Power consumer complaint risk prediction method and device

By obtaining and analyzing the power user's electricity preference information and telephone communication records, and combining the dialogue matching between power users and power customer service, the problem of low accuracy of power user complaint risk prediction in the existing technology is solved, and the accuracy of prediction is improved .

CN118917663APending Publication Date: 2024-11-08STATE GRID HEBEI ELECTRIC POWER CO LTD +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410960118.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing methods for predicting risk of complaints from power users are low, mainly due to the deviation in the emotional identification of power users and power customer service, and the failure to effectively consider the information and language differences in electricity usage preferences.

Method used

By obtaining power users' electricity preference information, telephone communication records and power user information, determining the Chinese system to which they belong and correct it, and complying with power usage preferences and initial sentiment to calculate the complaint risk data, and input it into the power complaint prediction model. The dialogue between power users and power customer service is matched to improve prediction accuracy.

Benefits of technology

The accuracy of forecasting of risk level for power users complaints is improved, and by considering power consumption preferences and dialogue matching, emotional recognition bias is reduced and prediction reliability is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917663A_ABST
    Figure CN118917663A_ABST
Patent Text Reader

Abstract

The invention provides a power consumer complaint risk prediction method and device, and belongs to the field of power service risk early warning. The method comprises the following steps: acquiring power consumption preference information of a to-be-predicted power user, power user information and a telephone communication record; wherein the electricity consumption preference information comprises electricity price information, power failure information and circuit maintenance information; the power consumer information comprises a power complaint record and a power supply area; determining the initial emotion data, the power processing items and the dialogue matching degree between the to-be-predicted power user and the power customer service according to the telephone communication record; determining first complaint risk data according to the power consumption preference information and the initial emotion; determining second complaint risk data according to the power processing item and the power user information; and inputting the first complaint risk data, the second complaint risk data and the dialogue matching degree into an electric power complaint prediction model to obtain a predicted electric power complaint risk level of the to-be-predicted electric power user. The accuracy of power complaint risk prediction can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electric power service risk early warning, and in particular to a method and device for predicting electric power user complaint risks. Background Art

[0002] The customer service hotline of the power grid is an important bridge to connect and serve customers, and is also the main source for customers to express their demands and suggestions. The number of customer service complaint tickets of the power grid directly reflects the quality of power supply services, and further explains whether the power supply company has created value for customers.

[0003] The baseline risk of power grid customer service complaints is mainly reflected in the following aspects: for the power service work orders being processed, there is a lack of reliable risk analysis and prediction methods for escalating power service order complaints; and for power service risk work orders that are not easily detected by humans, they are prone to develop into escalated complaint power work orders.

[0004] At present, for the prediction of power users' complaints, the existing power users' complaint risk prediction is based on the historical power complaint records, the power users' emotions during the communication process and the power customers' emotions, etc. In the process of identifying the emotions of power users and power customers, there are often deviations, which makes the prediction results inaccurate. Summary of the invention

[0005] The embodiments of the present invention provide a method and device for predicting the risk of power user complaints, so as to solve the problem of low accuracy in predicting the risk of power complaints at the current stage.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting the risk of power user complaints, comprising:

[0007] Obtain the power preference information, power user information and telephone communication records of the power user to be predicted; the power preference information includes power price information, power outage information and circuit maintenance information; the power user information includes power complaint records and the power supply area;

[0008] Determine the Chinese language system to which the power users to be predicted belong, power handling matters, and the degree of match between the conversations between the power users to be predicted and the power customer service based on the telephone communication records;

[0009] The telephone communication record is modified based on the Chinese language system to which the power user to be predicted belongs; and the initial emotion is determined based on the modified telephone communication record;

[0010] Determine the first complaint risk data based on the electricity preference information and initial emotions; determine the second complaint risk data based on the electricity handling matters and electricity user information;

[0011] The first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0012] In a possible implementation, the Chinese system includes Mandarin, Mandarin dialect, Jin dialect, Wu dialect, Min dialect, Hakka dialect, Yue dialect, Xiang dialect, Gan dialect, Hui dialect and Pinghua dialect;

[0013] Determine the Chinese language system of the power users to be predicted based on telephone communication records, including:

[0014] Extract the recorded text and voice records of the power users to be predicted from the telephone communication records;

[0015] Determine the Chinese language system to which the power users to be predicted belong based on the recorded text and voice records.

[0016] In a possible implementation, the telephone communication record is modified based on the Chinese system to which the power user to be predicted belongs, including:

[0017] Based on the Chinese language system to which the power user to be predicted belongs, determining the standard volume and standard speaking speed corresponding to the Chinese language system;

[0018] Determine the correction coefficient according to the standard volume and standard speaking speed corresponding to the Chinese system and the standard volume and standard speaking speed corresponding to Mandarin;

[0019] The voice information of the power user to be predicted in the telephone communication record is corrected based on the correction coefficient.

[0020] In a possible implementation, determining the conversation matching degree between the power user to be predicted and the power customer service according to the telephone communication records includes:

[0021] Extract the content of each question asked by the power user to be predicted in the communication record and the power customer service reply content corresponding to each question;

[0022] Extract keywords from each question, and search for standard responses corresponding to each question based on the extracted keywords;

[0023] The conversation matching degree between each standard reply content and the corresponding power customer service answer content is calculated to obtain the conversation matching degree between the power user to be predicted and the power customer service.

[0024] In a possible implementation, searching for each standard reply content corresponding to each question content according to the extracted keywords includes:

[0025] Mapping the keywords in each question content to the standard power customer service database to obtain the standard question content corresponding to each question content; wherein the standard power customer service database includes multiple standard question contents and standard reply content corresponding to each standard question content;

[0026] Based on the content of each standard question, determine the standard response content corresponding to each question.

[0027] In a possible implementation, determining the first complaint risk data according to the electricity preference information and the initial emotion includes:

[0028] Searching for the power public opinion information of the corresponding power supply area according to the power consumption preference information; wherein the power public opinion information includes positive information and negative information;

[0029] Determine the first complaint risk data based on power public opinion information and initial sentiment data.

[0030] In a possible implementation, determining initial emotion data based on telephone communication records includes:

[0031] Perform voice recognition on telephone communication records and extract the recorded text in the communication records;

[0032] Detecting a preset sentence of the power user to be predicted in the recording text to obtain first emotion data of the power user to be predicted; wherein the preset sentence is a specified sentence in the recording text;

[0033] Determine the second emotion data according to the speech speed and volume of the preset sentence of the power user to be predicted in the telephone communication record;

[0034] Initial emotion data of the power user to be predicted is obtained according to the first emotion data and the second emotion data.

[0035] In a possible implementation, the power complaint record includes the historical power complaints of the power user to be predicted and the number of complaints corresponding to each historical power complaint;

[0036] The second complaint risk data is determined based on power handling matters and power user information, including:

[0037] According to the power handling items, search the power complaint records for the historical power complaint items and the corresponding complaint times corresponding to the power handling items;

[0038] In the power system, query the complaint records of power handling matters in the power supply area within a preset time period; wherein the complaint records are used to indicate the number of times the power handling matters are complained by other users within the preset time period, and whether the complaints are handled;

[0039] The second complaint risk data is determined based on the complaint records, historical power complaint items corresponding to power handling items, and the corresponding number of complaints.

[0040] In a possible implementation, the first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted, including:

[0041] The first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the probability value corresponding to each predicted complaint risk level of the power user to be predicted;

[0042] The predicted complaint risk level corresponding to the maximum probability value is taken as the predicted complaint risk level of the power user to be predicted.

[0043] In a second aspect, an embodiment of the present invention provides a device for predicting the risk of complaints from power users, comprising:

[0044] An acquisition module is used to acquire the power preference information, power user information and telephone communication records of the power user to be predicted; wherein the power preference information includes power price information, power outage information and circuit maintenance information; the power user information includes power complaint records and the power supply area;

[0045] A matching degree calculation module is used to determine the Chinese language system to which the power user to be predicted belongs, the power handling matters, and the matching degree of the conversation between the power user to be predicted and the power customer service according to the telephone communication records;

[0046] The matching degree calculation module is also used to correct the telephone communication record based on the Chinese system to which the power user to be predicted belongs; and determine the initial emotion based on the corrected telephone communication record;

[0047] The risk data calculation module is used to determine the first complaint risk data according to the electricity preference information and the initial emotion; and determine the second complaint risk data according to the electricity handling matters and the electricity user information;

[0048] The prediction module is used to input the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0049] The embodiment of the present invention provides a method and device for predicting the risk of power user complaints. Compared with the traditional method, the embodiment of the present invention takes into account the matching degree between the power user and the power customer service dialogue, as well as the power preference information of the power user to be predicted, when predicting the risk level of power complaints. The power preference information is used to characterize the type of attention that the power user to be predicted pays to the power grid information. In addition, conventional methods assume that the power user speaks Mandarin when making predictions. However, due to language differences, power users in different regions have different speaking speeds and volumes from Mandarin. Therefore, in order to improve the accuracy of the prediction, this difference needs to be corrected. Taking these factors into account when predicting the risk of power complaints can improve the accuracy of the prediction of the risk level of power complaints. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is a flow chart for implementing the method for predicting the risk of power user complaints provided by an embodiment of the present invention;

[0052] Figure 2 It is a structural schematic diagram of a power user complaint risk prediction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0054] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0055] Figure 1 The following is a flow chart of the implementation of the method for predicting the risk of power user complaints provided by the embodiment of the present invention. Figure 1 As shown:

[0056] Step 110: Obtain the power usage preference information, power user information and telephone communication records of the power user to be predicted; wherein the power usage preference information includes electricity price information, power outage information and circuit maintenance information; the power user information includes power complaint records and the power supply area.

[0057] In this embodiment, the power usage preference information of the power user to be predicted can be determined based on the query records, browsing records, search records, etc. of the power user to be predicted in the power system; or can also be determined based on the records of the power grid information previewed by the power user to be predicted on the Internet platform, which is not limited here. Among them, the power usage preference information may include electricity price adjustment information, power outage information, line maintenance information, electricity charge recharge discount information, electricity charge overdue payment penalty information, etc., which is not limited here.

[0058] The electricity user information may include the ID (Identity document) of the electricity user to be predicted, the power supply area where the electricity user to be predicted is located, the electricity consumption of the electricity user to be predicted, and the electricity complaint record of the electricity user to be predicted; among them, the electricity complaint record of the electricity user to be predicted may include the historical electricity matters complained by the electricity user to be predicted, the number of complaints, and the handling of the complaints, etc., which are not limited here.

[0059] The telephone communication record is a telephone recording of the power user communicating with the power customer service via the phone. It includes the questions asked by the power user during the communication process and the corresponding replies from the power customer service. By analyzing the telephone communication record, it is possible to determine the issues that the power user needs to solve, the emotional information of the power user and the power customer service during the communication process, etc.

[0060] Step 120: Determine the Chinese language system to which the power user to be predicted belongs, power handling matters, and the degree of dialogue matching between the power user to be predicted and the power customer service according to the telephone communication records.

[0061] In this embodiment, telephone communication records are an important factor affecting whether electricity users complain. Therefore, by analyzing telephone communication records, the initial emotional data of the power user to be predicted, power handling matters, and the conversation matching data between the power user to be predicted and the power customer service can be determined, and the power complaint risk level of the power user to be predicted can be predicted.

[0062] The first few sentences in the telephone communication record can reflect the emotions of the power user at this time. For example, if the power user is in a relatively calm mood at this time, then the probability of a power complaint is medium; if the power user is in a relatively happy mood at this time, then the probability of a power complaint is low; if the power user is in an angry mood at this time, then the probability of a power complaint is high. In other words, the higher the negative emotions of the power user, the higher the probability of a power complaint. Therefore, determining the initial emotional data of the power user to be predicted by analyzing the telephone communication records can have an important impact on the prediction of the risk of power complaints.

[0063] However, in the process of determining the initial emotional data of the power users to be predicted by analyzing the telephone communication records, the different speaking styles of users in different regions will affect the analysis of the initial emotional data accordingly. For example, power users in the Northeast speak at a high volume and a fast speed. When they speak in a normal way, they may be considered emotionally excited, and the corresponding initial emotional data may be biased towards the negative; for another example, power users in some areas of the South speak at a low volume and a slow speed. They will only raise their volume and speed up when they are angry, but at this time, their speed and volume may be the same as Mandarin, so the initial emotional data of the power user may be misjudged as positive emotions.

[0064] In a specific embodiment, the Chinese system may include Mandarin, Mandarin dialect, Jin dialect, Wu dialect, Min dialect, Hakka dialect, Cantonese dialect, Xiang dialect, Gan dialect, Hui dialect and Pinghua dialect;

[0065] Accordingly, in step 120, determining the Chinese language system to which the power user to be predicted belongs according to the telephone communication record may include:

[0066] The recorded text and voice record of the power user to be predicted in the telephone communication record are extracted.

[0067] The Chinese language system to which the power user to be predicted belongs is determined according to the recorded text and the voice record.

[0068] In this embodiment, dialects of different regions and Mandarin can be collectively referred to as language systems. The user's words are determined based on the recorded text of the predicted power user during the communication process, and the pronunciation and intonation of the power user are determined based on the voice recording to determine which dialect the predicted power user speaks or whether it is Mandarin.

[0069] In a specific embodiment, determining the conversation matching degree between the power user to be predicted and the power customer service according to the telephone communication record in step 120 may include:

[0070] Step 2.1: Extract the content of each question asked by the power user to be predicted in the communication record and the power customer service reply content corresponding to each question.

[0071] Step 2.2: Extract keywords from each question, and search for the standard response content corresponding to each question based on the extracted keywords.

[0072] Step 2.3: Calculate the conversation matching degree between each standard reply content and the corresponding power customer service answer content to obtain the conversation matching degree between the power user to be predicted and the power customer service.

[0073] In this embodiment, during the communication process, whether the answer of the power customer service matches the question of the power user to be predicted is an important indicator affecting whether the power user to be predicted will complain. Specifically, the communication record can be divided into several groups of dialogue sentences according to the question content of each question asked by the power user to be predicted and the corresponding reply content of the power customer service. Each group of dialogue sentences includes a question content and at least one reply content.

[0074] For each group of dialogue sentences, extract the keywords of each question content, and find the corresponding standard reply sentences based on the keywords. According to the Euclidean distance formula, calculate the dialogue matching degree between the found standard reply sentences and the corresponding power customer service answer content. Fit the dialogue matching degree of each group of dialogue sentences to obtain the dialogue matching degree between the power user to be predicted and the power customer service.

[0075] In a specific embodiment, searching for each standard reply content corresponding to each question content according to the extracted keywords in step 2.2 may include:

[0076] The keywords in each question content are mapped to the standard power customer service database to obtain the standard question content corresponding to each question content; wherein the standard power customer service database includes multiple standard question contents and standard reply content corresponding to each standard question content.

[0077] Based on the content of each standard question, determine the standard response content corresponding to each question.

[0078] In this embodiment, multiple standard questions are determined based on historical telephone communication records, and corresponding standard replies are determined based on the multiple standard questions. Then, a standard power customer service database is established based on the multiple standard questions and their corresponding standard replies.

[0079] When searching for standard responses, the keywords in each question can be mapped to the standard power customer service database. The standard question corresponding to each question can be found based on the keywords. Each standard question corresponds to one or more standard responses. After determining each standard question, the standard response corresponding to each question can be determined.

[0080] Step 130: modifying the telephone communication record based on the Chinese language system to which the power user to be predicted belongs; and determining the initial emotion based on the modified telephone communication record.

[0081] In this embodiment, the telephone communication record is modified based on the Chinese system to which the power user to be predicted belongs, which may include:

[0082] Based on the Chinese language system to which the power user to be predicted belongs, a standard volume and a standard speaking speed corresponding to the Chinese language system are determined.

[0083] The correction coefficient is determined according to the standard volume and standard speaking speed corresponding to the Chinese system and the standard volume and standard speaking speed corresponding to Mandarin.

[0084] The voice information of the power user to be predicted in the telephone communication record is corrected based on the correction coefficient.

[0085] In this embodiment, voice information may include volume and speaking speed; by analyzing telephone communication records of power users in different regions, a language library is established, which includes standard volumes and standard speaking speeds corresponding to various language systems; wherein the standard volume and standard speaking speed may be the average volume and average speaking speed corresponding to the language system.

[0086] Therefore, the standard volume and standard speaking speed corresponding to the Chinese system can be searched in the language library, and the standard volume and standard speaking speed corresponding to Mandarin can be determined in the language library; the standard volume and standard speaking speed corresponding to the Chinese system and the standard volume and standard speaking speed corresponding to Mandarin are compared and calculated respectively to determine the percentage value. For example, in the Northeast dialect, the standard volume is 10% louder than the standard volume corresponding to Mandarin, and the standard speaking speed is 5% faster than the standard speaking speed corresponding to Mandarin. Then, when correcting the voice information, the volume corresponding to the power user needs to be reduced by 10%, and the speaking speed needs to be slowed down by 5%. It should be understood that this embodiment is only for illustration and does not constitute a specific limitation. In a specific embodiment, determining the initial emotional data based on the corrected telephone communication record in step 130 may include:

[0087] Step 3.1: Perform speech recognition on the telephone communication record and extract the recorded text in the communication record.

[0088] Step 3.2: Detect the preset sentence of the power user to be predicted in the recording text to obtain the first emotion data of the power user to be predicted; wherein the preset sentence is a specified sentence in the recording text.

[0089] Step 3.3: Determine the second emotion data according to the speech speed and volume of the preset sentence of the power user to be predicted in the telephone communication record.

[0090] Step 3.4: Obtain initial emotion data of the power user to be predicted based on the first emotion data and the second emotion data.

[0091] In this embodiment, since the telephone communication record is in voice form, it is necessary to convert the voice therein into the corresponding recorded text. After the conversion, the obtained recorded text can be proofread. After obtaining the recorded text, it can be detected according to the preset sentence of the power user to be predicted in the recorded text to obtain the first emotional data of the power user to be predicted. Among them, the first emotional data can be the negative emotion probability value of the power user to be predicted. Exemplarily, the negative emotion probability value can be divided into three levels, the first level can be 0≤T1≤33%; the second level can be 33%<T2≤66%; the third level can be 66%<T3≤100%, which is not specifically limited here.

[0092] For example, the preset sentence can refer to the first three sentences of the power user in the telephone communication record. The first three sentences of the power user to be predicted are: "Hello, I have a question to consult. I heard that there is a *** policy now. I want to ask about the details of this policy." The first three sentences of the user to be predicted can be input into the language analysis model, and the probability of negative emotion is 5%, that is, the first emotion data is 5%. Among them, the language analysis model can be trained through historical telephone communication records.

[0093] Or, for example, the first three sentences of the user to be predicted are: "Why do I use so much electricity this month? How do you calculate it? Did you deliberately overcharge me?" After analysis, the first emotion data is 70%. It should be understood that the above embodiment is only for illustration, and in actual use, the first emotion data is determined by the specific analysis results.

[0094] The second emotion data can be determined by speech speed and volume. When the power user speaks faster and louder, it means that the negative emotion is higher and the probability of power complaints will also increase. Specifically, the preset forecast of the power user to be predicted can be input into the language analysis model to obtain the second emotion data.

[0095] After determining the first emotion data and the second emotion data, if the first emotion data is not less than a preset threshold, the first emotion data is used as the initial emotion data of the power user. The preset threshold may be 10%, 12%, 15%, etc., which is not limited here.

[0096] If the first emotion data is less than the preset threshold, it is determined whether the second emotion data is not less than the preset threshold. If it is not less than the preset threshold, the second emotion data is used as the initial emotion data of the user.

[0097] If the second emotion data is less than a preset threshold, the first emotion data and the second emotion data are compared, and the data with a larger value is used as the initial emotion data of the power user.

[0098] Step 140: Determine first complaint risk data based on electricity preference information and initial emotions; determine second complaint risk data based on electricity handling matters and electricity user information.

[0099] In a specific embodiment, determining the first complaint risk data according to the electricity usage preference information and the initial emotion in step 140 may include:

[0100] Step 4.1: Search for the power public opinion information of the corresponding power supply area according to the power consumption preference information; wherein the power public opinion information includes positive information and negative information;

[0101] Step 4.2: Determine the first complaint risk data based on the power public opinion information and initial sentiment data.

[0102] In this embodiment, the power public opinion information may include the power system internal information and the power system external information. For example, the power public opinion information may include: the electricity price will be charged in a tiered manner in the near future, and the charging standard for each tier will increase; a * discount can be enjoyed by pre-depositing *** yuan of electricity fee; the ** area will be out of power on * month * day due to line repair, etc., without limitation here.

[0103] Exemplarily, the electricity preference information of the power user to be predicted is electricity price information. When a call is received from the power user to be predicted, it is possible to find out whether there is electricity public opinion information related to the electricity price information. If there is electricity public opinion information, determine whether the information is positive information (good news) or negative information (bad news) for the power user to be predicted. If the electricity public opinion information is positive information, then the public opinion data corresponding to the electricity public opinion information is a negative probability value. If the electricity public opinion information is negative information, then the public opinion data corresponding to the electricity public opinion information is a positive probability value. After determining the corresponding type of electricity public opinion information, the public opinion data can be determined according to the degree of influence of the electricity public opinion information.

[0104] When determining the first complaint risk data, we can first determine whether the content of the power public opinion information is related to the power handling matters of the power user to be predicted. If not, the initial sentiment data is used as the first complaint risk data.

[0105] If the content of the power public opinion information is related to the power handling matters of the power user to be predicted, and the power public opinion information is positive information, the sum of the public opinion data and the initial sentiment data is taken as the first complaint risk data.

[0106] If the content of the power public opinion information is related to the power handling matters of the power user to be predicted, and the power public opinion information is negative information, the average between the public opinion data and the initial sentiment data will be used as the first complaint risk data.

[0107] In a specific embodiment, the electricity complaint record includes historical electricity complaints of the electricity user to be predicted and the number of complaints corresponding to each historical electricity complaint.

[0108] Determining the second complaint risk data according to the power handling matters and the power user information in step 140 may include:

[0109] Step 140.1: According to the power handling matter, search the power complaint record for the historical power complaint matter and the corresponding complaint times corresponding to the power handling matter.

[0110] Step 140.2: In the power system, query the complaint records of power handling matters within a preset time period in the power supply area; wherein the complaint records are used to indicate the number of times the power handling matters are complained by other users within the preset time period, and whether the complaints are handled.

[0111] Step 140.3: Determine the second complaint risk data based on the complaint records, the historical power complaint items corresponding to the power handling items, and the corresponding number of complaints.

[0112] In this embodiment, the matters to be handled by the power user can be determined during the phone call. After the power handling matters are determined, it can be determined whether the power user to be predicted has made a complaint about the power handling matters according to the power complaint records of the power user to be predicted. If so, the risk of the complaint by the power user to be predicted will increase with the number of historical complaints.

[0113] In addition, if there are a large number of other complaint records for similar power handling matters, it can also provide a basis for predicting the complaint risk of the power users to be predicted. Therefore, the power supply area where the power users to be predicted are located can be queried in the power system to see whether the power handling matter has been complained by other power users in the recent period. If so, it is determined whether the power handling matter has been handled after other power users complained. If it has been handled, the probability of complaints from the power users to be predicted will be reduced.

[0114] Finally, the second complaint risk data is determined based on the complaint records, the historical power complaint items corresponding to the power handling items, and the corresponding number of complaints.

[0115] Step 150: input the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0116] In this embodiment, after the first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model, the probability value corresponding to each predicted complaint risk level of the power user to be predicted can be obtained. For example: low complaint risk 70%, medium complaint risk 20%, high complaint risk 10%. Among them, the maximum probability value is 70%, then the low complaint risk corresponding to 70% is used as the predicted complaint risk level of the power user to be predicted. It should be understood that the above embodiment is only for illustration, and the specific predicted complaint risk level can be set according to actual needs.

[0117] In a specific embodiment, the power complaint prediction model can be constructed in the following way:

[0118] Obtain the electricity usage preference information, electricity user information, telephone communication records and electricity complaint risk level of each electricity user in the historical time period.

[0119] The initial emotional data, power handling matters and conversation matching degree corresponding to each power user are determined based on the telephone communication records of each power user in the historical time period.

[0120] The first complaint risk data corresponding to each electricity user is determined based on the electricity usage preference information, power supply area and initial emotion of each electricity user; the second complaint risk data corresponding to each electricity user is determined based on the electricity handling matters and electricity user information of each electricity user.

[0121] The power complaint prediction model is trained based on the first complaint risk data, second complaint risk data, conversation matching degree and power complaint risk level corresponding to each power user.

[0122] In this embodiment, the method for determining data such as initial emotion data, power processing matters, conversation matching degree, first complaint risk data and second complaint risk data can refer to the other related embodiments mentioned above and will not be repeated here.

[0123] In a specific embodiment, before inputting the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model in step 150 to obtain the predicted power complaint risk level of the power user to be predicted, the following may also be included:

[0124] According to the telephone communication records, the emotional bias, speech speed, emotional bias and response speed of the power customer service during the telephone communication of the power user to be predicted are obtained to obtain the third complaint risk data.

[0125] The first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted, including:

[0126] The first complaint risk data, the second complaint risk data, the third complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0127] In this embodiment, in order to further improve the accuracy of the prediction of the electricity complaint risk level, the emotional bias, speaking speed, question keywords of the power users to be predicted in the telephone communication records, the emotional bias, reply speed, reply keywords, volume and other data of the power customer service can also be analyzed to determine the third complaint risk data.

[0128] Accordingly, the first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted, which may also include:

[0129] The first complaint risk data, the second complaint risk data, the third complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0130] Accordingly, if the third complaint risk data is used when predicting the power complaint risk level of the power user to be predicted, the third complaint risk data is also needed when constructing the power complaint prediction model.

[0131] In summary, compared with traditional methods, the embodiments of the present invention take into account the matching degree between the dialogue between the power user and the power customer service, as well as the electricity preference information of the power user to be predicted when predicting the risk level of electricity complaints. The electricity preference information is used to characterize the type of attention that the power user to be predicted pays to power grid information. Taking these factors into consideration when predicting the risk level of electricity complaints can improve the accuracy of the prediction of the risk level of electricity complaints.

[0132] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0133] The following is an embodiment of the device of the present invention. For details not described in detail, reference may be made to the corresponding method embodiments described above.

[0134] Figure 2 The following is a schematic diagram showing the structure of a device for predicting the risk of power user complaints provided by an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows:

[0135] like Figure 2 As shown, the power user complaint risk prediction device 2 includes:

[0136] The acquisition module 21 is used to acquire the power preference information, power user information and telephone communication records of the power user to be predicted; wherein the power preference information includes power price information, power outage information and circuit maintenance information; the power user information includes power complaint records and power supply area;

[0137] A matching degree calculation module 22 is used to determine the initial emotion data, power processing items, and the matching degree of the conversation between the power user to be predicted and the power customer service according to the telephone communication records;

[0138] The matching degree calculation module 22 is further used to modify the telephone communication record based on the Chinese system to which the power user to be predicted belongs; and determine the initial emotion based on the modified telephone communication record;

[0139] The risk data calculation module 23 is used to determine the first complaint risk data according to the electricity preference information and the initial emotion; and determine the second complaint risk data according to the electricity handling matters and the electricity user information;

[0140] The prediction module 24 is used to input the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0141] In a possible implementation, the Chinese system includes Mandarin, Mandarin dialect, Jin dialect, Wu dialect, Min dialect, Hakka dialect, Yue dialect, Xiang dialect, Gan dialect, Hui dialect and Pinghua dialect;

[0142] The matching degree calculation module 22 is specifically used for:

[0143] Extract the recorded text and voice records of the power users to be predicted from the telephone communication records;

[0144] Determine the Chinese language system to which the power users to be predicted belong based on the recorded text and voice records.

[0145] In a possible implementation, the matching degree calculation module 22 is specifically used for:

[0146] Based on the Chinese language system to which the power user to be predicted belongs, determining the standard volume and standard speaking speed corresponding to the Chinese language system;

[0147] Determine the correction coefficient according to the standard volume and standard speaking speed corresponding to the Chinese system and the standard volume and standard speaking speed corresponding to Mandarin;

[0148] The voice information of the power user to be predicted in the telephone communication record is corrected based on the correction coefficient.

[0149] In a possible implementation, the matching degree calculation module 22 is specifically used for:

[0150] Extract the content of each question asked by the power user to be predicted in the communication record and the power customer service reply content corresponding to each question;

[0151] Extract keywords from each question, and search for standard responses corresponding to each question based on the extracted keywords;

[0152] The conversation matching degree between each standard reply content and the corresponding power customer service answer content is calculated to obtain the conversation matching degree between the power user to be predicted and the power customer service.

[0153] In a possible implementation, the matching degree calculation module 22 is specifically used for:

[0154] Mapping the keywords in each question content to the standard power customer service database to obtain the standard question content corresponding to each question content; wherein the standard power customer service database includes multiple standard question contents and standard reply content corresponding to each standard question content;

[0155] Based on the content of each standard question, determine the standard response content corresponding to each question.

[0156] In a possible implementation, the risk data calculation module 23 is specifically used to:

[0157] Searching for the power public opinion information of the corresponding power supply area according to the power consumption preference information; wherein the power public opinion information includes positive information and negative information;

[0158] Determine the first complaint risk data based on power public opinion information and initial sentiment data.

[0159] In a possible implementation, the risk data calculation module 23 is specifically used to:

[0160] Perform voice recognition on telephone communication records and extract the recorded text in the communication records;

[0161] Detecting a preset sentence of the power user to be predicted in the recording text to obtain first emotion data of the power user to be predicted; wherein the preset sentence is a specified sentence in the recording text;

[0162] Determine the second emotion data according to the speech speed and volume of the preset sentence of the power user to be predicted in the telephone communication record;

[0163] Initial emotion data of the power user to be predicted is obtained according to the first emotion data and the second emotion data.

[0164] In a possible implementation, the power complaint record includes the historical power complaints of the power user to be predicted and the number of complaints corresponding to each historical power complaint;

[0165] The risk data calculation module 23 is specifically used for:

[0166] According to the power handling items, search the power complaint records for the historical power complaint items and the corresponding complaint times corresponding to the power handling items;

[0167] In the power system, query the complaint records of power handling matters in the power supply area within a preset time period; wherein the complaint records are used to indicate the number of times the power handling matters are complained by other users within the preset time period, and whether the complaints are handled;

[0168] The second complaint risk data is determined based on the complaint records, historical power complaint items corresponding to power handling items, and the corresponding number of complaints.

[0169] In a possible implementation, the prediction module 24 is specifically used for:

[0170] The first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the probability value corresponding to each predicted complaint risk level of the power user to be predicted;

[0171] The predicted complaint risk level corresponding to the maximum probability value is taken as the predicted complaint risk level of the power user to be predicted.

[0172] In a possible implementation, the power user complaint risk prediction device 2 further includes a construction module 25; the construction module 25 is specifically used for:

[0173] Obtain the electricity usage preference information, electricity user information, telephone communication records and electricity complaint risk level of each electricity user in the historical time period;

[0174] Determine the initial emotional data, power handling items and conversation matching degree corresponding to each power user based on the telephone communication records of each power user in the historical time period;

[0175] Determine the first complaint risk data corresponding to each power user based on the power preference information, power supply area and initial emotion corresponding to each power user; determine the second complaint risk data corresponding to each power user based on the power handling matters and power user information corresponding to each power user;

[0176] The power complaint prediction model is trained based on the first complaint risk data, second complaint risk data, conversation matching degree and power complaint risk level corresponding to each power user.

[0177] In a possible implementation, the risk data calculation module 23 is further used to:

[0178] According to the telephone communication records, the emotional bias, speech speed, emotional bias and response speed of the power customer service during the telephone communication of the power user to be predicted are obtained to obtain the third-party complaint risk data;

[0179] Accordingly, the prediction module 24 is specifically used for:

[0180] The first complaint risk data, the second complaint risk data, the third complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

[0181] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0182] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0183] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned embodiments of the power user complaint risk prediction method can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium.

[0184] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for predicting the risk of power user complaints, characterized in that: include: Obtaining the power preference information, power user information and telephone communication records of the power user to be predicted; wherein the power preference information includes power price information, power outage information and circuit maintenance information; the power user information includes power complaint records and power supply area; Determine the Chinese language system to which the power user to be predicted belongs, power handling matters, and the degree of conversation matching between the power user to be predicted and the power customer service according to the telephone communication records; Modifying the telephone communication record based on the Chinese language system to which the power user to be predicted belongs; and determining the initial emotion based on the modified telephone communication record; Determine first complaint risk data based on the electricity usage preference information and the initial emotion; determine second complaint risk data based on the electricity handling matters and the electricity user information; The first complaint risk data, the second complaint risk data and the dialogue matching degree are input into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

2. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The Chinese system includes Mandarin, Mandarin dialect, Jin dialect, Wu dialect, Min dialect, Hakka dialect, Cantonese dialect, Xiang dialect, Gan dialect, Hui dialect and Pinghua dialect; The determining the Chinese language system to which the power user to be predicted belongs according to the telephone communication record includes: Extracting the audio recording text and voice recording of the power user to be predicted in the telephone communication record; The Chinese language system to which the power user to be predicted belongs is determined according to the recorded text and the voice record.

3. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The modifying of the telephone communication record based on the Chinese system to which the power user to be predicted belongs includes: Based on the Chinese language system to which the power user to be predicted belongs, determining a standard volume and a standard speaking speed corresponding to the Chinese language system; Determine the correction coefficient according to the standard volume and standard speaking speed corresponding to the Chinese system and the standard volume and standard speaking speed corresponding to Mandarin; The voice information of the power user to be predicted in the telephone communication record is corrected based on the correction coefficient.

4. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The determining, according to the telephone communication record, the conversation matching degree between the power user to be predicted and the power customer service comprises: Extract the content of each question asked by the power user to be predicted in the communication record and the power customer service reply content corresponding to each question content; Extract keywords from each question, and search for standard responses corresponding to each question based on the extracted keywords; The conversation matching degree between each standard reply content and the corresponding power customer service answer content is calculated to obtain the conversation matching degree between the power user to be predicted and the power customer service.

5. The method for predicting the risk of power user complaints according to claim 4 is characterized in that: The step of searching for each standard reply content corresponding to each question content according to the extracted keywords includes: Mapping the keywords in each question content to a standard electric power customer service database to obtain the standard question content corresponding to each question content; wherein the standard electric power customer service database includes multiple standard question contents and standard reply content corresponding to each standard question content; Based on the content of each standard question, determine the standard response content corresponding to each question.

6. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The determining the first complaint risk data according to the electricity usage preference information and the initial emotion includes: Searching for corresponding power public opinion information of the power supply area according to the power consumption preference information; wherein the power public opinion information includes positive information and negative information; The first complaint risk data is determined based on the electric power public opinion information and the initial sentiment data.

7. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The determining of the initial emotion data based on the revised telephone communication record includes: Perform voice recognition on the corrected telephone communication records to extract the recorded text; Detecting a preset sentence of the power user to be predicted in the recording text to obtain first emotion data of the power user to be predicted; wherein the preset sentence is a designated sentence in the recording text; Determine the second emotion data according to the speech speed and volume of the preset sentence of the power user to be predicted in the revised telephone communication record; Initial emotion data of the power user to be predicted is obtained according to the first emotion data and the second emotion data.

8. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The power complaint record includes the historical power complaints of the power user to be predicted and the number of complaints corresponding to each historical power complaint; The determining the second complaint risk data according to the power processing matters and the power user information includes: According to the power handling matter, searching the power complaint record for a historical power complaint matter and a corresponding number of complaints corresponding to the power handling matter; In the power system, querying the complaint record of the power processing matter within a preset time period in the power supply area; wherein the complaint record is used to indicate the number of times the power processing matter is complained by other users within the preset time period, and whether the complaint is handled; The second complaint risk data is determined based on the complaint record, the historical power complaint items corresponding to the power processing items, and the corresponding number of complaints.

9. The method for predicting the risk of power user complaints according to claim 1, characterized in that: The step of inputting the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted includes: Inputting the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain a probability value corresponding to each predicted complaint risk level of the power user to be predicted; The predicted complaint risk level corresponding to the maximum probability value is used as the predicted complaint risk level of the power user to be predicted.

10. A device for predicting the risk of power user complaints, characterized in that: include: An acquisition module is used to acquire the power usage preference information, power user information and telephone communication records of the power user to be predicted; wherein the power usage preference information includes power price information, power outage information and circuit maintenance information; the power user information includes power complaint records and power supply area; A matching degree calculation module, used to determine the Chinese language system to which the power user to be predicted belongs, power handling matters, and the matching degree of the conversation between the power user to be predicted and the power customer service according to the telephone communication records; The matching degree calculation module is further used to modify the telephone communication record based on the Chinese system to which the power user to be predicted belongs; and determine the initial emotion based on the modified telephone communication record; A risk data calculation module, configured to determine first complaint risk data according to the electricity preference information and the initial emotion; and determine second complaint risk data according to the electricity handling matters and the electricity user information; The prediction module is used to input the first complaint risk data, the second complaint risk data and the dialogue matching degree into the power complaint prediction model to obtain the predicted power complaint risk level of the power user to be predicted.

Citation Information

Cited By

  • Power customer appeal upgrading risk assessment method and system based on causal reasoning

    CN122617152A