A method for identifying power customer calls for complaints

Abstract

The present invention provides a method for identifying power customer's incoming call complaints, including: when receiving a customer's incoming call, matching whether there is a business interaction behavior of the customer recently; when there is a business interaction behavior of the customer recently, determining the complaint cause category corresponding to the customer's recent business interaction behavior; counting the time interval and the number of incoming calls between this customer's incoming call and the most recent business interaction behavior, and taking the time interval and the number of incoming calls as input quantities and inputting them into a pre-trained statistical model to obtain the certainty interval of this incoming call; calculating the corresponding complaint risk score according to the certainty interval of this incoming call, and judging whether this incoming call behavior is a complaint incoming call according to a preset risk threshold. If it is determined to be a complaint incoming call, a reminder will be given to the customer service staff. The present invention extracts the distribution relationship between the time interval and the incoming call frequency between the incoming call complaint behavior and the initial cause service, calculates the complaint risk when the customer makes an incoming call, and identifies the incoming call complaint behavior of power customers.

Description

Technical Field

[0001] The present invention relates to the technical field of power system automation, and particularly to a method for identifying power customer calls for complaints. Background Art

[0002] With the continuous advancement of the power market reform, high-quality power supply has become the main demand of power customers, which also poses a test to the customer service and marketing methods of power enterprises. There is an urgent need for more intelligent methods to improve the initiative and effectiveness of customer service and establish an intelligent customer service system. As an important issue in the process of power customer service, identifying and predicting power customer calls for complaints can assist customer service personnel in organizing conversations, locating potential causes of complaints, calming customer emotions, etc., greatly improving the initiative and quality of customer service.

[0003] Customer complaint behaviors usually stem from specific business behaviors. The main causes include: direct power consumption business behaviors such as power outages and power quality that directly affect users' production and living electricity consumption, and power consumption business-related factors such as service efficiency and service quality in the process of electricity consumption and electricity charges or handling power consumption business processes. These characteristic factors related to customer complaint behaviors are reflected in data such as historical customer service work orders, power consumption data, business expansion work orders, and power outage event information. By analyzing the characteristics of customers in power consumption business and its related factors, the potential risk of customer calls for complaints can be mined and calculated.

[0004] In current research and related applications for power customer complaints, information related to customer complaints is often extracted through customer-related tags and fields or text mining in customer service work orders, and a typical machine learning classification model is constructed to predict and identify whether a customer service work order belongs to a complaint type work order. However, such methods are relatively broad in feature selection related to complaint classification, do not consider the characteristics of specific different types of businesses in detail, and some features require extracting text information of the call content in the customer service work order. In the current power customer service system, the call content text is usually obtained by customer service operators organizing relevant information and standardizing the record of the call content during the call service process, and it is difficult to obtain features in real time. Summary of the Invention

[0005] The purpose of the present invention is to propose a method for identifying power customer calls for complaints, and solve the technical problem that existing methods have a large demand for identifying power customer calls for complaints and cannot accurately identify the relationship between customers' power consumption business and their call behaviors for complaints.

[0006] On the one hand, a method for identifying power customer calls for complaints is provided, including:

[0007] When receiving a customer call, based on the pre-determined associated results of customer call complaints and the associated results of complaint cause business work orders, determine whether the customer has recent business interaction behaviors;

[0008] When the customer has recent business interaction behaviors, determine the complaint cause categories corresponding to the customer's recent business interaction behaviors according to the pre-determined associated results of complaint cause business work orders;

[0009] Statistically calculate the time interval and call times between this customer call and the most recent business interaction behavior, and use the time interval and the call times as input quantities to input into a pre-trained statistical model to obtain the certainty interval of this call;

[0010] Calculate the corresponding complaint risk score according to the certainty interval of this call, and judge whether this call behavior is a complaint call according to a preset risk threshold. If it is determined to be a complaint call, output the corresponding complaint risk score, potential complaint cause category, and initial associated complaint business, and give a reminder to the customer service staff.

[0011] Preferably, the associated results of customer call complaints are determined according to the following steps:

[0012] Obtain the power customer information table, electricity customer contact person table, power outage event notice table, and electricity bill information notice table, and determine the telephone number and corresponding user number of the power customer according to the power customer information table, electricity customer contact person table, power outage event notice table, and electricity bill information notice table;

[0013] Obtain the customer service historical work order data, and determine the work order call number corresponding to the user number according to the customer service historical work order data; and determine the call complaint number corresponding to the user number according to the historical complaint work orders in the customer service historical work order data;

[0014] Associate the obtained user number, the telephone number of the power customer, the call complaint number, and the work order call number to obtain the associated results of customer call complaints.

[0015] Preferably, the pre-determined associated results of complaint cause business work orders are determined according to the following steps:

[0016] Obtain the historical complaint work order data, and identify the user number, power outage order number, business expansion work order number, and customer service work order number in the historical complaint work order data;

[0017] Associate and match the identified user number, power outage order number, business expansion work order number, and customer service work order number with the corresponding power outage event list, business expansion work order list, customer service work order list, and complaint cause categories to obtain the initial business behavior information corresponding to the complaint cause categories;

[0018] Associate the user number, the category of complaint causes, and the initial business behavior information of the corresponding complaints to obtain the associated result of the complaint cause business work order.

[0019] Preferably, the category of complaint causes is determined according to the following steps:

[0020] Statistically calculate the proportion of complaint work orders in recent years in the customer service call work orders;

[0021] Statistically calculate the proportion of each customer service call complaint work order originating from each secondary business subclass, and calculate the average value of the proportions over the years to obtain the proportion value of each secondary business subclass; among them, the secondary subclasses at least include frequent power outages, fault power outages, service efficiency, business processes, abnormal electricity consumption, service errors.

[0022] Sort the proportion values of the secondary business subclasses from large to small to obtain the sorting result of the secondary business subclasses, and select the secondary business subclasses with a cumulative proportion exceeding the preset proportion threshold in the sorting result of the secondary business subclasses, and re-integrate the selected secondary business subclasses into power outage-related types, electricity charge-related types, business expansion business-related types, business service-related types and other complaint causes to obtain the category of complaint causes.

[0023] Preferably, matching whether the customer has recent business interaction behavior according to the pre-determined customer call complaint association result and the complaint cause business work order association result specifically includes:

[0024] When there is a business interaction behavior for the user number associated with the customer call number, it is determined that the customer has recent business interaction behavior;

[0025] When there is no business interaction behavior for the user number associated with the customer call number, it is determined that the customer has no recent business interaction behavior.

[0026] Preferably, the pre-trained statistical model is determined according to the following steps:

[0027] Obtain historical complaint work order data, and statistically calculate the time interval between the call time of each historical complaint work order and the time of the complaint cause business work order;

[0028] Statistically calculate the number of calls for the same business from the time of receiving the complaint cause business work order to the call time of the historical complaint work order for this call number;

[0029] Integrate the time interval data of each historical complaint work order and its corresponding complaint cause business according to different businesses and different call times to obtain a data group composed of the call times and the time intervals of the complaint work orders;

[0030] Input a data set composed of the number of incoming calls and the time interval of complaint work orders into a preset initial statistical model to calculate the corresponding correlation statistics and digital feature quantities, and obtain a trained statistical model.

[0031] Preferably, the initial statistical model includes:

[0032]

[0033]

[0034] Among them, T yw,k =[t 1 ,t 2 ,...,t n represents the time interval data set of complaint work orders with the number of incoming calls being k. represents the mean of the time interval data T yw,k with the number of complaint incoming calls being k, k represents the number of incoming calls, yw represents the cause of the complaint business, and B 1 represents the absolute central moment of the T yw,k data, α represents the hyperentropy scaling coefficient, and S 2 represents the variance of the T yw,k data, E n represents entropy, H e represents hyperentropy, and E x represents the expected value.

[0035] Preferably, calculating the corresponding complaint risk score according to the certainty interval of this incoming call includes:

[0036]

[0037] Among them, a′ m represents the business proportion, r represents the complaint risk value, t represents the time interval between the customer's incoming call and the potential complaint cause business, k represents the number of incoming calls within the time interval, represents the certainty interval, r 2 represents the lower bound of the risk value interval, and r 1 represents the upper bound of the risk value interval. represents the coefficient composed of the business proportion a′ m and yw represents the cause of the complaint business.

[0038] Preferably, calculating the corresponding complaint risk score according to the certainty interval of this incoming call further includes:

[0039] R = r 2 +β·(r 1 -r 2 )

[0040] Among them, R represents the complaint risk score, and r2 Represents the lower bound of the risk value range, r 1 Represents the upper bound of the risk value range, and β represents the complaint risk interval coefficient between 0 and 1.

[0041] Preferably, the determining whether the incoming call behavior is a complaint call according to a preset risk threshold includes:

[0042] When the corresponding complaint risk score is greater than or equal to the preset risk threshold, it is determined as a complaint call;

[0043] When the corresponding complaint risk score is less than the preset risk threshold, it is determined not to be a complaint call.

[0044] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0045] The method for identifying power customer incoming call complaints provided by the present invention focuses on the specific services handled by customers and related power consumption problems, excavates the complaint characteristics of different historical power consumption service causes, extracts the distribution relationship between the time interval and incoming call frequency between the incoming call complaint behavior and the initial cause service, calculates the complaint risk when the customer makes a call, and identifies the power customer incoming call complaint behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, obtaining other drawings without creative efforts still belongs to the scope of the present invention.

[0047] Figure 1 It is a main process schematic diagram of a method for identifying power customer incoming call complaints in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings.

[0049] As Figure 1 shown, it is a schematic diagram of an embodiment of a method for identifying power customer incoming call complaints provided by the present invention. In this embodiment, the method includes the following steps:

[0050] When receiving a customer call, match whether the customer has recent business interaction behaviors according to the pre-determined correlation results of customer call complaints and the correlation results of complaint cause business work orders; that is, integrate the correlation relationship between the user numbers related to the customer's electricity consumption business and the call numbers that initiate complaint behaviors from customer information tables such as the on-file archive information, customer service work order information, business expansion work order information, power outage event information, and customer service work order user association table in the marketing system and the data tables related to the customer's electricity consumption business. Statistically calculate the proportion of complaint work orders in the past 5 years and the average proportion of each secondary business subclass in the customer service call complaint work orders each year to obtain the complaint proportion factors for each business category. Re-integrate the complaint subclasses from largest to smallest proportion and divide them into power outage-related types, electricity quantity and electricity charge-related types, business expansion business-related types, business service-related types, and other complaint causes. Through the key information such as the call number of the complaint call work order, the user number involved, and the work order number and power outage order number involved in the call content text, associate with the customer service work order information table, business expansion work order, and power outage time information table, and associate each complaint business work order to the corresponding complaint cause business work order.

[0051] In a specific embodiment, the correlation result of the customer call complaint is determined according to the following steps: Obtain the power customer information table, electricity customer contact table, power outage event notice table, and electricity charge information notice table, and determine the telephone number of the power customer and the corresponding user number according to the power customer information table, electricity customer contact table, power outage event notice table, and electricity charge information notice table; Obtain the historical customer service work order data, and determine the call number of the work order corresponding to the user number according to the historical customer service work order data; And determine the call complaint number corresponding to the user number according to the historical complaint work order in the historical customer service work order data; Associate the obtained user number, the telephone number of the power customer, the call complaint number, and the call number of the work order to obtain the correlation result of the customer call complaint. That is, obtain the corresponding relationship between the telephone number (used to initiate customer behavior) and the user number (representing electricity consumption and business handling information) from system marketing forms such as the power customer information table, electricity customer contact table, power outage event notice table, and electricity charge information notice table. Use the customer service work order data to mine the user number in the work order content to match the call number of the work order, and mine the work order number in the work order content to match the user number in the business expansion work order information to match the call number of the work order.

[0052] Specifically, the pre-determined complaint cause work order association result is determined according to the following steps: Obtain historical complaint work order data, and identify the user number, power outage order number, business expansion work order number, and customer service work order number in the historical complaint work order data; Correlate and match the identified user number, power outage order number, business expansion work order number, and customer service work order number with the corresponding fields in the power outage event list, business expansion work order list, customer service work order list, and complaint cause categories to obtain the initial business behavior information corresponding to the complaint cause categories; Associate the user number, complaint cause category, and the corresponding initial business behavior information of the complaint to obtain the complaint cause work order association result. That is, conduct an initial investigation of the complaint causes according to the divided complaint cause categories, mine and match through the text of the incoming call content of the complaint work order. In the way of regular expressions, extract the user number, power outage order number, business expansion work order number, customer service work order number, etc. from the defined methods of different numbers, and conduct corresponding field association and matching with the power outage event list, business expansion work order list, and customer service work order list to find the initial business behavior information of the complaint. If no valid information can be mined from the text information of the complaint work order to match the relevant potential business, search respectively according to the association relationship between the complaint incoming call number and the user number: For power outage-related complaints, first find the user number associated with the incoming call number of the complaint work order from 1, and then use this user number to match the information in the power outage event table to find the user's most recent power outage event; For complaints related to electricity quantity and electricity charges, match the user's most recent payment behavior; For complaints related to business expansion services, similar to the power outage type, match the business expansion work order recently handled by the customer through the associated user number; For power outage complaints related to business services, according to the customer's most recent power outage event or business expansion service. If there is no relevant work order, use the customer service work order for the first consultation and query of the relevant business.

[0053] The complaint cause category is determined according to the following steps: Statistically analyze the proportion of complaint work orders in the customer service incoming call work orders in recent years; Statistically analyze the proportion of each secondary business subclass in the customer service incoming call complaint work orders each year, and calculate the average value of the proportions over the years to obtain the proportion values of each secondary business subclass; Among them, the secondary subclasses at least include frequent power outage class, fault power outage class, service efficiency class, business process class, abnormal electricity quantity class, service error class; Sort the proportion values of the secondary business subclasses from large to small to obtain the sorting result of the secondary business subclasses, and select the secondary business subclasses with a cumulative proportion exceeding the preset proportion threshold in the sorting result of the secondary business subclasses, and re-integrate the selected secondary business subclasses into power outage-related type, electricity quantity and electricity charge-related type, business expansion service-related type, business service-related type, and other complaint causes to obtain the complaint cause category. That is, statistically analyze the proportion a of complaint work orders in the customer service incoming call work orders in the recent 5 years 0Statistically count the proportion of customer service call complaint work orders originating from each secondary business subclass annually, and calculate the average of the proportions over multiple years to obtain the proportion of each secondary business subclass: a 1 , a 2 ,..., a n (n is the number of secondary business subclass categories of complaint services, and each subclass includes: frequent power outages, fault power outages, service efficiency, business processes, abnormal electricity consumption, service errors, etc.). Select the main complaint cause factor subclasses with a cumulative proportion exceeding 95% from largest to smallest, and according to the complaint preferences and customer service experience, re-integrate the business categories to convert the original proportion into a′ 1 , a′ 2 ,..., a′ m (m is the number of complaint service categories after integrating the secondary business subclass categories of complaint services). In this embodiment, the complaint causes after integration are divided into power outage-related type a 1 ′ = 0.495, electricity charge-related type a′ 2 = 0.107, business expansion service-related type a 3 ′ = 0.059, business service-related type a′ 4 = 0.182, and other complaint causes a′ 5 = 0.157.

[0054] More specifically, when there is a business interaction behavior for the user number associated with the customer's call number, it is determined that the customer has a recent business interaction behavior; when there is no business interaction behavior for the user number associated with the customer's call number, it is determined that the customer has no recent business interaction behavior.

[0055] Furthermore, when the customer has a recent business interaction behavior, determine the complaint cause category corresponding to the customer's recent business interaction behavior according to the pre-determined complaint cause business work order association result; that is, through the association relationship between the customer user number and the call number, and using the relevant business data tables in the system, match whether the customer has handled relevant services or experienced a power outage event recently.

[0056] Furthermore, count the time interval and the number of calls between this customer call and the most recent business interaction behavior, and use the time interval and the number of calls as input quantities to input into a pre-trained statistical model to obtain the certainty interval of this call; that is, if the customer has a business interaction behavior, calculate the time interval and the number of calls between this call and the relevant service, and according to the statistical model, obtain the membership interval of this call, and combine the relevant calculation parameters to comprehensively calculate the risk of this call.

[0057] In specific embodiments, the pre-trained statistical model is determined according to the following steps: Obtain historical complaint work order data, and count the time interval between the call time of each historical complaint work order and the time of the complaint cause business work order; Count the number of calls from the same business by the call number from the acceptance time of the complaint cause business work order to the call time of the historical complaint work order; Integrate the time interval data of each historical complaint work order and its corresponding complaint cause business according to different businesses and different call times to obtain a data group composed of the call number and the time interval of the complaint work order; Input the data group composed of the call number and the time interval of the complaint work order into a preset initial statistical model to calculate the corresponding correlation statistics and digital feature quantities, and obtain the trained statistical model. That is, calculate the time interval between the call time of each historical complaint work order and the time of the complaint cause business work order. Count the number of calls from the same business by the call number from the acceptance time of the complaint cause business work order to the call time of this complaint. Integrate the time interval data of each historical complaint work order and its corresponding complaint cause business according to different businesses and different call times to obtain the time interval data group T yw,k =[t 1 ,t 2 ,...,t n (k = 1, 2, 3…), where yw represents the cause of the complaint business, including the power outage related type, electricity quantity and electricity charge related type, business expansion business related type, business service related type, and other complaint causes mentioned above. From the time interval data T yw,k , calculate the correlation statistics and digital feature quantities, and generate a cloud model distribution.

[0058] Specifically, the initial statistical model includes:

[0059]

[0060] Among them, T yw,k =[t 1 ,t 2 ,...,t n represents the time interval data group of complaint work orders with a call number of k, represents the mean of the time interval data T yw,k with a complaint call number of k, k represents the call number, yw represents the cause of the complaint business, B 1 represents the absolute central moment of the T yw,k data, α represents the hyperentropy scaling coefficient, S 2 represents the variance of the T yw,k data, E n represents entropy, H e represents hyperentropy, E x represents the expected value.

[0061] Specifically, generating the cloud model (statistical model) distribution specifically includes the following steps:

[0062] Calculate the time interval data T for the business yw with the number of complaint calls being k yw,k for its mean value, as shown in the following formula:

[0063]

[0064] Calculate the absolute central moment B of the T yw,k data, with the formula being: 1

[0065] Calculate the variance S of the T yw,k data, with the formula being: 2

[0066] Further obtain the expectation entropy and hyperentropy Among them, the hyperentropy parameter is used to characterize the degree of data fluctuation and can reflect the uncertainty impact of the call time interval. α is the hyperentropy scaling coefficient, which is used to control the fluctuation degree of cloud droplet generation later. In this example, its value is taken as 0.2.

[0067] Generate cloud model cloud droplets using the forward cloud generator, and the steps include

[0068] 1) Generate a normal random number E' with an expectation of E n and a variance of H e 2 ; n

[0069] 2) Generate a normal random number x with an expectation of E x and a variance of ;

[0070] 3) Calculate the certainty degree, and the formula is as follows

[0071]

[0072] 4) Generate cloud droplets according to the normal random number x and the certainty degree μ;

[0073] 5) Repeatedly execute steps 1) to 4) to generate the required n cloud droplets.

[0074] 4.6: Calculate the upper and lower certainty degree boundaries of the cloud model through the parameters in 4.4, which are respectively:

[0075]

[0076] ​​​The upper and lower boundaries of the cloud model represent that: for business yw, among the incoming calls with the number of complaint calls being k, when the time interval is t, the certainty degree is within the interval Inside.

[0077] Furthermore, calculate the corresponding complaint risk score according to the certainty degree interval of this incoming call, and judge whether this incoming call behavior is a complaint call according to the preset risk threshold. If it is determined to be a complaint call, output the corresponding complaint risk score, potential complaint cause category, and initial associated complaint business, and remind the customer service staff.

[0078] In a specific embodiment, for a new customer service incoming call behavior, through the obtained correspondence between the incoming call number and the user number, query whether the user has potential complaint cause services such as power outage, business consultation and query, and business expansion work order in the near future. The query historical time node here is determined by multiplying the longest time limit of the business time of the cause service type and the business process specification in the customer service process by 1.5 times. If there is a potential complaint business, multiply the basic score by the coefficient m Composed of That is, the complaint risk value at this time is Among them, m1, m2 to mm are the types of potential complaint cause services that have been associated. The calculating the corresponding complaint risk score according to the certainty degree interval of this incoming call includes:

[0079]

[0080] Among them, a′ m Represents the business proportion, r represents the complaint risk value, t represents the time interval between the customer's incoming call and the potential complaint cause service, k represents the number of incoming calls within the time interval, Represents the certainty degree interval, r 2 Represents the lower bound of the risk value interval, r 1 Represents the upper bound of the risk value interval, Represents the coefficient composed of the business proportion a′ m Composed, yw represents the cause of the complaint business.

[0081] Specifically, the calculating the corresponding complaint risk score according to the certainty degree interval of this incoming call further includes:

[0082] R = r 2 +β·(r 1 -r 2 )

[0083] Among them, R represents the complaint risk score, r 2 Represents the lower bound of the risk value interval, r 1represents the upper bound of the risk value interval, and β represents the complaint risk interval coefficient between 0 and 1. Set the complaint risk threshold ε. When the complaint risk value satisfies R = r 2 +β·(r 1 -r 2 ), it is considered an incoming call complaint behavior, and an incoming call complaint warning is issued. Among them, β is the complaint risk interval coefficient between 0 and 1, and in this example, it takes 0.2.

[0084] When the corresponding complaint risk score is greater than or equal to the preset risk threshold, it is determined as a complaint incoming call; when the corresponding complaint risk score is less than the preset risk threshold, it is determined not to be a complaint incoming call. That is, set the complaint risk threshold and the complaint behavior determination condition, and judge whether the complaint risk value of this incoming call behavior is satisfied. If it is satisfied, it is considered a complaint behavior, and the corresponding complaint risk score, potential complaint cause category, and initial associated complaint business are output to remind the customer service staff, so that the customer service staff can quickly locate the potential cause and make a coping strategy in advance. Specifically: set the complaint risk threshold ε, that is, the complaint behavior determination condition r 2 +β·(r 1 -r 2 )≥ε, where β is the complaint risk interval coefficient between 0 and 1, and in this example, it takes 0.2.

[0085] When the complaint risk value satisfies r 2 +β·(r 1 -r 2 )≥ε, it is considered an incoming call complaint behavior, and an incoming call complaint warning reminder is issued, otherwise no complaint reminder is issued. For an incoming call with a business complaint risk degree greater than the threshold, it is considered a complaint work order to remind the customer service staff, and the complaint risk score, potential complaint cause category, and initially associated complaint business of this incoming call behavior are output and displayed, so that the customer service staff can quickly locate the potential cause, make a coping strategy in advance, and organize the conversation to soothe the customer's emotions and improve the customer service level.

[0086] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0087] The method for identifying power customer incoming call complaints provided by the present invention focuses on the specific business handled by the customer and related power consumption problems, excavates the complaint characteristics of different historical power consumption business causes, extracts the distribution relationship between the time interval and incoming call frequency between the incoming call complaint behavior and the initial cause business, calculates the complaint risk when the customer makes an incoming call, and identifies the incoming call complaint behavior of power customers. The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited by this. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A method for identifying power customer calls for complaints, characterized in that, it includes: When receiving a customer call, according to the pre-determined correlation result of customer call complaints and the correlation result of complaint cause service work orders, match whether the customer has business interaction behaviors recently; When the customer has business interaction behaviors recently, determine the complaint cause category corresponding to the customer's recent business interaction behaviors according to the pre-determined correlation result of complaint cause service work orders; Statistically calculate the time interval and call times between this customer call and the most recent business interaction behavior, and use the time interval and the call times as input quantities to input into a pre-trained statistical model to obtain the certainty interval of this call; Among them, the statistical model includes: Among them, T yw,k = [t 1 , t 2 ,..., t n represents the time interval data group of complaint work orders with the call count of k, k represents the call count, yw represents the cause of complaint business, B 1 represents the absolute central moment of the T yw,k data, α represents the hyperentropy scaling coefficient, S 2 represents the variance of the T yw,k data, E n represents entropy, H e represents hyperentropy, E x represents the expected value, μ 1 represents the upper certainty boundary of the model, μ 2 represents the lower certainty boundary of the model; Calculate the corresponding complaint risk score according to the certainty interval of this call, and judge whether this call behavior is a complaint call according to a preset risk threshold. If it is determined to be a complaint call, output the corresponding complaint risk score, potential complaint cause category, and initial associated complaint service, and remind the customer service staff; Among them, the calculation of the corresponding complaint risk score according to the certainty interval of this call includes: Among them, a′ m represents the business proportion, r represents the complaint risk value, and t represents the time interval between the customer's call and the business of the potential complaint cause represents the certainty interval, r 2 represents the lower bound of the risk value interval, r 1 represents the upper bound of the risk value interval represents the coefficient formed by the business proportion a′ m m1, m2...mm represent the types of potential complaint cause businesses, and a′ m1 +a′ m2 …+a′ mm represents the business proportion corresponding to the type of potential complaint cause business, and a 0 is the proportion of complaint work orders in the customer service call work orders 2. The method according to claim 1, characterized in that, The correlation result of customer call complaints is determined according to the following steps: Obtain the power customer information table, electricity customer contact table, power outage event notice table, and electricity bill information notice table, and determine the telephone number and corresponding user number of the power customer according to the power customer information table, electricity customer contact table, power outage event notice table, and electricity bill information notice table; Obtain the customer service historical work order data, and determine the work order call number corresponding to the user number according to the customer service historical work order data; and determine the call complaint number corresponding to the user number according to the historical complaint work orders in the customer service historical work order data; Associate the obtained user number, the telephone number of the power customer, the call complaint number, and the work order call number to obtain the correlation result of customer call complaints.

3. The method according to claim 2, characterized in that, The pre-determined correlation result of complaint cause service work orders is determined according to the following steps: Obtain historical complaint work order data, and identify the user number, power outage order number, business expansion work order number, and customer service work order number in the historical complaint work order data; Perform corresponding field association matching on the identified user number, power outage order number, business expansion work order number, customer service work order number with the corresponding power outage event list, business expansion work order list, customer service work order list, and complaint cause category to obtain the initial business behavior information corresponding to the complaint cause category; Associate the user number, complaint cause category, and the corresponding initial business behavior information of the complaint to obtain the correlation result of complaint cause service work orders.

4. The method according to claim 3, characterized in that, The complaint cause category is determined according to the following steps: Statistically calculate the proportion of complaint work orders in customer service call work orders in recent years; Count the proportion of each secondary business subclass in the customer service call complaint work orders every year, and calculate the average value of the proportions over the years to obtain the proportion values of each secondary business subclass; where the secondary subclasses at least include frequent power outages, fault power outages, service efficiency, business processes, abnormal electricity consumption, service errors. Sort the secondary business subclasses in descending order according to their proportion values to obtain the sorting result of the secondary business subclasses, and select the secondary business subclasses with a cumulative proportion exceeding the preset proportion threshold in the sorting result of the secondary business subclasses, and re-integrate the selected secondary business subclasses into power outage-related types, electricity charge-related types, business expansion service-related types, business service-related types and other complaint causes to obtain the complaint cause categories.

5. The method according to claim 4, characterized in that the matching of whether the customer has a recent business interaction behavior according to the pre-determined customer call complaint association result and the complaint cause business work order association result specifically includes: When there is a business interaction behavior for the user number associated with the customer call number, it is determined that the customer has a recent business interaction behavior; When there is no business interaction behavior for the user number associated with the customer call number, it is determined that the customer does not have a recent business interaction behavior.

6. The method according to claim 5, characterized in that the pre-trained statistical model is determined according to the following steps: Obtain historical complaint work order data, and count the time interval between the call time of each historical complaint work order and the time of the complaint cause business work order; Count the number of calls for the same business from the time of acceptance of the complaint cause business work order to the call time of the historical complaint work order for this call number; Integrate the time interval data of each historical complaint work order and its corresponding complaint cause business according to different businesses and different call times to obtain a data group composed of the call times and the time intervals of the complaint work orders; Input the data group composed of the call times and the time intervals of the complaint work orders into a preset initial statistical model to calculate the corresponding correlation statistics and digital feature quantities, and obtain the trained statistical model.

7. The method according to claim 1, characterized in that the calculation of the corresponding complaint risk score according to the certainty interval of this call further includes: R＝r 2 +β·(r 1 -r 2 ) Among them, R represents the complaint risk score, r 2 represents the lower bound of the risk value interval, r 1 represents the upper bound of the risk value interval, and β represents the complaint risk interval coefficient between 0 and 1.

8. The method according to claim 7, characterized in that the judgment of whether this call behavior is a complaint call according to the preset risk threshold includes: When the corresponding complaint risk score is greater than or equal to the preset risk threshold, it is determined to be a complaint call; When the corresponding complaint risk score is less than the preset risk threshold, it is determined not to be a complaint call.

Images