Association rule mining and traceability analysis-based appeal attribution optimization method

Abstract

The invention discloses an association rule mining and traceability analysis-based appeal attribution optimization method, and relates to the technical field of power marketing, and the method comprises the following steps: inputting the inconvenience of bill query and the unsmooth payment channel into an optimization module, outputting a comprehensive association degree and a comprehensive influence degree, and carrying out the analysis of the comprehensive association degree and the comprehensive influence degree; and an optimization measure is formulated based on the comprehensive correlation degree and the comprehensive influence degree, the measure is optimized in real time in an electric power marketing system, and then an optimization effect index is output. The method has the advantages that relevance among different appeals can be identified, follow-up analysis results are convenient to be comprehensive and accurate, a complete appeal chain can be formed to connect multiple appeal times in series, potential problem traceability analysis on power marketing is facilitated, artificial experience and judgment are reduced on the basis of a data driving mode, and power marketing efficiency is improved. Therefore, the accuracy is improved, the root cause of the customer appeal can be accurately identified, the corresponding optimization strategy can be formulated, the customer problem can be more effectively solved, and the customer satisfaction and loyalty are further improved.

Description

Technical Field

[0001] The present invention relates to the field of power marketing technology, and in particular to a demand attribution optimization method based on association rule mining and traceability analysis. Background Art

[0002] In the field of power marketing, power companies often receive customer complaints. Timely handling and optimization of customer demands are key to improving customer satisfaction and loyalty. With the intensification of market competition, power companies need to identify customer demands more accurately, understand the root causes behind them, and formulate effective optimization strategies accordingly.

[0003] However, existing customer complaint analysis is often based on single-dimensional data, such as complaint type and complaint time, which may ignore the potential connections between different complaints and make it difficult to fully grasp the true nature of customer needs.

[0004] Moreover, in the existing demand analysis and processing, it may be difficult to link various demand events, which makes it difficult to find the root cause of the demand. In addition, the demand processing often relies on manual experience and judgment, which may be highly subjective, resulting in inaccurate attribution results and affecting the effectiveness of the optimization strategy. Summary of the Invention

[0005] The purpose of the present invention is to provide a demand attribution optimization method based on association rule mining and traceability analysis, which solves the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a demand attribution optimization method based on association rule mining and traceability analysis, including the following optimization methods:

[0007] Collect customer demand data from power marketing systems, online forum comments, and customer online surveys through the data collection module;

[0008] The processing module cleans, removes noise, and formats the collected customer complaint data, outputting complaint 1: inconvenient bill query and complaint 2: unsmooth payment channels.

[0009] Input the first complaint (inconvenient bill query) and the second complaint (impeded payment channels) into the optimization module, output the comprehensive correlation and comprehensive impact, and formulate optimization measures based on the comprehensive correlation and comprehensive impact. The measures are optimized in real time in the power marketing system, and then the optimization effect index is output;

[0010] The comprehensive correlation, comprehensive influence and optimization effect index are output to the terminal display device through the output module. The actual effect of the optimization measures is analyzed based on the optimization effect index, and a power customer feedback mechanism is established to collect power customers' opinions and suggestions on the optimization measures.

[0011] Optionally, the optimization module includes: a correlation submodule, an influence submodule, an optimization index submodule and a weight adjustment submodule.

[0012] Optionally, the processing process of the relevance submodule is as follows:

[0013]

[0014] in:

[0015] SKT represents comprehensive correlation;

[0016] A means it is inconvenient to check bills, and B means payment channels are not smooth;

[0017] count(A∩B) represents the number of demands that contain both demands A and B, count(A) represents the number of demands that contain demand A, SKTA represents the total number of demands collected within a certain period of time, and w represents the adjustable factor;

[0018] a1 represents the weight coefficient of one, and a1 adjusts the contribution of count(A∩B) in the comprehensive correlation;

[0019] a2 represents the weight coefficient 2, a2 ​​adjustment The degree of contribution to the overall relevance;

[0020] Indicates the confidence level of the appeal;

[0021] The inconvenience of bill query A and the poor payment channel B are input into the correlation sub-module, and the demands of A and B are reflected based on count(A∩B) and count(A). The weight coefficient a1 and the weight coefficient a2 are used to adjust the weight of the inconvenience of bill query A and the poor payment channel B, and the comprehensive correlation SKT is output.

[0022] Optionally, the influence submodule processing process is as follows:

[0023]

[0024] in:

[0025] EDG(X, A) represents the comprehensive influence;

[0026] X indicates a potential problem, which is the poor stability of the online service platform;

[0027] EDGA represents the degree of influence of adjusting SKT based on X, EDGB(A) represents the frequency of occurrence of appeal A,

[0028] EDGC(A, X) represents the severity adjustment of the comprehensive impact EDG(X, A) based on demand A and potential problem X, and EDGA*EDGB(A) represents the prevalence of demand A.

[0029] Based on the comprehensive correlation SKT and bill query inconvenience A, and combined with the potential problem X and EDGC(A, X), the correlation between bill query inconvenience A and potential problem X is reflected, and the comprehensive impact EDG(X, A) is output.

[0030] Optionally, the optimization index submodule processing process is as follows:

[0031]

[0032] ΔSKT=SKT new -SKT old ;

[0033] ΔEDG(X, A) = EDG(X, A) new -EDG (X, A) old ;

[0034] in:

[0035] WE represents the optimization effect index;

[0036] SKT new and SKT old Represent the comprehensive correlation degree before and after the improvement measures;

[0037] EDG (X, A) new and EDG (X, A) old Respectively represent the comprehensive impact before and after the improvement measures;

[0038] ΔSKT represents the reduction in the comprehensive correlation degree SKT after the improvement measures, and ΔEDG(X, A) represents the reduction in the comprehensive influence degree EDG(X, A) after the improvement measures;

[0039] BSKT represents the benchmark value of the comprehensive correlation BSK before the improvement measures, and BEDG(X, A) represents the benchmark value of the comprehensive influence EDG(X, A) before the improvement measures;

[0040] Based on the comprehensive correlation SKT and the comprehensive influence EDG(X, A), the optimization effect index WE is output by evaluating the reduction of SKT and EDG(X, A) before and after the improvement measures.

[0041] Optionally, the adjustment submodule processing process is as follows:

[0042] S1, w new =w old+β*(WE T -WE);

[0043] S2. Set the iterative convergence termination condition:

[0044] Condition 1: Maximum number of iterations n = 15;

[0045] Condition 2: |WE n -WE n-1 |<0.5%;

[0046] When any one of the conditions 1 and 2 is met, the loop terminates;

[0047] in

[0048] w new Indicates the updated optimization effect index, w old represents the original optimization effect index in the correlation submodule, β represents the learning rate, and β controls w new Adjustment range, WE T Indicates the target value of the optimization effect index;

[0049] Based on the optimization effect index WE and the target value WE of the optimization effect index T The absolute value of the difference is used as an influencing factor, and w is controlled by β new The original optimization effect index in the correlation submodule is input into the adjustment submodule, and the updated optimization effect index w is output. new , the updated optimization effect index w new The original optimization effect index w in the correlation submodule is replaced to adjust the comprehensive correlation SKT, and the convergence is terminated by conditions one and two.

[0050] Optionally, the bill inquiry inconvenience A includes: problems with multiple inquiry channels, problems during the inquiry process, and problems with the accuracy of bill information;

[0051] Problems with various inquiry channels include: inconvenience in checking bills on the power marketing website, inconvenience in checking bills on the power marketing app, and inconvenience in checking bills via the power marketing customer service telephone;

[0052] Problems encountered during the inquiry process include: cumbersome bill inquiry procedures, cumbersome steps and unclear information;

[0053] Issues with the accuracy of billing information include: incorrect bill inquiry amount and inconsistent bill inquiry date.

[0054] Optionally, the poor payment channel B includes: online payment problems, offline payment problems, third-party payment platform problems, and insufficient payment notifications and reminders.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. The present invention calculates a comprehensive correlation through a correlation submodule. The calculation of the comprehensive correlation facilitates power companies to quickly identify which demands often occur simultaneously, and these combinations often reflect the deep-seated needs or problems of customers. Therefore, companies can infer that customers have common problems in electricity bill management and prioritize solving these problems.

[0057] 2. The present invention calculates the comprehensive impact through the impact sub-module, enabling enterprises to more objectively evaluate the priority and urgency of demands, avoiding subjective judgment deviations, and quantifying potential problems and demands together to achieve the purpose of tracing the root causes, fully reflecting the problems existing in the power marketing platform.

[0058] 3. The present invention calculates the optimization effect index through the optimization index sub-module. The optimization effect index can scientifically evaluate the actual effects of these measures on improving service quality and customer satisfaction. The optimization effect index provides enterprises with a quantifiable evaluation standard, which helps enterprises to objectively evaluate the effects of different improvement measures, thereby optimizing the decision-making process and improving resource utilization efficiency. The optimization effect index helps enterprises to objectively evaluate the effects of different improvement measures, thereby optimizing the decision-making process and improving resource utilization efficiency.

[0059] 4. The present invention iteratively influences the adjustable factors in the correlation submodule based on the optimization effect index, thereby continuously adjusting the comprehensive correlation, which helps power companies to more accurately identify key demands and potential problems, and thus take more targeted optimization measures. By adjusting the adjustable factors, it is convenient to try different parameter combinations, so as to find the best parameter settings that can improve the optimization effect index. By optimizing the parameters in the correlation submodule, customer demands can be more accurately identified and processed, thereby improving customer satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flowchart of the steps of the demand attribution optimization method based on association rule mining and traceability analysis;

[0061] Figure 2 This is a schematic diagram of the overall structure of the demand attribution optimization method based on association rule mining and traceability analysis;

[0062] Figure 3 Schematic diagram of the optimization module. DETAILED DESCRIPTION

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0064] The demand attribution optimization method based on association rule mining and traceability analysis is different from the existing demand organization optimization method. The existing demand organization optimization method mainly relies on single-dimensional data of customer feedback, such as complaint type and complaint time, and ignores the potential connections between different demands. It is difficult to fully grasp the true face of customer needs. In the demand attribution process, it may rely on manual experience and judgment, which is highly subjective and may lead to inaccurate attribution results, affecting the effectiveness of the optimization strategy. The algorithm unit of this demand attribution optimization system method based on association rule mining and traceability analysis processes multi-dimensional data through data mining technology, identifies complex association patterns between demands, makes the analysis results more comprehensive and accurate, and connects isolated demands with potential problems to form a complete demand network, so as to trace the root cause of the demands and provide a scientific basis for formulating targeted optimization strategies. In addition, this method is based on a data-driven attribution method, which reduces the interference of manual experience and judgment in the attribution process and improves the objectivity and accuracy of the attribution results.

[0065] Example 1: Please refer to Figures 1 to 3 This implementation provides the following demand attribution optimization method based on association rule mining and traceability analysis:

[0066] Collect customer demand data from power marketing systems, online forum comments, and customer online surveys through the data collection module;

[0067] The processing module cleans, removes noise, and formats the collected customer complaint data, outputting complaint 1: inconvenient bill query and complaint 2: unsmooth payment channels.

[0068] Input the first complaint (inconvenient bill query) and the second complaint (impeded payment channels) into the optimization module, output the comprehensive correlation and comprehensive impact, and formulate optimization measures based on the comprehensive correlation and comprehensive impact. The measures are optimized in real time in the power marketing system, and then the optimization effect index is output;

[0069] The output module outputs the comprehensive correlation, comprehensive impact and optimization effect index to the terminal display device. Based on the optimization effect index, the actual effect of the optimization measures is analyzed. A power customer feedback mechanism is established to collect power customers' opinions and suggestions on the optimization measures.

[0070] The optimization module includes a relevance submodule, an influence submodule, an optimization index submodule and a weight adjustment submodule.

[0071] In this embodiment, the method uses data mining technology to identify the correlation patterns between demands and traces the root causes of the demands through traceability analysis. The CRM system and data analysis software can be used in the data collection module to more efficiently collect and organize customer demand data. These tools can automatically record customer consultation, complaint and other information, and perform preliminary classification and organization. The collected data can be carefully reviewed in the form of screening and labeling to ensure the accuracy and reliability of the data, and the data can be classified and labeled for subsequent analysis. A database server can be used to store the collected raw data to ensure data security and accessibility. ETL tools are used in the processing module for data cleaning, and the processed data is stored on SSD solid-state drives to provide fast data reading and writing speeds, thereby improving data processing efficiency.

[0072] After the comprehensive correlation and comprehensive impact are calculated, a traceability analysis can be conducted to trace the root cause of the demand. Causal reasoning analysis tools can be used to mine and infer causal relationships, and knowledge graph construction tools can be used to build a knowledge graph for the business domain to assist in traceability analysis. Optimization measures are then formulated based on the traceability analysis and implemented in the power marketing system. Project management software such as Jira and Trello can be used to manage the implementation progress of optimization measures, and service optimization software CRM systems can be used to support the optimization and automation of service processes. Finally, the optimization effect index is calculated after the measures are implemented.

[0073] Compared with the existing demand optimization processing methods that may rely on manual experience and subjective judgment to identify the correlation between demands and evaluate the degree of impact, this method has the disadvantages of strong subjectivity, low efficiency, poor accuracy, and difficulty in comprehensively evaluating the effectiveness of improvement measures. This method combines the correlation submodule, the impact submodule, the optimization index submodule and the weight adjustment submodule to not only achieve accurate identification and quantitative evaluation of demand correlation and impact, but also provide a means to scientifically evaluate the optimization effect. This systematic approach can significantly improve the service quality and customer satisfaction of enterprises in the field of power marketing.

[0074] See also Figures 1 to 3 , the processing process of the correlation submodule is as follows:

[0075]

[0076] in:

[0077] SKT represents comprehensive correlation;

[0078] A means it is inconvenient to check bills, and B means payment channels are not smooth;

[0079] count(A∩B) represents the number of demands that contain both demands A and B, reflecting the number of times these two demands appear together. count(A) represents the number of demands that contain demand A, that is, the importance of demand A. SKTA represents the total number of demands collected within a certain period of time. w represents the adjustable factor.

[0080] a1 represents the weight coefficient of one, and a1 adjusts the contribution of count(A∩B) in the comprehensive correlation;

[0081] a2 represents the weight coefficient 2, a2 ​​adjustment The degree of contribution to the overall relevance;

[0082] Indicates the confidence level of the claim, Used to measure the degree of confidence that item set B will occur after item set A occurs;

[0083] The inconvenience of bill query A and the poor payment channel B are input into the correlation sub-module, and the demands of A and B are reflected based on count(A∩B) and count(A). The weight coefficient a1 and the weight coefficient a2 are used to adjust the weight of the inconvenience of bill query A and the poor payment channel B, and the comprehensive correlation SKT is output.

[0084] In this embodiment: the comprehensive correlation SKT calculates the co-occurrence frequency and confidence between different demands A and B, and can accurately quantify the correlation strength between these demands, which helps enterprises quickly identify which demands often appear at the same time, so as to understand the potential connection and mutual influence between them. The comprehensive correlation SKT provides a more objective and accurate correlation measurement, which helps enterprises to formulate more targeted marketing strategies and service improvement plans. The comprehensive correlation SKT reveals the correlation strength between demands A and B, which helps enterprises identify key demand combinations, understand customer needs and behavior patterns, and provide a basis for formulating marketing strategies and service plans. By analyzing the comprehensive correlation SKT, the industry can quickly identify which demands are These combinations often appear at the same time and often reflect the deep-seated needs or problems of customers. For example, if the SKT values ​​of the two demands, inconvenient bill inquiry and poor payment channels, are relatively high, the company can infer that customers have general problems in electricity bill management and give priority to solving these problems. The SKT value can help companies deeply understand customer behavior patterns and demand changes. By monitoring the changing trends of the SKT value, companies can discover the dynamic changes in customer needs and adjust marketing strategies and service models in time to meet customer expectations. Based on the calculation results of the SKT value, companies can formulate more accurate marketing strategies and service plans. For combinations of demands with high correlation, companies can launch combined service packages to improve customer satisfaction and loyalty.

[0085] See also Figures 1 to 3 ,The processing process of the influence submodule is as follows:

[0086]

[0087] in:

[0088] EDG(X, A) represents the comprehensive impact, that is, the degree of influence of different demands under potential problems;

[0089] X indicates a potential problem, which is the poor stability of the online service platform;

[0090] EDGA represents the impact of SKT adjusted based on X, EDGB(A) represents the frequency of occurrence of demand A, EDGC(A, X) represents the severity adjustment of the combined impact EDG(X, A) based on demand A and potential problem X, and EDGA*EDGB(A) represents the prevalence of demand A.

[0091] Based on the comprehensive correlation SKT and bill query inconvenience A, and combined with the potential problem X and EDGC(A, X), the correlation between bill query inconvenience A and potential problem X is reflected, and the comprehensive impact EDG(X, A) is output.

[0092] In this embodiment, the formula for the comprehensive impact EDG(X, A) considers the prevalence and importance of demand A to more carefully assess the impact of demand A under specific factor X. This detailed assessment enables enterprises to more comprehensively understand the actual impact of different demands in different scenarios, providing strong support for optimizing resource allocation and formulating emergency response plans. The comprehensive impact EDG(X, A) quantifies the impact of demand A under potential problem X, helping enterprises assess the priority and urgency of different demands and providing guidance for resource allocation and emergency response.

[0093] The calculation of the comprehensive impact index EDG(X, A) provides enterprises with a quantitative perspective to assess the impact of different demands under a specific potential problem X. This allows enterprises to more objectively assess the priority and urgency of demands and avoid subjective judgment bias. By calculating and analyzing the impact of different demands, enterprises can optimize resource allocation and invest limited resources in demands with higher impact. This helps improve service efficiency, reduce resource waste, and quickly resolve customers' most concerned issues. The calculation of the comprehensive impact index EDG(X, A) can provide guidance for enterprises to formulate emergency response plans. Enterprises can prioritize demands based on their impact to ensure that negative impacts are minimized within a limited time. When A refers to the inconvenience of bill inquiries, which is a specific customer demand, by calculating it together with X, which refers to the stability of the online service platform, we can assess the impact of these root causes on the specific demand of inconvenience of bill inquiries.

[0094] Potential problem X may represent poor stability of the online service platform. X is usually not a claim. The purpose is to link potential problem X with specific claims and to quantify the impact of each root cause on the customer's specific claim. In other embodiments, potential problem X may include power supply interruptions, equipment failures, errors in the bill generation process, cumbersome payment procedures, slow customer service response, poor service attitude, untimely notification of electricity price adjustments, and unclear policy explanations.

[0095] See also Figures 1 to 3 , the processing of the optimization index submodule is as follows:

[0096]

[0097] ΔSKT=SKT new -SKT old ;

[0098] ΔEDG(X, A) = EDG(X, A) new -EDG (X, A) old ;

[0099] in:

[0100] WE represents the optimization effect index;

[0101] SKT new and SKT old Represent the comprehensive correlation degree before and after the improvement measures;

[0102] EDG (X, A) new and EDG (X, A) old Respectively represent the comprehensive impact before and after the improvement measures;

[0103] ΔSKT represents the reduction in the comprehensive correlation degree SKT after the improvement measures, and ΔEDG(X, A) represents the reduction in the comprehensive influence degree EDG(X, A) after the improvement measures;

[0104] BSKT represents the benchmark value of the comprehensive correlation BSK before the improvement measures, and BEDG(X, A) represents the benchmark value of the comprehensive influence EDG(X, A) before the improvement measures;

[0105] Based on the comprehensive correlation SKT and the comprehensive influence EDG(X, A), the optimization effect index WE is output by evaluating the reduction of SKT and EDG(X, A) before and after the improvement measures.

[0106] In this embodiment: first, the optimization measures are described. Although the optimization measures are constantly adjusted according to actual conditions, some corresponding optimization measures will be described below.

[0107] In response to the frequent complaints about cumbersome service processes, companies can optimize service processes, simplify operational steps, and improve service efficiency. To enhance customer experience, companies can upgrade their technology platforms, such as updating mobile applications and optimizing website interfaces. By analyzing customer data, companies can develop personalized service strategies and provide products and services that better meet customer needs.

[0108] By comparing the changes in the comprehensive correlation and impact before and after improvement measures, this submodule can scientifically evaluate the actual effects of these measures on improving service quality and customer satisfaction. The optimization effect index (WE) provides enterprises with a quantifiable evaluation standard, helping them objectively evaluate the effects of different improvement measures, thereby optimizing the decision-making process and improving resource utilization efficiency. The optimization effect index (WE) reflects the impact of improvement measures on improving service quality, helping enterprises evaluate the effectiveness and cost-effectiveness of different improvement measures, and providing a basis for continuous optimization and improvement. By comparing the changes in the comprehensive correlation and impact before and after improvement measures, the optimization effect index (WE) scientifically evaluates the actual effects of these measures on improving service quality and customer satisfaction. This enables power companies to objectively evaluate the effectiveness of improvement measures and avoid blind investment and ineffective efforts. Based on the calculation results of the optimization effect index (WE), power companies can optimize the decision-making process to select more effective improvement measures. Through continuous iteration and optimization, companies can continuously improve service quality and customer satisfaction. The calculation results of the optimization effect index (WE) also provide power companies with the motivation for continuous improvement and innovation. By monitoring the changing trends of the optimization effect index (WE), companies can identify deficiencies and potential problems in their services and take timely measures to improve and innovate to meet customers' ever-changing needs and expectations.

[0109] See also Figures 1 to 3 , the processing of the adjustment submodule is as follows:

[0110] S1, w new =w old +β*(WE T -WE);

[0111] S2. Set the iterative convergence termination condition:

[0112] Condition 1: Maximum number of iterations n = 15;

[0113] Condition 2: |WE n -WE n-1 |<0.5%;

[0114] When any one of the conditions 1 and 2 is met, the loop terminates;

[0115] in

[0116] w new Indicates the updated optimization effect index, w old represents the original optimization effect index in the correlation submodule, β represents the learning rate, and β controls w new Adjustment range, WE T Indicates the target value of the optimization effect index;

[0117] Based on the optimization effect index WE and the target value WE of the optimization effect indexT The absolute value of the difference is used as an influencing factor, and w is controlled by β new The original optimization effect index in the correlation submodule is input into the adjustment submodule, and the updated optimization effect index w is output. new , the updated optimization effect index w new The original optimization effect index w in the correlation submodule is replaced to adjust the comprehensive correlation SKT, and the convergence is terminated by conditions one and two.

[0118] In this embodiment: through the feedback of WE, the optimization effect index w of the correlation degree can be accurately adjusted, so that the overall method system can more accurately reflect the demand correlation situation in the actual business, which helps power companies to more accurately identify key demands and potential problems, so as to take more targeted optimization measures, and through iteration, it can be dynamically adjusted based on the latest data and business changes, thereby maintaining the effectiveness and accuracy of the optimization method system. Compared with static systems, this method is more adaptable to complex and changing marketing environments. This iterative form combines data mining technology and traceability analysis to deeply explore the root causes of the demands, rather than just superficial phenomena. This helps companies to solve problems fundamentally and avoid similar problems from happening again. The process can be automated, reducing the possibility of manual intervention and errors. This method optimizes the effect index WE and the target value WE of the optimization effect index. T The absolute value of the difference is used to analyze and judge the optimization effect index WE and the target value WE of the normal level optimization effect index. T The difference between the two is calculated, and based on this, the adjustable factor w in the comprehensive correlation SKT is iterated to adjust the comprehensive correlation SKT. The optimization effect index WE is used as an evaluation indicator to quantify the correlation submodule and the influence submodule. Therefore, when the optimization effect index WE value is not in the normal range, that is, it is different from WE TAfter comparison, it shows that there is room for improvement in the system, and it is necessary to adjust parameters to improve the optimization effect. By iterating the adjustable factor w, the influence of different factors in the calculation of the comprehensive correlation SKT can be changed. This refined control helps us better adapt to different business scenarios and customer needs, thereby improving the overall optimization effect. When the WE value is not good, by adjusting the adjustable factor w, it is easy to try different parameter combinations to find the best parameter setting that can improve the WE value. This iterative process helps us discover and solve bottleneck problems in the system, thereby improving the overall performance of the system. In the field of power marketing, customer satisfaction is one of the important indicators for measuring service quality. By optimizing the parameters in the correlation sub-module, we can more accurately identify and handle customer demands, thereby improving customer satisfaction and loyalty. The iterative process not only helps to find the optimal parameter settings, but also helps us test the stability and robustness of the system under different parameter combinations. Through multiple iterations and tests, potential problems that may exist in the system can be discovered and fixed, thereby improving the overall performance and reliability of the system.

[0119] In the specific implementation process, multiple submodules in this method are used to form a system for the demand attribution optimization method. By substituting the two demands of inconvenient bill query A and poor payment channels B into the correlation submodule, the comprehensive correlation SKT is calculated. The calculation of the comprehensive correlation SKT can help power companies quickly identify which demands often occur simultaneously. These combinations often reflect the deep-seated needs or problems of customers. Then, companies can infer that customers have common problems in electricity bill management and give priority to solving these problems.

[0120] By substituting the inconvenience of bill querying (A) and the comprehensive correlation (SKT) into the impact submodule, the comprehensive impact (EDG(X, A)) is calculated. This calculation provides a quantitative perspective for enterprises to assess the impact of different demands under potential problem X. This allows enterprises to more objectively assess the priority and urgency of demands, avoiding subjective judgment bias. This quantitatively links potential problems and demands, tracing their root causes and fully reflecting the problems existing in the power marketing platform.

[0121] By substituting the comprehensive correlation SKT and the comprehensive influence EDG(X, A) into the optimization index submodule, the optimization effect index WE is calculated. The optimization effect index WE can scientifically evaluate the actual effect of these measures on improving service quality and customer satisfaction by comparing the changes in comprehensive correlation and influence before and after the improvement measures. The optimization effect index WE provides a quantifiable evaluation standard for enterprises, which helps them objectively evaluate the effects of different improvement measures, thereby optimizing the decision-making process and improving resource utilization efficiency. The optimization effect index WE helps enterprises objectively evaluate the effects of different improvement measures, thereby optimizing the decision-making process and improving resource utilization efficiency.

[0122] Based on w and WE, the weight adjustment submodule is substituted and w in the correlation submodule is iterated, thereby automatically adjusting the value of the comprehensive correlation SKT. This allows the overall method system to more accurately reflect the demand correlation in actual business, helping power companies to more accurately identify key demands and potential problems, thereby taking more targeted optimization measures. It can change the influence of different factors in the comprehensive correlation SKT calculation. When the WE value is not good, the adjustable factor w is adjusted to try different parameter combinations to find the optimal parameter setting that can improve the WE value. By optimizing the parameters in the correlation submodule, we can more accurately identify and handle customer demands, thereby improving customer satisfaction and loyalty.

[0123] Furthermore, the method coordinates calculations between multiple submodules in pairs, and can also perform overall cycles and iterations, so that the overall system has the effect of automatic optimization and updating, and thus has better adaptability.

[0124] Example 2: Please refer to Figure 1 、 Figure 2 and Figure 3 ,In this embodiment, the bill inquiry inconvenience A includes: ,problems with multiple inquiry channels, problems during the inquiry process, and ,problems with the accuracy of bill information;

[0125] Problems with various inquiry channels include: inconvenience in checking bills on the power marketing website, inconvenience in checking bills on the power marketing app, and inconvenience in checking bills via the power marketing customer service telephone;

[0126] Problems encountered during the inquiry process include: cumbersome bill inquiry procedures, cumbersome steps and unclear information;

[0127] Issues with the accuracy of billing information include: incorrect bill inquiry amount and inconsistent bill inquiry date.

[0128] In this embodiment, these components constitute the specific problems and needs of customers regarding the inconvenience of bill inquiry A. By analyzing these components, we can gain a deeper understanding of the nature and root causes of customer demands, thereby formulating more effective optimization strategies.

[0129] In this embodiment, the poor payment channels B include: online payment problems, offline payment problems, third-party payment platform problems, and insufficient payment notifications and reminders.

[0130] In this embodiment, online payment issues include: malfunctions of the power marketing website or power marketing app, that is, when customers pay through the power company's official website or mobile app, they encounter problems such as website crashes, the app cannot be opened, or it runs slowly; payment interface issues, that is, during the online payment process, the payment interface is unstable, the payment failure rate is high, or the supported payment methods are limited, resulting in customers being unable to complete the payment smoothly; information synchronization delays, that is, after a successful payment, there is a delay in the system updating the customer's account balance or payment status, causing inconvenience to the customer;

[0131] Offline payment issues include: long queues at business halls, meaning customers have to wait in line for a long time when paying their bills at electricity business halls, which affects payment efficiency; limited business hours, meaning some business halls have short business hours or are closed on weekends, making it impossible for customers to pay their bills during off-hours; and malfunctions of self-service payment machines, meaning that malfunctions or complex operations of the self-service payment machines in business halls prevent customers from completing their payments themselves.

[0132] Issues with third-party payment platforms include: limited platform coverage, meaning some third-party payment platforms do not support electricity bill payments or only support limited areas, limiting customers' payment options; handling fees or additional fees, meaning some third-party platforms may charge handling fees or additional fees during the payment process, increasing customers' payment costs; and information synchronization issues, meaning after payments are made through third-party platforms, there are delays or errors in the synchronization of information to the power company's system, resulting in inaccurate customer account status.

[0133] Insufficient payment notifications and reminders include: untimely payment notifications, that is, the power company failed to notify customers of payment in a timely manner through text messages and emails, causing customers to miss the payment deadline; single reminder method, that is, the power company only reminds customers to pay through one or a few methods, which cannot meet the personalized needs of different customers.

[0134] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A demand attribution optimization method based on association rule mining and traceability analysis, characterized by: The following steps are involved: (1) Collect customer demand data from the power marketing system, online forum comments, and customer online surveys through the data collection module; (2) The collected customer complaint data is cleaned, denoised, and formatted through the processing module, and complaint 1: inconvenient bill query and complaint 2: unsmooth payment channels are output; (3) Input the first claim, inconvenient bill query, and the second claim, unsmooth payment channels, into the optimization module, output the comprehensive correlation and comprehensive impact, formulate optimization measures based on the comprehensive correlation and comprehensive impact, optimize the measures in real time in the power marketing system, and then output the optimization effect index; (4) The comprehensive correlation, comprehensive influence and optimization effect index are output to the terminal display device through the output module, the actual effect of the optimization measures is analyzed based on the optimization effect index, and a power customer feedback mechanism is established to collect the opinions and suggestions of power customers on the optimization measures.

2. The demand attribution optimization method based on association rule mining and traceability analysis according to claim 1 is characterized by: The optimization module includes: Relevance submodule, influence submodule, optimization index submodule and weight adjustment submodule.

3. The demand attribution optimization method based on association rule mining and traceability analysis according to claim 2 is characterized by: The processing process of the relevance submodule is as follows: in: SKT represents comprehensive relevance; A means that it is inconvenient to check bills, and B means that the payment channels are not smooth; countA∩B represents the number of demands that include both demands A and B, countA represents the number of demands that include demands A, SKTA represents the total number of demands collected within a certain period of time, and w represents an adjustable factor; a1 represents the weight coefficient 1, and a1 adjusts the contribution of countA∩B in the comprehensive correlation; a2 represents the weight coefficient 2, a2 ​​adjustment The degree of contribution to the overall relevance; Indicates the confidence level of the appeal; The inconvenience of bill query A and the poor payment channel B are input into the correlation sub-module, and the demands of A and B are reflected based on countA∩B and countA. The weights of the inconvenience of bill query A and the poor payment channel B are adjusted in combination with the weight coefficient a1 and the weight coefficient a2, and the comprehensive correlation SKT is output.

4. The demand attribution optimization method based on association rule mining and traceability analysis according to claim 3 is characterized by: The influence submodule processing process is as follows: in: EDGX, A represents the comprehensive influence; X indicates a potential problem, which is the poor stability of the online service platform; EDGA indicates the impact of adjusting SKT based on X, EDBA indicates the frequency of occurrence of demand A, EDGCA,X indicates the severity adjustment of the combined impact EDGX,A based on demand A and potential problem X, and EDGA*EDGBA indicates the prevalence of demand A. Based on the comprehensive correlation SKT and bill query inconvenience A, and combined with the potential problem X and EDGCA,X, it reflects the correlation between bill query inconvenience A and potential problem X, and outputs the comprehensive impact EDGX,A.

5. The demand attribution optimization method based on association rule mining and traceability analysis according to claim 4 is characterized by: The optimization index submodule processing process is as follows: ΔSKI=SKT new -SKT old ; ΔEDG(X,A)=EDG(X,A) new -EDG(X,A) old ; in: WE represents the optimization effect index; SKT new and SKT old Respectively represent the comprehensive correlation after and before the improvement measures; EDG(X,A) new and EDG (X, A) old Respectively represent the comprehensive impact after and before the improvement measures; ΔSKT represents the reduction of the comprehensive correlation degree SKT after the improvement measures, and ΔEDG(X, A) represents the reduction of the comprehensive influence degree EDG(X, A) after the improvement measures; BSKT represents the benchmark value of the comprehensive correlation BSK before the improvement measures, and BEDG(X, A) represents the benchmark value of the comprehensive influence EDG(X, A) before the improvement measures; Based on the comprehensive correlation SKT and the comprehensive influence EDG(X, A), the optimization effect index WE is output by evaluating the reduction of SKT and EDG(X, A) before and after the improvement measures.

6. The method for optimizing demand attribution based on association rule mining and source tracing analysis according to claim 5 is characterized by: The processing process of the adjustment submodule is as follows: S1、w new =w old +β*(WE T -WE); S2. Set the iterative convergence termination condition: Condition 1: Maximum number of iterations n = 15; Condition 2: |WE n -WE n-1 |<0.5%; When any one of condition 1 and condition 2 is met, the loop terminates; in w new Indicates the updated optimization effect index, w old represents the original optimization effect index in the association submodule, β represents the learning rate, and β controls w new Adjustment range, WE T Indicates the target value of the optimization effect index; Based on the optimization effect index WE and the target value WE of the optimization effect index T The absolute value of the difference is used as an influencing factor, and w is controlled by β new The original optimization effect index in the correlation submodule is input into the adjustment submodule, and the updated optimization effect index w is output. new , the updated optimization effect index w new The original optimization effect index w in the correlation submodule is replaced to adjust the comprehensive correlation SKT, and the convergence is terminated by conditions one and two.

7. The demand attribution optimization method based on association rule mining and traceability analysis according to claim 3 is characterized by: The bill inquiry inconvenience A includes: problems with multiple inquiry channels, problems in the inquiry process and problems with the accuracy of bill information; Problems with multiple inquiry channels include: inconvenience in checking bills on the power marketing website, inconvenience in checking bills on the power marketing app, and inconvenience in checking bills via the power marketing customer service telephone; Problems during the inquiry process include: cumbersome bill inquiry procedures, cumbersome procedures and unclear information; Issues with the accuracy of bill information include: incorrect bill inquiry amount and inconsistent bill inquiry date.

8. The method for optimizing demand attribution based on association rule mining and source tracing analysis according to claim 3 is characterized by: The unsmooth payment channels B include: Problems with online payment, offline payment, third-party payment platforms, and insufficient payment notifications and reminders.

Images