Customer Complaint Form Processing Method Based on Big Data

Abstract

The present invention relates to the technical field of data processing, and particularly to a method for processing customer complaint forms based on big data. The method includes: receiving a number of customer text complaint forms within a preset period, dividing and sorting the number of customer text complaint forms based on a preset time period; analyzing a key complaint problem set corresponding to the number of customer text complaint forms in any initial classification result, so as to perform clustering processing on the number of customer text complaint forms based on the key problem analysis result; analyzing the processing complexity corresponding to a number of initial clusters to determine the initial processing priority; judging the change trend of the same initial cluster based on the sorting sequence to adjust the initial processing priority; determining the initial allocation of transmission resources based on the adjusted processing priority to transmit to corresponding respective processing terminals, obtaining feedback results of any processing terminal, and adjusting the adjusted processing priority based on the feedback results. The present invention improves the processing efficiency of customer complaint forms.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a method for processing customer complaint forms based on big data. Background Art

[0002] With the rapid development of big data technology, some enterprises have begun to try to apply big data technology to the field of customer complaint handling, and initially classify and extract keywords from customer text complaint forms through natural language processing (NLP) technology.

[0003] The patent document with the Chinese patent publication number CN111274380A discloses a method and related device for processing consultation and complaint information based on big data. The method includes: obtaining the work order information of the target customer, where the work order information includes the target customer identifier, the target work order content, and the target service classification; determining the first customer classification label and the target user portrait corresponding to the target customer according to the target customer identifier; determining the second customer classification label corresponding to the target customer according to the target work order content; determining the target consultation and complaint handling suggestion corresponding to the target customer according to the first customer classification label, the second customer classification label, and the target service classification, and displaying the target user portrait and the target consultation and complaint handling suggestion.

[0004] In the prior art, it still stays at the basic classification and statistics level, lacking in-depth analysis of the complexity of complaint form processing and a priority adjustment mechanism based on time dynamic changes, resulting in inaccurate allocation of processing resources and room for improvement in processing efficiency. Summary of the Invention

[0005] Therefore, the present invention provides a method for processing customer complaint forms based on big data, which can solve the problem of low processing efficiency of customer complaint forms by performing real-time analysis on a number of customer complaint forms, continuously adjusting the processing priority, and adjusting the transmission resources according to the priority.

[0006] To achieve the above object, the present invention provides a method for processing customer complaint forms based on big data, and the method includes:

[0007] The service center platform receives a number of customer text complaint forms within a preset period, divides the number of customer text complaint forms based on a preset time period, obtains a number of initial classification results, and sorts the number of initial classification results in chronological order to obtain a sorted sequence;

[0008] Analyze the key complaint problem sets corresponding to the number of customer text complaint forms in any of the initial classification results to obtain a key problem analysis result, and perform clustering processing on the number of customer text complaint forms based on the key problem analysis result to obtain a number of initial clusters;

[0009] Analyze the processing complexity corresponding to several of the initial clusters, and determine the initial processing priority based on the processing complexity;

[0010] Judge the change trend of the same initial cluster based on the sorting sequence, so as to adjust the initial processing priority based on the change trend to obtain the adjusted processing priority;

[0011] Determine the initial allocated transmission resources based on the adjusted processing priority, and transmit several of the initial clusters to the corresponding processing terminals based on the initial allocated transmission resources to obtain the feedback results of any processing terminal, so as to adjust the adjusted processing priority based on the feedback results to obtain the target processing priority.

[0012] Further, the step of analyzing the key complaint problem sets corresponding to several of the customer text complaint forms in any of the initial classification results includes:

[0013] Perform word segmentation processing on the customer text complaint form to obtain several initial word segments;

[0014] Process several of the initial word segments to obtain an effective vocabulary set;

[0015] Perform semantic similarity matching between the effective vocabulary set and a preset complaint text database to obtain a similarity matching result, and determine the key complaint problem set based on the similarity matching result.

[0016] Further, the step of clustering several of the customer text complaint forms based on the key problem analysis result includes:

[0017] Calculate the similarity between any two of the customer text complaint forms based on the key complaint problem set to obtain the calculated similarity;

[0018] Construct a similarity matrix based on several of the calculated similarities;

[0019] Cluster several of the customer text complaint forms based on the similarity matrix and a preset clustering algorithm to obtain several initial clusters.

[0020] Further, the step of analyzing the processing complexity corresponding to several of the initial clusters includes:

[0021] Count the number of the customer text complaint forms corresponding to any of the initial clusters as the actual quantity value, and calculate the quantity complexity based on the actual quantity value;

[0022] Identify the cluster center problem set corresponding to any of the initial clusters, analyze the cluster center problem set, and judge the problem complexity based on the problem set analysis result;

[0023] Calculate the processing complexity based on the quantity complexity and the problem complexity.

[0024] Further, the step of determining the problem complexity based on the analysis result of the problem set includes:

[0025] Determine the actual problem type based on the cluster center problem set;

[0026] Count the number of customer text complaint forms corresponding to any of the actual problem types as the statistical quantity;

[0027] Determine the calculation weights corresponding to several actual problem types based on several of the statistical quantities;

[0028] Calculate the problem complexity based on the calculation weights and the statistical quantities.

[0029] Further, the step of determining the initial processing priority based on the processing complexity includes:

[0030] Sort several of the processing complexities from largest to smallest to obtain the first sorting;

[0031] Determine the initial processing priority based on the first sorting.

[0032] Further, the step of determining the change trend of the same initial cluster based on the sorting sequence includes:

[0033] Determine the number of customer text complaint forms corresponding to the same initial cluster within different time periods based on the sorting sequence as several first quantities;

[0034] Determine the actual proportion of the initial cluster in several of the initial clusters based on any of the first quantities;

[0035] Determine the change trend of the actual proportion based on the sorting sequence.

[0036] Further, the step of adjusting the initial processing priority based on the change trend includes:

[0037] Analyze the change trend corresponding to any of the initial clusters to determine whether the change trend is an upward trend or a downward trend according to the analysis result;

[0038] Determine the adjustment strategy according to the trend judgment result, and adjust the initial processing priority according to the adjustment strategy to obtain the adjusted processing priority.

[0039] Further, the step of determining the initial allocation of transmission resources based on the adjusted processing priority includes:

[0040] Determine the resource allocation ratio based on the adjusted processing priority;

[0041] Determine the initial allocated transmission resources based on the resource allocation ratio and the preset total transmission resources.

[0042] Further, the step of adjusting the processing complexity corresponding to the initial clustering based on the feedback result includes:

[0043] Receive and parse the processing result of the customer complaint form corresponding to any of the processing terminals to obtain feedback data;

[0044] Analyze the feedback data to identify the number of successfully processed customer text complaint forms and the number of failed customer text complaint forms, as the successful number and the failed number;

[0045] Based on the successful number and the failed number, calculate the processing success rate corresponding to any of the initial clusterings;

[0046] Adjust the processing complexity of the corresponding initial clustering according to the processing success rate.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows. By receiving complaint forms at preset intervals and dividing them based on preset time periods, it is ensured that complaint handling can closely follow the time context and respond to customer needs in a timely manner. By sorting the initial classification results in chronological order, a clear timeline is provided for subsequent analysis, which helps to identify the evolution trend of complaint problems. By analyzing the key complaint problem set, the core demands and main problem points of customers can be accurately identified, providing a clear direction for subsequent handling. Based on the analysis results of key problems, clustering processing is performed to classify similar complaint forms, reducing repetitive labor in the processing process and improving processing efficiency. By analyzing the processing complexity to determine the initial processing priority, it is ensured that problems with high priority and high complexity are processed first, thereby optimizing resource allocation. Setting a clear processing priority for complaint forms helps each processing terminal to quickly respond and give priority to handling urgent or important problems. By analyzing the change trend of the initial clustering and dynamically adjusting the processing priority, the suddenness and evolution of complaint problems can be flexibly responded to, ensuring that the processing strategy always fits the actual needs. Based on the adjusted processing priority, the initial clustering is allocated to each processing terminal to ensure that each terminal can receive the task most suitable for its processing ability and resource status. By collecting the feedback results of the processing terminals and adjusting the processing priority again, a closed-loop optimization mechanism is formed, which can continuously correct and optimize the processing strategy and improve the overall processing effect.

[0048] In particular, by performing word segmentation on the customer text complaint form and splitting the long text into several initial word segments, it helps to analyze the text content more meticulously during subsequent processing, improve the accuracy of information processing. By processing the initial word segments to obtain an effective vocabulary set, the accuracy and efficiency of subsequent analysis are improved, providing a more precise and targeted vocabulary set for subsequent matching with the preset complaint text database. By performing semantic similarity matching between the effective vocabulary set and the preset complaint text database, the key complaint issues in the customer text complaint form can be accurately identified, providing a clear classification basis for subsequent processing and improving the accuracy and efficiency of complaint handling.

[0049] In particular, by counting the number of customer text complaint forms in each initial cluster, that is, the actual quantity value, the processing burden represented by each cluster can be directly quantified, providing an intuitive numerical basis for subsequent resource allocation and priority setting. By identifying and analyzing the cluster center problem set, the problem types and severity levels represented by each cluster can be deeply understood, helping to formulate targeted processing strategies, improve processing efficiency and quality. Based on the comprehensive calculation of quantity complexity and problem complexity, the calculation result of processing complexity is more comprehensive and accurate, ensuring the high efficiency of the entire processing process. Brief Description of the Drawings

[0050] Figure 1 It is a schematic flowchart of the method for processing customer complaint forms based on big data provided by an embodiment of the present invention;

[0051] Figure 2 It is a schematic flowchart of analyzing the key complaint problem set in the method for processing customer complaint forms based on big data provided by an embodiment of the present invention;

[0052] Figure 3 It is a schematic flowchart of calculating the processing complexity in the method for processing customer complaint forms based on big data provided by an embodiment of the present invention;

[0053] Figure 4 It is a schematic flowchart of determining the change trend in the method for processing customer complaint forms based on big data provided by an embodiment of the present invention. Detailed Embodiments

[0054] In order to make the objectives and advantages of the present invention clearer and more understandable, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0056] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0057] Please refer to Figure 1 As shown, an embodiment of the present invention provides a method for processing customer complaint forms based on big data. The method includes:

[0058] Step S100, the service center platform receives a number of customer text complaint forms within a preset period, divides the number of customer text complaint forms based on a preset time period, obtains a number of initial classification results, and sorts the number of initial classification results in chronological order to obtain a sorting sequence;

[0059] Step S200, analyze a key complaint problem set corresponding to a number of the customer text complaint forms in any of the initial classification results to obtain a key problem analysis result, and perform clustering processing on the number of customer text complaint forms based on the key problem analysis result to obtain a number of initial clusters;

[0060] Step S300, analyze the processing complexity corresponding to a number of the initial clusters, and determine an initial processing priority based on the processing complexity;

[0061] Step S400, judge the change trend of the same initial cluster based on the sorting sequence, and adjust the initial processing priority based on the change trend to obtain an adjusted processing priority;

[0062] Step S500, determine an initial allocated transmission resource based on the adjusted processing priority, and transmit a number of the initial clusters to corresponding respective processing terminals based on the initial allocated transmission resource to obtain a feedback result of any processing terminal, and adjust the adjusted processing priority based on the feedback result to obtain a target processing priority.

[0063] Specifically, the preset period in the embodiment of the present invention is one month;

[0064] The preset time period is one day.

[0065] Specifically, the embodiment of the present invention divides a number of customer text complaint forms based on a preset time period, including:

[0066] Determine the receiving moments corresponding to a number of the customer text complaint forms;

[0067] Divide several customer text complaint forms corresponding to the receiving moments within the same time period based on the preset time period to obtain several initial classification results.

[0068] Specifically, the embodiments of the present invention sort several initial classification results based on the time sequence, including:

[0069] Sort the time periods corresponding to several initial classification results from the earliest to the latest. The initial classification result corresponding to the earliest time within the preset period is sorted as 1, and so on, to obtain the sorting sequence.

[0070] Specifically, the embodiments of the present invention receive complaint forms through a preset period and divide them based on a preset time period to ensure that complaint handling can closely follow the time context and respond to customer needs in a timely manner. By sorting the initial classification results in chronological order, a clear timeline is provided for subsequent analysis, which helps to identify the evolution trend of complaint problems. By analyzing the key complaint problem set, the core demands and main problem points of customers can be accurately identified, providing a clear direction for subsequent processing. Based on the analysis results of key problems, clustering processing is performed to classify similar complaint forms, reducing repetitive labor in the processing process and improving processing efficiency. By analyzing the processing complexity to determine the initial processing priority, it can ensure that problems with high priority and high complexity are processed first, thereby optimizing resource allocation. Setting a clear processing priority for complaint forms helps each processing terminal to quickly respond and prioritize the handling of urgent or important problems. By analyzing the change trend of the initial clustering and dynamically adjusting the processing priority, it can flexibly respond to the suddenness and evolution of complaint problems, ensuring that the processing strategy always fits the actual needs. Based on the adjusted processing priority, the initial clustering is assigned to each processing terminal to ensure that each terminal can receive tasks that are most suitable for its processing ability and resource status. By collecting the feedback results of the processing terminal and adjusting the processing priority again, a closed-loop optimization mechanism is formed, which can continuously correct and optimize the processing strategy and improve the overall processing effect.

[0071] See Figure 2 As shown, the steps of analyzing the key complaint problem set corresponding to several customer text complaint forms in any of the initial classification results include:

[0072] Step S210, perform word segmentation processing on the customer text complaint form to obtain several initial word segments;

[0073] Step S220, process several initial word segments to obtain an effective vocabulary set;

[0074] Step S230, perform semantic similarity matching between the effective vocabulary set and a preset complaint text database to obtain a similarity matching result, and determine the key complaint problem set based on the similarity matching result.

[0075] Specifically, the preset complaint text library described in the embodiment of the present invention is constructed based on historical customer complaint data, covering the vocabulary and expressions of common complaint issues. The preset complaint text library includes word vectors of complaint issues and corresponding problem labels.

[0076] Specifically, the embodiment of the present invention processes several of the initial word segmentations, including filtering stop words and irrelevant words in the initial word segmentations, wherein the stop words may be preset commonly used words without practical meaning, such as "的", "了", etc., and may also include some common phrases, such as "在哪里", "实际", etc. The stop words can be determined through a predefined stop word list, and the irrelevant words may be commercial brand names, names, place names, etc. that are irrelevant to the complaint issue. By identifying several historical customer complaint texts, words that appear frequently but are irrelevant to the complaint issue are identified and used as an irrelevant word library, and irrelevant words are determined through the irrelevant word library.

[0077] Specifically, the step of matching the effective vocabulary set with the preset complaint text database for semantic similarity in the embodiment of the present invention includes:

[0078] Convert each valid word in the valid vocabulary set into a plurality of valid word vectors;

[0079] Similarity calculation is performed between the valid word vectors and the corresponding preset word vectors in the preset complaint text database to obtain a plurality of similarities.

[0080] Specifically, the embodiment of the present invention calculates the similarity between the several valid word vectors and the corresponding preset word vectors in the preset complaint text database, and can measure the similarity between the words by calculating the distance or angle between the vectors.

[0081] Specifically, embodiments of the present invention may include:

[0082] Original complaint text: "The product I ordered last week has not arrived yet. The logistics information shows that it is still waiting for delivery. I have contacted customer service many times but to no avail. I am very dissatisfied with this.";

[0083] After word segmentation: "The product I ordered last week has not arrived yet. The logistics information shows that it is still waiting for delivery. I have contacted customer service many times but to no avail. I am very dissatisfied with this.";

[0084] Filter out stop words and irrelevant words, such as "我", "的", "了", "," etc.

[0085] Remaining valid words: "last week", "scheduled", "goods", "not arrived", "logistics information", "display", "always", "wait", "delivery", "multiple times", "contact", "customer service", "fruitless", "very", "dissatisfied";

[0086] Convert the valid words into word vectors through a pre-trained word embedding model (the pre-trained word embedding model can include, for example, Word2Vec, GloVe, etc.);

[0087] Search for similar word vectors or phrases in the preset complaint text database. The preset database may contain word vectors and corresponding labels for complaint problems such as "delivery delay", "no response from customer service", "order not arrived", etc.;

[0088] Calculate the similarity (such as cosine similarity) between the valid word vectors and the word vectors in the preset complaint text database;

[0089] Suppose the word vector similarity between "not arrived" and "delivery delay" is very high, and the word vector combination of "contact customer service fruitlessly" and "no response from customer service" also has a relatively high similarity;

[0090] Based on the similarity matching results, it can be judged that this complaint form mainly involves two problems: one is "delivery delay" (corresponding to words such as "not arrived", "wait", "delivery", etc.), and the other is "no response from customer service" (corresponding to words such as "contact multiple times", "customer service", "fruitless", etc.);

[0091] Search for the corresponding labels of these two problems in the preset complaint text database, which are "logistics problem" and "service response problem" respectively;

[0092] Associate the two labels of "logistics problem" and "service response problem" with this complaint form as the key complaint problem set of this complaint form.

[0093] Specifically, in the embodiment of the present invention, by performing word segmentation processing on the customer text complaint form, splitting the long text into several initial word segments, it helps to analyze the text content more carefully during subsequent processing, improves the accuracy of information processing. By processing the initial word segments, an effective vocabulary set is obtained, which improves the accuracy and efficiency of subsequent analysis, provides a more accurate and targeted vocabulary set for subsequent matching with the preset complaint text database. By performing semantic similarity matching between the effective vocabulary set and the preset complaint text database, the key complaint problems in the customer text complaint form can be accurately identified, providing a clear classification basis for subsequent processing, and improving the accuracy and efficiency of complaint handling.

[0094] Specifically, the step of clustering several of the customer text complaint forms based on the key problem analysis results includes:

[0095] Calculate the similarity between any two of the customer text complaints based on the set of key complaint problems, and obtain the calculated similarity;

[0096] Construct a similarity matrix based on a number of the calculated similarities;

[0097] Perform clustering on a number of the customer text complaints based on the similarity matrix and a preset clustering algorithm to obtain a number of initial clusters.

[0098] Specifically, the preset clustering algorithm in the embodiments of the present invention may be K-means clustering or hierarchical clustering. When it is K-means clustering, the number K of clustering centers needs to be specified in advance. The value of K can be determined according to the actual situation, such as the diversity of complaint problems, the limitations of processing resources, etc. When performing hierarchical clustering, an agglomerative method from bottom to top (such as AGNES) or a divisive method from top to bottom (such as DIANA) can be selected.

[0099] Specifically, in the embodiments of the present invention, a similarity matrix is constructed based on a number of the calculated similarities. Among them, each element in the matrix represents the similarity value between two customer text complaints.

[0100] Specifically, the clustering process of a number of the customer text complaints based on the similarity matrix and a preset clustering algorithm in the embodiments of the present invention includes:

[0101] Preprocess the similarity matrix to ensure that the data in the matrix can be applied to the selected clustering algorithm. For example, in K-means clustering, it may be necessary to normalize the similarity values to better calculate the distance during the clustering process.

[0102] Taking K-means clustering as an example, first randomly select K customer text complaints as the initial clustering centers. Then, for each customer text complaint, calculate its similarity with each initial clustering center and assign it to the cluster where the clustering center with the highest similarity is located. After completing one round of assignment, recalculate the center of each cluster, that is, the point corresponding to the average similarity of all complaints in the cluster. This process is continuously iterated until a certain termination condition is met, such as the change in the clustering center is less than a preset threshold, or the preset number of iterations is reached. Among them, the preset threshold can be 0.1, and the preset number of iterations can be 100 times.

[0103] In hierarchical clustering, if the bottom-up aggregation method is adopted, initially each customer text complaint form is regarded as a separate cluster. Then, the similarity between all pairs of clusters is calculated, and the two clusters with the highest similarity are merged. This process is repeated until the preset number of clusters is reached or other termination conditions are met. Among them, the preset number of clusters can be adjusted according to the actual situation. If it is necessary to divide the customer text complaint forms into five main complaint categories, then the preset number of clusters is set to five. If the top-down splitting method is adopted, initially all complaint forms are regarded as a single cluster, and then it is gradually split into smaller clusters until the conditions are met.

[0104] See Figure 3 As shown, the steps of analyzing the processing complexity corresponding to several of the initial clusters include:

[0105] Step S310, count the number of customer text complaint forms corresponding to any of the initial clusters as the actual quantity value, and calculate the quantity complexity based on the actual quantity value;

[0106] Step S320, identify the cluster center problem set corresponding to any of the initial clusters, analyze the cluster center problem set, and judge the problem complexity based on the analysis result of the problem set;

[0107] Step S330, calculate the processing complexity based on the quantity complexity and the problem complexity.

[0108] Specifically, the steps of calculating the quantity complexity based on the actual quantity value in the embodiments of the present invention include:

[0109] Set a quantity threshold;

[0110] Take the ratio of the actual quantity value to the quantity threshold as the quantity complexity.

[0111] Specifically, the quantity threshold in the embodiments of the present invention is the total number of several customer text complaint forms collected within the same preset time period divided by the total number of several initial clusters.

[0112] Specifically, in the embodiments of the present invention, the cluster center problem set corresponding to any of the initial clusters can be identified by vectorizing each customer text complaint form. For example, a bag-of-words model or a word vector model can be used. Then, calculate the centroid of all vectors in each cluster, that is, the average value on each dimension, as the vector representation of the cluster center, and select the key complaint problem set corresponding to the customer text complaint form that is closest to the vector of the cluster center among each customer text complaint form as the cluster center problem set.

[0113] Specifically, in the embodiments of the present invention, by counting the number of customer text complaint forms in each initial cluster, that is, the actual quantity value, the processing burden represented by each cluster can be directly quantified, providing an intuitive numerical basis for subsequent resource allocation and priority setting. By identifying the cluster center problem set and analyzing it, the problem types and severity levels represented by each cluster can be deeply understood, which helps to formulate targeted processing strategies and improve the processing efficiency and quality. Based on the comprehensive calculation of the quantity complexity and the problem complexity, the calculation result of the processing complexity is more comprehensive and accurate, ensuring the high efficiency of the entire processing process.

[0114] Specifically, the step of judging the problem complexity based on the analysis result of the problem set includes:

[0115] Determine the actual problem type based on the cluster center problem set;

[0116] Count the number of customer text complaint forms corresponding to any of the actual problem types as the statistical quantity;

[0117] Determine the calculation weights corresponding to several actual problem types based on several of the statistical quantities;

[0118] Calculate the problem complexity based on the calculation weights and the statistical quantities.

[0119] Specifically, a possible example of the embodiments of the present invention is:

[0120] Determine the actual problem types as "logistics problems" and "service response problems";

[0121] Count the number of customer text complaint forms. Among them, the total number of customer complaint text forms is 100, there are 60 complaints about "logistics problems", and 80 complaints about "service response problems". It can be seen that the number of customer complaint text forms that repeatedly contain "logistics problems" and "service response problems" is 40;

[0122] The weight corresponding to "logistics problems" is (60 - 40 / 2) / 100 = 0.4;

[0123] The weight corresponding to "service response problems" is (80 - 40 / 2) / 100 = 0.6;

[0124] Problem complexity = 0.4×60 / (60 + 80) + 0.6×80 / (60 + 80) = 0.17 + 0.34 = 0.51.

[0125] Specifically, in the embodiments of the present invention, the processing complexity = α×quantity complexity + β×problem complexity, where α is the weight corresponding to the quantity complexity, β is the weight corresponding to the problem complexity, and α + β = 1.

[0126] Specifically, the steps of determining the initial processing priority based on the processing complexity include:

[0127] Sort several of the processing complexities from largest to smallest to obtain a first sorting;

[0128] Determine the initial processing priority based on the first sorting.

[0129] Specifically, in the embodiments of the present invention, by sorting the processing complexities, it is possible to clearly identify which clusters need to be processed first. The clear priority setting helps to reasonably allocate limited resources, ensuring that problems with high complexity and high urgency can be solved first, and avoiding wasting too many resources on low-complexity problems, thereby achieving the optimal allocation and efficient utilization of resources.

[0130] See Figure 4 As shown, the steps of judging the change trend of the same initial cluster based on the sorting sequence include:

[0131] Step S410, determine the number of customer text complaint forms corresponding to the same initial cluster in different time periods based on the sorting sequence, as several first quantities;

[0132] Step S420, determine the actual proportion of the initial cluster in several of the initial clusters based on any of the first quantities;

[0133] Step S430, determine the change trend of the actual proportion based on the sorting sequence.

[0134] Specifically, in the embodiments of the present invention, by counting the number of customer text complaint forms corresponding to the same initial cluster in different time periods, a solid data foundation is provided for subsequent analysis. By calculating the actual proportion of the initial cluster in the overall situation, the relative importance of this cluster problem in the overall complaints can be intuitively understood, eliminating the comparison obstacles caused by the absolute quantity differences between different clusters, making the comparison between different clusters fairer and more accurate, and the change trend based on the actual proportion provides strong decision-making support for adjusting resource allocation.

[0135] Specifically, the steps of adjusting the initial processing priority based on the change trend include:

[0136] Analyze the change trend corresponding to any of the initial clusters to judge whether the change trend is an upward trend or a downward trend according to the analysis result;

[0137] Determine an adjustment strategy according to the trend judgment result, and adjust the initial processing priority according to the adjustment strategy to obtain an adjusted processing priority.

[0138] Specifically, the adjustment strategy described in the embodiments of the present invention includes: increasing the processing priority of the corresponding initial clustering when the change trend is an upward trend;

[0139] When the change trend is a downward trend, the processing priority of the corresponding initial clustering is reduced.

[0140] Specifically, if the actual proportion of a certain initial clustering shows an upward trend in two consecutive time periods, and the upward amplitude exceeds the preset upward amplitude threshold, the initial processing priority of this initial clustering is raised by one level;

[0141] If the actual proportion of a certain initial clustering shows a downward trend in two consecutive time periods, and the downward amplitude exceeds the preset downward amplitude threshold, the initial processing priority of this initial clustering is lowered by one level.

[0142] Specifically, the preset upward amplitude threshold described in the embodiments of the present invention is 5%, and the preset downward amplitude threshold is 3%.

[0143] Specifically, the embodiments of the present invention can determine whether the change trend is an upward trend or a downward trend by calculating the difference in the actual proportion within consecutive time periods. If the difference is greater than 0, it is determined that the adjacent time period is an upward trend; if the difference is less than 0, it is determined that the adjacent time period is a downward trend. The proportion of the area corresponding to the upward trend or the downward trend within the preset period is comprehensively judged. When the upward proportion or the downward proportion is greater than the preset proportion, it is determined that the trend is an upward trend or a downward trend.

[0144] Specifically, the steps for determining the initial allocated transmission resources based on the adjusted processing priority include:

[0145] Determining the resource allocation ratio based on the adjusted processing priority;

[0146] Determining the initial allocated transmission resources based on the resource allocation ratio and the preset total transmission resources.

[0147] Specifically, the embodiments of the present invention determining the resource allocation ratio based on the adjusted processing priority include:

[0148] Normalizing the adjusted processing priorities of each initial clustering to obtain the normalized priority values of each initial clustering;

[0149] Allocating the corresponding resource ratios according to the normalized priority values of each initial clustering.

[0150] Specifically, the higher the normalized priority value of an initial clustering, the higher the allocated resource ratio, so as to ensure the efficient processing of high-priority complaint sheets.

[0151] Specifically, the transmission resources in the embodiments of the present invention can be CPU and network bandwidth resources.

[0152] Specifically, the steps of adjusting the processing complexity corresponding to the initial clustering based on the feedback results include:

[0153] Receiving and parsing the processing results of the customer complaint form corresponding to any of the processing terminals to obtain feedback data;

[0154] Analyzing the feedback data to identify the number of successfully processed customer text complaint forms and the number of unsuccessfully processed customer text complaint forms as the successful number and the failed number;

[0155] Calculating the processing success rate corresponding to any of the initial clusterings based on the successful number and the failed number;

[0156] Adjusting the processing complexity of the corresponding initial clustering according to the processing success rate.

[0157] Specifically, when adjusting the processing complexity in the embodiments of the present invention, if the processing success rate of a certain initial clustering is higher than the preset success rate threshold, the adjustment processing priority of this initial clustering is reduced by one level to release some resources to other clusters that may be more in need of processing; conversely, if the processing success rate is lower than the preset success rate threshold, the adjustment processing priority of this initial clustering is increased by one level to improve the processing priority and the allocated resources, thereby improving the processing effect.

[0158] Specifically, the preset success rate threshold in the embodiments of the present invention is 80%.

[0159] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0160] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for processing customer complaints based on big data, characterized in that: include: The service center platform receives a number of customer text complaint forms within a preset period, divides the number of customer text complaint forms based on a preset time period, obtains a number of initial classification results, and sorts the number of initial classification results based on a time sequence to obtain a sorting sequence; Analyze the key complaint problem sets corresponding to the plurality of customer text complaint forms in any of the initial classification results to obtain key problem analysis results, and perform clustering processing on the plurality of customer text complaint forms based on the key problem analysis results to obtain a plurality of initial clusters; Analyzing the processing complexity corresponding to the initial clusters, and determining the initial processing priority based on the processing complexity; Determining a change trend of the same initial cluster based on the sorting sequence, so as to adjust the initial processing priority based on the change trend to obtain an adjusted processing priority; Determine the initial allocation of transmission resources based on the adjustment processing priority, transmit the initial clusters to the corresponding processing terminals based on the initial allocation of transmission resources, obtain feedback results from any processing terminal, adjust the adjustment processing priority based on the feedback results, and obtain the target processing priority; The step of determining the initial processing priority includes: sorting the plurality of processing complexities from large to small to obtain a first sorting; determining the initial processing priority based on the first sorting; Wherein, the processing complexity = α×quantity complexity + β×problem complexity, α is the weight corresponding to the quantity complexity, β is the weight corresponding to the problem complexity, α+β=1; The process of calculating the quantity complexity is as follows: counting the quantity of the customer text complaint sheets corresponding to any of the initial clusters as the actual quantity value, and calculating the quantity complexity based on the actual quantity value; The process of calculating the complexity of the problem is: identifying the cluster center problem set corresponding to any of the initial clusters, and determining the actual problem type based on the cluster center problem set; counting the number of customer text complaint sheets corresponding to any of the actual problem types as a statistical quantity; determining the calculation weights corresponding to several actual problem types based on several of the statistical quantities; and calculating the problem complexity based on the calculation weights and the statistical quantities.

2. The method for processing customer complaints based on big data according to claim 1 is characterized in that: The step of analyzing the key complaint problem set corresponding to the plurality of customer text complaint sheets in any of the initial classification results comprises: Perform word segmentation on the customer text complaint form to obtain a number of initial word segments; Processing a number of the initial word segmentations to obtain a valid vocabulary set; The effective vocabulary set is matched with a preset complaint text database for semantic similarity to obtain a similarity matching result, and the key complaint problem set is determined based on the similarity matching result.

3. The method for processing customer complaints based on big data according to claim 2 is characterized in that: The step of clustering the plurality of customer text complaint sheets based on the key problem analysis results comprises: Calculate the similarity between any two customer text complaint sheets based on the key complaint question set to obtain the calculated similarity; Constructing a similarity matrix based on the several calculated similarities; Based on the similarity matrix and a preset clustering algorithm, a plurality of the customer text complaint sheets are clustered to obtain a plurality of initial clusters.

4. The method for processing customer complaints based on big data according to claim 1 is characterized in that: The step of determining the change trend of the same initial cluster based on the sorting sequence comprises: Determine, based on the sorting sequence, the number of a plurality of customer text complaint sheets corresponding to the same initial cluster in different time periods as a plurality of first numbers; Determining an actual proportion of the initial cluster in the plurality of initial clusters based on any of the first numbers; A change trend of the actual proportion is determined based on the sorting sequence.

5. The method for processing customer complaints based on big data according to claim 4 is characterized in that: The step of adjusting the initial processing priority based on the change trend includes: Analyze the change trend corresponding to any of the initial clusters to determine whether the change trend is an upward trend or a downward trend according to the analysis result; An adjustment strategy is determined according to the trend determination result, and the initial processing priority is adjusted according to the adjustment strategy to obtain an adjusted processing priority.

6. The method for processing customer complaints based on big data according to claim 5 is characterized in that: The step of determining the initial allocation of transmission resources based on the adjustment processing priority includes: Determining a resource allocation ratio based on the adjustment processing priority; The initial allocated transmission resources are determined based on the resource allocation ratio and the preset total transmission resources.

7. The method for processing customer complaints based on big data according to claim 6 is characterized in that: The step of adjusting the processing complexity corresponding to the initial clustering based on the feedback result includes: Receive and analyze the processing results of the customer complaint form corresponding to any of the processing terminals to obtain feedback data; Analyze the feedback data to identify the number of customer text complaint tickets that are successfully processed and the number of customer text complaint tickets that are unsuccessfully processed as the number of successes and the number of failures; Based on the number of successes and the number of failures, calculating a processing success rate corresponding to any of the initial clusters; According to the processing success rate, the processing complexity of the corresponding initial clustering is adjusted.

Images