Customer call prediction method, device, computer equipment and storage medium

Abstract

This application relates to a customer call prediction method, apparatus, computer device, and storage medium. The method comprises: obtaining historical business data for a target in-transit waybill; obtaining corresponding business feature data based on the historical business data for each target in-transit waybill; extracting the customer call habit feature data corresponding to each target in-transit waybill from a pre-stored customer call habit feature set; obtaining corresponding target feature data based on the business feature data corresponding to each target in-transit waybill and the customer call habit feature data; and predicting customer calls based on the target feature data using a trained customer call prediction model to obtain a list of incoming calls within a preset future time period. This method can improve the accuracy of customer call prediction.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, apparatus, computer equipment, and storage medium for predicting customer incoming calls. Background Art

[0002] In the logistics industry, customer service centers are one of the logistics operations closest to customers. Customers can express their service needs through customer service agents at the call center. Logistics companies can gain a deeper understanding of customer needs and even predict them by analyzing call data. However, customer service agents present high costs. For example, these agents incur not only labor costs for customer service representatives and work order handlers, but also call charges. Therefore, predicting customer calls and providing more accurate and proactive services based on these predictions is a significant concern, as is customer service costs.

[0003] Currently, sales personnel typically manually analyze the relevant business data for waybills and use the results to predict whether the corresponding customer will call. However, this customer call prediction method is limited by the sales personnel's prediction experience, resulting in unstable customer call prediction results and low accuracy. Summary of the Invention

[0004] Based on this, it is necessary to provide a customer call prediction method, device, computer equipment and storage medium that can improve the accuracy of customer call prediction in response to the above technical problems.

[0005] A method for predicting incoming customer calls, comprising:

[0006] Obtain historical business data of the target in-transit waybill;

[0007] Obtain corresponding business characteristic data based on the historical business data of each target in-transit waybill;

[0008] Extracting customer call habit feature data corresponding to each target in-transit waybill from a pre-stored customer call habit feature set;

[0009] Obtaining corresponding target feature data based on the business feature data corresponding to each target in-transit waybill and the customer's call habit feature data;

[0010] By using the trained customer call prediction model, customer call prediction is performed based on the target feature data to obtain a list of calling customers within a preset time period in the future.

[0011] In one embodiment, obtaining historical business data of the target in-transit waybill includes:

[0012] Get the current time;

[0013] When the current time is consistent with the pre-configured customer call forecast time, determining the target in-transit waybill for the current time; two customer call forecast times are pre-configured in each forecast period;

[0014] Obtain historical business data for each target in-transit waybill.

[0015] In one embodiment, the method of performing customer call prediction based on the target feature data using a trained customer call prediction model to obtain a list of incoming call customers within a preset time period in the future includes:

[0016] Input the target feature data into the trained customer call prediction model to obtain the corresponding target waybill call label and call intent;

[0017] According to the target waybill call tag and the calling intention, a list of calling customers within a future preset time period and the calling intention corresponding to each customer in the calling customer list are obtained.

[0018] In one embodiment, the step of training the customer call prediction model includes:

[0019] Obtain a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill call label;

[0020] Model training is performed based on the training sample set to obtain a trained customer incoming call prediction model.

[0021] In one embodiment, obtaining a training sample set includes:

[0022] Obtain historical business data corresponding to the sample in-transit waybills, as well as customer call data within a preset time period in the future of the corresponding customer call forecast time;

[0023] Obtaining corresponding sample feature data based on the historical business data of the sample in-transit waybill and the customer's call habit feature set;

[0024] Determine a corresponding sample waybill call label based on customer call data of the sample in-transit waybill;

[0025] A training sample set is obtained according to the sample feature data and the sample waybill call label.

[0026] In one embodiment, the performing model training based on the training sample set to obtain a trained customer call prediction model includes:

[0027] Performing model training based on the training sample set to obtain a customer call prediction model;

[0028] Get a test sample set;

[0029] Testing the customer incoming call prediction model according to the test sample set to obtain a test result;

[0030] When the test result meets the preset test condition, the customer incoming call prediction model is determined as a trained customer incoming call prediction model.

[0031] In one embodiment, the method further comprises:

[0032] Regularly obtain the target incoming call data and target waybill data corresponding to each customer according to the pre-configured update cycle;

[0033] The pre-stored customer incoming call habit feature set is updated according to the target incoming call data and the target waybill data.

[0034] A device for predicting incoming customer calls, comprising:

[0035] Business data acquisition module, used to obtain historical business data of the target in-transit waybill;

[0036] A business feature extraction module is used to obtain corresponding business feature data based on the historical business data of each target in-transit waybill;

[0037] A habit feature extraction module, configured to extract customer call habit feature data corresponding to each target in-transit waybill from a pre-stored customer call habit feature set;

[0038] A feature data acquisition module, configured to obtain corresponding target feature data based on the business feature data corresponding to each target in-transit waybill and the customer's call habit feature data;

[0039] The customer call prediction module is used to predict customer calls based on the target feature data using the trained customer call prediction model to obtain a list of calling customers within a preset time period in the future.

[0040] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0041] A computer-readable storage medium stores a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0042] The above-mentioned customer call prediction method, device, computer equipment and storage medium obtain the historical business data of the target in-transit waybill, obtain the corresponding business feature data based on the historical business data of each target in-transit waybill, extract the customer call habit feature data corresponding to each target in-transit waybill from the pre-stored customer call habit feature set, and based on the business feature data and customer call habit feature data corresponding to each target in-transit waybill, can comprehensively obtain the target feature data that affects customer calls. Furthermore, through the trained customer call prediction model, it is predicted based on the target feature data whether the corresponding customer will call within a preset time period in the future, and a list of calling customers within the preset time period in the future is obtained. This can improve the accuracy of customer call prediction, so that proactive services can be provided to the corresponding customers based on the customer call list with higher accuracy, thereby reducing customer service costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 1 is a flow chart of a method for predicting incoming customer calls in one embodiment;

[0044] Figure 2 1 is a flow chart of a method for predicting incoming customer calls in another embodiment;

[0045] Figure 3 A schematic diagram of the principle of a method for predicting incoming customer calls in one embodiment;

[0046] Figure 4 A structural block diagram of a device for predicting incoming customer calls in one embodiment;

[0047] Figure 5 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0049] In one embodiment, Figure 1 As shown, a method for predicting incoming customer calls is provided. This embodiment uses the method applied to a server as an example for illustration. It is understood that the method can also be applied to a terminal, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0050] Step 102: Obtain historical business data of the target in-transit waybill.

[0051] The target in-transit waybill refers to the in-transit waybill for which customer call prediction is being made. Specifically, it may refer to the in-transit waybill for which a call from the corresponding customer is being predicted within a preset time period. An in-transit waybill refers to a waybill that has not yet been signed for. The target in-transit waybill refers to the waybill that has not yet been signed for at the time the customer call prediction process is triggered. Historical business data refers to the business data corresponding to the in-transit waybill prior to the current time, that is, the actual business data of the in-transit waybill. Historical business data may include basic waybill information, routing information, complaint information, call information, and client user operation data corresponding to the in-transit waybill. Basic waybill information includes the origin and destination, shipping time, timeliness type, declared value, product code, gross weight, freight, insurance status, and promised time. Routing information includes dynamic routing time, dynamic routing type, exception reason, and exception status. Complaint information includes complaint creation time, complaint level, escalation, timeout, holdup, points of concern, customer feedback, and work order status. Call information includes whether a call was made and the time of the call. Clients, such as WeChat, have user operation data including the initiating channel, abnormal tasks, and initiating role.

[0052] Specifically, when the customer call prediction conditions are met, the server determines the target in-transit waybills at the current time and retrieves the historical transaction data for each target in-transit waybills. Customer call prediction conditions are the basis or conditions for triggering the customer call prediction process, such as when the current time matches a preconfigured customer call prediction time or when a user-triggered customer call prediction trigger instruction is obtained. These conditions are not detailed here.

[0053] In one embodiment, the server obtains waybills that were created within a preset time period in the past and have not yet been signed for at the current time, and determines the obtained waybills as target in-transit waybills. The preset time period in the past refers to a time period before the current time, and specifically may refer to a time period or time interval with the current time as the end time and a preset time length, where the preset time length is a pre-set time length. For example, a preset time period in the past within 3 days refers to a time interval with the current time as the end time and a time length of 3 days. The current time refers to the time point at which the customer call prediction process is currently triggered.

[0054] Step 104: Obtain corresponding business characteristic data based on the historical business data of each target in-transit waybill.

[0055] Business feature data refers to data corresponding to business features, specifically data extracted from business data according to the business feature definition. Business features include time features, waybill statistics, and shipment volume statistics, and may also include whether the shipping address is in the same city. Accordingly, business feature data includes time feature data, waybill statistics, and shipment volume statistics, and may also include business data such as whether the shipping address is in the same city. Time features are time difference features derived by pairwise combining the collection time, routing time, shipping time, complaint time, commitment time, call time, and client operation time for the target in-transit waybill. Examples include "commitment time - collection time," "collection time - shipping time," "collection time - call time," "call time - client operation time," and "routing time - complaint time," which are not listed here. Waybill statistics features refer to features derived from statistical analysis of the historical business data of the target in-transit waybill, such as the number of complaints, calls, and client operations for the target in-transit waybill over a preset time period. The statistical features of the shipment volume include, for example, the number of waybills of the product type with the largest shipment volume and the average time difference between the incoming calls to the waybills, as well as the shipping address with the largest shipment volume, and the number of waybills in the shipping area and the average time difference between the incoming calls to the waybills.

[0056] Specifically, the server locally pre-configures multiple business features for in-transit waybills. After acquiring the historical business data for each in-transit waybill, the server determines the business feature data corresponding to each business feature based on the historical business data of each target in-transit waybill and the pre-configured feature definition of each business feature, thereby obtaining the current business feature data corresponding to each target in-transit waybill.

[0057] Step 106 : extracting the customer's incoming call habit feature data corresponding to each target in-transit waybill from the pre-stored customer's incoming call habit feature set.

[0058] The customer calling habit feature set is a collection of calling habit feature data for multiple customers, with each customer corresponding to a set of customer calling habit feature data. Customer calling habit feature data is used to characterize a customer's calling habits and can be determined by matching the customer's historical calling data with historical waybill data. Customer calling habit features refer to the customer's calling habit characteristics, including the customer's preferred or preferred calling times, shipping locations, and shipping products. Examples include the average time difference between calls for waybills, the average time difference between calls within and outside the same city, the average time difference between calls as a recipient or sender, the volume of shipments, the volume of shipments received, the volume of third-party shipments, the volume of same-city shipments, the volume of non-same-city shipments, the volume of standard express and same-day shipments and their corresponding average time difference, the volume of preferential transfer and small shipments and their corresponding average time difference, the volume of shipments for other product types and their corresponding average time difference, the product type with the highest volume, the time-sensitive type with the highest volume and its corresponding volume of waybills, and the average time difference between calls for waybills of that time-sensitive type.

[0059] Specifically, the server is locally pre-configured with a customer call habit feature set. The customer call habit feature data in the customer call habit feature set is associated with the corresponding customer via a customer identifier. After acquiring the historical business data for each target in-transit waybill, the server obtains the customer identifier corresponding to each target in-transit waybill. Based on each acquired customer identifier, the server searches the pre-configured customer call habit feature set for the associated customer call habit feature data, which is used as the customer call habit feature data corresponding to the target in-transit waybill.

[0060] In one embodiment, the server determines the customer identifier corresponding to each target in-transit waybill based on the historical business data of the target in-transit waybill. Each target in-transit waybill may also be directly associated with a customer identifier, so that the server can directly determine the corresponding customer identifier based on the target in-transit waybill.

[0061] In one embodiment, the server regularly updates the client's call habit feature set according to a pre-configured update cycle. The update cycle refers to the time interval between two consecutive updates of the client's call habit feature set, which can be customized, such as 6 months.

[0062] Step 108: Obtain corresponding target feature data based on the service feature data corresponding to each target in-transit waybill and the customer's incoming call habit feature data.

[0063] Specifically, the server combines the service feature data corresponding to each target in-transit waybill with the customer's incoming calling habits feature data to obtain the target feature data corresponding to the target in-transit waybill. Thus, the target feature data corresponding to each target in-transit waybill includes the service feature data corresponding to the target in-transit waybill and the customer's incoming calling habits feature data.

[0064] Step 110 , using the trained customer call prediction model, predict customer calls based on target feature data to obtain a list of incoming call customers within a preset time period in the future.

[0065] The customer call prediction model is trained based on a pre-acquired training sample set and can predict whether a customer will call within a preset future time period based on target feature data. The training sample set includes the sample feature data and call labels for each sample in-transit waybill. The future preset time period refers to a period after the current time, specifically a time period or time interval starting at a point after the current time and ending at a point after the start time. For example, if the current time is 8:00, the start and end times of the future preset time period are 9:00 and 12:00, respectively. Therefore, the future preset time period is from 9:00 to 12:00. For example, if the current time is 13:00, the start and end times of the future preset time period are 14:00 and 17:00, respectively. Therefore, the future preset time period is from 14:00 to 17:00. The caller list is a collection of customers expected to call within the preset future time period and specifically includes the customer identifiers corresponding to each customer expected to call within the preset future time period.

[0066] Specifically, the server inputs the target feature data corresponding to each target in-transit waybill into a trained customer call prediction model. The customer call prediction model then uses the target feature data to predict customer calls for each target in-transit waybill, obtaining a target waybill call label corresponding to each target in-transit waybill within a preset future time period. Specifically, the customer call prediction model determines, based on the target feature data for each target in-transit waybill, whether the corresponding customer will call within the preset future time period and, based on the determination, obtains the corresponding target waybill call label. Furthermore, based on the target waybill call label corresponding to each target in-transit waybill, the server filters out customers who are expected to call within the preset future time period from among the customers corresponding to each target in-transit waybill. Based on the filtered customers, the server obtains a list of customers who will call within the preset future time period. The target waybill call label is a label that indicates whether the customer corresponding to the target waybill will call within the preset future time period and can specifically include a yes or no call label.

[0067] In one embodiment, the server uses a customer call prediction model to predict customer calls for the corresponding target in-transit waybill based on the target feature data, obtains the corresponding target waybill call tag, and for the target in-transit waybill whose target waybill call tag indicates a call will come, also obtains the corresponding call intention. The server uses a customer call prediction model to predict customer calls for the corresponding target in-transit waybill based on the target feature data, and can obtain the target waybill call tag and call intention corresponding to each target in-transit waybill, wherein when the target waybill call tag indicates a call will not come, the corresponding call intention is empty. In this way, based on the target waybill call tag and call intention corresponding to the target in-transit waybill, more accurate proactive services can be provided to users, thereby further reducing customer service costs.

[0068] The above-mentioned customer call prediction method obtains the historical business data of the target in-transit waybill, obtains the corresponding business feature data based on the historical business data of each target in-transit waybill, extracts the customer call habit feature data corresponding to each target in-transit waybill from the pre-stored customer call habit feature set, and based on the business feature data and customer call habit feature data corresponding to each target in-transit waybill, can comprehensively obtain the target feature data that affects customer calls. Furthermore, through the trained customer call prediction model, it is predicted based on the target feature data whether the corresponding customer will call within a preset time period in the future, and a list of calling customers within the preset time period in the future is obtained. This can improve the accuracy of customer call prediction, so as to provide proactive services to the corresponding customers based on the customer call list with higher accuracy, and can reduce customer service costs.

[0069] In one embodiment, step 102 includes: obtaining the current time; when the current time is consistent with a preconfigured customer call prediction time, determining a target in-transit waybill for the current time; preconfiguring two customer call prediction times in each prediction cycle; and obtaining historical business data for each target in-transit waybill.

[0070] The customer call prediction time refers to the time point at which the customer call prediction process is triggered. The prediction cycle refers to the period or duration of time that triggers the same customer call prediction process. The customer call prediction time that triggers the customer call prediction process is consistent within each prediction cycle. For example, a daily prediction cycle can have two pre-configured customer call prediction times, such as 8:00 and 1:00. Therefore, the customer call prediction process will be triggered at these two customer call prediction times every day.

[0071] Specifically, the server pre-configures two predicted customer call times for each prediction period. The server obtains the current time in real time and compares it with the pre-configured predicted customer call times. If the current time matches either of the pre-configured predicted customer call times, the server identifies the target in-transit waybill for the current time and obtains historical transaction data for each target in-transit waybill.

[0072] In one embodiment, for each target in-transit waybill, the server obtains the newly added business data during the time period from the last time the customer call prediction process was triggered to the current time the customer call prediction process is triggered, and obtains the historical business data when the customer call prediction process is triggered this time based on the historical business data when the customer call prediction process was triggered last time and the newly added business data, that is, obtains the historical business data corresponding to the target in-transit waybill at the current time.

[0073] In one embodiment, taking a prediction cycle of one day or every day as an example, the customer call prediction process is triggered twice a day. In this way, each waybill will trigger the customer call prediction process multiple times during its life cycle, which can improve the flexibility and accuracy of customer call prediction. Moreover, triggering the customer call process twice a day will result in the output of two lists of incoming callers. In other words, two time periods will be used every day to predict customer calls for the waybill in transit to predict whether the customer corresponding to the waybill in transit will call within the corresponding time period. In this way, based on the output list of incoming callers, the time range of customer calls can be more accurately captured, so as to provide proactive services in a timely manner. The life cycle of a waybill refers to the entire process from the creation to the receipt of the waybill. For example, if the life cycle of a waybill is 4 days, the customer call prediction process will be triggered 8 times for the waybill.

[0074] In the above embodiment, the customer call prediction process is triggered twice for the target in-transit waybill in each prediction cycle. Each time, the customer call prediction process combines the existing business data with the dynamically added business data to determine the current historical business data, so as to improve the accuracy of customer call prediction when performing customer call prediction based on historical business data.

[0075] In one embodiment, step 110 includes: inputting the target feature data into a trained customer call prediction model to obtain the corresponding target waybill call tag and call intention; based on the target waybill call tag and call intention, obtaining a list of calling customers within a preset time period in the future, and the calling intention corresponding to each customer in the calling customer list.

[0076] Among them, call intent refers to the intention or purpose of the customer's call, such as urging the order, or inquiring about the logistics or delivery status of the waybill, etc., which are not listed here one by one.

[0077] Specifically, the server inputs the target feature data corresponding to each target in-transit waybill into a trained customer call prediction model. The model then uses the target feature data to predict customer calls for the corresponding target in-transit waybill, obtaining the corresponding target waybill call label and call intent. Based on the target waybill call label corresponding to each target in-transit waybill, the server determines customers expected to call within a preset future time period, thereby obtaining a list of customers calling within that preset future time period. Based on the call intent corresponding to each target in-transit waybill, the server determines the call intent of customers expected to call within that preset future time period, thereby obtaining the call intent of each customer in the caller list.

[0078] In the above embodiment, while predicting whether a customer will call within a preset time period in the future based on target feature data, the customer's calling intention is also predicted, so that more accurate proactive services can be provided to customers who will call within the preset time period in the future based on the calling intention, thereby further reducing customer service costs.

[0079] In one embodiment, the training steps of the customer call prediction model include: obtaining a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill call label; and performing model training based on the training sample set to obtain a trained customer call prediction model.

[0080] Specifically, during the model training phase, the server identifies multiple sample waybills in transit, obtains the sample feature data and call labels corresponding to each sample waybill, and generates a training sample set based on the sample feature data and call labels corresponding to each sample waybill. This allows for model training based on the training sample set to produce a trained customer call prediction model. The server uses the sample feature data in the training sample set as input features and the corresponding call labels as the desired output features for model training, resulting in a trained customer call prediction model.

[0081] In one embodiment, the server determines the in-transit waybills at each customer call prediction time within multiple prediction cycles as sample in-transit waybills, and obtains sample feature data corresponding to the sample in-transit waybills based on the historical business data of each sample in-transit waybills before the corresponding customer call prediction time and the customer call habit feature set pre-stored at the corresponding customer call prediction time, and obtains the sample waybill call label corresponding to the sample in-transit waybills based on the customer call data corresponding to each sample in-transit waybills in a future preset time period after the corresponding customer call prediction time, and then obtains a training sample set based on the sample feature data and sample waybill call labels corresponding to each sample in-transit waybills.

[0082] Taking a prediction period of one day or every day, and two customer call prediction times pre-configured in each prediction period as an example, multiple prediction periods are, for example, 4 days, specifically the most recent 4 days, and two customer call prediction times are pre-configured in each prediction period, such as 8:00 and 13:00. When the customer call prediction time is 8:00, the corresponding future preset time period is from 9:00 to 12:00 on the same day. In this way, the in-transit waybills at 8:00 and 13:00 every day in the recent 4 days are all determined as sample in-transit waybills, and the historical business data of each sample in-transit waybills and the customer call data of each sample in-transit waybills in the corresponding future preset time period are obtained. Based on the historical business data, corresponding sample feature data are obtained, and based on the customer call data, corresponding sample waybill call labels are obtained. Taking the customer call prediction time of 8:00 as an example, the customer call data corresponding to the sample in-transit waybills corresponding to the customer call prediction time refers to the customer call data of the sample in-transit waybills in the future time period from 9:00 to 12:00 on the day when the corresponding historical business data was collected.

[0083] In one embodiment, the machine learning algorithm involved in training the customer call prediction model includes but is not limited to GBDT (Gradient Boosting Decision Tree), specifically LGBM (LightGBM, lightweight and efficient gradient boosting tree).

[0084] In one embodiment, the model parameters involved in training the customer call prediction model and the parameter values corresponding to each model parameter include but are not limited to: the classification target is binary classification, the loss function is binary_logloss (binary classification logarithmic loss), the number of leaf nodes is 31, the minimum number of data on a leaf is 1, the learning rate is 0.1, the proportion of randomly selected features in each iteration is 0.9, the proportion of randomly selected part of the data without resampling is 0.8, the number of bagging is 5, lambda_l1 (L1 regularization) and lambda_l2 (L2 regularization) are both 0.2, and the threshold for stopping splitting leaf nodes is -1. It can be understood that the model parameters and corresponding parameter values listed in this embodiment are only examples and are not used for specific limitations.

[0085] In one embodiment, the trained customer call prediction model is a multi-classification model, so that the target waybill call label and call intention corresponding to the corresponding target in-transit waybill can be predicted based on the target feature data through the multi-classification customer call prediction model.

[0086] In the above embodiment, a trained customer call prediction model is obtained by pre-training based on a training sample set, so that in the actual application of customer call prediction, the trained customer call prediction model can be used to quickly and accurately predict whether the corresponding customer will call within a preset time period in the future based on the target feature data of each target in-transit waybill.

[0087] In one embodiment, obtaining a training sample set includes: obtaining historical business data corresponding to a sample in-transit waybill, and customer call data within a preset time period in the future of a corresponding customer call prediction time; obtaining corresponding sample feature data based on the historical business data of the sample in-transit waybill and a customer call habit feature set; determining a corresponding sample waybill call label based on the customer call data of the sample in-transit waybill; and obtaining a training sample set based on the sample feature data and the sample waybill call label.

[0088] Among them, the customer call data corresponding to the sample in-transit order refers to the call data of the customer corresponding to the sample in-transit order within a future preset time period after the predicted call time of the customer corresponding to the sample in-transit order, which may specifically include whether there is a call and the call time of each call.

[0089] Specifically, after determining multiple sample waybills in transit, the server obtains the historical business data corresponding to each sample waybill in transit, as well as the customer call data for each sample waybill in the future preset time period after the corresponding customer call prediction time. The server obtains the corresponding business feature data based on the historical business data of each sample waybill in transit, and obtains the customer call habit feature data corresponding to each sample waybill in transit from the preconfigured customer call habit feature set, and then obtains the corresponding sample feature data based on the business feature data and customer call habit feature data corresponding to each sample waybill in transit. The server determines the corresponding sample waybill call label based on the customer call data corresponding to each sample waybill in transit, and obtains the training sample set based on the sample feature data and the sample waybill call label corresponding to each sample waybill in transit.

[0090] It can be understood that during the model training phase, the server obtains the corresponding sample feature data based on the historical business data corresponding to each sample in-transit waybill and the pre-configured customer call habit feature set. This process is similar to the corresponding process provided for the customer call prediction phase in one or more embodiments of the present application and will not be repeated here.

[0091] In the above embodiment, by obtaining the historical business data and customer call data corresponding to each sample in-transit waybill, the sample feature data is determined based on the historical business data and the preconfigured customer call habit feature set, the sample waybill call label is determined based on the customer call data, and the training sample set is obtained based on the sample feature data and the sample waybill call label corresponding to each sample in-transit waybill, so that a customer call prediction model with higher accuracy can be trained based on the training sample set.

[0092] In one embodiment, model training is performed based on a training sample set to obtain a trained customer call prediction model, including: performing model training based on the training sample set to obtain a customer call prediction model; obtaining a test sample set; testing the customer call prediction model based on the test sample set to obtain a test result; when the test result meets a preset test condition, determining the customer call prediction model as the trained customer call prediction model.

[0093] The test results are used to characterize the prediction accuracy of the trained customer call prediction model, specifically including recall and precision. The preset test conditions are used to compare the test results to determine whether the trained customer call prediction model should be considered the trained customer call prediction model. Specifically, they can be conditions or criteria such as a recall greater than or equal to a recall threshold and a precision greater than or equal to a precision threshold. The recall threshold is customizable, such as 73%, and the precision threshold is customizable, such as 88%. The recall rate is the ratio of the number of waybills in the test sample set that were predicted to receive a call and actually received a call to the number of waybills that actually received a call. The precision rate is the ratio of the number of waybills in the test sample set that were predicted to receive a call and actually received a call to the number of waybills that were predicted to receive a call. For example, if the number of waybills in the test sample set that were predicted to receive a call and actually received a call is denoted as P, the number of waybills in the test sample set that actually received a call is denoted as A, and the number of waybills in the test sample set that were predicted to receive a call is denoted as B, then the recall rate is P / A and the precision rate is P / B.

[0094] Specifically, the server performs model training based on the training sample set to obtain a corresponding customer call prediction model. Accordingly, the server obtains a test sample set in a similar manner to the training sample set, and the test sample set includes test feature data and test waybill call labels corresponding to multiple test waybills in transit. After the server obtains the customer call prediction model based on the training sample set, it inputs the test feature data corresponding to each test waybill in transit in the test sample set into the customer call prediction model to perform customer call prediction and obtain the corresponding predicted waybill call label. Based on the test waybill call label and the predicted waybill call label corresponding to each test waybill in transit in the test sample set, the server calculates the corresponding test result and compares the test result with the pre-configured preset test conditions. When it is determined that the test result meets the preset test conditions, the server determines the customer call prediction model obtained by training based on the training sample set as the trained customer call prediction model.

[0095] In one embodiment, when the test result does not meet the preset test conditions, the server continues to iteratively train the customer call prediction model whose test result does not meet the preset test conditions based on the pre-acquired training sample set or the re-acquired training sample set, and retests the customer call prediction model obtained by continued training based on the test sample set until the test result obtained meets the preset test conditions, or when the number of tests is greater than or equal to the test number threshold, the customer call prediction model obtained by the current training is determined as the trained customer call prediction model.

[0096] In the above embodiment, the customer call prediction model trained based on the training sample set is tested based on the test sample set, and when the test result meets the preset test conditions, the trained customer call prediction model is determined as the trained customer call prediction model, so that when the customer call prediction is performed based on the trained customer call prediction model, the accuracy of the customer call prediction result can be improved.

[0097] In one embodiment, the above-mentioned customer incoming call prediction method also includes: regularly obtaining target incoming call data and target waybill data corresponding to each customer according to a preconfigured update cycle; updating the pre-stored customer incoming call habit feature set based on the target incoming call data and target waybill data.

[0098] Target incoming call data refers to newly added incoming call data within a single update cycle, specifically, the period from the last time the customer incoming call habit feature set update process was triggered to the current time the process was triggered. Target waybill data refers to newly added or updated waybill data within a single update cycle, specifically, the period from the last time the customer incoming call habit feature set update process was triggered to the current time the process was triggered.

[0099] Specifically, the server counts the time interval since the last time the update process of the customer's incoming call habit feature set was triggered, and compares the counted time interval with the pre-configured update period in real time. When the counted time interval is greater than or equal to the update period, the server obtains the newly added incoming call data and waybill data of each customer within the time interval, and uses them as the target incoming call data and target waybill data corresponding to the customer. The server obtains the current corresponding historical incoming call data and historical waybill data based on the historical incoming call data and historical waybill data corresponding to each customer when the update process of the customer's incoming call habit feature set was triggered last time, as well as the currently obtained target incoming call data and target waybill data. The server analyzes the current corresponding historical incoming call data and historical waybill data of each customer to obtain the current corresponding customer incoming call habit feature data of the customer, and updates the pre-stored customer incoming call habit feature set based on the current corresponding customer incoming call habit feature data of each customer.

[0100] In the above embodiment, the customer call habit feature set is regularly updated according to a preset update cycle, so that when the customer call prediction is performed based on the dynamically updated customer call habit feature set, the accuracy of the customer call prediction can be improved.

[0101] like Figure 2 As shown, a method for predicting customer incoming calls is provided, which specifically includes the following steps:

[0102] Step 202: Get the current time.

[0103] Step 204 : When the current time is consistent with the pre-configured customer incoming call forecast time, determine the target in-transit waybill for the current time; two customer incoming call forecast times are pre-configured in each forecast period.

[0104] Step 206: Obtain historical business data of each target in-transit waybill.

[0105] Step 208: Obtain corresponding business characteristic data based on the historical business data of each target in-transit waybill.

[0106] Step 210 : extracting the customer's incoming call habit feature data corresponding to each target in-transit waybill from the pre-stored customer's incoming call habit feature set.

[0107] Step 212: Obtain corresponding target feature data based on the service feature data corresponding to each target in-transit waybill and the customer's incoming call habit feature data.

[0108] Step 214: input the target feature data into the trained customer call prediction model to obtain the corresponding target waybill call label and call intention.

[0109] Step 216 , obtaining a caller list within a preset time period in the future and the caller intention corresponding to each customer in the caller list based on the caller tag and the caller intention of the target waybill.

[0110] In the above embodiment, two customer call prediction times are set within each prediction cycle. At each customer call prediction time, customer call predictions are performed on all currently in-transit waybills to predict whether each in-transit waybill will receive a call within a preset time period in the future. This allows for real-time acquisition of dynamically changing feature information about the waybill, and timely prediction of customer call intentions, thereby enabling effective proactive service. Furthermore, corresponding target feature data is derived based on the historical business data of the target in-transit waybill and the feature data of the customer's call habits. This allows for comprehensive acquisition of feature data that influences customer calls, thereby improving the accuracy of customer call predictions when predicting customer call situations based on the target feature data. Furthermore, by using a trained customer call prediction model, customer call predictions are performed on each target in-transit waybill based on its target feature data, further improving the accuracy of customer call predictions.

[0111] Figure 3 FIG. 1 is a schematic diagram showing the principle of a method for predicting incoming customer calls in one embodiment. Figure 3 Two customer call prediction times are set daily, at 8:00 AM and 1:00 PM. Thus, the customer call prediction process is triggered at 8:00 AM and 1:00 PM each day. After the customer call prediction process is triggered, historical business data for the currently unsigned waybills is first extracted. Specifically, historical business data for the target in-transit waybills is extracted. This historical business data includes basic waybill information, routing information, complaint information, call information, and client user operation information. Then, based on the historical business data for each target in-transit waybill and a preconfigured customer call habit feature set, target feature data corresponding to each target in-transit waybill is calculated through feature engineering. The steps of calculating the target feature data for each target in-transit waybill through feature engineering include: obtaining business feature data based on historical business data, such as how long after shipping a package a complaint occurs; extracting customer call habit feature data corresponding to each target in-transit waybill from the preconfigured customer call habit feature set, such as the average time after shipping a package within the same city for a customer to call; and obtaining target feature data based on the business feature data and the customer call habit feature data.

[0112] Furthermore, using the trained customer call prediction model, customer call predictions are performed based on the target feature data to obtain corresponding target waybill call labels, and then a list of callers who will call within the next three hours is obtained and output. The customer call prediction model uses "target feature data corresponding to waybills at 8:00 or 13:00" - "waybill call labels indicating whether the waybill will call within the next three hours" as a set of data, extracts eight sets of data from the last four days as a training sample set, and trains the model based on this training sample set. It will be understood that the above embodiment, which uses a prediction period of one day, two customer call prediction times within each prediction period, and a length of three hours for each future preset time period, is not intended to be a specific limitation.

[0113] In one embodiment, the historical business data corresponding to each target in-transit waybill or sample in-transit waybill may specifically be business data within a specified time period closest to the predicted time of a corresponding customer call. The specified time period may be, for example, the last three days, which is not specifically limited here.

[0114] It should be understood that although Figure 1-2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1-2 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0115] In one embodiment, Figure 4 As shown, a customer call prediction device 400 is provided, comprising: a service data acquisition module 401, a service feature extraction module 402, a habit feature extraction module 403, a feature data acquisition module 404 and a customer call prediction module 405, wherein:

[0116] Business data acquisition module 401, used to obtain historical business data of the target in-transit waybill;

[0117] A business feature extraction module 402 is used to obtain corresponding business feature data based on the historical business data of each target in-transit waybill;

[0118] The habit feature extraction module 403 is used to extract the customer call habit feature data corresponding to each target in-transit waybill from the pre-stored customer call habit feature set;

[0119] The characteristic data acquisition module 404 is used to obtain corresponding target characteristic data based on the business characteristic data corresponding to each target in-transit waybill and the customer's calling habit characteristic data;

[0120] The customer incoming call prediction module 405 is used to predict customer incoming calls based on the target feature data using the trained customer incoming call prediction model to obtain a list of incoming call customers within a preset time period in the future.

[0121] In one embodiment, the business data acquisition module 401 is also used to obtain the current time; when the current time is consistent with the preconfigured customer call prediction time, determine the target in-transit waybill at the current time; two customer call prediction times are preconfigured in each prediction cycle; and obtain historical business data for each target in-transit waybill.

[0122] In one embodiment, the customer call prediction module 405 is also used to input the target feature data into a trained customer call prediction model to obtain the corresponding target waybill call label and call intention; based on the target waybill call label and call intention, a list of calling customers within a preset time period in the future is obtained, as well as the calling intention corresponding to each customer in the calling customer list.

[0123] In one embodiment, the customer incoming call prediction device 400 further includes: a model training module;

[0124] The model training module is used to obtain a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill call label; the model is trained based on the training sample set to obtain a trained customer call prediction model.

[0125] In one embodiment, the model training module is also used to obtain historical business data corresponding to the sample in-transit waybill, as well as customer call data within a preset time period in the future of the corresponding customer call prediction time; obtain corresponding sample feature data based on the historical business data of the sample in-transit waybill and the customer call habit feature set; determine the corresponding sample waybill call label based on the customer call data of the sample in-transit waybill; and obtain a training sample set based on the sample feature data and the sample waybill call label.

[0126] In one embodiment, the model training module is further used to perform model training based on the training sample set to obtain a customer call prediction model; obtain a test sample set; test the customer call prediction model based on the test sample set to obtain a test result; when the test result meets the preset test conditions, the customer call prediction model is determined to be a trained customer call prediction model.

[0127] In one embodiment, the habit feature extraction module 403 is further used to regularly obtain target incoming call data and target waybill data corresponding to each customer according to a preconfigured update cycle; and update the pre-stored customer incoming call habit feature set based on the target incoming call data and target waybill data.

[0128] The specific definition of the customer call prediction device can be found in the definition of the customer call prediction method above and will not be repeated here. Each module in the aforementioned customer call prediction device can be implemented in whole or in part via software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0129] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store a customer call habit feature set and business data of the waybill. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a customer call prediction method is implemented.

[0130] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0131] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0132] Obtain historical business data of the target in-transit waybill; obtain corresponding business feature data based on the historical business data of each target in-transit waybill; extract the customer call habit feature data corresponding to each target in-transit waybill from the pre-stored customer call habit feature set; obtain corresponding target feature data based on the business feature data and customer call habit feature data corresponding to each target in-transit waybill; use the trained customer call prediction model to predict customer calls based on the target feature data, and obtain a list of calling customers within a preset time period in the future.

[0133] In one embodiment, when the processor executes the computer program, the following steps are further implemented: obtaining the current time; when the current time is consistent with the preconfigured customer call prediction time, determining the target in-transit waybill at the current time; preconfiguring two customer call prediction times in each prediction cycle; and obtaining historical business data for each target in-transit waybill.

[0134] In one embodiment, when the processor executes the computer program, it also implements the following steps: inputting the target feature data into a trained customer call prediction model to obtain the corresponding target waybill call tag and call intention; based on the target waybill call tag and call intention, obtaining a list of calling customers within a preset time period in the future, as well as the calling intention corresponding to each customer in the calling customer list.

[0135] In one embodiment, when the processor executes the computer program, it also implements the following steps: obtaining a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill incoming call label; performing model training based on the training sample set to obtain a trained customer call prediction model.

[0136] In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining historical business data corresponding to the sample in-transit waybill, and customer call data within a preset time period in the future of the corresponding customer call prediction time; obtaining corresponding sample feature data based on the historical business data of the sample in-transit waybill and the customer call habit feature set; determining the corresponding sample waybill call label based on the customer call data of the sample in-transit waybill; and obtaining a training sample set based on the sample feature data and the sample waybill call label.

[0137] In one embodiment, when the processor executes the computer program, the following steps are further implemented: performing model training based on the training sample set to obtain a customer call prediction model; obtaining a test sample set; testing the customer call prediction model based on the test sample set to obtain a test result; when the test result meets a preset test condition, determining the customer call prediction model as a trained customer call prediction model.

[0138] In one embodiment, when the processor executes the computer program, it also implements the following steps: regularly obtaining the target incoming call data and target waybill data corresponding to each customer according to a preconfigured update cycle; and updating the pre-stored customer incoming call habit feature set based on the target incoming call data and target waybill data.

[0139] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0140] Obtain historical business data of the target in-transit waybill; obtain corresponding business feature data based on the historical business data of each target in-transit waybill; extract the customer call habit feature data corresponding to each target in-transit waybill from the pre-stored customer call habit feature set; obtain corresponding target feature data based on the business feature data and customer call habit feature data corresponding to each target in-transit waybill; use the trained customer call prediction model to predict customer calls based on the target feature data, and obtain a list of calling customers within a preset time period in the future.

[0141] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented: obtaining the current time; when the current time is consistent with a preconfigured customer call prediction time, determining a target in-transit waybill for the current time; preconfiguring two customer call prediction times within each prediction cycle; and obtaining historical business data for each target in-transit waybill.

[0142] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: inputting the target feature data into a trained customer call prediction model to obtain the corresponding target waybill call label and call intention; based on the target waybill call label and call intention, obtaining a list of calling customers within a preset time period in the future, as well as the calling intention corresponding to each customer in the calling customer list.

[0143] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill incoming call label; performing model training based on the training sample set to obtain a trained customer call prediction model.

[0144] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining historical business data corresponding to the sample in-transit waybill and customer call data within a preset time period in the future of the corresponding customer call prediction time; obtaining corresponding sample feature data based on the historical business data of the sample in-transit waybill and the customer call habit feature set; determining the corresponding sample waybill call label based on the customer call data of the sample in-transit waybill; and obtaining a training sample set based on the sample feature data and the sample waybill call label.

[0145] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: performing model training based on a training sample set to obtain a customer call prediction model; obtaining a test sample set; testing the customer call prediction model based on the test sample set to obtain a test result; when the test result meets a preset test condition, determining the customer call prediction model as a trained customer call prediction model.

[0146] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: regularly obtaining the target incoming call data and target waybill data corresponding to each customer according to a preconfigured update cycle; and updating the pre-stored customer incoming call habit feature set based on the target incoming call data and target waybill data.

[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0148] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The above embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for predicting incoming customer calls, characterized in that: The method comprises: Obtaining historical business data of the target in-transit waybill, wherein the historical business data includes at least one of basic waybill information, routing information, complaint information, incoming call information, and client user operation data corresponding to the in-transit waybill; Obtain corresponding business characteristic data based on the historical business data of each target in-transit order, wherein the business characteristic data includes time characteristic data, which is used to represent the time difference between the collection time, routing time, shipping time, complaint time, commitment time, call time, and client operation time corresponding to the target in-transit order; Extracting customer call habit feature data corresponding to each target in-transit waybill from a pre-stored customer call habit feature set, wherein the customer call habit feature data includes at least one of the customer's customary or preferred call time, the customer's customary or preferred shipping location, the customer's customary or preferred shipping product, and an average call time difference; Obtaining corresponding target feature data based on the business feature data corresponding to each target in-transit waybill and the customer's call habit feature data; The target feature data is input into a trained customer call prediction model to obtain the corresponding target waybill call tag and call intention; based on the target waybill call tag and the call intention, a list of calling customers within a preset time period in the future is obtained, as well as the call intention corresponding to each customer in the calling customer list, wherein the call intention includes at least one of the intention to urge the order and the intention to inquire about the waybill status.

2. The method according to claim 1, characterized in that The acquisition of historical business data of the target in-transit waybill includes: Get the current time; When the current time is consistent with the pre-configured customer call forecast time, determining the target in-transit waybill for the current time; two customer call forecast times are pre-configured in each forecast period; Obtain historical business data for each target in-transit waybill.

3. The method according to claim 1, characterized in that The training steps of the customer call prediction model include: Obtain a training sample set; the training sample set includes sample feature data corresponding to the sample in-transit waybill and the sample waybill call label; Model training is performed based on the training sample set to obtain a trained customer incoming call prediction model.

4. The method according to claim 3, characterized in that The obtaining of the training sample set includes: Obtain historical business data corresponding to the sample in-transit waybills, as well as customer call data within a preset time period in the future of the corresponding customer call forecast time; Obtaining corresponding sample feature data based on the historical business data of the sample in-transit waybill and the customer's call habit feature set; Determine a corresponding sample waybill call label based on customer call data of the sample in-transit waybill; A training sample set is obtained according to the sample feature data and the sample waybill call label.

5. The method according to claim 3, characterized in that The performing model training based on the training sample set to obtain a trained customer call prediction model includes: Performing model training based on the training sample set to obtain a customer call prediction model; Get a test sample set; Testing the customer incoming call prediction model according to the test sample set to obtain a test result; When the test result meets the preset test condition, the customer incoming call prediction model is determined as a trained customer incoming call prediction model.

6. The method according to any one of claims 1 to 5, characterized in that The method further comprises: Regularly obtain the target incoming call data and target waybill data corresponding to each customer according to the pre-configured update cycle; The pre-stored customer incoming call habit feature set is updated according to the target incoming call data and the target waybill data.

7. A customer call prediction device, characterized in that: The device comprises: A business data acquisition module, configured to acquire historical business data of a target in-transit waybill, wherein the historical business data includes at least one of basic waybill information, routing information, complaint information, incoming call information, and client user operation data corresponding to the in-transit waybill; A business feature extraction module is used to obtain corresponding business feature data based on the historical business data of each target in-transit order, wherein the business feature data includes time feature data, which is used to represent the time difference between the collection time, routing time, shipping time, complaint time, commitment time, call time and client operation time corresponding to the target in-transit order; a habit feature extraction module, configured to extract customer call habit feature data corresponding to each target in-transit waybill from a pre-stored customer call habit feature set, wherein the customer call habit feature data includes at least one of the customer's habitual or preferred call time, the customer's habitual or preferred shipping location, the customer's habitual or preferred shipping product, and an average call time difference; A feature data acquisition module, configured to obtain corresponding target feature data based on the business feature data corresponding to each target in-transit waybill and the customer's call habit feature data; The customer call prediction module is used to input the target feature data into the trained customer call prediction model to obtain the corresponding target waybill call label and call intention; based on the target waybill call label and the call intention, a list of calling customers within a preset time period in the future is obtained, as well as the call intention corresponding to each customer in the calling customer list, wherein the call intention includes at least one of the intention to urge the order and the intention to inquire about the waybill status.

8. The device according to claim 7, characterized in that The business data acquisition module is also used to obtain the current time; when the current time is consistent with the pre-configured customer call prediction time, determine the target in-transit waybill at the current time; two customer call prediction times are pre-configured in each prediction cycle; and obtain historical business data for each target in-transit waybill.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images