Intelligent call center resource optimization method and system based on predictive analysis

Abstract

The invention provides an intelligent call center resource optimization method and system based on predictive analysis, and relates to the technical field of call center management, and the method comprises the steps: obtaining to-be-mediated customer data, mediator data and external economic indicators; processing according to the to-be-conciliated customer data to obtain a customer emotion feature matrix; performing feature selection and quantitative evaluation processing according to the mediator data to obtain a mediator capability feature matrix; performing scene simulation according to the external economic indexes, the customer emotion feature matrix and the mediator ability feature matrix to obtain a mediation effect score list; performing time sequence analysis and behavior pattern recognition according to the historical mediation information to obtain a prediction result; and performing optimization according to the mediation effect score list and the prediction result to obtain a final mediator scheduling scheme. According to the method, emotion analysis, behavior pattern recognition and scheduling optimization technologies are integrated, the processing efficiency of the mediation request is improved by using data analysis and a prediction model, the waiting time of a customer is shortened, and resource allocation is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of call center management, and in particular to an intelligent call center resource optimization method and system based on prediction analysis. Background Art

[0002] In the financial services industry, call centers are facing increasing pressure as customer demands increase, especially when call volumes fluctuate greatly. Current call center scheduling methods rely mainly on historical data and manual experience for planning, and usually adopt a simple static scheduling model, which makes it difficult to accurately predict demand fluctuations during peak and off-peak periods, resulting in insufficient flexibility in resource allocation. During peak periods, long customer waiting times can lead to decreased satisfaction, while during off-peak periods, call center resources are often idle, resulting in a waste of resources. This traditional scheduling method is unable to cope with sudden peak call demands and is difficult to meet the efficient response requirements of modern financial mediation services.

[0003] In order to solve the above problems, the present invention proposes an intelligent call center resource optimization method and system based on predictive analysis. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for optimizing intelligent call center resources based on predictive analysis to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, the present application provides an intelligent call center resource optimization method based on predictive analysis, comprising:

[0006] Obtaining data of clients to be mediated, data of mediators, and external economic indicators, wherein the data of clients to be mediated includes historical mediation information and repayment information of clients, and the data of mediators includes background information of mediators and successful mediation cases;

[0007] Processing the customer data to be mediated, analyzing the text data of the customer in social media and customer service conversations based on a preset sentiment analysis model, and constructing a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility;

[0008] Perform feature selection and quantitative evaluation processing based on the mediator data, and quantitatively evaluate the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate by using correlation analysis to construct a mediator capability feature matrix;

[0009] Scenario simulation is performed based on the external economic indicators, the customer emotion feature matrix, and the mediator capability feature matrix. By identifying the weights of the economic and emotional impacts on mediation needs, the dynamic trajectory of customer emotions is generated in combination with the chaos mapping theory. The decision-making process of different mediators in different scenarios is simulated and evaluated using a deep reinforcement learning algorithm to obtain a mediation effect score list.

[0010] Perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain a prediction result, wherein the prediction result includes the number of mediation requests and customer behavior patterns within a preset time in the future;

[0011] The shift scheduling is optimized according to the mediation effect score list and the prediction results, and the final mediator shift scheduling plan is obtained by optimizing the shift scheduling combination using a simulated annealing algorithm.

[0012] In a second aspect, the present application also provides an intelligent call center resource optimization system based on predictive analysis, including:

[0013] An acquisition module, used to acquire data of clients to be mediated, data of mediators, and external economic indicators, wherein the data of clients to be mediated includes historical mediation information and repayment information of clients, and the data of mediators includes background information of mediators and successful mediation cases;

[0014] An analysis module is used to process the customer data to be mediated, analyze the text data of the customer in social media and customer service dialogue based on a preset sentiment analysis model, and construct a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility;

[0015] An evaluation module is used to perform feature selection and quantitative evaluation processing based on the mediator data, and to quantitatively evaluate the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate by using correlation analysis to construct a mediator capability feature matrix;

[0016] A simulation module is used to perform scenario simulation based on the external economic indicators, the customer emotion feature matrix and the mediator capability feature matrix, generate customer emotion dynamic trajectories by identifying the weights of the impact of economy and emotion on mediation demand, and use the chaos mapping theory to simulate and evaluate the decision-making process of different mediators in different scenarios using a deep reinforcement learning algorithm to obtain a mediation effect score list;

[0017] A prediction module, used to perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain prediction results, wherein the prediction results include the number of mediation requests and customer behavior patterns within a preset time in the future;

[0018] The optimization module is used to optimize the scheduling according to the mediation effect score list and the prediction results, and obtain the final mediator scheduling plan by optimizing the scheduling combination using the simulated annealing algorithm.

[0019] The beneficial effects of the present invention are:

[0020] The present invention integrates sentiment analysis, behavioral pattern recognition and scheduling optimization technology, and utilizes data analysis and prediction models to improve the processing efficiency of mediation requests, reduce customer waiting time and optimize resource allocation.

[0021] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 It is a flow chart of the intelligent call center resource optimization method based on predictive analysis described in an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of an intelligent call center resource optimization system based on predictive analysis described in an embodiment of the present invention.

[0025] Markings in the figure: 1. Acquisition module; 2. Analysis module; 21. First extraction unit; 22. First analysis unit; 23. First calculation unit; 24. First integration unit; 3. Evaluation module; 31. Second extraction unit; 32. Second analysis unit; 33. Third analysis unit; 34. First construction unit; 4. Simulation module; 41. Second construction unit; 42. First simulation unit; 421. First screening unit; 422. First setting unit; 423. Third simulation unit; 424. Third extraction unit; 43. Second simulation unit; 44. First evaluation unit; 5. Prediction module; 51. Third construction unit; 52. First decomposition unit; 53. First identification unit; 54. First prediction unit; 6. Optimization module; 61. Fourth analysis unit; 62. First generation unit; 63. First definition unit; 64. First optimization unit. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0028] Embodiment 1:

[0029] This embodiment provides an intelligent call center resource optimization method based on predictive analysis.

[0030] See also Figure 1 , the figure shows that the method includes steps S100 to S600.

[0031] Step S100, obtaining data of the client to be mediated, data of the mediator, and external economic indicators, wherein the data of the client to be mediated includes the client's historical mediation information and repayment information, and the data of the mediator includes the mediator's background information and successful mediation cases;

[0032] It is understandable that the data of customers to be mediated includes historical mediation information and repayment records, which can reflect the repayment behavior of customers, the frequency of mediation requests, and repayment performance under different economic environments. By analyzing this information, the risk characteristics and urgency of the customer's needs can be identified. At the same time, obtaining mediator data including their background information (such as professional field, years of experience) and successful mediation cases will help evaluate the performance of each mediator in a specific mediation situation. The introduction of external economic indicators (such as market interest rates, inflation rates, etc.) provides important parameters for analyzing the potential impact of the economic environment on customer behavior. It should be noted that the data acquisition method strictly follows data privacy and legal compliance requirements to ensure that customer privacy is not violated. The collection of customer data to be mediated, mediator data, and external economic indicators is based on legal authorization and strict data management processes. The customer's historical mediation information and repayment records are only used with the consent of the customer or in compliance with the legal compliance standards of the financial institution to ensure that the data is used for legitimate mediation service purposes. The mediator's data comes from internal employee files and business performance data, which complies with the provisions of human resource management and does not involve the over-scope use of personal privacy. In addition, external economic indicator data comes from public databases or legal channels of financial institutions and is public information and does not involve personal privacy.

[0033] Step S200: Process the customer data to be mediated, analyze the text data of the customer in social media and customer service conversations based on a preset sentiment analysis model, and construct a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility;

[0034] It can be understood that the emotional feature values ​​in the matrix not only help to describe the emotional trends of customers during the mediation cycle, but also reveal the emotional sensitivity of customers in specific situations (such as economic pressure or financial difficulties).

[0035] Step S300: feature selection and quantitative evaluation processing are performed based on the mediator data. The mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate are quantitatively evaluated by correlation analysis to construct a mediator capability feature matrix;

[0036] It should be noted that the constructed mediator competency characteristic matrix quantifies the performance of all mediators in specific competency dimensions so that in the subsequent allocation process, appropriate mediators can be flexibly matched according to client needs and mediation situations.

[0037] Step S400: Scenario simulation is performed based on external economic indicators, customer emotion feature matrix and mediator capability feature matrix. By identifying the weights of the impact of economy and emotion on mediation demand, the dynamic trajectory of customer emotion is generated in combination with chaos mapping theory. The decision-making process of different mediators in different scenarios is simulated and evaluated using a deep reinforcement learning algorithm to obtain a mediation effect score list.

[0038] Step S500: Perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain prediction results, which include the number of mediation requests and customer behavior patterns within a preset time in the future;

[0039] It should be noted that this step, through a comprehensive assessment of economic, emotional and mediator characteristics, enables mediation resources to be efficiently allocated in a complex and ever-changing environment, ensuring the provision of high-quality mediation services at critical moments, and improving mediation success rates and client satisfaction.

[0040] Step S600: Perform scheduling optimization processing according to the mediation effect score list and the prediction results, and obtain the final mediator scheduling plan by optimizing the scheduling combination using the simulated annealing algorithm.

[0041] It should be noted that the optimized scheduling scheme obtained by the simulated annealing algorithm in this step can minimize customer waiting time, improve mediation efficiency, and balance the workload of mediators, thereby achieving optimal resource allocation under different demand peaks.

[0042] Further, step S200 includes step S210 to step S240.

[0043] Step S210, extracting data based on the customer data to be mediated to obtain original text data;

[0044] It is understood that this step involves extracting text content related to the mediation request from social media and customer service conversation records, including the customer's comments, feedback, and emotions and needs expressed in the conversation. Through API interfaces, database queries or data crawling technology, the customer's mediation-related text information on different platforms is centrally collected, and the consistency and integrity of the data are ensured. The extracted raw text data fully reflects the customer's emotional state, expression method and potential needs during the mediation process.

[0045] Step S220: Analyze the original text data based on the BERT sentiment classification model, identify the sentiment category of each text and calculate the intensity of each sentiment to obtain the sentiment labeling result;

[0046] It should be noted that the BERT model has a strong language understanding ability based on deep learning and can capture complex emotional details in the text. In this step, each text is classified by the BERT model, identifying its emotional category (such as positive, negative, neutral, etc.) and further quantifying the intensity of each emotion. In order to accurately reflect customer emotions, an emotional intensity scoring mechanism is also combined to assign corresponding numerical values ​​to each type of emotion to ensure the depth and accuracy of the emotional analysis results. These emotional labeling results construct the emotional characteristics of the customer, so that the subsequent emotional volatility calculation and the establishment of the emotional feature matrix have a specific emotional dimension.

[0047] Step S230: According to the emotion labeling result, by introducing the volatility index, using the sliding window method to calculate the frequency and intensity of changes in customer emotions during the mediation cycle, and obtain the emotion volatility feature;

[0048] Specifically, the volatility index is first introduced, and the frequency and intensity of changes in customer emotions are used as a measurement standard to identify the severity of emotional changes. Preferably, the sliding window method is used to segment the emotion labeling results within the mediation cycle, and the average amplitude and frequency of emotional changes within each window are gradually calculated. The settings of the sliding window (such as window size and sliding step) can be adjusted according to the length of the mediation cycle and analysis needs to ensure that the dynamic trend of customer emotional changes is captured. By calculating the standard deviation or rate of change of emotional intensity to reflect emotional volatility, the volatility characteristics can reveal the emotional sensitivity of customers at different mediation stages.

[0049] Step S240: perform data integration processing according to the emotion labeling results and the emotion volatility characteristics, and encode the integrated results to construct a customer emotion feature matrix.

[0050] It is understandable that in this integration process, the emotion category (such as positive, negative, neutral) of each customer is first combined with the emotion intensity to form complete emotion annotation data. Subsequently, the emotion volatility characteristics (such as change frequency and intensity) are integrated into the emotion annotation data to reflect the dynamic change trend of customer emotions. In order to ensure the consistency and ease of processing of the data in the matrix, the integrated emotion features are further encoded. Preferably, through unique hot encoding and normalization, different features have a unified numerical scale and structure. The final customer emotion feature matrix comprehensively describes the emotional state and volatility of the customer during the mediation cycle.

[0051] Furthermore, step S300 includes step S310 to step S340.

[0052] Step S310: Perform feature extraction processing based on the mediator data, generate continuous features by mapping the mediator's professional field with the average value of the mediation success rate, and convert the years of experience into categorical features by segmentation processing, so as to obtain a mediator feature set;

[0053] Specifically, the mediator's professional field is first mapped to the average value of his / her mediation success rate to generate a continuous feature that reflects professional ability. This mapping allows the success rate of mediators in a specific field in their field of strength to be quantified by calculating the correlation between different professional fields and mediation success rates. Secondly, the mediator's years of experience are segmented and converted into categorical features. For example, the years of experience are divided into categories such as "0-2 years", "3-5 years", and "more than 6 years" to capture the nonlinear impact of experience on mediation performance. Through this combination of continuous and categorical features, a mediator feature set is obtained, which covers the mediator's professional ability and experience depth. This step provides an accurate data basis for subsequent feature selection and capability matrix construction by converting the mediator's complex capability data into a structured and quantified feature set, thereby supporting the intelligent allocation of mediation resources and the implementation of efficient mediation strategies.

[0054] Step S320: performing correlation analysis based on the mediator feature set, evaluating the relationship between the mediator's professional field, years of experience, case types and mediation success rate by calculating mutual information, and combining the chi-square test to evaluate the independence of the feature and the mediation success rate, to obtain the correlation analysis result;

[0055] It should be noted that this step first uses mutual information to calculate the information correlation between each feature in the mediator feature set (such as professional field, years of experience, and type of case handled) and the mediation success rate. The mutual information method can capture nonlinear relationships and identify which features have significant contributions to the mediation success rate. Furthermore, the independence of each feature and the mediation success rate is evaluated through the chi-square test, and it is verified from a statistical point of view whether these features are significantly associated with the success rate. Combining the results of the mutual information and chi-square tests, the correlation analysis results of the mediator characteristics can be obtained, clarifying which features have an important influence in predicting the mediation success rate. The calculation formula is:

[0056]

[0057] Where X represents the specific characteristics of the mediator (such as professional field, years of experience, and types of cases handled); Y represents the mediation success rate (after discretization, it is divided into success and failure categories); x represents the eigenvalue number; y represents the mediation success rate category; I(X; Y) represents the mutual information, which is used to measure the nonlinear dependency between two variables X and Y; p(x, y) represents the joint probability of the eigenvalue x and the mediation success rate y occurring at the same time; p(x) represents the marginal probability of the eigenvalue x; p(y) represents the marginal probability of the mediation success rate y; χ represents the chi-square statistic, which is used to measure the degree of deviation between the actual observed value and the expected value; n represents the number of eigenvalue categories; m represents the number of mediation result categories; i represents the eigenvalue category number; j represents the result category number; O ij represents the actual observed frequency under feature category i and result category j; E ij represents the expected frequency of feature category i and result category j; N represents the total number of samples; R i represents the total observed frequency of feature category i; C j represents the total observed frequency of outcome category j.

[0058] Step S330, performing principal component analysis based on the correlation analysis results, extracting principal components by performing dimensionality reduction processing on the mediator features to capture the main variability in the data and obtain a main feature set;

[0059] It should be noted that, first of all, based on the previous correlation analysis results, mediator features with high correlation are selected as the input of principal component analysis. These high-dimensional feature sets are linearly transformed by principal component analysis to extract a few principal components that can explain most of the information in the data. Principal component analysis combines highly correlated variables in the original feature set into new principal components, reducing data redundancy while retaining the most explanatory features for mediator ability evaluation. The main feature set obtained contains the core ability performance of the mediator in a specific situation, and reduces the feature dimension, simplifying subsequent analysis.

[0060] Step S340: Matrix construction is performed based on the main feature set, and the mediator capability feature matrix is ​​constructed by standardizing the features. The mediator capability feature matrix includes the numerical performance of each mediator on each feature.

[0061] Specifically, the main feature set is standardized so that each feature value is in the same scale range (such as through Z-score standardization or Min-Max normalization) to ensure that the features are comparable in value and avoid weight imbalance due to scale differences between features. Subsequently, the numerical performance of each mediator on each main feature is integrated into row data to form a mediator capability feature matrix. Each row of the matrix represents the feature data of a mediator, covering his or her comprehensive performance in professional field, years of experience, and case types.

[0062] Further, step S400 includes step S410 to step S440.

[0063] Step S410: Scenario construction is performed based on external economic indicators and the customer emotion feature matrix. By applying a weighted linear combination method, the market interest rate, inflation rate, customer emotion category and volatility are integrated to calculate the customer emotion impact weights under different economic scenarios, and obtain a comprehensive scenario feature matrix.

[0064] First, external economic indicators (such as market interest rates, inflation rates, etc.) are integrated with the emotion categories and emotion volatility in the customer emotion feature matrix through the weighted linear combination method. Specifically, each economic indicator and emotion feature is assigned a corresponding weight to reflect its potential impact on customer emotions. For example, when market interest rates rise, customers with high volatility emotions are given a higher weight to highlight their emotional sensitivity. Based on these weight combinations, the impact value of customer emotions in each economic scenario is calculated to generate a comprehensive scenario feature matrix. This matrix can describe the expected changes in customer emotional states under different economic conditions and provide dynamic emotion input for subsequent scenario simulations. The effect of this step is to accurately quantify the interaction between economy and emotion through scenario construction, so that the adaptability of mediation strategies in different economic scenarios is enhanced, thereby more effectively meeting customer needs.

[0065] Step S420: Perform simulation processing according to the comprehensive situational feature matrix, perform nonlinear dynamic simulation of customer emotions through chaos mapping theory, identify the urgency and sensitivity of customers to mediation services, and generate dynamic trajectories of customer emotions under different economic conditions;

[0066] It should be noted that this step first recursively calculates the numerical input of customer emotions in the comprehensive situational feature matrix through chaos mapping theory (such as Logistic mapping or Henon mapping) to simulate the nonlinear change trajectory of emotions. Chaos mapping theory is suitable for capturing subtle fluctuations and mutations in complex systems, and can effectively reveal the nonlinear response of customers' emotional states under economic fluctuations or specific situations. During the simulation process, emotional states fluctuate and mutate under different economic pressures, and the generated emotional dynamic trajectory can reflect the speed and amplitude of customer emotional changes and their sensitivity to the demand for mediation services. The effect of this step is to provide a time-series dynamic diagram of emotional changes, accurately depict the emotional response patterns of customers under economic fluctuations, and support mediators to take priority countermeasures among customers with high emotional fluctuations, thereby achieving a more personalized mediation service strategy.

[0067] Step S430, performing decision simulation according to the dynamic trajectory of customer emotions and the mediator's ability feature matrix, using the policy gradient algorithm to take the dynamic trajectory of customer emotions as the environment state, taking the mediator's features as the input of the intelligent agent, and training the mediator's response strategy through reinforcement learning to obtain the decision simulation result;

[0068] Specifically, the dynamic trajectory of customer emotions is first used as the input of the environmental state to describe the changes in customer emotions under different economic and emotional conditions. At the same time, the mediator's ability feature matrix is ​​used as the input of the intelligent agent, representing the mediator's performance in various mediation capabilities. Then, by using the policy gradient algorithm (such as PPO or A3C), the reinforcement learning model is trained so that the mediator can choose the optimal response strategy under different customer emotional states. The policy gradient algorithm allows the model to adjust the mediation strategy through gradient optimization in each scenario to maximize the mediation success rate and customer satisfaction. The model is continuously iterated, and the decision-making effect of the mediator in complex emotional scenarios is strengthened through positive feedback to obtain optimized decision simulation results.

[0069] Step S440: Based on the decision simulation results, the fuzzy logic algorithm is used to evaluate the performance of each mediator in different economic and emotional situations, and the decision effect is converted into a quantitative mediation effect score to obtain a mediation effect score list.

[0070] It should be noted that this step inputs the performance of each mediator in various scenarios into the fuzzy logic system based on the decision-making simulation results. The fuzzy logic algorithm handles the uncertainty and nonlinear changes in these different scenarios, and converts the mediator's performance (such as response speed, emotional coping ability, mediation success rate, etc.) into a quantitative score through fuzzy rules. The fuzzy logic system scores the mediator's performance according to preset rules, such as "response success rate of customers with high emotional fluctuations under high economic pressure" or "mediation success rate of customers with low economic fluctuations." In this way, the decision-making effect in each scenario is converted into a quantitative mediation effect score, and a complete mediation effect score list is constructed.

[0071] Further, step S420 includes step S421 to step S424.

[0072] Step S421: Variable screening is performed according to the comprehensive situational feature matrix, and the contribution of market interest rate, customer repayment pressure and mediation history to customer sentiment change is calculated by applying the information gain algorithm, so as to obtain a set of key variables;

[0073] It is understandable that for each variable in the comprehensive scenario feature matrix (such as market interest rate, customer repayment pressure and mediation history, etc.), the information gain algorithm is applied to calculate the contribution of each variable to the change of customer emotions. The information gain algorithm is used here to measure the correlation between each variable and customer emotional fluctuations, that is, the role of each variable in reducing customer emotional uncertainty. Through this analysis, variables that have a significant impact on emotional changes can be screened out, thereby obtaining a set of key variables. These key variables will be given higher weights in the scenario simulation to ensure the accuracy of the dynamic simulation of customer emotions.

[0074] Step S422: Perform initial emotional state setting processing according to the key variable set, generate the initial emotional state of each customer in a specific situation by combining the customer's historical emotional fluctuations, mediation success rate and current economic status through a nonlinear autoregressive model, and obtain the customer's emotional initial state matrix;

[0075] Specifically, this step uses a nonlinear autoregressive model (NARX) to input data such as the customer's historical emotional fluctuations, mediation success rate, and current economic status into the model to generate an accurate estimate of the customer's initial emotional state. The nonlinear autoregressive model is suitable for processing nonlinear relationships in time series data and can capture the complex dependencies between customer emotional fluctuations and mediation results. Through model calculation, the initial emotional state value of each customer under the current economic situation and historical behavior is obtained, thereby forming an emotional initial state matrix. The effect of this step is to provide an accurate starting state for the simulation of the emotional trajectory, so that the subsequent emotional dynamic simulation can more realistically reflect the trend of emotional changes of customers during the mediation process, thereby supporting the optimization of contextualized mediation strategies.

[0076] Step S423: Perform nonlinear dynamic simulation processing according to the initial state matrix of emotions, introduce fractal dimension analysis, and recursively simulate the subtle fluctuations of customer emotions in a specific economic context in combination with fractal geometry methods to obtain the trajectory of customer emotion changes;

[0077] It should be noted that this step uses fractal dimension analysis, which regards the emotional state in the emotional initial state matrix as a dynamic system, and captures the subtle structural characteristics of customer emotional fluctuations through fractal geometry methods. Fractal dimension analysis can reveal the self-similarity and complexity of customer emotional changes, and is particularly suitable for subtle fluctuations and mutations in emotions in different economic situations. In the recursive simulation, each emotional state is input into the fractal model for iteration to generate an emotional change trajectory that conforms to nonlinear laws. The final customer emotional change trajectory shows in detail the dynamic response of customer emotions under economic and mediation pressures.

[0078] Step S424, generate an emotion fluctuation pattern based on the emotion change trajectory, extract the instantaneous fluctuation frequency and fluctuation intensity in the customer emotion trajectory through Hilbert-Huang transform, identify the urgency and sensitivity of the customer emotion to the mediation service, and finally obtain the dynamic trajectory of the customer emotion.

[0079] It is understandable that the Hilbert-Huang transform can extract instantaneous features from nonlinear and non-stationary data, which is suitable for analyzing subtle change patterns in customer emotional trajectories. Through the Hilbert-Huang decomposition, the frequency and intensity of emotional fluctuations at each moment are obtained, and the urgency (frequency of fluctuations) and sensitivity (intensity of fluctuations) of emotions are identified. The final generated dynamic trajectory of customer emotions contains the instantaneous characteristics of emotional changes, which provides an important basis for the personalized response of mediation services. The specific transformation process is:

[0080] First, perform Hilbert transform on the emotion change trajectory x(t) to obtain its analytical signal z(t):

[0081]

[0082] Where t represents the current time point; z(t) represents the analytical signal at the current time point, which contains the instantaneous amplitude and instantaneous phase information of the emotion; x(t) represents the original emotion change signal; k represents the imaginary unit; represents Hilbert transform; PV represents principal value, which is used for definition in integration to ensure that the problem of singularity is handled in Hilbert transform calculation; τ is the integral variable, which represents an offset or delay in time.

[0083] By analyzing the signal z(t), the instantaneous amplitude A(t) and instantaneous phase φ(t) can be calculated:

[0084]

[0085] The instantaneous frequency f(t) is the time derivative of the instantaneous phase and is used to indicate the urgency of emotional fluctuations:

[0086]

[0087] Further, step S500 includes step S510 to step S540.

[0088] Step S510: sort out the data based on the historical mediation information of all customers, and construct a time series data set including the customer's credit status, repayment ability and mediation success rate by integrating the customer's mediation request records, repayment information and timestamps;

[0089] It is understandable that this step integrates historical data such as the customer's mediation request record, repayment information, and timestamp to ensure the temporal continuity and consistency of the data. Mediation request records are used to track the customer's demand frequency and mediation process, reflecting the customer's mediation preference and urgency; repayment information, including repayment amount, overdue records, and number of defaults, can quantify the customer's repayment ability and financial reliability. By integrating these data, the customer's credit status (such as credit score changes), repayment ability, and historical mediation success rate are calculated, and finally a time series data set is formed.

[0090] Step S520, performing decomposition processing according to the time series data set, by decomposing the time series of the number of customer mediation requests into a trend component, a seasonal component and a residual, and smoothing the trend component using a weighted moving average method, to obtain a decomposed time series data set;

[0091] Specifically, this step uses time series decomposition techniques (such as STL decomposition) to decompose the number of customer mediation requests into three parts: the trend component represents the long-term trend of the data, the seasonal component reflects the cyclical fluctuations of mediation requests (such as quarterly or monthly peaks), and the residual component captures short-term random fluctuations and outliers. Subsequently, the trend component is smoothed using the weighted moving average method to eliminate the noise caused by short-term fluctuations and make the trend clearer and more stable. This decomposed time series data set can help identify the long-term trend and cyclical characteristics of customer mediation requests, and provide structured input data for subsequent demand forecasting.

[0092] Step S530, performing behavior pattern recognition based on the decomposed time series data set, and obtaining a behavior pattern recognition result by analyzing the pattern similarity of customer mediation requests under different economic environments;

[0093] It can be understood that this step performs cluster analysis on the trend and seasonal components in the decomposed time series data set to identify similar mediation request patterns. This analysis method can identify the behavior patterns of different customers under specific economic conditions, such as whether some customers show a more frequent mediation request trend during economic downturns or rising interest rates. By clustering the similarity of patterns, customers are divided into different behavior type groups, such as a high-frequency mediation request group, an economically sensitive group, etc., to generate behavioral pattern recognition results.

[0094] Step S540: predict future requests based on the behavior pattern recognition results, and obtain prediction results by analyzing the long-term dependency of customer behaviors and predicting customer behavior patterns under different economic environments.

[0095] Specifically, this step uses the identified customer behavior patterns and economic scenario data as input, and uses a prediction model suitable for processing time dependencies, preferably a long short-term memory network, to capture the long-term trend and periodicity of customer behavior. When training the model, special attention is paid to the changing response of customer behavior under different economic conditions, so that the model can predict the customer's mediation needs under specific future economic environments. This prediction not only takes into account the customer's historical behavior, but also combines the impact of external economic factors on mediation requests to generate mediation demand forecasts under different scenarios.

[0096] Further, step S600 includes step S610 to step S640.

[0097] Step S610: Perform demand analysis based on the mediation effect score list and the prediction results, and obtain demand analysis results by comprehensively calculating the mediator's score and the corresponding customer demand. The demand analysis results include the service demand of each mediator in different time periods;

[0098] It should be noted that the mediator score reflects the mediator's professional ability and historical performance, while the customer demand is dynamically adjusted according to the peak and trough periods of mediation requests. Through weighted calculation, the customer demand is matched with the mediator's score to ensure that high-scoring mediators can be arranged during peak periods, optimizing the mediation success rate and customer satisfaction. The demand analysis results provide each mediator with a specific time period demand distribution, providing key data support for subsequent scheduling optimization.

[0099] Step S620: Generate a preliminary scheduling plan based on the demand analysis results and the availability, professional skills and historical performance of each mediator;

[0100] Understandably, this process maximizes mediation success and client satisfaction by prioritizing mediators with high scores and relevant skills during specific time periods, while ensuring reasonable resource allocation. For peak hours and complex situations, experienced and high-scoring mediators are prioritized to cope with high demand and challenging mediation tasks. The generated preliminary scheduling plan best matches the mediators' capabilities, availability, and needs, laying the foundation for subsequent optimization steps.

[0101] Step S630: defining optimization objectives based on a linear programming model, the optimization objectives including maximizing customer satisfaction and minimizing mediator load, and setting constraints in combination with the mediator's working hours and mediation success rate to obtain a fitness function;

[0102] It should be noted that the customer satisfaction target is directly related to the mediator's mediation success rate and response speed, while the minimization of the mediator's load ensures that the working hours are reasonably distributed and avoids excessive fatigue. Based on this, specific constraints are set, including daily and weekly working hours limits for mediators, the lower limit of mediation success rate, and the need for specific mediation skills during peak hours. These constraints ensure the quality of mediation services and a healthy working environment for mediators. After the goals and constraints are clear, a fitness function is constructed to quantify the pros and cons of each scheduling combination. The fitness function converts the comprehensive indicators of customer satisfaction and mediator load into optimizable values, providing a quantitative basis for the subsequent optimization process.

[0103] Step S640: Perform scheduling optimization processing according to the fitness function and the preliminary scheduling plan, and gradually optimize through the simulated annealing algorithm to obtain the final mediator scheduling plan.

[0104] Specifically, the preliminary scheduling scheme is input into the simulated annealing algorithm as the starting point. The algorithm allows a large range of exploration in the initial stage by setting the initial temperature. Even slightly inferior schemes have a certain probability of being accepted to prevent falling into the local optimum. In each iteration, the temperature is gradually lowered to reduce the probability of accepting inferior solutions, so as to gradually converge to the global optimal solution. After each scheme adjustment, the score of the scheduling scheme is recalculated by the fitness function to evaluate its performance in maximizing customer satisfaction and minimizing the mediator load. If the new scheme scores higher, it is directly accepted; if the score is slightly lower, it is decided whether to accept it according to the current temperature. After multiple rounds of iterations and adjustments, the optimal scheduling scheme that meets the requirements of the fitness function is finally obtained. The effect of this step is that through the gradual optimization of the simulated annealing algorithm, the scheduling scheme is ensured to achieve the optimal mediation resource allocation under different requirements and constraints, which improves the efficiency of mediation services and customer satisfaction, while avoiding excessive load on mediators.

[0105] Embodiment 2:

[0106] like Figure 2 As shown, this embodiment provides an intelligent call center resource optimization system based on predictive analysis, and the system includes:

[0107] Acquisition module 1 is used to acquire data of clients to be mediated, data of mediators and external economic indicators. The data of clients to be mediated includes historical mediation information and repayment information of clients, and the data of mediators includes background information of mediators and successful mediation cases;

[0108] Analysis module 2 is used to process the customer data to be mediated, analyze the text data of the customer in social media and customer service dialogue based on the preset sentiment analysis model, and construct a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility;

[0109] Evaluation module 3 is used to perform feature selection and quantitative evaluation based on mediator data. It uses correlation analysis to quantitatively evaluate the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate, and constructs a mediator's ability feature matrix;

[0110] Simulation module 4 is used to conduct scenario simulation based on external economic indicators, customer emotion feature matrix and mediator capability feature matrix. By identifying the weights of the impact of economy and emotion on mediation demand, combining chaos mapping theory to generate dynamic trajectory of customer emotion, and using deep reinforcement learning algorithm to simulate and evaluate the decision-making process of different mediators in different scenarios, a mediation effect score list is obtained;

[0111] Prediction module 5, used to perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain prediction results, including the number of mediation requests and customer behavior patterns within a preset time in the future;

[0112] The optimization module 6 is used to optimize the scheduling according to the mediation effect score list and the prediction results, and obtain the final mediator scheduling plan by optimizing the scheduling combination using the simulated annealing algorithm.

[0113] In a specific embodiment disclosed in the present application, the analysis module 2 includes:

[0114] The first extraction unit 21 is used to perform data extraction processing according to the customer data to be mediated to obtain original text data;

[0115] The first analysis unit 22 analyzes the original text data based on the BERT sentiment classification model, and obtains sentiment labeling results by identifying the sentiment category of each text and calculating the intensity of each sentiment;

[0116] The first calculation unit 23 is used to calculate the change frequency and intensity of the customer's emotions in the mediation cycle by introducing the volatility index according to the emotion labeling result, and using the sliding window method to obtain the emotion volatility feature;

[0117] The first integration unit 24 is used to perform data integration processing according to the emotion labeling results and the emotion volatility characteristics, and encode the integrated results to construct a customer emotion feature matrix.

[0118] In a specific implementation disclosed in the present application, the evaluation module 3 includes:

[0119] The second extraction unit 31 is used to perform feature extraction processing based on the mediator data, generate continuous features by mapping the professional field of the mediator with the average value of the mediation success rate, and convert the years of experience into categorical features by segmentation processing, so as to obtain a mediator feature set;

[0120] The second analysis unit 32 is used to perform correlation analysis based on the mediator feature set, evaluate the relationship between the mediator's professional field, years of experience, and case types handled and the mediation success rate by calculating mutual information, and combine the chi-square test to evaluate the independence of the feature and the mediation success rate to obtain the correlation analysis result;

[0121] The third analysis unit 33 is used to perform principal component analysis based on the correlation analysis results, extract the principal components to capture the main variability in the data by performing dimensionality reduction processing on the mediator features, and obtain the main feature set;

[0122] The first construction unit 34 is used to perform matrix construction processing according to the main feature set, and construct a mediator capability feature matrix by standardizing the features. The mediator capability feature matrix includes the numerical performance of each mediator on each feature.

[0123] In a specific implementation disclosed in the present application, the simulation module 4 includes:

[0124] The second construction unit 41 is used to perform scenario construction processing according to the external economic indicators and the customer emotion feature matrix, and integrate the market interest rate, inflation rate, customer emotion category and volatility by applying the weighted linear combination method, calculate the customer emotion impact weights under different economic scenarios, and obtain a comprehensive scenario feature matrix;

[0125] The first simulation unit 42 is used to perform simulation processing according to the comprehensive situational feature matrix, perform nonlinear dynamic simulation of customer emotions through chaos mapping theory, identify the urgency and sensitivity of customers to mediation services, and generate dynamic trajectories of customer emotions under different economic conditions;

[0126] The second simulation unit 43 is used to perform decision simulation according to the dynamic trajectory of customer emotions and the mediator's ability feature matrix, and use the dynamic trajectory of customer emotions as the environment state and the mediator's features as the input of the intelligent agent by using a policy gradient algorithm, and train the mediator's response strategy by reinforcement learning to obtain a decision simulation result;

[0127] The first evaluation unit 44 is used to evaluate the performance of each mediator in different economic and emotional situations according to the decision simulation results by using a fuzzy logic algorithm, convert the decision effect into a quantitative mediation effect score, and obtain a mediation effect score list.

[0128] In a specific implementation disclosed in the present application, the first simulation unit 42 includes:

[0129] The first screening unit 421 is used to perform variable screening processing according to the comprehensive situational feature matrix, calculate the contribution of market interest rate, customer repayment pressure and mediation history impact to customer sentiment change by applying information gain algorithm, and screen out a key variable set;

[0130] The first setting unit 422 is used to perform initial emotional state setting processing according to the key variable set, and generate the initial emotional state of each customer in a specific situation by combining the customer's historical emotional fluctuations, mediation success rate and current economic status through a nonlinear autoregressive model to obtain the customer's emotional initial state matrix;

[0131] The third simulation unit 423 is used to perform nonlinear dynamic simulation processing according to the initial state matrix of emotions, and recursively simulate the subtle fluctuations of customer emotions in a specific economic context by introducing fractal dimension analysis and combining fractal geometry methods to obtain the trajectory of customer emotions changes;

[0132] The third extraction unit 424 is used to generate and process the emotion fluctuation pattern according to the emotion change trajectory, extract the instantaneous fluctuation frequency and fluctuation intensity in the customer emotion trajectory through Hilbert-Huang transform, identify the urgency and sensitivity of the customer emotion to the mediation service, and finally obtain the dynamic trajectory of the customer emotion.

[0133] In a specific implementation disclosed in the present application, the prediction module 5 includes:

[0134] The third construction unit 51 is used to organize data based on the historical mediation information of all customers, and to construct a time series data set including the customer's credit status, repayment ability and mediation success rate by integrating the customer's mediation request records, repayment information and timestamps;

[0135] The first decomposition unit 52 is used to perform decomposition processing according to the time series data set, by decomposing the time series of the number of customer mediation requests into a trend component, a seasonal component and a residual, and smoothing the trend component by using a weighted moving average method to obtain a decomposed time series data set;

[0136] A first recognition unit 53 is used to perform behavior pattern recognition based on the decomposed time series data set, and obtain a behavior pattern recognition result by analyzing the pattern similarity of customer mediation requests under different economic environments;

[0137] The first prediction unit 54 is used to predict future requests based on the behavior pattern recognition result, and obtain the prediction result by analyzing the long-term dependency of customer behavior and predicting the behavior pattern of customers under different economic environments.

[0138] In a specific implementation disclosed in the present application, the optimization module 6 includes:

[0139] The fourth analysis unit 61 is used to perform demand analysis based on the mediation effect score list and the prediction result, and obtain demand analysis results by comprehensively calculating the mediator's score and the corresponding customer demand. The demand analysis results include the service demand of each mediator in different time periods;

[0140] A first generating unit 62 is used to generate a preliminary scheduling plan based on the availability, professional skills and historical performance of each mediator according to the demand analysis results;

[0141] A first definition unit 63 defines an optimization target based on a linear programming model, wherein the optimization target includes maximizing customer satisfaction and minimizing mediator load, and sets constraints in combination with the working hours and mediation success rate of the mediator to obtain a fitness function;

[0142] The first optimization unit 64 is used to perform scheduling optimization processing according to the fitness function and the preliminary scheduling plan, and gradually optimize through the simulated annealing algorithm to obtain the final mediator scheduling plan.

[0143] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An intelligent call center resource optimization method based on predictive analysis, characterized in that: include: Obtaining data of clients to be mediated, data of mediators, and external economic indicators, wherein the data of clients to be mediated includes historical mediation information and repayment information of clients, and the data of mediators includes background information of mediators and successful mediation cases; Processing the customer data to be mediated, analyzing the text data of the customer in social media and customer service conversations based on a preset sentiment analysis model, and constructing a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility; Perform feature selection and quantitative evaluation processing based on the mediator data, and quantitatively evaluate the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate by using correlation analysis to construct a mediator capability feature matrix; Scenario simulation is performed based on the external economic indicators, the customer emotion feature matrix, and the mediator capability feature matrix. By identifying the weights of the economic and emotional impacts on mediation needs, the dynamic trajectory of customer emotions is generated in combination with the chaos mapping theory. The decision-making process of different mediators in different scenarios is simulated and evaluated using a deep reinforcement learning algorithm to obtain a mediation effect score list. Perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain a prediction result, wherein the prediction result includes the number of mediation requests and customer behavior patterns within a preset time in the future; The shift scheduling is optimized according to the mediation effect score list and the prediction results, and the final mediator shift scheduling plan is obtained by optimizing the shift scheduling combination using a simulated annealing algorithm.

2. The intelligent call center resource optimization method based on predictive analysis according to claim 1 is characterized in that: According to the customer data to be mediated, the text data of the customer in social media and customer service dialogue is analyzed based on the preset sentiment analysis model, and the customer sentiment feature matrix is ​​constructed by identifying the customer's sentiment category and sentiment volatility, including: Performing data extraction processing according to the customer data to be mediated to obtain original text data; The original text data is analyzed based on the BERT sentiment classification model, and the sentiment labeling result is obtained by identifying the sentiment category of each text and calculating the intensity of each sentiment; According to the emotion annotation results, by introducing the volatility index, the changing frequency and intensity of customer emotions in the mediation cycle are calculated using the sliding window method to obtain the emotion volatility characteristics; Data integration processing is performed according to the emotion labeling results and the emotion volatility characteristics, and the integrated results are coded to construct a customer emotion feature matrix.

3. The intelligent call center resource optimization method based on predictive analysis according to claim 1 is characterized in that: Based on the mediator data, feature selection and quantitative evaluation are performed. By using correlation analysis, the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate are quantitatively evaluated to construct a mediator capability feature matrix, including: Performing feature extraction processing based on the mediator data, generating continuous features by mapping the mediator's professional field with the average mediation success rate, and converting the years of experience into categorical features by segmentation processing, thereby obtaining a mediator feature set; A correlation analysis is performed based on the mediator feature set, and the relationship between the mediator's professional field, years of experience, and type of cases handled and the mediation success rate is evaluated by calculating mutual information, and the independence of the features and the mediation success rate is evaluated by combining the chi-square test to obtain the correlation analysis results; According to the correlation analysis results, principal component analysis is performed to extract principal components to capture the main variability in the data by reducing the dimension of the mediator features, thereby obtaining the main feature set; A matrix construction process is performed based on the main feature set, and a mediator capability feature matrix is ​​constructed by standardizing the features. The mediator capability feature matrix includes the numerical performance of each mediator on each feature.

4. The intelligent call center resource optimization method based on predictive analysis according to claim 1 is characterized in that: A scenario simulation is performed based on the external economic indicators, the customer emotion feature matrix, and the mediator capability feature matrix. By identifying the weights of the impact of economy and emotion on mediation demand, the dynamic trajectory of customer emotion is generated in combination with the chaos mapping theory. A deep reinforcement learning algorithm is used to simulate and evaluate the decision-making process of different mediators in different scenarios, and a mediation effect score list is obtained, including: Scenario construction is performed based on the external economic indicators and the customer emotion feature matrix, and by applying a weighted linear combination method, the market interest rate, inflation rate, customer emotion category and volatility are integrated to calculate the customer emotion impact weights under different economic scenarios, thereby obtaining a comprehensive scenario feature matrix; According to the comprehensive situational feature matrix, simulation processing is performed, and nonlinear dynamic simulation of customer emotions is performed through chaos mapping theory to identify the urgency and sensitivity of customers to mediation services, and generate dynamic trajectories of customer emotions under different economic conditions; Perform decision simulation according to the customer's emotional dynamic trajectory and the mediator's ability feature matrix, use the customer's emotional dynamic trajectory as the environment state and the mediator's features as the input of the intelligent agent by using a policy gradient algorithm, and train the mediator's response strategy through reinforcement learning to obtain a decision simulation result; According to the decision simulation results, the fuzzy logic algorithm is used to evaluate the performance of each mediator in different economic and emotional situations, and the decision effect is converted into a quantitative mediation effect score to obtain a mediation effect score list.

5. The intelligent call center resource optimization method based on predictive analysis according to claim 4 is characterized in that: According to the comprehensive situational feature matrix, simulation processing is performed, and nonlinear dynamic simulation of customer emotions is performed through chaos mapping theory to identify the urgency and sensitivity of customers to mediation services, and generate dynamic trajectories of customer emotions under different economic conditions, including: Variable screening is performed according to the comprehensive situational feature matrix, and the contribution of market interest rates, customer repayment pressure, and mediation history impact to customer sentiment changes is calculated by applying an information gain algorithm to obtain a set of key variables; Performing initial emotional state setting processing according to the key variable set, generating each customer's initial emotional state in a specific situation by combining the customer's historical emotional fluctuations, mediation success rate and current economic status through a nonlinear autoregressive model, and obtaining the customer's emotional initial state matrix; According to the initial state matrix of emotion, nonlinear dynamic simulation processing is performed, and by introducing fractal dimension analysis and combining fractal geometry method, the subtle fluctuations of customer emotion in a specific economic situation are recursively simulated to obtain the trajectory of customer emotion change; Emotional fluctuation pattern generation processing is performed according to the emotion change trajectory, and the instantaneous fluctuation frequency and fluctuation intensity in the customer's emotional trajectory are extracted through Hilbert-Huang transform, the urgency and sensitivity of customer emotions to mediation services are identified, and finally the customer's emotional dynamic trajectory is obtained.

6. The intelligent call center resource optimization method based on predictive analysis according to claim 1 is characterized in that: Based on the historical mediation information of all customers, time series analysis and behavior pattern recognition are performed to obtain prediction results, including: Data is collated based on the historical mediation information of all customers. By integrating the customer's mediation request records, repayment information, and timestamps, a time series dataset containing the customer's credit status, repayment ability, and mediation success rate is constructed. Decomposing the time series data set, decomposing the time series of the number of customer mediation requests into a trend component, a seasonal component and a residual, and smoothing the trend component using a weighted moving average method to obtain a decomposed time series data set; Performing behavioral pattern recognition based on the decomposed time series data set, and obtaining behavioral pattern recognition results by analyzing the pattern similarity of customer mediation requests under different economic environments; Future request prediction is performed based on the behavior pattern recognition results, and the prediction results are obtained by analyzing the long-term dependence of customer behavior and predicting the behavior patterns of customers under different economic environments.

7. The intelligent call center resource optimization method based on predictive analysis according to claim 1 is characterized in that: The scheduling optimization process is performed according to the mediation effect score list and the prediction results, and the final mediator scheduling plan is obtained by optimizing the scheduling combination using the simulated annealing algorithm, including: Performing demand analysis based on the mediation effect score list and the prediction results, and obtaining demand analysis results by comprehensively calculating the mediator's score and the corresponding customer demand, wherein the demand analysis results include the service demand of each mediator in different time periods; Based on the needs analysis, generate a preliminary scheduling plan based on each mediator’s availability, expertise, and historical performance; Based on the linear programming model, the optimization goal is defined, wherein the optimization goal includes maximizing customer satisfaction and minimizing mediator load, and constraint conditions are set in combination with the working hours of the mediator and the mediation success rate to obtain a fitness function; The shift scheduling optimization process is performed according to the fitness function and the preliminary shift scheduling plan, and the final mediator shift scheduling plan is obtained by gradually optimizing through a simulated annealing algorithm.

8. An intelligent call center resource optimization system based on predictive analysis, characterized in that: include: An acquisition module, used to acquire data of clients to be mediated, data of mediators, and external economic indicators, wherein the data of clients to be mediated includes historical mediation information and repayment information of clients, and the data of mediators includes background information of mediators and successful mediation cases; An analysis module is used to process the customer data to be mediated, analyze the text data of the customer in social media and customer service dialogue based on a preset sentiment analysis model, and construct a customer sentiment feature matrix by identifying the customer's sentiment category and sentiment volatility; An evaluation module is used to perform feature selection and quantitative evaluation processing based on the mediator data, and to quantitatively evaluate the mediator's professional field, years of experience, types of cases handled in the past, and mediation success rate by using correlation analysis to construct a mediator capability feature matrix; A simulation module is used to perform scenario simulation based on the external economic indicators, the customer emotion feature matrix and the mediator capability feature matrix, generate customer emotion dynamic trajectories by identifying the weights of the impact of economy and emotion on mediation demand, and use the chaos mapping theory to simulate and evaluate the decision-making process of different mediators in different scenarios using a deep reinforcement learning algorithm to obtain a mediation effect score list; A prediction module, used to perform time series analysis and behavior pattern recognition based on the historical mediation information of all customers to obtain prediction results, wherein the prediction results include the number of mediation requests and customer behavior patterns within a preset time in the future; The optimization module is used to optimize the scheduling according to the mediation effect score list and the prediction results, and obtain the final mediator scheduling plan by optimizing the scheduling combination using the simulated annealing algorithm.

9. The intelligent call center resource optimization system based on predictive analysis according to claim 8, characterized in that: The analysis module comprises: A first extraction unit, configured to perform data extraction processing based on the customer data to be mediated to obtain original text data; A first analysis unit analyzes the original text data based on the BERT sentiment classification model, and obtains sentiment labeling results by identifying the sentiment category of each text and calculating the intensity of each sentiment; A first calculation unit is used to calculate the change frequency and intensity of the customer's emotions in the mediation cycle by introducing a volatility index according to the emotion labeling result, and using a sliding window method to obtain an emotion volatility feature; The first integration unit is used to perform data integration processing according to the emotion labeling results and the emotion volatility characteristics, and encode the integrated results to construct a customer emotion feature matrix.

10. The intelligent call center resource optimization system based on predictive analysis according to claim 8, characterized in that: The evaluation module includes: A second extraction unit is used to perform feature extraction processing based on the mediator data, generate continuous features by mapping the professional fields of the mediators with the average value of the mediation success rate, and convert the years of experience into categorical features by segmentation processing, so as to obtain a mediator feature set; A second analysis unit is used to perform a correlation analysis based on the mediator feature set, evaluate the relationship between the mediator's professional field, years of experience, and case types handled and the mediation success rate by calculating mutual information, and combine the chi-square test to evaluate the independence of the feature and the mediation success rate to obtain the correlation analysis result; A third analysis unit is used to perform principal component analysis based on the correlation analysis result, extract principal components to capture the main variability in the data by performing dimensionality reduction processing on the mediator features, and obtain a main feature set; The first construction unit is used to perform matrix construction processing according to the main feature set, and construct a mediator capability feature matrix by standardizing the features. The mediator capability feature matrix includes the numerical performance of each mediator on each feature.

Images