Dynamic scheduling method and device for massive call requests and medium

Abstract

The invention discloses a dynamic scheduling method and device for mass call requests and a medium, and relates to the technical field of information management, and the method comprises the steps: receiving mass call request information through an information sending device, carrying out the language processing to generate an analysis data set, and obtaining the request types through deep learning classification; performing multi-objective optimization on the analysis data set to generate a dynamic call queue; constructing a time sequence prediction model based on historical records, predicting a call peak, carrying out resource scheduling analysis, and formulating a scheduling strategy; and generating a feedback result after executing the scheduling strategy, synchronizing the feedback result to the call center, and updating the strategy according to response information to realize intelligent dynamic scheduling. The technical problem that in the prior art, the resource scheduling efficiency of the call center in the mass request processing process is low is solved, and the technical effect of optimizing the resource scheduling efficiency of the call center is achieved.

Description

Technical Field

[0001] The present application relates to the field of information management technology, and in particular to a dynamic scheduling method, device and medium for massive call requests. Background Art

[0002] With the rapid growth of communication volume, call centers face challenges in data complexity and resource allocation when processing massive call requests. Traditional methods rely on fixed rules for scheduling, but it is difficult to process large-scale requests in a timely manner during peak hours or complex scenarios, which can easily lead to request backlogs or resource waste. At the same time, the diversity and uncertainty of call requests have put forward higher requirements on the real-time response capabilities of existing scheduling systems, especially in the optimization of request processing order and resource allocation, where traditional methods are insufficient. In this context, how to respond to diverse requests, allocate limited resources and improve overall processing efficiency has become a key issue that needs to be solved.

[0003] In the current related technologies, there is a technical problem of low resource scheduling efficiency in the process of handling massive requests in call centers. Summary of the invention

[0004] The present application solves the technical problem of low resource scheduling efficiency in the process of massive request processing in the call center in the prior art by providing a dynamic scheduling method, device and medium for massive call requests.

[0005] The present application provides a dynamic scheduling method for massive call requests, including: Information is received through a multi-party information sending device to obtain massive call request information; language processing is performed based on the massive call request information to generate a request parsing data set, and the request parsing data set is classified by deep learning to obtain multiple request types; multi-objective optimization is performed on the request parsing data set according to the multiple request types to generate multiple dynamic call queues; historical call record logs are retrieved to build a time series prediction model, and prediction is performed through the time series prediction model to determine call request peak prediction data; the multiple dynamic call queues are mapped to the call request peak prediction data to perform resource scheduling analysis and formulate scheduling strategies; the scheduling strategies are executed to process feedback and generate scheduling feedback results, the scheduling feedback results are synchronized to the call center for response, and the scheduling strategies are updated according to the response information to perform intelligent dynamic scheduling on the massive call request information.

[0006] The present application also provides an electronic device, including: The memory is used to store executable instructions; the processor is used to implement a dynamic scheduling method for massive call requests when executing the executable instructions stored in the memory.

[0007] The present application also provides a computer-readable storage medium, comprising: A computer program is stored thereon, and when the program is executed by a processor, a dynamic scheduling method for massive call requests is implemented.

[0008] The present application proposes a dynamic scheduling method, device and medium for massive call requests. First, massive call request information is received through an information sending device, language processing is performed to generate a parsed data set, and the request type is obtained through deep learning classification; multi-objective optimization is performed on the parsed data set to generate a dynamic call queue; a time series prediction model is constructed based on historical records to predict call peaks and perform resource scheduling analysis to formulate a scheduling strategy; feedback results are generated after the scheduling strategy is executed, synchronized to the call center and the strategy is updated according to the response information to realize intelligent dynamic scheduling, thereby achieving the technical effect of optimizing the resource scheduling efficiency of the call center. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0010] Figure 1 A flowchart of a dynamic scheduling method for massive call requests provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0011] Description of the reference numerals: input device 301 , processor 302 , memory 303 , output device 304 . DETAILED DESCRIPTION

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0013] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0014] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0015] The present application embodiment provides a dynamic scheduling method for massive call requests, such as Figure 1 As shown, the method includes: Step S100, receiving information through a multi-party information sending device to obtain massive call request information. Specifically, the multi-party information sending device receives massive call request information from multiple channels such as telephone networks, online platforms, mobile applications, etc. in real time by monitoring multiple communication protocols (such as SIP protocol for voice requests and HTTP protocol for network requests), and generates a source tag, time tag and format tag for each message to indicate the source and type of the information. The requests received by the device include voice, text and multimedia content, and are stored through a data cache module, using distributed storage technology to ensure data integrity and efficient management. At the same time, the received call requests are pre-processed, including denoising, filtering invalid or duplicate requests, and converting requests in multiple formats into a unified standard format, providing a high-quality data foundation for subsequent language processing, request parsing and dynamic scheduling.

[0016] Step S200, based on the massive call request information, language processing is performed to generate a request parsing data set, and the request parsing data set is classified by deep learning to obtain multiple request types. Specifically, for the massive call request information, the text information and voice information are first separated and processed, and the voice information is converted into text data by voice-to-text technology. Then, the natural language processing (NLP) technology is used to extract semantic features and context information from the text data, and the sentiment analysis algorithm is used to extract the emotional state (such as anger, anxiety, calmness, etc.) from the voice data. These semantic features, emotional features and context-related factors are integrated to generate a request parsing data set. Based on this data set, a deep learning algorithm (such as CNN or LSTM) is used to classify the request information, identify specific request types (such as technical support, business consultation, complaints and suggestions, etc.), and assign category labels and weights to each request to provide precise support for subsequent dynamic scheduling and resource matching.

[0017] In one possible implementation, language processing is performed based on the massive call request information to generate a request parsing data set, deep learning classification is performed on the request parsing data set, and multiple request types are obtained. Step S200 further includes step S210, performing text processing based on the massive call request information to obtain a text feature set. Specifically, the text content in the massive call request information is preprocessed, including removing noise data and irrelevant characters, such as special symbols and HTML tags; using natural language processing technology to perform word segmentation and part-of-speech tagging to extract meaningful keywords; and removing stop words to reduce the interference of irrelevant data and enhance the significance of key information. Further, methods such as TF-IDF or Word2Vec are used to calculate the importance of words, generate feature vectors that describe the text topic, and combine dependency syntax or semantic analysis to capture logical relationships in the context. Finally, the extracted text features are converted into a structured vector format that can be used for machine learning to form a complete text feature set, providing basic support for subsequent request parsing and classification.

[0018] Step S220, context analysis is performed based on the text feature set to capture text correlation factors. Specifically, by performing context analysis on the text feature set, first use natural language processing technology (such as dependency syntax analysis or BERT model) to build a context relationship graph and extract semantic dependencies within the sentence. Subsequently, the context window technology is used to analyze the semantic connection between sentences within a multi-sentence range, and the strength of the correlation is quantified by the repetition, reference relationship and semantic similarity of keywords. At the same time, a clustering algorithm is used to aggregate the correlation factors, and different weights are given to keywords in combination with the semantic weight model to construct a multi-level correlation feature model. Finally, the captured text correlation factors are represented as a weighted feature matrix, which serves as the basic input for subsequent emotion recognition and intent recognition, providing semantic support for dynamic scheduling.

[0019] Step S230, based on the massive call request information, voice processing is performed to obtain a voice feature set. Specifically, the voice signals of the massive call requests are preprocessed, including noise suppression, signal enhancement, and silent segment removal, to ensure the quality of the voice signal. Subsequently, the time-frequency characteristics of the voice are analyzed by short-time Fourier transform (STFT), and features such as spectrogram, Mel-frequency cepstral coefficient (MFCC), fundamental frequency, and pitch change are extracted to capture the timbre, intonation, and rhythm of the voice. Combined with speech recognition technology (ASR), the voice is converted into text to provide support for semantic analysis. At the same time, the emotion recognition model is used to analyze the emotional characteristics of the voice, such as the anger, calmness, or anxiety of the tone. Finally, all features are integrated into a high-dimensional voice feature vector to provide data support for subsequent intent recognition and scheduling analysis.

[0020] Step S240, emotion recognition is performed according to the speech feature set to capture speech emotion factors. Specifically, emotion recognition is performed according to the speech feature set. First, the key features of the speech are extracted, including fundamental frequency changes, energy intensity, speech rate rhythm, and spectrogram features. Preprocessing and dimensionality reduction are used to remove redundant information and optimize feature expression. Subsequently, the processed speech features are input into the emotion classification model, and the support vector machine (SVM) or deep learning model (such as CNN combined with LSTM) is used to classify the emotional state and identify factors such as emotion categories (such as anger, joy, sadness) and emotion intensity in the speech. Finally, the extracted emotion factors are integrated into high-dimensional vectors, providing key support for subsequent user intent recognition and resource scheduling optimization.

[0021] Step S250, normalize the text association factor and the voice emotion factor to generate a processing result, perform intent recognition on the massive call request information based on the processing result, and generate multiple request intent labels. Specifically, the text association factor and the voice emotion factor are normalized to ensure that different features are in the same numerical range and eliminate the impact of feature weight differences. The normalized feature matrix is ​​input into the intent recognition model, and a deep learning algorithm is used to combine text context and voice emotion features to identify the user's main intentions, such as service requests, complaints or consultations, and generate a confidence score for each intent label. Based on the scoring results, high-confidence intent labels are screened out and added to the request parsing data set to provide support for subsequent classification and dynamic scheduling.

[0022] Step S260, adding the multiple request intent tags to the request resolution data set. Specifically, when adding the generated multiple request intent tags to the request resolution data set, first perform feature analysis and filtering on the intent tags to ensure their accuracy and representativeness; then update the structure of the request resolution data set so that it can adapt to the newly added tags. By binding each intent tag to the corresponding request data, improving the recorded feature information, and verifying the consistency of the data, conflicting or duplicate tags are filtered out. Finally, the updated data set is stored in the database to provide comprehensive and accurate data support for subsequent classification, optimization and scheduling links.

[0023] In a possible implementation, language processing is performed based on the massive call request information to generate a request parsing data set, and deep learning classification is performed on the request parsing data set to obtain multiple request types. Step S200 further includes step S270, performing impact analysis on the massive call request information according to the multiple request intent tags to generate multiple impact levels. Specifically, by analyzing multiple request intent tags, the massive call request information is parsed one by one, and key attributes related to the request are extracted, such as urgency (based on completion time requirements), resource consumption (such as the number of seats or hardware requirements), service priority (such as the distinction between VIP users and ordinary users), and user expectations (such as historical service evaluation or satisfaction). Subsequently, the actual value ranges of these impact factors are standardized so that they are within the same quantitative scale for subsequent analysis. For example, the urgency can be divided into high, medium, and low levels, and resource consumption is graded according to complexity. The standardized impact factors are dynamically divided into different priority levels according to importance, for example, requests with urgent time and large resource requirements are marked as high priority, while requests with loose time and low resource requirements are marked as low priority. This division is dynamically adjusted based on historical data and business needs to ensure that the classification results accurately reflect the importance of the request, providing a reliable basis for subsequent clustering analysis and resource allocation.

[0024] Step S280, cluster analysis is performed on the request resolution data set based on the multiple impact levels, multiple data clustering centers are determined, and the request resolution data set is divided into multiple resolution data clusters according to the multiple data clustering centers. Specifically, by analyzing each record in the request resolution data set, the request is marked as high priority, medium priority or low priority according to the impact level, and key attributes such as response time requirements, resource consumption and user type are extracted as cluster feature vector input. Using a clustering algorithm (such as K-means or DBSCAN), the request is divided into multiple cluster candidate groups according to the similarity of data features, and the cluster center of each group is determined. These centers reflect the common characteristics of the corresponding category requests. In the clustering process, the clustering parameters are dynamically adjusted to adapt to the data distribution, and a soft allocation strategy is adopted for boundary points or data that is difficult to classify to enhance flexibility. Finally, the request resolution data set is divided into multiple resolution data clusters, each cluster corresponds to a cluster center, and a request set with specific features is identified, providing a basis for subsequent resource scheduling and optimization strategies.

[0025] Step S290, the massive call request information is learned and allocated according to the multiple parsed data clusters in combination with the multiple request intent labels, and multiple weight coefficients are generated. Specifically, the association between the parsed data clusters and the request intent labels is achieved through feature matching. Each data cluster contains requests with similar features. By analyzing features such as priority and service type, the matching strength with the request intent labels is evaluated. Combined with historical call records, a supervised learning model is used for learning allocation to establish a weight relationship between the parsed data clusters and the request intent labels. The weight coefficient reflects the priority of each label in the data cluster. The weight coefficient output by the model is optimized and adjusted to ensure allocation accuracy and resource scheduling efficiency, thereby providing targeted processing solutions for different types of call requests.

[0026] Step S2100, the request parsing data set is processed in descending order according to the multiple weight coefficients to determine the request data sequence. Specifically, each request data in the request parsing data set is sorted in descending order according to the generated weight coefficient, and the request data with higher priority is located at the front of the sequence, so as to clarify the processing priority of the request. After sorting, priority labels are assigned to the request data in turn, and requests with high weight coefficients are marked as high priority, and requests with low weight coefficients are marked as low priority, to ensure that the classification of requests is clear. Subsequently, the sorted request data is arranged according to priority to generate a request data sequence sorted from high to low, providing a clear processing order for subsequent resource allocation and processing. Finally, the generated request data sequence is checked for consistency to ensure the accuracy of the sorting results, and re-sorted by adjusting the weight coefficients when necessary to output a stable and reliable request data sequence.

[0027] Step S2110, segment the request data sequence according to the multiple parsed data clusters, and determine the multiple request types. Specifically, by matching the parsed data cluster with the request data sequence, each request data in the request data sequence is compared with the central feature of the parsed data cluster to determine its affiliation, ensuring that each data is assigned to the most matching cluster. The segmentation is based on the central feature of the parsed data cluster, and the data is divided by similarity judgment to ensure classification accuracy. After the segmentation is completed, the request type represented by each cluster is determined according to the overall feature label of the parsed data cluster, such as a high-priority technical support request or a common consultation request. Subsequently, the segmentation results are checked for consistency, and the data features of the same request type are analyzed to see if they are consistent. If necessary, the segmentation is adjusted to ensure that the definition of each request type is clear and representative. Finally, the segmentation results and request type classification are output to provide clear basic data support for subsequent resource scheduling and policy formulation.

[0028] Step S300, multi-objective optimization is performed on the request parsing data set according to the multiple request types to generate multiple dynamic call queues. Specifically, in the process of multi-objective optimization of the request parsing data set according to the multiple request types, the optimization goals are first set according to different request types, such as reducing the waiting time of high-priority requests or improving resource utilization, and the request parsing data is encoded to mark the request type, priority and other information. Then, by analyzing the conflicts between different goals, the importance of each goal is defined using weight balance calculation, and a multi-objective optimization processing strategy is established. Subsequently, a multi-objective optimization algorithm (such as genetic algorithm or particle swarm optimization) is used for iterative calculation to generate an initial call queue that meets the optimization goals. For the generated queues, compatibility analysis is further performed to detect whether there are resource conflicts or queue overlap problems, and the problem queues are dynamically adjusted and optimized to ensure the compatibility and rationality of the queues. Finally, multiple dynamic call queues are output, each of which contains request priority and resource allocation plan to meet the request processing requirements in a dynamic environment.

[0029] In a possible implementation, the request parsing data set is multi-objectively optimized according to the multiple request types to generate multiple dynamic call queues, and step S300 further includes step S310, traversing the request parsing data set according to the multiple request types to encode and generate multiple encoded data. Specifically, according to the request type in the request parsing data set, the data is traversed and category features (such as priority, resource type, processing time period, etc.) are extracted, and these features are encoded according to predefined rules. For example, the priority can be encoded as high (1), medium (2), and low (3), and the resource type can be encoded as voice (101), text (102), etc. The consistency of feature conversion is ensured during the encoding process, so that each request information is digitally represented as encoded data. The generated encoded data is then stored in a structured form (such as a table or matrix) to provide efficient and clear input support for subsequent multi-objective processing and dynamic queue generation.

[0030] Step S320, analyzing the massive call request information based on the multiple coded data, and setting multiple processing targets. Specifically, based on the multiple coded data, the massive call request information is deeply analyzed to extract key features, such as priority, request type, resource requirements and processing time. Different processing requirements of requests are identified by classification, for example, high-priority requests require quick responses, while low-priority requests can be processed with appropriate delays. At the same time, multiple processing targets are set according to the request characteristics, including response time, resource utilization efficiency, queue balance and user satisfaction. According to the request types in different scenarios, the target weights are dynamically adjusted, such as giving priority to high-priority requests in emergency assistance or optimizing resource allocation efficiency to avoid backlogs. Finally, these targets are standardized into parameter templates to provide a basis for the formulation of multi-objective optimization and dynamic scheduling strategies.

[0031] Step S330, performing processing balance calculations according to the multiple processing targets, and determining the multi-target processing balance ratio. Specifically, according to the set multiple processing targets (such as response time, resource utilization efficiency, queue balance, etc.), firstly, the importance of each target is prioritized, and through historical data and system demand analysis, it is determined that high-priority targets are assigned higher weights, and secondary targets are assigned lower weights. Subsequently, by quantitatively analyzing the deviation between the actual value and the expected target value, combined with the weight of each target, the multi-target processing balance ratio is calculated to ensure that the deviation of the high-priority target is optimized first. Finally, through the dynamic adjustment mechanism, the balance ratio is updated in real time according to the system operation status, for example, the response time weight is prioritized during peak hours, and the resource utilization efficiency is optimized when resources are idle, so as to achieve a dynamic balance between multiple targets and meet the comprehensive needs of request processing efficiency and resource utilization.

[0032] Step S340, iteratively evaluate the request data sequence according to the multi-objective processing balance ratio until the request data sequence meets the multi-objective processing balance ratio, and determine multiple initial call queues. Specifically, by iteratively evaluating the request data sequence, according to the set multi-objective processing balance ratio (such as response time, resource utilization efficiency, etc.), the request sorting and allocation strategy in the sequence is gradually optimized. First, load the initial request data sequence and set the evaluation parameters according to the multi-objective balance ratio. Through traversal and dynamic adjustment algorithms, calculate the fitness of each request and reorder the requests that do not meet the ratio. After each round of iteration, combine the dynamic weight adjustment and real-time evaluation results to determine whether the request data sequence reaches the set ratio. If the deviation is less than the preset threshold, the iteration is terminated and the initial call queue is generated. Finally, according to the optimized sequence, multiple initial call queues are divided to ensure task load balancing and provide an efficient basis for subsequent resource scheduling.

[0033] Step S350, perform call compatibility analysis on the multiple initial call queues, generate queue compatibility coefficients, adjust and optimize the multiple initial call queues according to the queue compatibility coefficients, and generate the multiple dynamic call queues. Specifically, perform call compatibility analysis on the initial call queues, judge the compatibility between queues by matching the time windows, resource requirements and priority information of the tasks, and mark the conflicting task pairs as "incompatible combinations". Based on the conflict statistics and priority impact, calculate the compatibility coefficient of each queue to quantify the degree of compatibility. Adjust the queues according to the compatibility coefficients, reduce task conflicts and improve the compatibility of queues by reallocating high-conflict tasks or optimizing resource allocation. After iterative optimization, generate dynamic call queues to ensure that the queues are balanced in resources and time, while improving the efficiency and orderliness of task processing.

[0034] In one possible implementation, call compatibility analysis is performed on the multiple initial call queues to generate queue compatibility coefficients, and the multiple initial call queues are adjusted and optimized according to the queue compatibility coefficients to generate the multiple dynamic call queues. Step S350 further includes step S351, based on the multiple initial call queues, random combinations are performed in pairs to generate multiple queue combinations. Specifically, based on the multiple initial call queues, queue combinations are generated in a pairwise random combination manner. Specifically, two queues are randomly selected from the initial queues, their task sets are merged to form a new queue combination, and each combination is assigned a unique identifier to record the source queue and the combination method. By repeating this process, a sufficient number of queue combinations are generated to provide a diverse data basis for subsequent call overlap detection and compatibility analysis, while maintaining the flexibility of task allocation and the independence of the initial queues.

[0035] Step S352, analyze whether there is call overlap among the multiple queue combinations. If there is call overlap among the multiple queue combinations, the multiple queue combinations with call overlap are identified as call incompatible combinations. Specifically, analyze the multiple queue combinations generated one by one to check whether there is overlap in the time periods of call tasks in each queue combination. If it is detected that the same task is repeatedly assigned to different queues, the queue combination is marked as a call incompatible combination, and the detailed information of the overlapping tasks is recorded, including the task identification, time period and related queue combinations. The analysis results are stored in the compatibility database to provide a basis for subsequent splitting and reorganization, thereby optimizing the allocation of call tasks and ensuring the uniqueness of task allocation and efficient use of resources.

[0036] Step S353, a call occupation instruction is generated according to the call incompatible combination, and multiple queue combinations with call overlap are split and reorganized through the call occupation instruction to generate multiple reorganized queue combinations. Specifically, for the identified call incompatible combination, the system first generates a call occupation instruction, extracts the specific information of the conflicting tasks and related queues, including task time periods, priorities and resource requirements. According to the instruction, the conflicting tasks in the incompatible queues are split out, and these tasks are processed and temporarily stored separately to ensure that the conflicts are isolated. Subsequently, the split tasks are reorganized and optimized in combination with the task priority, time period and resource requirements to generate multiple new queue combinations. The reorganization process adopts a resource matching algorithm to ensure reasonable and efficient allocation while avoiding task duplication or omission. Finally, the compatibility of the generated reorganized queue is verified to ensure that all conflicts have been resolved, and the optimized queue solution is output for subsequent resource scheduling.

[0037] Step S354, synchronize the multiple reorganized queue combinations to the queue compatibility coefficient. Specifically, after completing the splitting and reorganization of incompatible queues, the system analyzes the generated reorganized queue combination, extracts its key feature information, including the number of tasks, priority distribution, resource requirements and time allocation, etc., and performs compatibility evaluation on each combination to calculate the compatibility score. The evaluation results are combined with the conflict elimination effect of the reorganized queue and the resource relationship between tasks to ensure that the compatibility evaluation is accurate. Subsequently, the compatibility score is synchronized to the queue compatibility coefficient module, and the system compatibility database is updated to ensure real-time update and storage in the queue management database, providing accurate support for subsequent resource scheduling and strategy optimization.

[0038] Step S400, retrieve the historical call record log to build a time series prediction model, perform predictions through the time series prediction model, and determine the call request peak prediction data. Specifically, the system retrieves the historical call record log, including data such as call time, duration, agent allocation, user category and request type, and cleans and normalizes it to ensure data consistency. By extracting time-related features (such as periodicity and holiday effects), a time series prediction model (such as ARIMA or LSTM) is constructed, and the model is supervised and verified using training data to improve prediction performance. With the support of real-time data, the model predicts the call request trend for a period of time in the future, and outputs peak prediction data including peak time period, request volume and request type distribution, to provide support for resource allocation and scheduling strategy optimization, and store the prediction results for subsequent feedback and optimization.

[0039] Step S500, mapping the multiple dynamic call queues to the call request peak forecast data for resource scheduling analysis and formulating scheduling strategies. Specifically, for multiple dynamic call queues, first extract their core attributes, such as request type, priority, processing time requirements, etc., and map them to the call request peak forecast data to establish a correspondence between queues and peak hours. Subsequently, perform resource demand analysis based on the mapping results, evaluate the number of seats, skill types and equipment configurations required for each queue, identify resource matching and potential shortage or surplus problems. Combined with the queue priority, set scheduling rules, give priority to emergency requests, and dynamically adjust resource allocation strategies to maximize resource utilization and avoid backlogs. By comprehensively considering service timeliness and customer satisfaction, formulate specific scheduling strategies, including adjusting seat scheduling, optimizing queue processing order, etc., and output the strategies to the execution system. At the same time, monitor the scheduling execution effect, optimize the model based on real-time feedback, and continuously improve the accuracy and adaptability of the scheduling strategy.

[0040] In one possible implementation, the multiple dynamic call queues are mapped to the call request peak forecast data to perform resource scheduling analysis and formulate a scheduling strategy. Step S500 further includes step S510, performing request analysis based on the multiple dynamic call queues in combination with the multiple request types to determine multiple request attributes. Specifically, by analyzing the association between the multiple dynamic call queues and the request types, the key feature information of the request is extracted, and request attributes such as priority, complexity, skill requirements and time constraints are defined. By analyzing the request content, historical records and semantic features, these attribute dimensions are identified and mapped to generate a request set containing complete attributes. For example, the request type is determined by keyword extraction, the priority is inferred by combining the queue historical records, the skill requirements are annotated by semantic analysis, and the time constraints are identified according to system rules. After the attribute set is generated, the unclear attributes are supplemented through model verification and global consistency adjustment to ensure the completeness and accuracy of the request attributes, laying the foundation for subsequent resource scheduling.

[0041] Step S520, perform call resource analysis according to the call request peak prediction data and the request data sequence to determine the agent resource status prediction data. Specifically, based on the call request peak prediction data, extract the time series information related to the call demand, such as the call volume distribution and the proportion of request types, and combine the priority, skill requirements and processing time limit and other characteristics in the request data sequence to parse the specific request requirements. Retrieve the call center's agent resource data, including the number of agents, skill distribution, work status and historical efficiency, and perform preliminary matching. Predict the future utilization of agent resources through time series analysis methods, generate agent resource status prediction data, including possible idle time, resource shortage time period and key deployment nodes, and output the integrated prediction table to provide data support for the formulation of resource scheduling strategies.

[0042] Step S530, performing resource matching according to the multiple request attributes in combination with the agent resource status prediction data, and obtaining multiple resource matching pairs. Specifically, according to the multiple request attributes and the agent resource status prediction data, firstly analyze the type, priority, skill requirement and processing time limit of the request, and combine the status information such as the skill distribution, idle time, and work efficiency of the agent to set matching rules, including skill matching, priority matching and load balancing. Subsequently, an optimization algorithm is used to match each request with the most suitable agent resource, generate multiple resource matching pairs, and verify the matching results to ensure that high-priority requests are allocated to the best resources, while avoiding low-priority requests from occupying key resources. The resource matching pairs finally output provide support for the efficient allocation of dynamic call queues.

[0043] Step S540, based on mapping the multiple resource matching pairs with the multiple dynamic call queues, determine multiple priority allocation queues. Specifically, based on the association analysis of the multiple resource matching pairs with the dynamic call queues, firstly accurately match the request attributes in each call queue with the agent characteristics (such as skills, idle status, etc.) in the resource matching pair, and realize dynamic mapping through the resource allocation algorithm to ensure the optimal association between resources and queue requests. In the mapping process, according to the urgency of the request, the priority level and business rules, the association between resources and queues is dynamically adjusted to give priority to high-priority requests while maximizing the overall resource utilization efficiency. Finally, a priority allocation queue is generated according to the mapping results, and the requests are processed in order of priority. The high-priority queue is responded first, and the low-priority queue enters the standby state, ensuring the efficient configuration of call resources and the overall improvement of service quality.

[0044] Step S550, record resource scheduling according to the multiple priority allocation queues and generate a resource scheduling trajectory information set. Specifically, when recording resource scheduling according to the priority allocation queues, first analyze the request attributes (such as priority and urgency) in each queue and the status of currently available resources (such as agent skills and idle time), complete resource allocation one by one, and record key information of the allocation process in real time, including the source of the request, the allocated resources, and the external conditions during scheduling. Subsequently, these scheduling logs are integrated into a resource scheduling trajectory information set to record the dynamic change data of resource allocation (such as queue adjustment and completion status) to ensure that it has time series traceability and analysis value. This trajectory information set not only fully records the execution of the scheduling strategy, but also provides a basis for subsequent strategy optimization. By analyzing indicators such as scheduling efficiency and resource utilization, more accurate and efficient resource scheduling management can be achieved.

[0045] Step S560, adding the resource scheduling trajectory information set to the scheduling strategy. Specifically, the process of adding the resource scheduling trajectory information set to the scheduling strategy includes: first, organizing and parsing the trajectory information set, extracting key data points such as resource allocation efficiency, response time and unprocessed reasons, etc., as the basis for strategy optimization. Next, compare the trajectory information with the existing scheduling strategy, analyze the shortcomings of the current strategy, and adjust key parameters accordingly, such as reallocating queue priorities or optimizing resource matching rules. Subsequently, update the policy logic through iteration to make it more adaptable to the actual operation needs of the call center, and update it synchronously to the system to ensure that the application is effective. Through this process, the scheduling strategy can be dynamically optimized to improve resource utilization and response efficiency.

[0046] Step S600, execute the scheduling strategy to process feedback, generate scheduling feedback results, synchronize the scheduling feedback results to the call center for response, and update the scheduling strategy according to the response information to perform intelligent dynamic scheduling on the massive call request information. Specifically, the process of executing the scheduling strategy includes that the system performs resource scheduling on the massive call request information according to the optimized scheduling strategy, dynamically allocates seat resources and priority processing queues, and generates feedback results in real time, and records indicators such as processing success rate, number of unprocessed requests and resource utilization. The generated scheduling feedback results are synchronized to the call center, and the request processing status and system response speed are displayed in an intuitive visual interface, which is convenient for managers to evaluate the scheduling effect. The call center analyzes the problems in the processing based on the feedback results, such as insufficient resource allocation or request processing delay, and collects seat manual response data and abnormal reports to provide a basis for scheduling strategy optimization. Finally, the system dynamically adjusts the strategy according to the feedback information, such as optimizing priority settings, resource allocation ratios or adding new processing resources, and the updated strategy is put into real-time operation to ensure more efficient response to dynamic changes in call requests.

[0047] In a possible implementation, the scheduling strategy is executed to process feedback, generate a scheduling feedback result, synchronize the scheduling feedback result to the call center for response, and update the scheduling strategy according to the response information to perform intelligent dynamic scheduling on the massive call request information. Step S600 further includes step S610, executing the scheduling strategy to perform resource scheduling on the massive call request information to obtain a resource scheduling result. Specifically, when executing the scheduling strategy to perform resource scheduling on the massive call request information, the system first parses the scheduling strategy, determines key parameters such as queue priority and resource allocation rules, and initializes the resource scheduling module. Subsequently, according to the priority, content type and urgency of the request, the resource allocation rules in the scheduling strategy are matched with the call request information to determine the best resource allocation plan. After the matching is completed, the system dynamically allocates resources according to the scheduling strategy to ensure that high-priority requests are processed in a timely manner while maintaining a balanced allocation of system resources. During the process, the system monitors the resource usage status and queue load in real time. If overload or idle resources occur, adjustments and optimizations will be made according to the strategy rules. Finally, the system generates resource scheduling results, recording in detail the allocation path, estimated processing time, and response status of each request, providing data support for subsequent feedback and optimization.

[0048] Step S620, make a resource allocation judgment based on the resource scheduling result to determine whether the resource scheduling result meets the expected scheduling threshold. Specifically, after the resource scheduling is completed, the system inputs the generated resource scheduling result into the judgment module, extracts key indicators such as the processing path, resource allocation and response time of each call request, and compares them with the preset expected scheduling threshold. The expected scheduling threshold includes indicators such as response time, resource utilization and priority satisfaction. The system checks item by item whether the scheduling result meets the requirements, such as whether it is within the response time range, whether the resource allocation is efficient, and whether there is no priority conflict. After comparison, if all conditions are met, the scheduling is determined to be successful; if there are unsatisfied items, such as timeout or resource conflict, it is recorded as a failure and a detailed reason report is generated. The judgment result is synchronized to the feedback module for generating positive feedback or negative feedback signals, and storing unsatisfied records to support subsequent scheduling strategy optimization.

[0049] Step S630, if the resource scheduling result meets the expected scheduling threshold, the resource scheduling result satisfies the massive call request information, generates a positive feedback signal, and adds the positive feedback signal to the scheduling feedback result. Specifically, when the resource scheduling result is verified by the resource allocation determination module and meets the preset expected scheduling threshold, the system confirms that the current scheduling scheme successfully meets the processing requirements of the massive call request information. The scheduling system marks this successful resource allocation as positive feedback, specifically including recording key information such as the efficiency of resource matching, the compliance of response time, and the correct processing of priority. The system then generates a positive feedback signal to identify the correctness and feasibility of this resource scheduling result. The positive feedback signal is added to the scheduling feedback result as a historical basis for optimizing the scheduling strategy, while updating core parameters such as resource utilization and response speed, providing reference data for subsequent scheduling processes, thereby further improving the overall resource scheduling efficiency of the system.

[0050] Step S640, if the resource scheduling result does not meet the expected scheduling threshold, the resource scheduling result does not meet the massive call request information, a negative feedback signal is generated, and the negative feedback signal is added to the scheduling feedback result. Specifically, when the resource scheduling result is verified by the resource allocation determination module, if the preset expected scheduling threshold is not reached, it indicates that the current resource scheduling fails to effectively meet the processing requirements of the massive call requests. At this time, the system will analyze the deficiencies in the scheduling results, such as uneven resource allocation, long response time or priority processing errors. Based on the analysis results, the system generates a negative feedback signal to clearly mark the deficiencies in the execution of this scheduling strategy. The negative feedback signal is recorded and added to the scheduling feedback result for subsequent optimization of the scheduling strategy. The addition of the negative feedback signal not only provides the system with a basis for further analyzing scheduling anomalies, but also provides data support for real-time adjustment of resource allocation strategies. By summarizing the negative feedback signal, the system can gradually locate the source of the problem, optimize the scheduling rules, and finally realize the dynamic adjustment of resource configuration, thereby gradually improving the scheduling efficiency and the accuracy of call request processing.

[0051] Step S650, if the scheduling feedback result is the negative feedback signal, the resource scheduling result is synchronized to the call center for scheduling abnormality alarm, a scheduling troubleshooting instruction is generated, and the scheduling troubleshooting instruction is added to the response information. Specifically, when the scheduling feedback result contains a negative feedback signal, the system synchronizes the resource scheduling result to the call center in real time to warn of the abnormal situation. This process transmits the resource allocation situation that does not meet the expected scheduling threshold to the management module of the call center through the scheduling abnormality alarm mechanism. Subsequently, the system generates a detailed scheduling troubleshooting instruction, which includes a specific description of the abnormal situation, an analysis of possible causes, and task items that need to be processed as a priority. The troubleshooting instruction is integrated into the response information as an operational reference for call center operators. Through this mechanism, the call center can quickly identify and handle scheduling problems to ensure the timeliness of services and the efficiency of resource utilization.

[0052] The embodiment of the present application adopts a method of receiving massive call request information through an information sending device, performing language processing to generate a parsed data set, and obtaining the request type through deep learning classification; performing multi-objective optimization on the parsed data set to generate a dynamic call queue; constructing a time series prediction model based on historical records, predicting call peaks and performing resource scheduling analysis, and formulating scheduling strategies; generating feedback results after executing the scheduling strategies, synchronizing them to the call center and updating the strategies according to the response information, realizing intelligent dynamic scheduling, and achieving the technical effect of optimizing the resource scheduling efficiency of the call center.

[0053] Based on the foregoing embodiments, the embodiments of the present application further provide an electronic device and a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of the electronic device, it can implement the method described in any of the foregoing embodiments.

[0054] Figure 2 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing an implementation of the present invention. Figure 2 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. The electronic device is in the form of a general computing device, and its components may include but are not limited to an input device 301, a processor 302, a memory 303, and an output device 304. Among them, the processor 302 may be one or more; the memory 303 may include a computer-readable medium and at least one program product, and the program product has a set (at least one) of program modules, which are configured to perform the functions of each embodiment of the present application.

[0055] The memory 303 shown in the embodiment of the present invention may adopt any combination of one or more computer-readable media; the computer-readable storage medium may be, but is not limited to, an infrared, semiconductor system, device or component, or any combination of the above, for storing software programs, computer executable programs and modules, such as program instructions / modules corresponding to a dynamic scheduling method for massive call requests in an embodiment of the present invention. The processor 302 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 303, thereby realizing the above-mentioned dynamic scheduling method for massive call requests.

[0056] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A dynamic scheduling method for massive call requests, characterized in that: The method comprises: Receive information through multi-party information sending devices to obtain massive call request information; Performing language processing based on the massive call request information to generate a request parsing data set, and performing deep learning classification on the request parsing data set to obtain multiple request types; Performing multi-objective optimization on the request parsing data set according to the multiple request types to generate multiple dynamic call queues; Retrieving historical call record logs to build a time series prediction model, and performing predictions through the time series prediction model to determine call request peak prediction data; Mapping the multiple dynamic call queues to the call request peak forecast data to perform resource scheduling analysis and formulate a scheduling strategy; The scheduling strategy is executed to process feedback, generate a scheduling feedback result, synchronize the scheduling feedback result to the call center for response, and update the scheduling strategy according to the response information to perform intelligent dynamic scheduling on the massive call request information.

2. A dynamic scheduling method for massive call requests as claimed in claim 1, characterized in that: Based on the massive call request information, language processing is performed to generate a request parsing data set, and the method includes: Performing text processing based on the massive call request information to obtain a text feature set; Performing context analysis based on the text feature set to capture text relevance factors; Perform voice processing based on the massive call request information to obtain a voice feature set; Performing emotion recognition based on the speech feature set to capture speech emotion factors; Normalizing the text association factor and the voice emotion factor to generate a processing result, and performing intent recognition on the massive call request information according to the processing result to generate a plurality of request intent labels; The multiple request intent tags are added to the request parsing dataset.

3. A dynamic scheduling method for massive call requests as claimed in claim 2, characterized in that: Performing deep learning classification on the request parsing data set to obtain multiple request types, the method comprising: Performing impact analysis on the massive call request information according to the multiple request intent tags to generate multiple impact levels; performing cluster analysis on the request parsing data set based on the multiple impact levels, determining multiple data clustering centers, and dividing the request parsing data set into multiple parsing data clusters according to the multiple data clustering centers; Performing learning and allocation on the massive call request information according to the multiple parsed data clusters combined with the multiple request intent labels to generate multiple weight coefficients; Processing the request parsing data set in descending order according to the multiple weight coefficients to determine a request data sequence; The request data sequence is segmented according to the multiple parsed data clusters to determine the multiple request types.

4. A dynamic scheduling method for massive call requests as claimed in claim 3, characterized in that: Performing multi-objective optimization on the request parsing data set according to the multiple request types to generate multiple dynamic call queues, the method comprising: Traversing the request parsing data set according to the multiple request types to perform encoding and generate multiple encoded data; Analyzing the massive call request information based on the multiple coded data and setting multiple processing targets; Performing a processing balance calculation according to the multiple processing targets to determine a multi-target processing balance ratio; Iteratively evaluating the request data sequence according to the multi-objective processing balance ratio until the request data sequence satisfies the multi-objective processing balance ratio, and determining a plurality of initial call queues; Call compatibility analysis is performed on the multiple initial call queues to generate queue compatibility coefficients, and the multiple initial call queues are adjusted and optimized according to the queue compatibility coefficients to generate the multiple dynamic call queues.

5. A dynamic scheduling method for massive call requests as claimed in claim 4, characterized in that: Performing call compatibility analysis on the multiple initial call queues to generate queue compatibility coefficients, the method comprising: Based on the multiple initial call queues, random combinations are performed in pairs to generate multiple queue combinations; Analyze whether there is call overlap among the multiple queue combinations, and if there is call overlap among the multiple queue combinations, mark the multiple queue combinations with call overlap as call incompatible combinations; Generate a call occupation instruction according to the call incompatible combination, split and reorganize multiple queue combinations with call overlap through the call occupation instruction to generate multiple reorganized queue combinations; The plurality of reorganized queue combinations are synchronized to the queue compatibility coefficient.

6. A dynamic scheduling method for massive call requests as claimed in claim 3, characterized in that: Mapping the multiple dynamic call queues to the call request peak forecast data to perform resource scheduling analysis and formulate a scheduling strategy, the method comprising: Performing request analysis based on the multiple dynamic call queues in combination with the multiple request types to determine multiple request attributes; Perform call resource analysis according to the call request peak prediction data and the request data sequence to determine agent resource status prediction data; Perform resource matching according to the multiple request attributes combined with the agent resource status prediction data to obtain multiple resource matching pairs; Determine a plurality of priority allocation queues based on mapping the plurality of resource matching pairs with the plurality of dynamic call queues; Perform resource scheduling records according to the multiple priority allocation queues to generate a resource scheduling track information set; The resource scheduling trajectory information set is added to the scheduling strategy.

7. A dynamic scheduling method for massive call requests as claimed in claim 1, characterized in that: Executing the scheduling strategy to process feedback, generating a scheduling feedback result, and synchronizing the scheduling feedback result to a call center for response, the method comprising: Executing the scheduling strategy to perform resource scheduling on the massive call request information to obtain a resource scheduling result; Perform resource allocation determination based on the resource scheduling result to determine whether the resource scheduling result meets the expected scheduling threshold; If the resource scheduling result meets the expected scheduling threshold, the resource scheduling result satisfies the massive call request information, a positive feedback signal is generated, and the positive feedback signal is added to the scheduling feedback result; If the resource scheduling result does not meet the expected scheduling threshold, the resource scheduling result does not meet the massive call request information, a negative feedback signal is generated, and the negative feedback signal is added to the scheduling feedback result; If the scheduling feedback result is the negative feedback signal, the resource scheduling result is synchronized to the call center for scheduling abnormality alarm, a scheduling troubleshooting instruction is generated, and the scheduling troubleshooting instruction is added to the response information.

8. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement a dynamic scheduling method for massive call requests as described in any one of claims 1 to 7 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a dynamic scheduling method for massive call requests as described in any one of claims 1 to 7 is implemented.

Images