A system for correlation analysis of feedback data

Abstract

The application discloses a kind of associated analysis systems of feedback data, to improve feedback data processing efficiency and quality;System covers multiple functional modules: subject classification module uses bag-of-words model and clustering algorithm to classify feedback theme;Data acquisition and processing module collects and processes the feedback data in target area within a year, generates standard data set;Association rule module analyzes data by mining algorithm, mines association relationship, and filters out value association rules;Department analysis module evaluates department service, and carries out training for low-performance department;Seasonal analysis module identifies season high incidence feedback and early warning, and forms feedback team for region;Age salary analysis module analyzes the relationship between age, income and feedback category, and forms a study group for high-frequency feedback category, thereby providing data support and decision basis for feedback work in all directions and depth.

Description

Technical Field

[0001] The present invention relates to the field of feedback data analysis, and in particular to a feedback data correlation analysis system. Background Art

[0002] In the field of feedback work, with the development of society and the transformation of information dissemination methods, the amount of feedback data has shown explosive growth. Traditional feedback processing models face many difficulties:

[0003] On the one hand, the feedback data structure is complex, covering multiple aspects such as the feedback provider, feedback topic, time, channel, and processing results. The different data formats and sources make data integration and unified processing extremely difficult, making it difficult to quickly extract key information from massive amounts of data, which hinders subsequent analysis.

[0004] On the other hand, feedback work lacks effective data correlation analysis methods, making it difficult for staff to understand the potential connections between feedback data and develop targeted processing strategies based on data correlation. For example, it is impossible to accurately determine the high-incidence types and changing patterns of feedback issues in different regions and seasons, and it is also difficult to analyze the impact of factors such as age and income on feedback categories. As a result, resource allocation lacks scientific basis, departmental work efficiency is low, and it is difficult to meet the growing feedback demands of the public, making it difficult to achieve efficient and accurate management of feedback work.

[0005] Therefore, a correlation analysis system for feedback data is introduced to improve the quality and efficiency of feedback work. Summary of the Invention

[0006] In view of this, the present invention provides a feedback data correlation analysis system to solve the problems raised by the above background technology.

[0007] The object of the present invention can be achieved by the following technical solution: A feedback data association analysis system, comprising:

[0008] Data acquisition module: used to collect various feedback data of the target area within one year;

[0009] Topic classification module: Cluster analysis is performed on the feedback topics in the target area within one year, thereby classifying each feedback topic and obtaining k cluster centers and corresponding feedback categories;

[0010] Data processing module: processes all types of feedback data collected in the target area within five years, converts them into the same format, and forms a feedback data set;

[0011] Association rule module: Use association rule mining algorithm to analyze feedback data sets and mine the association relationships between feedback data;

[0012] Department analysis module: Extracts association rules from all departments → processing result scores from value association rules, which are recorded as department association rules. A scoring threshold is preset, and the departments corresponding to department association rules whose processing result scores are lower than the scoring threshold are extracted to form an analysis department set. Each department in the analysis department set is then analyzed.

[0013] Regional Season Analysis Module: Extracts association rules of all regions and seasons from the value association rules, records them as regional association rules, and analyzes the data in the regional association rules;

[0014] Age-salary analysis module: extracts association rules of all ages, monthly income → feedback categories from value association rules, and analyzes the feedback categories of each age range and monthly income range.

[0015] Furthermore, the various feedback data of the data acquisition module are:

[0016] Feedback data includes: basic information of the feedback person, feedback topic, feedback time, feedback channel, handling department, handling personnel, processing plan and processing result score. Basic information includes name, occupation, age, contact information, region and monthly income. Feedback channels include letters, telephone, online platforms and visits.

[0017] Furthermore, the specific steps for the topic classification module to obtain k cluster centers and corresponding feedback categories are:

[0018] S1: Use the bag-of-words model to convert the feedback topic into a vector as the topic vector, calculate the clustering error under different cluster numbers, draw a curve of clustering error changing with the number of clusters, and select the cluster number at the inflection point of the curve as the cluster k value;

[0019] S2-1: Randomly select k data points as initial cluster centers, calculate the Euclidean distance between the topic vector and each initial cluster center, and select the value with the smallest Euclidean distance as the cluster corresponding to the topic vector;

[0020] S2-2: Use the mean formula to calculate the topic vectors in each cluster to form a new cluster center. Calculate the Euclidean distance between the topic vector and the new cluster center again, and select the minimum value of the Euclidean distance between the topic vector and the new cluster center as the new cluster;

[0021] S2-3: Repeat S2-1 and S2-2 continuously. During this process, new cluster centers are continuously generated, and the Euclidean distance between the cluster centers and the corresponding new cluster centers is calculated. The iteration threshold is set. When it is detected that the Euclidean distance between all cluster centers and the corresponding new cluster centers in a certain iteration is less than the iteration threshold, the iteration is stopped, thereby obtaining k cluster centers;

[0022] S3: According to the defined classification standards, arrange professionals to analyze the various cluster centers formed and determine the specific feedback category to which each cluster center belongs.

[0023] Furthermore, the specific steps of the data processing module forming the feedback data set are:

[0024] Extract the region where the feedback provider is located, and extract the district or county where the feedback provider is located; use the bag-of-words model to convert the feedback topic into a vector, recorded as the topic vector, calculate the Euclidean distance between the topic vector and each cluster center, and select the feedback category corresponding to the cluster center with the smallest Euclidean distance as the feedback category of the feedback topic; convert the feedback time by season, converting it to the four seasons of spring, summer, autumn and winter; discretize the age by constructing age ranges; and construct salary ranges to discretize the monthly income;

[0025] The basic information of the feedback recipient, feedback topic, feedback time, feedback channel, processing department, and satisfaction with the processing results after processing are integrated to form a standard feedback data set.

[0026] Furthermore, the specific steps of the association rule module for mining the association relationship between the feedback data are as follows:

[0027] Calculate the support of the item set. An item set refers to a single feedback data or a combination of multiple feedback data. Support refers to the frequency of an item set appearing in the total number of feedbacks.

[0028] A support threshold is preset, and the support of individual feedback data in the item set is calculated. The individual feedback data whose support is greater than the support threshold is selected as the candidate 1-item set. Based on the data in the candidate 1-item set, a merging operation is performed to generate a 2-item set. The 2-item set corresponds to the combination of two feedback data in the item set and belongs to the item set. The support of each 2-item set is calculated, and the 2-item set whose support is greater than the support threshold is selected as the candidate 2-item set. Based on the candidate 2-item set, a 3-item set is generated, and so on. When the item x+1 set is generated based on the candidate x-item set, the support of each item x+1 set is less than the preset support threshold, and the merging is stopped.

[0029] Based on the generated candidate n sets, n=2,3,...,x, each candidate n set is divided into two non-empty subsets, U and V, where the intersection of U and V is an empty set and the union of U and V is the corresponding candidate n set. Based on the results of each division, the confidence of the candidate association rules U→V and V→U is calculated. The confidence calculation formula is ,in represents the confidence of the candidate association rule U→V, represents the support of the corresponding candidate n set, It represents the support of the item set formed by the corresponding feedback data combination in the set U, sets the confidence threshold, and selects the candidate association rules whose confidence is greater than the confidence threshold as the value association rules.

[0030] Furthermore, the specific steps of the department analysis module are:

[0031] Obtain all types of feedback data from the department within a year, extract feedback themes from the department's feedback data, vectorize them, calculate the Euclidean distance from each cluster center, select the feedback category corresponding to the cluster center with the minimum Euclidean distance as the feedback category of the feedback theme, classify the feedback by category, and calculate the scores of each feedback category and the total number of acceptances (ZCS);

[0032] Take the average of the scores of each feedback category as the average score JFZ of the corresponding feedback category, preset the weighted impact factors corresponding to the average score JFZ and the total number of acceptances ZCS, multiply the average score JFZ and the total number of acceptances ZCS by the corresponding weighted impact factors, and then add them together to obtain the result as the comprehensive evaluation index of the corresponding feedback category;

[0033] A comprehensive evaluation index threshold is preset, and feedback categories with a comprehensive evaluation index lower than the comprehensive evaluation index threshold are extracted from each feedback category as promotion categories. Professional personnel are arranged to provide promotion category training for the department’s staff.

[0034] Furthermore, the specific steps of the regional seasonal analysis module are:

[0035] Extract feedback categories from regional association rules and classify them by season to form a four-season feedback set. De-duplicate the four-season feedback set and calculate the number of feedback types in the four-season feedback set. Divide the number of feedback types in the four-season feedback set by the total number of counties and districts in the target region to obtain the frequency ratios corresponding to the four seasons.

[0036] A frequency ratio threshold is preset. If the frequency ratio of a season is less than the threshold, the feedback category in the current seasonal feedback set is determined to be seasonal high-incidence feedback. For seasonal high-incidence feedback, each department develops a treatment plan for each region. A warning time is preset, and the system issues a feedback warning x days before the corresponding season arrives. Each department then prepares in advance according to the formulated treatment plan.

[0037] The feedback categories corresponding to the four seasons in each region are classified by region, and duplicates are removed. The number of times each feedback category appears in the four seasons is counted. The ratio between the number of times each feedback category appears in the four seasons in the deduplicated set and 4 is calculated. The result is used as the corresponding regional proportion set. A regional proportion threshold is preset. If a data point in a regional proportion set for a region is greater than the regional proportion threshold, the feedback category corresponding to that data point is used as the main feedback category for that region.

[0038] Extract each staff member who has handled the main feedback category from the preset database, obtain the number of times each staff member handled the main feedback category CS, the duration of each processing, and the score of each processing result; perform mean and normalization calculation on each processing score, and obtain the results as the sub-mean value F1 and sub-wave value F2. After the sub-mean value F1 and sub-wave value F2 are normalized, they are inserted into the formula: Thus, the comprehensive score PF of the staff is obtained, where q1 and q2 are the weighted influencing factors corresponding to the mean value F1 and the wave value F2 respectively; the processing time of each staff member is plotted on a line graph according to the time series, and the ployfit function of MATLAB is called to perform straight line fitting, and the absolute value of the slope of the fitting line is extracted as the duration trend value F3;

[0039] Preset the weight influence factors corresponding to the number of times CS, comprehensive score PF and duration trend value F3 that the staff handles the main feedback category, normalize the number of times CS, comprehensive score PF and duration trend value F3 that the staff handles the main feedback category, multiply them with the corresponding weight influence factors and sum them up. The result obtained is the comprehensive processing value of the staff for the main feedback category; select the top three staff members with the top three comprehensive processing values ​​to form a feedback team and send them to work in the area.

[0040] Furthermore, the specific steps of the age-salary analysis module are:

[0041] First, classify each age range and perform frequency statistics on the feedback categories corresponding to each monthly income in each age range. Take the feedback category with the highest frequency among the feedback categories corresponding to each monthly income in each age range as the high-frequency feedback category for the corresponding age range. If there are multiple feedback categories with the same frequency in a certain age range, take the feedback category corresponding to the interval with the higher salary as the high-frequency feedback category for the corresponding age range.

[0042] Perform frequency statistics on the high-frequency feedback categories in each age range, divide the frequency of each high-frequency feedback category by the number of age ranges to obtain the age coverage ratio of each high-frequency feedback category, preset an age coverage ratio threshold, compare the age coverage ratio of each high-frequency feedback category with the age coverage ratio threshold, and extract high-frequency feedback categories with age coverage ratios greater than the age coverage ratio threshold to form a broad feedback category set;

[0043] For feedback categories within the broad feedback category set, an analysis team consisting of experts in relevant fields and feedback staff will be formed to develop targeted solutions.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] This invention uses the bag-of-words model and clustering algorithm to cluster and analyze feedback topics. Professionals determine categories according to standards, and the data processing module processes various types of feedback data into a unified format. This allows for efficient sorting and accurate classification of feedback data, laying a solid foundation for subsequent analysis.

[0046] The present invention starts from the perspective of a combination of single or multiple feedback data, and identifies potential patterns hidden in the feedback data by calculating support and confidence. These patterns can reveal the connections between different feedback data and provide data support for decision-making in feedback work;

[0047] This method analyzes the association rules between department and processing result scores, determines improvement categories based on comprehensive evaluation index thresholds, and provides targeted training for department staff. This data-driven training arrangement can effectively address departmental shortcomings in handling specific feedback categories, enhance staff capabilities, standardize processing procedures, and thereby improve the overall service quality of the feedback processing department and enhance public satisfaction.

[0048] This method analyzes the association rules between region, season, and feedback category, calculates the frequency ratio, and can accurately identify seasonal high-frequency feedback. By calculating the ratio set for each region and comparing it with a set ratio threshold, the main feedback category of each region is obtained. Based on the main feedback category, the comprehensive processing value of each staff member who has handled the main feedback category is calculated. Based on the comprehensive processing value, staff members are selected and assigned to the region to resolve feedback issues, thereby improving the efficiency of feedback resolution.

[0049] The present invention analyzes the association rules of age, monthly income, and feedback categories, classifies age intervals to obtain high-frequency feedback categories for each age interval, further calculates the age coverage ratio based on the high-frequency feedback categories, and generates a broad feedback category set through threshold comparison. This defines key areas of focus for subsequent work and helps to grasp the feedback situation from a macro perspective. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0051] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0052] Several embodiments of the present application will be described in more detail below with reference to the accompanying drawings so that those skilled in the art can implement the present application. The present application can be embodied in many different forms and for many different purposes and should not be limited to the embodiments described herein. These embodiments are provided to make the present application comprehensive and complete and to fully convey the scope of the present application to those skilled in the art. The embodiments do not limit the present application.

[0053] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the relevant art and / or the context of this specification, and will not be interpreted in an idealized or overly formal sense unless expressly defined as such herein.

[0054] See also Figure 1 As shown, a feedback data association analysis system includes: a data acquisition module, a subject classification module, a data processing module, an association rule module, a department analysis module, a region and season analysis module, and an age and salary analysis module;

[0055] Data acquisition module: used to collect various feedback data of the target area within one year;

[0056] Feedback data includes: basic information of the feedback person, feedback topic, feedback time, feedback channel, handling department, handling personnel, processing plan and processing result score. Basic information includes name, occupation, age, contact information, region and monthly income. Feedback channels include letters, telephone, online platforms and visits.

[0057] Topic classification module: Cluster analysis is performed on the feedback topics in the target area within one year, thereby classifying each feedback topic and obtaining k cluster centers and corresponding feedback categories;

[0058] S1: Use the bag-of-words model to convert the feedback topic into a vector as the topic vector, calculate the clustering error under different cluster numbers, draw a curve of clustering error changing with the number of clusters, and select the cluster number at the inflection point of the curve as the cluster k value;

[0059] S2-1: Randomly select k data points as initial cluster centers, calculate the Euclidean distance between the topic vector and each initial cluster center, and select the value with the smallest Euclidean distance as the cluster corresponding to the topic vector;

[0060] The feedback topic vectors are converted using the bag-of-words model, and a curve is plotted showing the clustering error versus the number of clusters. The number of clusters at the inflection point of the curve is selected as the k value. This helps find a more appropriate number of clusters based on data characteristics, avoiding unreasonable classification caused by too many or too few clusters, making subsequent analysis more scientific and accurate.

[0061] S2-2: Use the mean formula to calculate the topic vectors in each cluster to form a new cluster center. Calculate the Euclidean distance between the topic vector and the new cluster center again, and select the minimum value of the Euclidean distance between the topic vector and the new cluster center as the new cluster;

[0062] S2-3: Repeat S2-1 and S2-2 continuously. During this process, new cluster centers are continuously generated, and the Euclidean distance between the cluster centers and the corresponding new cluster centers is calculated. The iteration threshold is set. When it is detected that the Euclidean distance between all cluster centers and the corresponding new cluster centers in a certain iteration is less than the iteration threshold, the iteration is stopped, thereby obtaining k cluster centers;

[0063] S3: Arrange professionals to analyze each cluster center formed according to the defined classification standards and determine the specific feedback category to which each cluster center belongs;

[0064] According to the defined standards, professionals are arranged to analyze the cluster centers and determine the feedback categories to ensure that the classification complies with official standards, improve the authority and consistency of the classification, and facilitate the statistics, analysis and connection of feedback data with other related work.

[0065] Data processing module: processes all types of feedback data collected in the target area within five years, converts them into the same format, and forms a feedback data set;

[0066] Extract the region where the feedback provider is located and find the district or county where the feedback provider is located. Use the bag-of-words model to convert the feedback topic into a vector, recorded as the topic vector. Calculate the Euclidean distance between the topic vector and each cluster center, and select the feedback category corresponding to the cluster center with the smallest Euclidean distance as the feedback category of the feedback topic. Convert the feedback time to the four seasons of spring, summer, autumn and winter. Discretize age by constructing age intervals, such as (0,18], (18,22], (22,30], (30,40], (40,50], (50,60], (60,70], (70,200], where (60,70] represents people aged between 60 and 70 (excluding 60 but including 70), and (70,200] represents people over 70 years old. Construct salary intervals to discretize monthly income.

[0067] Discretizing age and salary, converting continuous data into discrete intervals, can simplify the data structure and reduce data complexity. It also highlights the characteristic differences of data in different intervals, which helps analyze the feedback tendencies and characteristics of people of different age groups and income levels.

[0068] The basic information of the feedback recipient, feedback topic, feedback time, feedback channel, processing department, and satisfaction with the processing results after processing are integrated to form a standard feedback data set.

[0069] Association rule module: Use association rule mining algorithm to analyze feedback data sets and mine the association relationships between feedback data;

[0070] Calculate the support of the item set. An item set refers to a single feedback data item or a combination of multiple feedback data items. For example, a single feedback data item is {Occupation = Worker}, {Age = (50, 60]}, and a combination of multiple feedback data items is {Occupation = Student, Age = (18, 22]}, {Feedback Time = Autumn, Age = (30, 40], Processing Result Score = 75}. Support refers to the frequency of an item set appearing in the total number of feedback items.

[0071] A support threshold is preset, and the support of individual feedback data in the item set is calculated. The individual feedback data whose support is greater than the support threshold is selected as a candidate 1-item set. Based on the data in the candidate 1-item set, a merging operation is performed to generate an item 2 set. The item 2 set corresponds to the combination of two feedback data in the item set and belongs to the item set. For example, if the candidate 1-item set is: {occupation = worker}, {age = (50, 60]}, {feedback time = autumn}, then a total of three item 2 sets can be generated, namely: {occupation = worker, age = (50, 60]}, {occupation = worker, feedback time = autumn}, and {age = (50, 60], feedback time = autumn}. The support of each item 2 set is calculated, and the item 2 set whose support is greater than the support threshold is selected as a candidate 2-item set. Based on the candidate 2-item set, an item 3 set is generated, and so on. When the item x+1 set is generated based on the candidate x-item set, the support of each item x+1 set is less than the preset support threshold, and the merging is stopped.

[0072] By calculating item set support and setting thresholds for filtering, we can extract frequently occurring single or multiple feedback data combinations, known as item sets, from massive amounts of feedback data. These item sets represent important and common information combinations in the feedback data, helping to focus on key feedback features and patterns.

[0073] Based on the generated candidate n sets, n=2,3,...,x, each candidate n set is divided into two non-empty subsets, U and V, where the intersection of U and V is an empty set and the union of U and V is the corresponding candidate n set. Based on the results of each division, the confidence of the candidate association rules U→V and V→U is calculated. The confidence calculation formula is ,in represents the confidence of the candidate association rule U→V, represents the support of the corresponding candidate n set, Represents the support of the item set formed by the corresponding feedback data combination in the set U, sets the confidence threshold, and selects the candidate association rules with a confidence greater than the confidence threshold from the candidate association rules as the value association rules;

[0074] By calculating the confidence of candidate association rules and screening by setting confidence thresholds, we can screen out valuable association rules with higher reliability from a large number of candidate association rules; these rules can more truly reflect the actual connection between feedback data, provide a more valuable reference basis for decision-making in feedback work, and help to formulate targeted feedback processing strategies and improvement measures.

[0075] Department analysis module: Extract all department → processing result score association rules from the value association rules, record them as department association rules; extract the departments corresponding to the department association rules whose processing result scores are lower than 75, form an analysis department set, and analyze each department in the analysis department set;

[0076] By screening out departments with low processing result scores from value association rules and forming an analysis department set, we can accurately locate departments with poor performance in feedback processing, making subsequent analysis and improvement work more targeted, focusing on departments with actual problems and avoiding resource waste.

[0077] Obtain all types of feedback data from the department within a year, extract feedback themes from the data, vectorize them, calculate the Euclidean distance from each cluster center, select the feedback category corresponding to the cluster center with the minimum Euclidean distance as the feedback category of the feedback theme, classify the feedback by category, and calculate the scores of each feedback category and the total number of times received (ZCS);

[0078] Take the average of the scores of each feedback category as the average score JFZ of the corresponding feedback category, preset the weighted impact factors corresponding to the average score JFZ and the total number of acceptances ZCS, multiply the average score JFZ and the total number of acceptances ZCS by the corresponding weighted impact factors, and then add them together to obtain the result as the comprehensive evaluation index of the corresponding feedback category;

[0079] A comprehensive evaluation index threshold is preset, and feedback categories with a comprehensive evaluation index lower than the comprehensive evaluation index threshold are extracted from each feedback category as promotion categories. Professional personnel are arranged to provide promotion category training for the department’s staff.

[0080] By presetting comprehensive evaluation index thresholds, we can identify categories that need improvement and arrange targeted training for department staff. This makes the training objectives clear and can effectively help department staff improve their ability to handle specific feedback categories, thereby improving the overall feedback processing level and quality, and enhancing public satisfaction.

[0081] Regional Season Analysis Module: Extracts association rules of all regions and seasons from the value association rules, records them as regional association rules, and analyzes the data in the regional association rules;

[0082] Extract feedback categories from regional association rules and classify them by season to form a four-season feedback set. De-duplicate the four-season feedback set and calculate the number of feedback types in the four-season feedback set. Divide the number of feedback types in the four-season feedback set by the total number of counties and districts in the target region to obtain the frequency ratios corresponding to the four seasons.

[0083] A frequency ratio threshold is preset. If the frequency ratio of a season is less than the threshold, the feedback category in the current seasonal feedback set is determined to be seasonal high-incidence feedback. For seasonal high-incidence feedback, each department develops a treatment plan for each region. A warning time is preset, and the system issues a feedback warning x days before the corresponding season arrives. Each department then prepares in advance according to the formulated treatment plan.

[0084] By extracting association rules between regions, seasons, and feedback categories, analyzing and forming a four-season feedback set, and calculating the frequency ratio, we can clearly understand the distribution of various feedback issues in different regions in different seasons and grasp the seasonal patterns of feedback issues.

[0085] The feedback categories corresponding to the four seasons in each region are classified by region, and duplicates are removed. The number of times each feedback category appears in the four seasons is counted. The ratio between the number of times each feedback category appears in the four seasons in the deduplicated set and 4 is calculated. The result is used as the corresponding regional proportion set. A regional proportion threshold is preset. If a data point in a regional proportion set for a region is greater than the regional proportion threshold, the feedback category corresponding to that data point is used as the main feedback category for that region.

[0086] By classifying feedback categories by region, counting the number of occurrences and calculating regional proportions, and determining the primary feedback categories in each region, this helps us gain a deeper understanding of the unique feedback characteristics of each region and provides a basis for developing more targeted regional feedback handling strategies.

[0087] Extract each staff member who has handled the main feedback category from the preset database, obtain the number of times each staff member handled the main feedback category CS, the duration of each processing, and the score of each processing result; perform mean and normalization calculation on each processing score, and obtain the results as the sub-mean value F1 and sub-wave value F2. After the sub-mean value F1 and sub-wave value F2 are normalized, they are inserted into the formula: Thus, the comprehensive score PF of the staff member is obtained, where q1 and q2 are the weighted influence factors corresponding to the mean value F1 and the wave value F2 respectively; the processing time of each staff member is plotted on a line graph according to the time series, and the ployfit function of MATLAB is called to perform straight line fitting, and the absolute value of the slope of the fitting line is extracted as the duration trend value F3. MATLAB is a software specially used for mathematical analysis, and the ployfit function is a function in MATLAB software for straight line fitting.

[0088] Preset the weight influence factors corresponding to the number of times CS, comprehensive score PF and duration trend value F3 that the staff handles the main feedback category, normalize the number of times CS, comprehensive score PF and duration trend value F3 that the staff handles the main feedback category, multiply them with the corresponding weight influence factors and sum them up. The result obtained is the comprehensive processing value of the staff for the main feedback category; select the top three staff members with the top three comprehensive processing values ​​to form a feedback team and send them to work in the area.

[0089] By calculating the comprehensive processing value, the top three staff members are selected to form a feedback team and dispatched to the region. This approach can accurately allocate human resources based on the characteristics of regional feedback, improving the professional level and efficiency of feedback processing.

[0090] Age and salary analysis module: extracts association rules for all ages, monthly income, and feedback categories from value association rules, and analyzes the feedback categories for each age range and monthly income range;

[0091] First, classify each age range and perform frequency statistics on the feedback categories corresponding to each monthly income in each age range. Take the feedback category with the highest frequency among the feedback categories corresponding to each monthly income in each age range as the high-frequency feedback category for the corresponding age range. If there are multiple feedback categories with the same frequency in a certain age range, take the feedback category corresponding to the interval with the higher salary as the high-frequency feedback category for the corresponding age range.

[0092] By classifying by age range, we counted the frequency of feedback categories corresponding to each monthly income and identified the most frequently reported feedback categories for each age range. This helps us accurately grasp the primary feedback demands of people in different age groups, clearly demonstrates the inherent connection between age and feedback categories, and provides a clear direction for subsequent targeted analysis and processing.

[0093] Perform frequency statistics on the high-frequency feedback categories in each age range, divide the frequency of each high-frequency feedback category by the number of age ranges to obtain the age coverage ratio of each high-frequency feedback category, preset an age coverage ratio threshold, compare the age coverage ratio of each high-frequency feedback category with the age coverage ratio threshold, and extract high-frequency feedback categories with age coverage ratios greater than the age coverage ratio threshold to form a broad feedback category set;

[0094] By calculating the age coverage ratio for each high-frequency feedback category, we can intuitively understand the prevalence of different feedback categories in different age groups. After comparing with the preset age coverage ratio threshold, we extract a set of broad feedback categories with high age coverage ratios, screening out feedback issues that involve a wide range of people and have a greater impact, providing a strong basis for resource allocation and work priority determination;

[0095] For feedback categories within the broad feedback category set, an analysis team consisting of experts in relevant fields and feedback staff will be formed to develop targeted solutions.

[0096] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A feedback data correlation analysis system, characterized in that: include: Data acquisition module: used to collect various feedback data of the target area within one year; Topic classification module: Cluster analysis is performed on the feedback topics in the target area within one year, thereby classifying each feedback topic and obtaining k cluster centers and corresponding feedback categories; The specific steps of the topic classification module to obtain k cluster centers and corresponding feedback categories are: S1: Use the bag-of-words model to convert the feedback topic into a vector as the topic vector, calculate the clustering error under different cluster numbers, draw a curve of clustering error changing with the number of clusters, and select the cluster number at the inflection point of the curve as the cluster k value; S2-1: Randomly select k data points as initial cluster centers, calculate the Euclidean distance between the topic vector and each initial cluster center, and select the value with the smallest Euclidean distance as the cluster corresponding to the topic vector; S2-2: Use the mean formula to calculate the topic vectors in each cluster to form a new cluster center. Calculate the Euclidean distance between the topic vector and the new cluster center again, and select the minimum value of the Euclidean distance between the topic vector and the new cluster center as the new cluster; S2-3: Repeat S2-1 and S2-2 continuously. During this process, new cluster centers are continuously generated, and the Euclidean distance between the cluster centers and the corresponding new cluster centers is calculated. The iteration threshold is set. When it is detected that the Euclidean distance between all cluster centers and the corresponding new cluster centers in a certain iteration is less than the iteration threshold, the iteration is stopped, thereby obtaining k cluster centers; S3: Arrange professionals to analyze each cluster center formed and determine the specific feedback category to which each cluster center belongs; Data processing module: processes all types of feedback data collected in the target area within five years, converts them into the same format, and forms a feedback data set; Association rule module: Use association rule mining algorithm to analyze feedback data sets and mine the association relationships between feedback data; Department analysis module: Extracts association rules from all departments → processing result scores from value association rules, which are recorded as department association rules. A scoring threshold is preset, and the departments corresponding to department association rules whose processing result scores are lower than the scoring threshold are extracted to form an analysis department set. Each department in the analysis department set is then analyzed. Regional Season Analysis Module: Extracts association rules of all regions and seasons from the value association rules, records them as regional association rules, and analyzes the data in the regional association rules; Age-salary analysis module: extracts association rules of all ages, monthly income → feedback categories from value association rules, and analyzes the feedback categories of each age range and monthly income range.

2. The feedback data correlation analysis system according to claim 1, characterized in that: The various feedback data of the data acquisition module are: Feedback data includes: basic information of the feedback person, feedback topic, feedback time, feedback channel, handling department, handling personnel, processing plan and processing result score. Basic information includes name, occupation, age, contact information, region and monthly income. Feedback channels include letters, telephone, online platforms and visits.

3. The feedback data correlation analysis system according to claim 1, characterized in that: The specific steps of the data processing module forming the feedback data set are: Extract the region where the feedback provider is located, and extract the district or county where the feedback provider is located; use the bag-of-words model to convert the feedback topic into a vector, recorded as the topic vector, calculate the Euclidean distance between the topic vector and each cluster center, and select the feedback category corresponding to the cluster center with the smallest Euclidean distance as the feedback category of the feedback topic; convert the feedback time by season, converting it to the four seasons of spring, summer, autumn and winter; discretize the age by constructing age ranges; and construct salary ranges to discretize the monthly income; The basic information of the feedback recipient, feedback topic, feedback time, feedback channel, processing department, and satisfaction with the processing results after processing are integrated to form a standard feedback data set.

4. The feedback data correlation analysis system according to claim 1, characterized in that: The specific steps of mining the association relationship between feedback data by the association rule module are as follows: Calculate the support of the item set. An item set refers to a single feedback data or a combination of multiple feedback data. Support refers to the frequency of an item set appearing in the total number of feedbacks. Preset a support threshold, calculate the support of individual feedback data in the item set, and select individual feedback data with support greater than the support threshold as candidate 1-item set; Based on the data in the candidate 1-item set, a merging operation is performed to generate an item 2 set. The item 2 set corresponds to the combination of two pieces of feedback data in the item set and belongs to the item set. The support of each item 2 set is calculated, and the item 2 sets in the item 2 set whose support is greater than the support threshold are selected as candidate 2-item sets. Based on the candidate 2 item sets, generate item 3 sets, and so on; until the item x+1 set is generated based on the candidate x item set, the support of each item x+1 set is less than the preset support threshold, and the merging stops; Based on the generated candidate n sets, n=2,3,...,x, each candidate n set is divided into two non-empty subsets, U and V, where the intersection of U and V is an empty set and the union of U and V is the corresponding candidate n set. Based on the results of each division, the confidence of the candidate association rules U→V and V→U is calculated. The confidence calculation formula is ,in represents the confidence of the candidate association rule U→V, represents the support of the corresponding candidate n set, It represents the support of the item set formed by the corresponding feedback data combination in the set U, sets the confidence threshold, and selects the candidate association rules whose confidence is greater than the confidence threshold as the value association rules.

5. The feedback data correlation analysis system according to claim 1, characterized in that: The specific steps of the department analysis module are: Obtain all types of feedback data from the department within a year, extract feedback themes from the department's feedback data, vectorize them, calculate the Euclidean distance from each cluster center, select the feedback category corresponding to the cluster center with the minimum Euclidean distance as the feedback category of the feedback theme, classify the feedback by category, and calculate the scores of each feedback category and the total number of acceptances (ZCS); Take the average of the scores of each feedback category as the average score JFZ of the corresponding feedback category, preset the weighted impact factors corresponding to the average score JFZ and the total number of acceptances ZCS, multiply the average score JFZ and the total number of acceptances ZCS by the corresponding weighted impact factors, and then add them together to obtain the result as the comprehensive evaluation index of the corresponding feedback category; A comprehensive evaluation index threshold is preset, and feedback categories with a comprehensive evaluation index lower than the comprehensive evaluation index threshold are extracted from each feedback category as promotion categories. Professional personnel are arranged to provide promotion category training for the department’s staff.

6. The feedback data correlation analysis system according to claim 1, characterized in that: The specific steps of the regional seasonal analysis module are: Extract feedback categories from regional association rules and classify them by season to form a four-season feedback set. De-duplicate the four-season feedback set and calculate the number of feedback types in the four-season feedback set. Divide the number of feedback types in the four-season feedback set by the total number of counties and districts in the target region to obtain the frequency ratios corresponding to the four seasons. A frequency ratio threshold is preset. If the frequency ratio of a season is less than the threshold, the feedback category in the current seasonal feedback set is determined to be seasonal high-frequency feedback. For seasonal high-frequency feedback, each department will develop a treatment plan for each region. The pre-set warning time is x days before the corresponding season arrives. The system will issue a feedback warning and each department will prepare in advance according to the established treatment plan. The feedback categories corresponding to the four seasons in each region are classified by region, and duplicates are removed. The number of times each feedback category appears in the four seasons is counted. The ratio between the number of times each feedback category appears in the four seasons in the deduplicated set and 4 is calculated. The result is used as the corresponding regional proportion set. A regional proportion threshold is preset. If a data point in a regional proportion set for a region is greater than the regional proportion threshold, the feedback category corresponding to that data point is used as the main feedback category for that region. Extract each staff member who has handled the main feedback category from the preset database, and obtain the number of times each staff member has handled the main feedback category, the duration of each processing, and the score of each processing result; The scores of each treatment were averaged and standardized, and the results were used as the mean value F1 and the wave value F2. After the normalization process, the mean value F1 and the wave value F2 were input into the formula: Thus, the comprehensive score PF of the staff is obtained, where q1 and q2 are the weighted influencing factors corresponding to the mean value F1 and the wave value F2 respectively; the processing time of each staff member is plotted on a line graph according to the time series, and the ployfit function of MATLAB is called to perform straight line fitting, and the absolute value of the slope of the fitting line is extracted as the duration trend value F3; Preset the weighted impact factors corresponding to the number of times the staff member handles the main feedback category CS, the comprehensive score PF, and the duration trend value F3. Normalize the number of times the staff member handles the main feedback category CS, the comprehensive score PF, and the duration trend value F3, multiply them by the corresponding weighted impact factors, and then sum them. The result is used as the staff member's comprehensive processing value for the main feedback category; The top three staff members in comprehensive processing values ​​are selected to form a feedback team and sent to work in the area.

7. The feedback data association analysis system according to claim 1, characterized in that: The specific steps of the age-salary analysis module are: First, classify each age range and perform frequency statistics on the feedback categories corresponding to each monthly income in each age range. Take the feedback category with the highest frequency among the feedback categories corresponding to each monthly income in each age range as the high-frequency feedback category for the corresponding age range. If there are multiple feedback categories with the same frequency in a certain age range, take the feedback category corresponding to the interval with the higher salary as the high-frequency feedback category for the corresponding age range. Perform frequency statistics on the high-frequency feedback categories in each age range, divide the frequency of each high-frequency feedback category by the number of age ranges to obtain the age coverage ratio of each high-frequency feedback category, preset an age coverage ratio threshold, compare the age coverage ratio of each high-frequency feedback category with the age coverage ratio threshold, and extract high-frequency feedback categories with age coverage ratios greater than the age coverage ratio threshold to form a broad feedback category set; For feedback categories within the broad feedback category set, an analysis team consisting of experts in relevant fields and feedback staff will be formed to develop targeted solutions.

Images