Customer service incoming line quantity processing method and device

Abstract

The invention discloses a customer service incoming line quantity processing method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the following steps: acquiring a customer service incoming record of each time point in a historical time period; using a time-delay analysis model to analyze time-delay influence among the index data at different time points so as to update customer service incoming line records; processing the updated customer service incoming line record by using a causal discovery model to obtain a causal graph, taking an index directly connected with customer service incoming line quantity data in the graph as a direct feature, and taking an index directly connected with the direct feature as an indirect feature; and for the updated customer service incoming record, removing other index data except the direct features and the indirect features to obtain a final customer service incoming record, and training a model through the final customer service incoming record to predict the customer service incoming quantity at the target time point by using the trained model. According to the embodiment, the key indexes are found for modeling through causal discovery and hysteresis analysis, so that the accuracy of incoming line quantity prediction is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for processing customer service incoming line volume. Background Art

[0002] In customer service scenarios, it is usually chosen to use all operational indicators for modeling to predict future incoming calls (i.e., the number of incoming calls received by customer service), and then recruit and schedule customer service personnel based on this number. In this process, the prediction can be abstracted as an attribution analysis process. However, the existing attribution analysis methods all have a conceptual ambiguity problem, that is, the relationship found may be an association rather than a causal relationship, and the association relationship does not necessarily have a "time sequence characteristic", and it cannot be determined in essence that it is a causal relationship. This makes it impossible to confirm whether these operational indicators can play a positive role in the prediction of incoming calls, and there may be some negative factors, resulting in a decrease in the accuracy of the prediction. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a method and device for processing customer service incoming call volume, which can at least solve the problem in the prior art that it is impossible to use operational indicators that are irrelevant or have a negative effect on causal analysis, resulting in a low accuracy rate in incoming call volume prediction.

[0004] To achieve the above-mentioned purpose, according to one aspect of an embodiment of the present invention, a method for processing customer service call volume is provided, including: obtaining customer service call records at each time point in a historical time period; wherein the customer service call records include indicator data and customer service call volume data; using a time lag analysis model to analyze the time lag impact between indicator data at different time points, and updating the customer service call records based on the time lag impact; using a causal discovery model to process the updated customer service call records to obtain a causal graph, using the indicators directly connected to the customer service call volume data in the causal graph as direct features, and using the indicators directly connected to the direct features as indirect features; for the updated customer service call records, removing other indicator data except the direct features and indirect features to obtain the final customer service call records, and training the model through the final customer service call records to use the trained model to predict the customer service call volume at the target time point.

[0005] Optionally, the use of a time-lag analysis model to analyze the time-lag effects between indicator data at different time points, and updating customer service call records based on the time-lag effects, includes: using the time-lag analysis model, taking the analyzed time point as the current time point, and identifying influencing indicators of historical time points that have time-lag effects on the current time point; replacing the indicator value of the influencing indicator at the current time point with the indicator value of the influencing indicator at the historical time point, to obtain an updated customer service call record.

[0006] Optionally, the method takes the analyzed time point as the current time point and identifies the influencing indicators of the historical time points that have a time-lag effect on the current time point, including: obtaining indicator value distribution information according to the indicator value of each indicator at different time points; determining the mutual information, conditional mutual information and partial correlation coefficient between the two indicators according to the indicator value distribution information of each indicator in the two indicators; determining the indicators with causal relationships according to the mutual information, conditional mutual information and partial correlation coefficients between all two indicators; taking the analyzed time point as the current time point, for the indicators with causal relationships, analyzing the changes in the mutual information and partial correlation coefficients between the historical time points and the current time points, and determining the causal direction according to the changes; according to the causal direction of the indicators with causal relationships, taking the indicators with priority changes at the historical time points as influencing indicators.

[0007] Optionally, the customer service call record includes the customer service call volume data at each time point and the index data at the previous time point of each time point, and the index data includes an index and an index value; the use of the trained model to predict the customer service call volume at the target time point includes: obtaining the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point, inputting the obtained index value into the trained model, and obtaining the predicted customer service call volume at the target time point; wherein, the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained in one of the following ways: determining the direct index value at the previous time point of the target time point according to the index value distribution information of the direct feature; determining the indirect index value at the previous time point of the target time point according to the index value distribution information of the indirect feature; obtaining the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point through a collection method.

[0008] Optionally, the use of a causal discovery model to process updated customer service call records to obtain a causal graph includes: using indicators as feature variables, using the customer service call volume as the target variable, and using the feature variables and the target variable as nodes in a directed acyclic graph; using a statistical method to process the indicator value of each feature variable and the call volume of the target variable in different customer service call records to obtain the correlation between the feature variable and the target variable, and between different feature variables; using a conditional independence test method to process the correlation between the different feature variables, obtain the direct causal relationship between the feature variables, and determine the direction of the direct causal relationship; and constructing a causal graph based on the correlation between the feature variables and the target variable, as well as the direct causal relationship and the direction of the causal relationship between the feature variables.

[0009] Optionally, the training model through the final customer service call records includes: splitting the final customer service call records of the historical time period into a training set and a test set; inputting the final customer service call records at each time point in the training set into the original model for training to obtain a trained model; inputting the indicator data of the final customer service call records at each time point in the test set into the trained model to obtain the predicted customer service call volume data at each time point; calculating the accuracy and achievement rate through the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set; when the accuracy is greater than or equal to the preset accuracy threshold and the achievement rate is within the preset range, determining that the trained model is available, otherwise adjusting the model parameters for iterative training.

[0010] Optionally, the accuracy rate and the achievement rate are calculated by using the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set, including: for each time point in the test set, calculating the difference between the actual customer service call volume data and the predicted customer service call volume data, calculating the ratio of the difference to the actual customer service call volume data, taking the absolute value of the ratio, determining the sum of the absolute values ​​of all time points, and taking the ratio of the sum of the absolute values ​​to the number of time points as the accuracy rate; calculating the quotient of the actual customer service call volume data and the predicted customer service call volume data, determining the sum of the quotients of all time points, and taking the ratio of the sum of the quotients to the number of time points as the achievement rate.

[0011] To achieve the above object, according to another aspect of an embodiment of the present invention, a customer service incoming line processing device is provided, comprising:

[0012] An acquisition module is used to acquire customer service call records at each time point in a historical time period; wherein the customer service call records include indicator data and customer service call volume data;

[0013] The time lag analysis module is used to use the time lag analysis model to analyze the time lag impact between indicator data at different time points, and update customer service call records based on the time lag impact;

[0014] A causal analysis module, used to process the updated customer service call record using a causal discovery model to obtain a causal graph, and use the indicator directly connected to the customer service call volume data in the causal graph as a direct feature, and use the indicator directly connected to the direct feature as an indirect feature;

[0015] The modeling module is used to remove other indicator data except direct features and indirect features from the updated customer service call records to obtain the final customer service call records. The model is trained through the final customer service call records to use the trained model to predict the customer service call volume at the target time point.

[0016] Optionally, the time-lag analysis module is used to: use the time-lag analysis model, take the analysis time point as the current time point, identify the influencing indicators of the historical time point that have a time-lag effect on the current time point; replace the indicator value of the influencing indicator at the current time point with the indicator value of the influencing indicator at the historical time point, and obtain an updated customer service call record.

[0017] Optionally, the time-lag analysis module is used to: obtain indicator value distribution information according to the indicator value of each indicator at different time points; determine the mutual information, conditional mutual information and partial correlation coefficient between the two indicators according to the indicator value distribution information of each indicator in the pairwise indicators; determine the indicators with causal relationship according to the mutual information, conditional mutual information and partial correlation coefficient between all two indicators; take the analysis time point as the current time point, and for the indicators with causal relationship, analyze the changes in the mutual information and partial correlation coefficients between the historical time point and the current time point, and determine the causal direction according to the changes; according to the causal direction of the indicators with causal relationship, take the indicators with priority changes at the historical time point as the influencing indicators.

[0018] Optionally, the customer service call record includes the customer service call volume data at each time point and the index data at the previous time point of each time point, and the index data includes an index and an index value; the modeling module is used to: obtain the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point, input the obtained index value into the trained model, and obtain the predicted customer service call volume at the target time point; wherein, the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained in one of the following ways: determining the direct index value at the previous time point of the target time point according to the index value distribution information of the direct feature; determining the indirect index value at the previous time point of the target time point according to the index value distribution information of the indirect feature; obtaining the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point through a collection method.

[0019] Optionally, the causal analysis module is used to: use the indicator as the characteristic variable, the customer service call volume as the target variable, and the characteristic variable and the target variable as nodes in a directed acyclic graph; use a statistical method to process the indicator value of each characteristic variable and the call volume of the target variable in different customer service call records to obtain the correlation between the characteristic variable and the target variable, and between different characteristic variables; use a conditional independence test method to process the correlation between the different characteristic variables, obtain the direct causal relationship between the characteristic variables, and determine the direction of the direct causal relationship; construct a causal graph based on the correlation between the characteristic variables and the target variable, as well as the direct causal relationship and the direction of the causal relationship between the characteristic variables.

[0020] Optionally, the modeling module is used to: split the final customer service call records of the historical time period into a training set and a test set; input the final customer service call records at each time point in the training set into the original model for training to obtain a trained model; input the indicator data of the final customer service call records at each time point in the test set into the trained model to obtain the predicted customer service call volume data at each time point; calculate the accuracy and achievement rate through the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set; when the accuracy is greater than or equal to the preset accuracy threshold and the achievement rate is within a preset range, determine that the trained model is available, otherwise adjust the model parameters for iterative training.

[0021] Optionally, the modeling module is used to: for each time point in the test set, calculate the difference between the actual customer service call volume data and the predicted customer service call volume data, calculate the ratio of the difference to the actual customer service call volume data, take the absolute value of the ratio, determine the sum of the absolute values ​​of all time points, and use the ratio of the sum of the absolute values ​​to the number of time points as the accuracy rate; calculate the quotient of the actual customer service call volume data and the predicted customer service call volume data, determine the sum of the quotients of all time points, and use the ratio of the sum of the quotients to the number of time points as the achievement rate.

[0022] To achieve the above objective, according to another aspect of an embodiment of the present invention, a customer service incoming line processing electronic device is provided.

[0023] The electronic device of an embodiment of the present invention includes: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned customer service incoming call processing methods.

[0024] To achieve the above-mentioned purpose, according to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, any of the above-mentioned customer service incoming call processing methods is implemented.

[0025] To achieve the above object, according to another aspect of an embodiment of the present invention, a computer program product is provided. A computer program product of an embodiment of the present invention includes a computer program, and when the program is executed by a processor, the customer service incoming call processing method provided by an embodiment of the present invention is implemented.

[0026] According to the solution provided by the present invention, one embodiment of the above invention has the following advantages or beneficial effects: by integrating causal discovery and hysteresis analysis, it is possible to accurately locate the causal influence and hysteresis effect of operational indicators on changes in customer service call volume, thereby guiding the operation and maintenance optimization of customer service scenarios, and providing customer service operators with key indicators that can be used as references to affect changes in call volume, so that they can proactively adjust recruitment and scheduling strategies. This method eliminates data hysteresis interference, ensures the accuracy of feature selection and time positioning, and only focuses on indicators that have a direct or indirect causal relationship with customer service call volume, thereby greatly improving the accuracy and practicality of the prediction model, thereby supporting more efficient and flexible business decisions.

[0027] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0029] Figure 1 This is a schematic diagram of the main process of a method for processing customer service incoming calls according to an embodiment of the present invention;

[0030] Figure 2 It is a flowchart of an optional customer service incoming call volume processing method according to an embodiment of the present invention;

[0031] Figure 3 is a flowchart of another optional customer service incoming call processing method according to an embodiment of the present invention;

[0032] Figure 4 is a flowchart of another optional customer service incoming call processing method according to an embodiment of the present invention;

[0033] Figure 5 2 is a schematic diagram of the overall process of processing customer service incoming calls according to an embodiment of the present invention;

[0034] Figure 6 This is a schematic diagram of main modules of a customer service incoming line processing device according to an embodiment of the present invention;

[0035] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0036] Figure 8 It is a schematic diagram of the structure of a computer system of a mobile device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0038] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0039] Attribution analysis is to find the reasons why things produce results through some methods. It is usually based on some historical observation data, modeling the target based on features, and finding the correlation or causal relationship. In the existing attribution analysis, most of them use statistical models or machine learning models, such as linear regression, to construct a linear model between input features (such as operating indicators, etc.) and target values ​​(customer service call volume), and then analyze it based on the coefficients learned for each feature. The larger coefficients are more important factors, and the smaller coefficients are less important factors. Alternatively, a machine learning model, such as XGBoost, is used to model the data, and then analyze the feature importance, and analyze the specific feature importance based on the nodes where the tree splits.

[0040] Causality is the phenomenon that truly determines that the change of one factor leads to the change of another factor. It is when a certain variable X changes, how X causes another variable Y to change under the premise of keeping other variables unchanged. In data analysis, it may not be possible to find the causal relationship between all variables. At the same time, due to the existence of hidden variables (Cofounder, that is, unobservable variables), only the relationship between some features can be determined. However, if causally related features are found, better analysis can be performed. For example, finding features related to the target value can better optimize the model to predict the target value.

[0041] In statistics, correlation and causality are often discussed together. As mentioned above, causality means that the existence of one variable leads to the generation of another variable. In essence, it has a temporal relationship, that is, the "cause" changes and the "effect" changes; while in correlation, when one variable X changes, another variable Y will also change, but it cannot be determined that X is the cause of the change in Y. It can only be said that there is a strong correlation between the two, so it cannot constitute a causal relationship.

[0042] This solution can be applied to all scenarios that require customer service. For example, the operational indicators used by logistics customer service are related to express delivery, such as delivery timeliness rate, terminal backlog, package damage rate, customer receipt confirmation rate, etc.; the operational indicators used by multilingual customer service are related to language, such as language conversion accuracy, multilingual response time, language coverage, and multilingual satisfaction score; the operational indicators used by member customer service are related to member services, such as member activity, member retention rate, member upgrade conversion rate, member service response time, and member exclusive discount utilization rate; the operational indicators used by after-sales customer service are related to after-sales service, such as return rate, exchange processing time, customer complaint resolution rate, and customer satisfaction survey results; the operational indicators used by technical support customer service are related to technical support services, such as technical problem resolution rate, average repair time, and knowledge base usage rate. The operational indicators of each customer service scenario are designed to ensure service quality, improve customer experience, and optimize internal processes. Selecting appropriate indicators and continuously monitoring them can help customer service teams identify opportunities for improvement, thereby continuously improving service quality and efficiency.

[0043] See also Figure 1 , which shows a main flow chart of a customer service incoming call processing method provided by an embodiment of the present invention, including the following steps:

[0044] S101: Obtain customer service call records at each time point in a historical time period; wherein the customer service call records include indicator data and customer service call volume data;

[0045] S102: Use a time lag analysis model to analyze the time lag impact between indicator data at different time points, and update customer service call records based on the time lag impact;

[0046] S103: Use the causal discovery model to process the updated customer service call record to obtain a causal graph, and use the indicator directly connected to the customer service call volume data in the causal graph as a direct feature, and use the indicator directly connected to the direct feature as an indirect feature;

[0047] S104: For the updated customer service call record, remove other indicator data except direct features and indirect features to obtain the final customer service call record, train the model through the final customer service call record, and use the trained model to predict the customer service call volume at the target time point.

[0048] In the above implementation, for step S101, this solution mainly uses the incoming call volume analysis scenario of logistics customer service as an example for explanation. The same is true for other scenarios, except that the operating indicators are different, but the target value is the incoming call volume of customer service, such as the incoming call volume of customer service in a certain branch center. The operating indicators used in the logistics customer service scenario mainly include the following categories: timeliness (such as order timeliness), backlog (such as terminal transportation backlog), weather (such as wind force level), loss (such as damage volume, damage rate), date (such as whether it is a holiday), and waybill (such as the number of successfully delivered orders).

[0049] This solution can give priority to obtaining customer service call records for a certain period of time, such as customer service call records from January 1, 2023 to May 1, 2024. Customer service call records include: customer service call volume data and corresponding operational indicator data. Considering that using the operational indicator data at the current time point to predict the call volume at the current time point has a lag, this solution prefers to use the operational indicator data at the previous time point, such as using the operational indicator data of the previous day to predict the customer service call volume of the day. The time point can also be in hours without restriction. By predicting the customer service call volume, the relevant departments will reasonably arrange the recruitment and scheduling of customer service personnel according to the predicted call volume. For example, in March, it is predicted that the call volume during the 618 promotion will reach more than 10,000. In this case, it is necessary to consider the number of calls that each customer service can handle to calculate the number of customer service personnel required, and then schedule the shifts.

[0050] For step S102, no matter which scenario, there is a certain hysteresis, that is, the recorded time of the order received by the customer service is not necessarily the time when the customer service receives the related call of the order. Therefore, this solution not only uses the causal discovery method to perform causal analysis, but also analyzes the hysteresis effect of different indicators. This solution preferably uses PCMCI (full name Partial Correlation and Mutual Information) type algorithms (i.e., hysteresis analysis models) to analyze the hysteresis effect between all operating indicators at different time points.

[0051] If time lag analysis is not performed, all operational indicators will affect each other at the time point of T-1. For example, the terminal backlog rate at T-1 will affect the delivery timeliness rate at T-1. However, after time lag analysis, it is found that the terminal backlog rate at T-2 will truly affect the delivery timeliness rate at T-1. Therefore, the terminal backlog rate at T-1 is replaced with the terminal backlog rate at T-2, so that the operational indicators use the features of more accurate time points, thereby eliminating the hidden influence of time lags and making the time points of the features more accurate. At the same time, the causal relationship found in subsequent causal discovery is more accurate, and it can also better assist in using operational indicators to analyze changes in incoming line volume.

[0052] For step S103, considering all operational indicators, some indicators may have no effect or counter-effect, resulting in low prediction accuracy. Therefore, it is necessary to screen relevant and more important features for prediction, thereby improving the accuracy. Causal discovery refers to determining the causal relationship between variables from a large amount of data. In this solution, by using the causal discovery method, the features directly or indirectly related to the target value of incoming line volume and their lags are found. Based on these analysis results, business personnel can be warned in advance. For example, the increase in the backlog of goods on the day can be used to judge that the incoming line volume may increase in 3 days, so that personnel adjustments can be made in advance.

[0053] This solution preferably uses a causal analysis algorithm (i.e., causal discovery model) to construct a directed acyclic graph (DAG), a process that aims to reveal the true causal relationship between the variables in the data set. Other types of causal discovery methods, such as LinGAM, can also be used to adapt to different types of data. The operating indicators that are directly connected to the target value of the incoming line are selected as direct features, and the upper-level features that are connected to the direct features are selected as indirect features as the results of the attribution analysis. The difference between direct and indirect lies mainly in whether there is a direct relationship. For example, if it is direct: terminal backlog rate->incoming line, then the terminal backlog rate is a direct feature; if it is an indirect relationship: terminal backlog rate->transportation efficiency->incoming line, then the terminal backlog rate is an indirect feature. The difference lies in whether it is a "direct" impact.

[0054] For step S104, by using a causal discovery model, the potential causal structure / relationship between a series of variables is understood from the data set, and features that have no effect on the customer service call volume are removed, that is, the obtained features are either direct features or indirect features that have caused an impact through other factors.

[0055] This application updates the customer service call record twice, once for the replacement of the indicator data after the time lag analysis, and the other time for feature screening to obtain the final customer service call record. Only the "causal" features of the final customer service call record are used for modeling, and the model is trained based on direct features, indirect features, and the target value "call volume". For example, for T-1, the delivery timeliness rate of T-1, the terminal backlog rate of T-2, and the customer service call volume are used. After the model training is completed, it can be used to predict the customer service call volume at the target time point, such as predicting the customer service call volume on June 18 in March, or using today's indicator data to predict tomorrow's customer service call volume.

[0056] The method provided in the above embodiment uses causal discovery technology to ensure the impact of "causality" on the change in target value during the attribution analysis process of using operational indicators to analyze changes in customer service call volume, and then accurately analyzes the direct or indirect indicators of the change in call volume to guide the operation and maintenance of customer service scenarios, thereby removing irrelevant or counterproductive indicators and improving prediction accuracy.

[0057] See also Figure 2 , shows a schematic flow chart of an optional customer service incoming call processing method according to an embodiment of the present invention, comprising the following steps:

[0058] S201: Obtaining indicator value distribution information according to the indicator value of each indicator at different time points;

[0059] S202: determining mutual information, conditional mutual information and partial correlation coefficient between the two indicators according to the indicator value distribution information of each indicator in the two indicators;

[0060] S203: Determine the indicators with causal relationship according to the mutual information, conditional mutual information and partial correlation coefficient between all the indicators;

[0061] S204: Taking the analyzed time point as the current time point, for indicators with causal relationships, analyzing the changes in mutual information and partial correlation coefficients between historical time points and the current time point, and determining the causal direction according to the changes;

[0062] S205: According to the causal direction of the indicators with causal relationship, the indicators with priority changes at the historical time point are used as influencing indicators.

[0063] In the above implementation, this implementation is used to describe how to use PCMCI to analyze the time lag data between operating indicators. Consider that the value of each operating indicator changes over time. Treat each operating indicator as a time series, record its changes over time, and obtain the indicator value distribution information. In this process, the time series can be smoothed to reduce the impact of noise. The whole process needs to calculate three values, namely:

[0064] 1. Mutual Information (MI): Mutual information measures the degree of dependence between two variables X and Y. It is based on the concept of information entropy, which indicates how much the information about X can reduce the uncertainty about Y, and vice versa. In PCMCI, the mutual information between all variables is calculated. This step is to find out which variables may have a relationship.

[0065] 2. Conditional Mutual Information (CMI): Conditional mutual information is the mutual information between X and Y given another random variable Z. It measures the remaining degree of dependence between X and Y given that Z is known. This step is more important because there are often many variables in real-world data that affect each other, and the direct relationship between two variables is not indirectly generated through a third variable.

[0066] 3. Partial Correlation (PC): The partial correlation coefficient is another statistic that measures the linear relationship between two variables X and Y after controlling other variables Z. This step is also to remove indirect effects and retain only direct causal relationships.

[0067] After calculating the mutual information, conditional mutual information, and partial correlation coefficients of all variables, it is necessary to determine which variables have causal relationships. This process is similar to a puzzle, that is, putting all the variables together to determine which variables are directly connected and which variables are indirectly connected through other variables. You can set a threshold or use statistical tests (such as hypothesis testing) to determine which relationships are significant.

[0068] Taking the time point of analysis as the current time point, combined with the changes in mutual information and partial correlation coefficients between historical time points and the current time point, try to identify the causal direction. If the change in one indicator precedes the change in another indicator, and this change is statistically significant, then the former may be the cause of the latter. This process can use the conditional independence test method, which is an existing method and will not be described here. Based on the results of the conditional independence test, a causal network between variables is constructed. The edges in the network represent causal relationships, the directions represent causal directions, and the weights can be the size of the mutual information or partial correlation coefficients. Based on the determined causal relationships and directions, analyze how the operating indicators at historical time points affect the operating indicators at the current time point.

[0069] The method provided in the above embodiment uses the PCMCI algorithm to analyze the time lag effects between operating indicators, which can help to better understand the dynamic relationship between different operating indicators, determine the influencing indicators at historical time points, and provide a data basis for how to eliminate the time lag effects in the future.

[0070] See also Figure 3 , shows a schematic flow chart of another optional customer service incoming call processing method according to an embodiment of the present invention, comprising the following steps:

[0071] S301: taking the indicator as a characteristic variable, taking the customer service call volume as a target variable, and taking the characteristic variable and the target variable as nodes in a directed acyclic graph;

[0072] S302: Using a statistical method to process the index value of each characteristic variable and the number of incoming calls of the target variable in different customer service call records, and obtain the correlation between the characteristic variable and the target variable and between different characteristic variables;

[0073] S303: Using a conditional independence test method, processing the correlation between the different feature variables, obtaining a direct causal relationship between the feature variables, and determining the direction of the direct causal relationship;

[0074] S304: Construct a causal graph based on the correlation between the feature variables and the target variables, as well as the direct causal relationship and the direction of the causal relationship between the feature variables.

[0075] In the above implementation, in the causal discovery method, observation data is used to construct the potential causal structure between variables, and finally a directed acyclic graph is formed. This process aims to reveal the true causal relationship between the variables in the data set. Directed: The edges in the graph are directed, indicating the causal direction from one node (variable) to another. If the edge points from A to B, it means that A may be the direct cause of B. Acyclic: There is no circular path in the graph, that is, no variable A can return to itself through a series of directed edges. This reflects the irreversibility of causality, that is, the causal chain cannot form a closed loop.

[0076] The process of constructing DAG: All operating indicators are used as feature variables, and "incoming line volume" is used as the target variable. These variables constitute the nodes in DAG. Determine the values ​​of feature variables and target variables at different time points. Use statistical methods (such as Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information, etc.) to detect the correlation between variables, specifically, the correlation between feature variables and target variables, and between different feature variables. Use conditional independence tests to identify direct causal relationships between variables and eliminate interference from intermediate variables. Determine the direction of causal relationships based on the directionality of time series data, prior knowledge, or causal inference criteria. Draw the correlation between the feature variables and target variables and the causal relationship between feature variables determined in the above steps into a DAG, where nodes represent variables and directed edges represent causal directions.

[0077] The method provided in the above embodiment clearly demonstrates the causal relationship between variables through DAG, which helps to understand and communicate, identify and control confounding factors, reduce bias in analysis, facilitate a deeper understanding of the intrinsic relationship between operating indicators, and provide a scientific basis for optimizing operating strategies.

[0078] See also Figure 4 , shows a schematic flow chart of another optional customer service incoming call processing method according to an embodiment of the present invention, comprising the following steps:

[0079] S401: split the final customer service call records in the historical time period into a training set and a test set;

[0080] S402: Input the final customer service call record at each time point in the training set into the original model for training to obtain a trained model;

[0081] S403: Input the index data of the final customer service call record at each time point in the test set into the trained model to obtain the predicted customer service call volume data at each time point;

[0082] S404: Calculate the accuracy rate and the achievement rate by testing the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set;

[0083] S405: When the accuracy is greater than or equal to a preset accuracy threshold and the achievement rate is within a preset range, it is determined that the trained model is usable; otherwise, the model parameters are adjusted for iterative training.

[0084] In the above implementation mode, this implementation mode describes how to train the model. Like the existing one, this solution also splits the data set into a training set and a test set. For example, the period from January 1, 2023 to March 1, 2024 is split into two parts: January 1, 2023 to March 1, 2024, and March 1, 2024 to May 1, 2024. The corresponding final customer service call records are the training set and the test set.

[0085] The indicator data from January 1, 2023 to March 1, 2024 is used as the independent variable, and the customer service call volume data is used as the dependent variable, and is input into the original model to be trained to obtain the trained model. The indicator data from March 1, 2024 to May 1, 2024 is also input into the trained model as the independent variable to obtain the daily predicted customer service call volume data from March 1, 2024 to May 1, 2024. Here, it is assumed that the time point is in days.

[0086] Due to the particularity of customer service scenarios, it is necessary not only to ensure the accuracy of the prediction, but also to ensure that the predicted volume meets the actual volume required on the day. Therefore, the accuracy rate and the achievement rate (the degree of conformity between the actual value and the target value (the degree of achievement of the target)) are used together as evaluation indicators. The accuracy rate represents the accuracy of the prediction, where the accuracy rate is a decimal from 0 to 1, the larger the better; the achievement rate represents whether the predicted incoming call volume can "cover" the actual incoming call volume on the day. If it is exactly covered, it is 100%. If it is less than 100%, some customer service staff will need to handle more incoming calls to make up for it. If it is greater than 100%, it will result in too many customer service staff being hired in advance, resulting in cost waste.

[0087] The achievement rate is a percentage, and the acceptable range is between 90% and 110%. That is, the cost of the predicted value exceeding the actual value by less than 10% is acceptable, and the predicted value less than the actual value can be adjusted by customer service personnel to cover the bottom line; if it is outside the range, it is unacceptable, as mentioned above, either due to excessive cost waste or due to the inability to actually complete the incoming line connection. When comparing, compare the degree of deviation of the achievement rate from 100%, the closer the better.

[0088] Among them, the accuracy uses the mean absolute percentage error Mape as an indicator, and its formula is:

[0089] The achievement rate formula is as follows:

[0090] in, This is the actual customer service call volume data. To predict customer service call volume data. This solution sets a preset accuracy threshold, such as 85%. If the calculated accuracy is greater than or equal to 85%, and the achievement rate is in the range of [90%~110%], it means that the trained model is usable. Otherwise, it is unusable and it is necessary to change the model parameters and iterate the training.

[0091] For subsequent predictions, you can collect the index values ​​of indirect indicators and direct indicators at the time point before the target time point, or use "the index value of each indicator at different time points to obtain the index value distribution information" to predict the index values ​​of indirect indicators and direct indicators at the previous time point. This is because the customer service call record of this solution includes the customer service call volume data at each time point and the index data at the time point before each time point, and the index data includes indicators and index values. Take the customer service call record of 2024.04.25 as an example, it includes the customer service call volume data of 2024.04.25 and the index data of 2024.04.24.

[0092] The method provided in the above embodiment, due to the particularity of the customer service scenario, must not only ensure the prediction accuracy, but also ensure that the predicted amount meets the actual amount needed on the day. The accuracy rate and the achievement rate are used together as evaluation indicators to double-judge whether the trained model is usable. Compared with single-indicator evaluation, this can avoid the situation where the prediction result is inconsistent with the actual situation.

[0093] At present, it is impossible to prove from the perspective of counterfactuals that the causal relationship found is better than the case where this method is not used. When verifying the effect, the accuracy rate and achievement rate of predicted incoming calls are selected as evaluation criteria. See the experimental design for details. This scheme is tested based on the customer service incoming call data of three sub-centers. The data set contains the customer service incoming call data and corresponding operation indicators for a certain period of time, such as January 1, 2023 to May 1, 2024. Experimental plan: Use the method in this paper and the traditional method to conduct experiments respectively:

[0094] Option 1: Do not use this method. Use all operating indicators from January 1, 2023 to March 1, 2024 to build a model, and backtest the accuracy of customer service call volume from March 1, 2024 to May 1, 2024.

[0095] Solution 2: Use this method to build a model using all operational indicators from January 1, 2023 to March 1, 2024, use the PCMCI algorithm for time lag analysis, and then use the FGES algorithm for causality analysis to obtain the direct and indirect causal relationships described above. Only use the "causal" features obtained in the previous step to build a model, and backtest the accuracy of customer service call volume from March 1, 2024 to May 1, 2024. For the overall Figure 5 shown.

[0096] The specific results are shown in Table 1 and Table 2:

[0097]

[0098]

[0099] It can be seen from Table 1 and Table 2 that the present invention (Scheme 2) is compared with Scheme 1:

[0100] 1) Higher prediction accuracy. Compared with Solution 1, the accuracy of the present invention is 0.87, which is 10 points higher. Therefore, it can be concluded that the use of the causal discovery method can better find the operating indicators that have a causal relationship with the target incoming line volume, exclude other irrelevant or indirect operating indicators, and improve the accuracy. At the same time, when performing attribution analysis, more emphasis can be placed on these causal operating indicators to better help analyze the reasons for changes in incoming line volume.

[0101] 2) The predicted achievement rate is higher. Compared with Solution 1, this solution is also better in terms of achievement rate. In practical terms, it is closer to 100% achievement, and a more accurate number of customer service staff can be prepared to handle customer service calls, which can save costs to a certain extent. At the same time, not only as mentioned in 1), we can also focus on operational indicators, but also estimate the impact in advance based on the results of time lag analysis. For example, if there is a backlog indicator at time T-3 that has a direct impact, we can estimate how big the impact will be 3 days later at the current time.

[0102] See also Figure 6 , shows a schematic diagram of main modules of a customer service incoming line processing device 600 provided in an embodiment of the present invention, including:

[0103] The acquisition module 601 is used to acquire the customer service call record at each time point in the historical time period; wherein the customer service call record includes indicator data and customer service call volume data;

[0104] The time lag analysis module 602 is used to use the time lag analysis model to analyze the time lag impact between indicator data at different time points, and update the customer service line record based on the time lag impact;

[0105] The causal analysis module 603 is used to process the updated customer service call record using the causal discovery model to obtain a causal graph, and use the indicator directly connected to the customer service call volume data in the causal graph as a direct feature, and use the indicator directly connected to the direct feature as an indirect feature;

[0106] Modeling module 604 is used to remove other indicator data except direct features and indirect features from the updated customer service call records to obtain the final customer service call records, train the model through the final customer service call records, and use the trained model to predict the customer service call volume at the target time point.

[0107] In the implementation device of the present invention, the time lag analysis module 602 is used to:

[0108] Using the time-lag analysis model, taking the time point of analysis as the current time point, identify the influencing indicators of the historical time point on the current time point with time lag;

[0109] The indicator value of the influencing indicator at the current time point is replaced with the indicator value of the influencing indicator at the historical time point to obtain an updated customer service line record.

[0110] In the implementation device of the present invention, the time lag analysis module 602 is used to:

[0111] According to the indicator value of each indicator at different time points, the indicator value distribution information is obtained;

[0112] Determine the mutual information, conditional mutual information and partial correlation coefficient between the two indicators according to the indicator value distribution information of each indicator in the two indicators;

[0113] According to the mutual information, conditional mutual information and partial correlation coefficient between all two indicators, the indicators with causal relationship are determined;

[0114] Taking the time point of analysis as the current time point, for indicators with causal relationships, analyze the changes in mutual information and partial correlation coefficients between historical time points and current time points, and determine the causal direction based on the changes;

[0115] According to the causal direction of the indicators with causal relationship, the indicators with priority changes at historical time points are used as influencing indicators.

[0116] In the implementation device of the present invention, the customer service incoming call record includes the customer service incoming call volume data at each time point and the index data at the previous time point of each time point, and the index data includes the index and the index value;

[0117] The modeling module 604 is used to:

[0118] Obtain the index values ​​of the indirect features and the direct features at the time point before the target time point, input the obtained index values ​​into the trained model, and obtain the predicted customer service call volume at the target time point;

[0119] The index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained by one of the following methods:

[0120] According to the index value distribution information of the direct feature, the direct index value of the previous time point of the target time point is determined; according to the index value distribution information of the indirect feature, the indirect index value of the previous time point of the target time point is determined;

[0121] Through the collection method, the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained.

[0122] In the implementation device of the present invention, the cause-effect analysis module 603 is used to:

[0123] The indicator is used as the characteristic variable, the customer service call volume is used as the target variable, and the characteristic variable and the target variable are used as nodes in a directed acyclic graph;

[0124] Use statistical methods to process the index value of each characteristic variable and the number of incoming calls of the target variable in different customer service lines, and obtain the correlation between the characteristic variable and the target variable, and between different characteristic variables;

[0125] Using a conditional independence test method, the correlation between the different feature variables is processed to obtain a direct causal relationship between the feature variables and determine the direction of the direct causal relationship;

[0126] A causal graph is constructed based on the correlation between feature variables and target variables, as well as the direct causal relationship and the direction of the causal relationship between feature variables.

[0127] In the implementation device of the present invention, the modeling module 604 is used to:

[0128] Split the final customer service call records in the historical time period into a training set and a test set;

[0129] Input the final customer service call record at each time point in the training set into the original model for training to obtain the trained model;

[0130] Input the index data of the final customer service call record at each time point in the test set into the trained model to obtain the predicted customer service call volume data at each time point;

[0131] Calculate the accuracy and achievement rate by testing the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set;

[0132] When the accuracy is greater than or equal to the preset accuracy threshold and the achievement rate is within a preset range, it is determined that the trained model is usable, otherwise the model parameters are adjusted for iterative training.

[0133] In the implementation device of the present invention, the modeling module 604 is used to:

[0134] For each time point in the test set, calculate the difference between the actual customer service call volume data and the predicted customer service call volume data, calculate the ratio of the difference to the actual customer service call volume data, take the absolute value of the ratio, determine the sum of the absolute values ​​of all time points, and take the ratio of the sum of the absolute values ​​to the number of time points as the accuracy rate;

[0135] Calculate the quotient of the actual customer service call volume data and the predicted customer service call volume data, determine the sum of the quotients at all time points, and use the ratio of the sum of the quotients to the number of time points as the achievement rate.

[0136] In addition, the specific implementation content of the device described in the embodiment of the present invention has been described in detail in the method described above, so the repeated content will not be described again here.

[0137] Figure 7 An exemplary system architecture 700 to which embodiments of the present invention may be applied is shown, including terminal devices 701 , 702 , 703 , a network 704 and a server 705 (only an example).

[0138] Terminal devices 701, 702, 703 can be various electronic devices with display screens and supporting web browsing, and various communication client applications are installed. Users can use terminal devices 701, 702, 703 to interact with server 705 through network 704 to receive or send messages, etc.

[0139] The network 704 is used to provide a medium for communication links between the terminal devices 701, 702, 703 and the server 705. The network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0140] The server 705 may be a server that provides various services, such as a backend management server that provides support for shopping websites browsed by users using terminal devices 501, 502, and 503 (for example only). The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - for example only) to the terminal device. It should be noted that the method provided in the embodiment of the present invention is generally executed by the server 705, and accordingly, the device is generally set in the server 705.

[0141] It should be understood that Figure 7The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0142] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing an embodiment of the present invention. Figure 8 The terminal 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.

[0143] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0144] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0145] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0146] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0147] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0148] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor, for example, it may be described as: a processor includes an acquisition module, a time lag analysis module, a causal analysis module, and a modeling module. The names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the modeling module may also be described as a "training module".

[0149] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device executes any of the above customer service incoming line processing methods.

[0150] The computer program product of the present invention includes a computer program, and when the computer program is executed by a processor, the method for processing customer service incoming calls in the embodiment of the present invention is implemented.

[0151] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for processing customer service incoming calls, characterized in that: include: Obtain customer service call records at each time point in the historical time period; the customer service call records include indicator data and customer service call volume data; Use the time lag analysis model to analyze the time lag impact between indicator data at different time points, and update customer service call records based on the time lag impact; Use the causal discovery model to process the updated customer service call record to obtain a causal graph, use the indicator directly connected to the customer service call volume data in the causal graph as a direct feature, and use the indicator directly connected to the direct feature as an indirect feature; For the updated customer service call records, remove other indicator data except direct features and indirect features to obtain the final customer service call records. The model is trained with the final customer service call records to use the trained model to predict the customer service call volume at the target time point.

2. The method according to claim 1, characterized in that The time lag analysis model is used to analyze the time lag impact between indicator data at different time points, and customer service call records are updated based on the time lag impact, including: Using the time-lag analysis model, taking the time point of analysis as the current time point, identify the influencing indicators of the historical time point on the current time point with time lag; The indicator value of the influencing indicator at the current time point is replaced with the indicator value of the influencing indicator at the historical time point to obtain an updated customer service line record.

3. The method according to claim 2, characterized in that The analysis time point is taken as the current time point, and the impact indicators of the historical time point on the current time point with time lag are identified, including: According to the indicator value of each indicator at different time points, the indicator value distribution information is obtained; Determine the mutual information, conditional mutual information and partial correlation coefficient between the two indicators according to the indicator value distribution information of each indicator in the two indicators; According to the mutual information, conditional mutual information and partial correlation coefficient between all two indicators, the indicators with causal relationship are determined; Taking the time point of analysis as the current time point, for indicators with causal relationships, analyze the changes in mutual information and partial correlation coefficients between historical time points and current time points, and determine the causal direction based on the changes; According to the causal direction of the indicators with causal relationship, the indicators with priority changes at historical time points are used as influencing indicators.

4. The method according to claim 3, characterized in that The customer service call record includes the customer service call volume data at each time point and the index data at the previous time point of each time point, and the index data includes the index and the index value; The method of using the trained model to predict the customer service call volume at a target time point includes: Obtain the index values ​​of the indirect features and the direct features at the time point before the target time point, input the obtained index values ​​into the trained model, and obtain the predicted customer service call volume at the target time point; The index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained by one of the following methods: According to the index value distribution information of the direct feature, the direct index value of the previous time point of the target time point is determined; according to the index value distribution information of the indirect feature, the indirect index value of the previous time point of the target time point is determined; Through the collection method, the index value of the indirect feature and the index value of the direct feature at the previous time point of the target time point are obtained.

5. The method according to claim 1, characterized in that The causal discovery model is used to process the updated customer service line record to obtain a causal graph, including: The indicator is used as the characteristic variable, the customer service call volume is used as the target variable, and the characteristic variable and the target variable are used as nodes in a directed acyclic graph; Use statistical methods to process the index value of each characteristic variable and the number of incoming calls of the target variable in different customer service lines, and obtain the correlation between the characteristic variable and the target variable, and between different characteristic variables; Using a conditional independence test method, the correlation between the different feature variables is processed to obtain a direct causal relationship between the feature variables and determine the direction of the direct causal relationship; A causal graph is constructed based on the correlation between feature variables and target variables, as well as the direct causal relationship and the direction of the causal relationship between feature variables.

6. The method according to claim 1, characterized in that The model training through the final customer service line record includes: Split the final customer service call records in the historical time period into a training set and a test set; Input the final customer service call record at each time point in the training set into the original model for training to obtain the trained model; Input the index data of the final customer service call record at each time point in the test set into the trained model to obtain the predicted customer service call volume data at each time point; Calculate the accuracy and achievement rate by testing the actual customer service call volume data and the predicted customer service call volume data at each time point in the test set; When the accuracy is greater than or equal to the preset accuracy threshold and the achievement rate is within a preset range, it is determined that the trained model is usable, otherwise the model parameters are adjusted for iterative training.

7. The method according to claim 6, characterized in that The actual customer service call volume data and the predicted customer service call volume data at each time point in the test set are used to calculate the accuracy rate and the achievement rate, including: For each time point in the test set, calculate the difference between the actual customer service call volume data and the predicted customer service call volume data, calculate the ratio of the difference to the actual customer service call volume data, take the absolute value of the ratio, determine the sum of the absolute values ​​of all time points, and take the ratio of the sum of the absolute values ​​to the number of time points as the accuracy rate; Calculate the quotient of the actual customer service call volume data and the predicted customer service call volume data, determine the sum of the quotients at all time points, and use the ratio of the sum of the quotients to the number of time points as the achievement rate.

8. A customer service incoming line processing device, characterized in that: include: An acquisition module is used to acquire customer service call records at each time point in a historical time period; wherein the customer service call records include indicator data and customer service call volume data; The time lag analysis module is used to use the time lag analysis model to analyze the time lag impact between indicator data at different time points, and update customer service call records based on the time lag impact; A causal analysis module, used to process the updated customer service call record using a causal discovery model to obtain a causal graph, and use the indicator directly connected to the customer service call volume data in the causal graph as a direct feature, and use the indicator directly connected to the direct feature as an indirect feature; The modeling module is used to remove other indicator data except direct features and indirect features from the updated customer service call records to obtain the final customer service call records. The model is trained through the final customer service call records to use the trained model to predict the customer service call volume at the target time point.

9. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images