noted that the embodiments of this disclosure can also be further described in the following ways: Figure 2 This is a flowchart illustrating a method for predicting the risk of overdue electricity bills for individual users, as described in an embodiment of this disclosure. Figure 2As shown, the method for predicting the risk of overdue electricity bills for individual users includes: acquiring individual users' electricity consumption data, payment data, and user attribute data; integrating and preprocessing the individual users' electricity consumption data, payment data, and user attribute data; determining basic behavioral profiles (basic electricity consumption characteristics), payment capability profiles (basic payment characteristics), behavioral fluctuation profiles (fluctuation characteristics), behavioral correlation profiles (behavioral correlation characteristics), trend profiles (electricity consumption trend characteristics), and electricity consumption pattern label profiles (electricity consumption pattern sequence characteristics) based on the preprocessed data; determining the individual user's user value level through a first-stage clustering approach, and determining the individual user's overdue electricity bill risk level through a second-stage clustering approach; verifying the risk identification accuracy and abnormal risk recall rate through time series cross-validation, and finally outputting the individual user's overdue electricity bill risk level.

[0117] Figure 3 This is a flowchart illustrating a method for predicting the risk of overdue electricity bills for merchant users, as described in this embodiment of the disclosure. Figure 3 As shown, the method for predicting the risk of overdue electricity bills for merchant users includes: acquiring the electricity consumption and payment data of merchant users; integrating and preprocessing the electricity consumption and payment data of merchant users; determining the electricity consumption density during business hours (business electricity consumption density), the stability index of basic electricity consumption (basic electricity consumption stability characteristics), the stability index of electricity consumption during business hours (business electricity consumption stability characteristics), the electricity consumption activity during non-business hours (electricity consumption activity characteristics), the energy consumption per unit operating area of ​​the merchant (energy consumption characteristics), and the peak-valley electricity price response coefficient (peak-valley electricity consumption change rate) based on the preprocessed data; determining the user value level of merchant users through the first-stage clustering, and determining the overdue electricity bill risk level of merchant users through the second-stage clustering; verifying the accuracy of risk identification and the recall rate of abnormal risks through time series cross-validation, and finally outputting the overdue electricity bill risk level of merchant users.

[0118] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0119] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0120] Based on the same inventive concept, corresponding to any of the above-described embodiments, this disclosure also provides a device for predicting the risk of overdue electricity bills.

[0121] refer to Figure 4 The device for predicting the risk of overdue electricity bills includes: The acquisition module 301 is configured to acquire the electricity consumption data and payment data of the target user. The feature determination module 302 is configured to determine the target electricity consumption behavior features of the target user based on the user type of the target user and according to the electricity consumption data and the payment data. The risk prediction module 303 is configured to use a pre-trained risk prediction model to determine the user value level of the target user based on the target electricity consumption behavior characteristics, and to determine the overdue electricity bill risk level of the target user based on the user value level.

[0122] In some embodiments, the feature determination module 302 includes: The user type determination unit is configured to determine the user type of the target user; A personal electricity consumption behavior characteristic determination unit is configured to, in response to determining that the target user's user type is an individual user, determine personal electricity consumption behavior characteristics based on the electricity consumption data and the payment data, and use the personal electricity consumption characteristics as the target electricity consumption behavior characteristics of the target user; or... The merchant electricity consumption behavior characteristic determination unit is configured to, in response to determining that the user type of the target user is a merchant user, determine the merchant electricity consumption behavior characteristics based on the electricity consumption data and the payment data, and use the merchant electricity consumption behavior characteristics as the target electricity consumption behavior characteristics of the target user.

[0123] In some embodiments, the personal electricity consumption behavior characteristic determination unit includes: A basic electricity consumption characteristic determination subunit is configured to determine basic electricity consumption characteristics from the electricity consumption data; wherein, the basic electricity consumption characteristics include: total electricity consumption, average electricity consumption duration, average electricity consumption amount, and electricity consumption behavior attributes; The basic payment feature determination subunit is configured to determine basic payment features from the payment data; wherein, the basic payment features include: arrears amount, number of arrears, electricity bill notification method, payment method, number of times payment method has been changed, payment frequency, payment duration, and payment amount fluctuation features; The fluctuation characteristic determination subunit is configured to determine the electricity fluctuation rate and peak-valley difference based on the electricity consumption data, determine the standard deviation of payment delay and electricity load parameters based on the payment data, and use the electricity fluctuation rate, the peak-valley difference, the standard deviation of payment delay, and the electricity load parameters as fluctuation characteristics; The behavior association feature determination subunit is configured to use the ratio of the arrears amount to the total electricity consumption as the payment ability feature, determine the month-on-month change in electricity consumption based on the electricity consumption of two adjacent periods, and use the payment ability feature and the month-on-month change in electricity consumption as the behavior association feature. The electricity consumption trend feature determination subunit is configured to convert the electricity consumption data into an electricity consumption sequence and determine the electricity consumption trend feature based on the electricity consumption sequence. The electricity consumption pattern sequence feature determination subunit is configured to analyze and process the electricity consumption data through a sliding window to obtain electricity consumption pattern sequence features; The personal electricity consumption behavior characteristic determination subunit is configured to use the basic electricity consumption characteristics, the basic payment characteristics, the fluctuation characteristics, the behavior correlation characteristics, the electricity consumption trend characteristics, and the electricity consumption pattern sequence characteristics as personal electricity consumption behavior characteristics.

[0124] In some embodiments, the merchant electricity consumption behavior characteristic determination unit includes: The business electricity consumption density determination subunit is configured to determine the business electricity consumption of merchants during business hours and the total daily electricity consumption from the electricity consumption data, and to use the ratio of the business electricity consumption to the total daily electricity consumption as the business electricity consumption density. The basic electricity consumption stability characteristic determination subunit is configured to determine a first change parameter of the average daily electricity consumption of merchant users within a preset time period based on the electricity consumption data, and determine the basic electricity consumption stability characteristic based on the first change parameter. The subunit for determining the stability characteristics of business electricity consumption is configured to determine a second variation parameter of the average electricity consumption of merchants during business hours based on the business electricity consumption, and to determine the stability characteristics of business electricity consumption based on the second variation parameter. The electricity consumption activity feature determination subunit is configured to determine the non-business hours of a merchant, determine the active hours during which the electricity consumption exceeds a preset electricity threshold from the non-business hours based on the electricity consumption data, and use the ratio of the active hours to the non-business hours as the electricity consumption activity feature. The energy consumption characteristic determination subunit is configured to determine the monthly electricity consumption of the merchant user based on the electricity consumption data, and use the ratio of the monthly electricity consumption to the merchant's business area as the energy consumption characteristic. An energy efficiency synergy characteristic determination subunit is configured to determine the daily average load rate and average power based on the electricity consumption data, and to determine energy efficiency synergy characteristics based on the daily average load rate and the average power. The peak-valley electricity consumption change rate determination subunit is configured to determine the peak-valley electricity consumption of merchants and users during peak-valley periods this month and last month based on the electricity consumption data, and to determine the peak-valley electricity consumption change rate based on the peak-valley electricity consumption of this month and last month. The electricity cost matching degree determination subunit is configured to determine the electricity consumption and electricity cost within a preset time period based on the electricity consumption data, and to determine the electricity cost matching degree based on the electricity consumption and electricity cost; The payment delay duration determination subunit is configured to determine the payment method distribution entropy and payment delay duration based on the payment data; The merchant electricity consumption behavior characteristic determination subunit is configured to use the business electricity consumption density, the basic electricity consumption stability characteristic, the business electricity consumption stability characteristic, the electricity consumption activity characteristic, the energy consumption characteristic, the energy efficiency synergy characteristic, the peak-valley electricity consumption change rate, the electricity price matching degree, the payment method distribution entropy, and the payment delay duration as merchant electricity consumption behavior characteristics.

[0125] In some embodiments, the risk prediction module 303 includes: The feature segmentation unit is configured to segment the target electricity consumption behavior feature into target electricity consumption feature and target payment feature; The value level determination unit is configured to use a pre-trained risk prediction model to determine clustered value clusters based on the target electricity consumption characteristics, and to determine the user value level of the target user from the clustered value clusters; The risk level determination unit is configured to determine the payment characteristics corresponding to the target value cluster from the target payment characteristics, and determine the overdue electricity bill risk level of the target user based on the payment characteristics corresponding to the target value cluster.

[0126] In some embodiments, the risk level determination unit includes: The input feature determination subunit is configured to normalize the continuous features in the target electricity consumption features to obtain the first input feature, and to encode the discrete features in the target electricity consumption features to obtain the second input feature; Cluster value cluster determination subunits are configured to use a pre-trained risk prediction model to perform clustering processing on the first input feature and the second input feature to obtain cluster value clusters; The value level determination subunit is configured to determine a target value cluster from the clustered value clusters using Bayesian information criteria, and to use the target value cluster as the user value level of the target user.

[0127] In some embodiments, the apparatus further includes a model training module, the model training module comprising: The sample data partitioning unit is configured to acquire sample data of user electricity consumption and partition the sample data into a first sample sequence, a second sample sequence, and a third sample sequence in chronological order. The first training unit is configured to train a neural network using the first sample sequence and the second sample sequence to obtain a first initial prediction model, and to verify the first initial prediction model using the third sample sequence to obtain a first target prediction model. The second training unit is configured to train the neural network using the first sample sequence and the third sample sequence to obtain a second initial prediction model, and to verify the second initial prediction model using the second sample sequence to obtain a second target prediction model. The third training unit is configured to train the neural network using the second sample sequence and the third sample sequence to obtain a third initial prediction model, and to verify the third initial prediction model using the first sample sequence to obtain a third target prediction model. The risk prediction model determination unit is configured to determine a risk prediction model based on the first target prediction model, the second target prediction model, and the third target prediction model.

[0128] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0129] The apparatus described above is used to implement the corresponding method for predicting the risk of overdue electricity bills in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0130] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for predicting the risk of overdue electricity bills as described in any of the above embodiments.

[0131] Figure 5This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0132] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0133] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0134] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0135] The communication interface 1040 is used to connect the communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB (Universal Serial Bus), network cable, etc.) or wireless means (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0136] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0137] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0138] The electronic devices described above are used to implement the corresponding methods for predicting the risk of overdue electricity bills in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0139] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the method for predicting the risk of overdue electricity bills as described in any of the above embodiments.

[0140] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0141] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the method for predicting the risk of overdue electricity bills as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0142] Based on the same inventive concept, corresponding to any of the above embodiments, this application also provides a computer program product, including computer program instructions. When the computer program instructions are run on a computer, the computer executes the method for predicting the risk of overdue electricity bills as described in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0143] It is understood that before using the technical solutions of the various embodiments in this disclosure, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0144] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations of this disclosed technical solution.

[0145] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0146] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0147] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0148] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0149] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0150] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this disclosure. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the protection scope of this disclosure.

Claims

1. A method for predicting the risk of overdue electricity bills, characterized in that, The method includes: Obtain the target users' electricity consumption and payment data; Based on the user type of the target user, the target electricity consumption behavior characteristics of the target user are determined according to the electricity consumption data and the payment data. Using a pre-trained risk prediction model, the user value level of the target user is determined based on the target electricity consumption behavior characteristics, and the overdue electricity bill risk level of the target user is determined based on the user value level.

2. The method according to claim 1, characterized in that, The step of determining the target electricity consumption behavior characteristics of the target user based on the user type of the target user, according to the electricity consumption data and the payment data, includes: Determine the user type of the target user; In response to determining that the target user's user type is an individual user, the individual's electricity consumption behavior characteristics are determined based on the electricity consumption data and the payment data, and these individual electricity consumption behavior characteristics are used as the target electricity consumption behavior characteristics of the target user; or... In response to determining that the target user's user type is a merchant user, the merchant's electricity consumption behavior characteristics are determined based on the electricity consumption data and the payment data, and the merchant's electricity consumption behavior characteristics are used as the target electricity consumption behavior characteristics of the target user.

3. The method according to claim 2, characterized in that, The process of determining individual electricity consumption behavior characteristics based on the electricity consumption data and the payment data includes: Basic electricity consumption characteristics are determined from the electricity consumption data; wherein, the basic electricity consumption characteristics include: total electricity consumption, average electricity consumption duration, average electricity consumption amount, and electricity consumption behavior attributes; Basic payment characteristics are determined from the payment data; wherein, the basic payment characteristics include: amount in arrears, number of arrears, electricity bill notification method, payment method, number of times payment method is changed, payment frequency, payment duration, and fluctuation characteristics of payment amount; The electricity consumption volatility and peak-valley difference are determined based on the electricity consumption data, and the standard deviation of payment delay and electricity load parameters are determined based on the payment data. The electricity consumption volatility, peak-valley difference, standard deviation of payment delay, and electricity load parameters are used as fluctuation characteristics. The ratio of the outstanding amount to the total electricity consumption is used as the payment ability feature. The month-on-month change in electricity consumption is determined based on the electricity consumption of two adjacent periods. The payment ability feature and the month-on-month change in electricity consumption are used as behavioral association features. The electricity consumption data is converted into an electricity consumption sequence, and electricity consumption trend characteristics are determined based on the electricity consumption sequence; The electricity consumption data is analyzed and processed using a sliding window to obtain electricity consumption pattern sequence characteristics; The basic electricity consumption characteristics, the basic payment characteristics, the fluctuation characteristics, the behavioral correlation characteristics, the electricity consumption trend characteristics, and the electricity consumption pattern sequence characteristics are used as personal electricity consumption behavior characteristics.

4. The method according to claim 2, characterized in that, The step of determining the merchant's electricity consumption behavior characteristics based on the electricity consumption data and the payment data includes: The electricity consumption data is used to determine the electricity consumption of merchants during business hours and the total electricity consumption for the whole day. The ratio of the electricity consumption during business hours to the total electricity consumption for the whole day is used as the electricity consumption density. Based on the electricity consumption data, a first variation parameter of the average daily electricity consumption of merchant users within a preset time period is determined, and the basic electricity consumption stability characteristics are determined based on the first variation parameter. Based on the electricity consumption of the business, a second variation parameter is determined to be the average electricity consumption of the business users during the business hours, and the stability characteristics of the electricity consumption of the business are determined based on the second variation parameter. Determine the non-business hours of the merchant, and based on the electricity consumption data, determine the active time during which the electricity consumption exceeds a preset electricity threshold within the non-business hours. Use the ratio of the active time to the non-business hours as the electricity consumption activity feature. The monthly electricity consumption of merchants is determined based on the electricity consumption data, and the ratio of the monthly electricity consumption to the business area of ​​the merchants is used as an energy consumption characteristic. The daily average load rate and average power are determined based on the electricity consumption data, and the energy efficiency synergy characteristics are determined based on the daily average load rate and average power. Based on the electricity consumption data, determine the peak and off-peak electricity consumption of merchants and users during peak and off-peak hours for the current month and the previous month, and determine the peak and off-peak electricity consumption change rate based on the peak and off-peak electricity consumption for the current month and the previous month. Based on the electricity consumption data, determine the electricity consumption and electricity cost within a preset time period, and determine the electricity cost matching degree based on the electricity consumption and electricity cost; The payment method distribution entropy and payment delay duration are determined based on the payment data. The following are taken as characteristics of merchant electricity consumption behavior: the density of operating electricity consumption, the stability characteristics of basic electricity consumption, the stability characteristics of operating electricity consumption, the characteristics of electricity consumption activity, the characteristics of energy consumption, the characteristics of energy efficiency synergy, the rate of change of peak and valley electricity consumption, the matching degree of electricity charges, the distribution entropy of payment method, and the duration of payment delay.

5. The method according to claim 1, characterized in that, The process of using a pre-trained risk prediction model to determine the user value level of the target user based on the target electricity consumption behavior characteristics, and then determining the target user's electricity bill overdue risk level based on the user value level, includes: The target electricity consumption behavior characteristics are divided into target electricity consumption characteristics and target payment characteristics; Using a pre-trained risk prediction model, cluster value clusters are determined based on the target electricity consumption characteristics, and the user value level of the target user is determined from the cluster value clusters; The payment characteristics corresponding to the target value cluster are determined from the target payment characteristics, and the overdue electricity bill risk level of the target user is determined based on the payment characteristics corresponding to the target value cluster.

6. The method according to claim 5, characterized in that, The step of using a pre-trained risk prediction model to determine cluster value clusters based on the target electricity consumption characteristics, and determining the user value level of the target user from the cluster value clusters, includes: The continuous features in the target electricity consumption characteristics are normalized to obtain the first input feature, and the discrete features in the target electricity consumption characteristics are encoded to obtain the second input feature; Using a pre-trained risk prediction model, clustering is performed on the first input feature and the second input feature to obtain cluster value clusters; The target value cluster is determined from the clustered value clusters using Bayesian information criteria, and the target value cluster is used as the user value level of the target user.

7. The method according to claim 1, characterized in that, The pre-training process of the risk prediction model includes: Obtain sample data of user electricity consumption, and divide the sample data into a first sample sequence, a second sample sequence, and a third sample sequence according to time sequence; A first initial prediction model is obtained by training a neural network using the first sample sequence and the second sample sequence, and a first target prediction model is obtained by validating the first initial prediction model using the third sample sequence. The neural network is trained using the first sample sequence and the third sample sequence to obtain a second initial prediction model, and the second initial prediction model is validated using the second sample sequence to obtain a second target prediction model. The neural network is trained using the second sample sequence and the third sample sequence to obtain a third initial prediction model, and the third initial prediction model is validated using the first sample sequence to obtain a third target prediction model. A risk prediction model is determined based on the first target prediction model, the second target prediction model, and the third target prediction model.

8. A device for predicting the risk of overdue electricity bills, characterized in that, include: The acquisition module is configured to acquire the target user's electricity consumption data and payment data. The feature determination module is configured to determine the target electricity consumption behavior features of the target user based on the user type of the target user, according to the electricity consumption data and the payment data. The risk prediction module is configured to use a pre-trained risk prediction model to determine the user value level of the target user based on the target electricity consumption behavior characteristics, and to determine the overdue electricity bill risk level of the target user based on the user value level.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the program, implements the method as claimed in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to perform the method according to any one of claims 1 to 7.

Images