Predicting outbound call method, device, electronic device and storage medium

Abstract

The present application provides a method, device, electronic device and storage medium for predicting outbound calls. The method for predicting outbound calls includes: determining a first probability of successfully calling a customer after a first time; determining a second probability that the current agent has not ended the call after the first time; and determining whether to make an outbound call for the current agent after the first time based on the first probability and the second probability. In this way, it can be ensured to the greatest extent that an outbound call is made to the customer after the time, and after the customer answers the call, the current agent just ends the previous call, so that the current agent can seamlessly provide services to the next customer, reduce harassment to the customer, and then accurately predict outbound calls for each agent, and improve the service efficiency of each agent.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, electronic device and storage medium for predicting outbound calls. Background Art

[0002] Predictive outbound calling means that the call center makes calls to customers in advance through prediction, and after connecting with the customer, the connected call is assigned to the agent, who can communicate with the customer as soon as he picks up the phone. This can greatly improve the service efficiency of the agent.

[0003] The current predictive outbound calls are mainly based on the number of idle seats that are not currently in a call state, the number of idle relay channels, the average time to reduce an idle seat, and the average time to generate an idle seat from the seats that are currently in a call state.

[0004] However, the connection time for different customers varies, and the time spent by different agents in communicating with customers also varies. Therefore, the current predictive outbound call method cannot accurately predict outbound calls for each agent. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a method, device, electronic device and storage medium for predicting outbound calls, so as to improve the accuracy of predicting outbound calls and reduce harassment to customers.

[0006] In order to solve the above technical problems, the embodiments of the present application provide the following technical solutions:

[0007] A first aspect of the present application provides a method for predicting outbound calls, the method comprising: determining a first probability of a successful outbound call to a customer after a first time; determining a second probability that the current agent has not ended the call after the first time; and determining whether to make an outbound call for the current agent after the first time based on the first probability and the second probability.

[0008] The second aspect of the present application provides a predictive outbound call device, which includes: a first determination module, used to determine a first probability of a successful outbound call to a customer after a first time; a second determination module, used to determine a second probability that the current agent has not ended the call after the first time; and a predictive outbound call module, used to determine whether to make an outbound call for the current agent after the first time based on the first probability and the second probability.

[0009] The third aspect of the present application provides a predictive outbound call system, which includes: a business system for generating outbound call demands; a call center connected to the business system, for obtaining the outbound call demands and making outbound calls to customers based on the outbound call demands; a predictive outbound call device connected to the call center, for controlling whether the call center makes outbound calls to customers based on the outbound call demands after a first time; a customer connected to the call center, for receiving calls from the call center; and an agent connected to the call center, for connecting to customers connected by the call center.

[0010] The fourth aspect of the present application provides an electronic device, which includes: a processor, a memory, and a bus; wherein the processor and the memory communicate with each other through the bus; and the processor is used to call program instructions in the memory to execute the method in the first aspect.

[0011] A fifth aspect of the present application provides a computer-readable storage medium, the storage medium comprising: a stored program; wherein, when the program is executed, the device where the storage medium is located is controlled to execute the method in the first aspect.

[0012] Compared with the prior art, the predictive outbound call method provided in the first aspect of the present application first determines the first probability of a successful outbound call to a customer after the first time; then, determines the second probability that the current agent has not ended the call after the first time, thus determining the probability that the customer will answer the call after the first time and the probability that the current agent cannot provide service to the customer who answered the call; finally, based on the first probability and the second probability, it is possible to know the probability that the customer has answered the call but the current agent cannot provide service, thereby determining whether to make an outbound call for the current agent after the first time based on the probability. In this way, it is possible to ensure to the greatest extent that an outbound call is made to the customer after the time, and after the customer answers the call, the current agent just ends the previous call, so that the current agent can seamlessly provide service to the next customer, reducing harassment to the customer, and then accurately predicting outbound calls for each agent, and improving the service efficiency of each agent.

[0013] The predictive outbound call device provided in the second aspect of the present application, the predictive outbound call system provided in the third aspect, the electronic device provided in the fourth aspect, and the computer-readable storage medium provided in the fifth aspect have the same or similar beneficial effects as the predictive outbound call method provided in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0015] Figure 1 This is a schematic diagram of the architecture of the outbound call prediction method in the embodiment of the present application;

[0016] Figure 2 The process diagram of the method for predicting outbound calls in the embodiment of the present application is as follows Figure 1 ;

[0017] Figure 3 The process diagram of the method for predicting outbound calls in the embodiment of the present application is as follows Figure 2 ;

[0018] Figure 4 The structure of the device for predicting outbound calls in the embodiment of the present application is shown in FIG. Figure 1 ;

[0019] Figure 5 The structure of the device for predicting outbound calls in the embodiment of the present application is shown in FIG. Figure 2 ;

[0020] Figure 6 Schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0022] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by technicians in the field to which this application belongs.

[0023] In the related art, based on the historical data of outbound calls, the average time of reducing an idle seat and the average time of generating an idle seat from the seats currently in the call state are used to predict outbound calls. Since the time of success of each outbound call is different. That is to say, after making an outbound call to a customer, some customers answer the call quickly, so the time of success of the outbound call is shorter. And some customers take a long time to answer the call, so the time of success of the outbound call is longer. Even, not every outbound call to a customer can be successful. For example: the customer is on the phone, the customer is turned off, the customer is out of the service area, etc., which will lead to the failure of the outbound call to the customer. Therefore, the situation of each outbound call is different, and the average time of success of the outbound call cannot be accurately calculated. In addition, when the agent and the customer have started to communicate, the business capabilities of different agents, the understanding capabilities of different customers, and the complexity of the business of communication will be different, which will lead to different call durations between different agents and different customers, and thus the average time of the call cannot be accurately calculated. In the case where the average time of success of the outbound call and the average time of the call cannot be accurately calculated, there will be a situation where the customer has answered the phone, but there is no idle seat to provide service. This will cause harassment to users. Therefore, the current outbound call prediction method cannot accurately predict outbound calls for each agent.

[0024] In view of this, the embodiment of the present application provides a method for predicting outbound calls, which determines whether to make an outbound call for the current agent after a certain time based on the probability of successfully making an outbound call to the customer after a certain time, and the second probability that the current agent has not ended the call after a certain time. In this way, it can be ensured to the greatest extent that the outbound call to the customer is made after the time, and after the customer answers the call, the current agent just ends the previous call, so that the current agent can seamlessly provide services to the next customer, reduce harassment to the customer, and then accurately predict outbound calls for each agent, and improve the service efficiency of each agent.

[0025] First, the architecture of the outbound call prediction method provided in the embodiment of the present application is described.

[0026] Figure 1 This is a schematic diagram of the architecture of the outbound call prediction system in the embodiment of the present application, see Figure 1 As shown, the predictive outbound call system includes: a business system 101, a call center 102, a predictive outbound call device 103, an agent 105, and a customer 104.

[0027] Business system 101 is used to generate outbound call requirements.

[0028] Generally speaking, after the business system 101 generates an outbound call demand, it can trigger the task corresponding to the demand to be executed by the call center 102. For example, the business system 101 generates a return call demand, which in turn triggers the return call task to be executed in the call center 102.

[0029] The call center 102 is connected to the business system 101 and is used to obtain outbound call demands and make outbound calls to customers based on the outbound call demands.

[0030] Typically, the call center 102 receives a call from the customer 104 and then transfers the call to the corresponding agent 105 for processing. Alternatively, according to a task triggered by the business system 101, the call center 102 proactively makes an outbound call to the customer 104, and after the customer 104 answers the call, the call is transferred to the corresponding agent 105 for processing.

[0031] The outbound call prediction device 103 is connected to the call center 102 and is used to control whether the call center 102 makes an outbound call to the customer based on the outbound call demand after the first time.

[0032] The predictive outbound calling device 103 is the execution subject of the predictive outbound calling method provided in the embodiment of the present application. Before the call center 102 makes an outbound call to the customer 104, it is necessary to determine whether it is currently appropriate to make an outbound call to the customer 104 through the predictive outbound calling device 103, and then feed back the prediction result to the call center 102, so that the call center 102 can determine whether to make an outbound call to the customer 104 according to the feedback result of the predictive outbound calling device 103.

[0033] Client 104 is connected to call center 102 to receive calls from the call center.

[0034] Here, the customer 104 is a person who has service needs and needs corresponding services from the agent 105. Alternatively, the customer 104 is a person that the call center 102 actively calls out for return visits or business promotion.

[0035] The seat 105 is connected to the call center 102 and is used to connect to the customers connected by the call center 102 .

[0036] The seat 105 here can also be called a desk seat, a center seat, or a seat representative. It is generally composed of a seat computer, seat software, a seat headset, and a service staff. Generally, the service staff provides services to the customer 104 through the seat computer, seat software, and seat headset.

[0037] When customer 104 has business needs such as inquiries, consultations, complaints, etc., customer 104 will make an incoming call. After receiving the incoming call from customer 104, call center 102 will transfer the call from customer 104 to the corresponding agent 105 for processing. Alternatively, when it is necessary to conduct a return visit, satisfaction survey, etc. for customer 104, the business system 101 will trigger an outbound call task. Before calling customer 104, call center 102 determines whether it is appropriate to make an outbound call to customer 104 through the prediction outbound call device 103. After the prediction outbound call device 103 determines that it is appropriate to make an outbound call to customer 104, call center 102 makes an outbound call to customer 104. After customer 104 is connected, call center 102 transfers the call from customer 104 to the corresponding agent 105 for processing.

[0038] Next, the predictive outbound call method provided in the embodiment of the present application is described in detail.

[0039] Figure 2 The process diagram of the method for predicting outbound calls in the embodiment of the present application is as follows Figure 1 , see Figure 2 As shown, the method may include:

[0040] S201: Determine a first probability of successfully calling the customer after the first time.

[0041] Since the outbound call prediction is not to make an outbound call at the current time, but to predict a suitable time in the future, it is necessary to select a time in the future for calculation to determine whether the time is suitable for making an outbound call. The time here is the first time mentioned above.

[0042] It should be noted that the outbound call prediction method provided by the embodiment of the present application is performed in real time. That is, at the current moment, the outbound call prediction method provided by the embodiment of the present application is used to predict whether a certain time in the future is suitable for outbound calling. If it is suitable, then an outbound call is made when the time arrives; if it is not suitable, then another time in the future is selected, and the judgment is continued until a suitable time is found, and an outbound call is made at the suitable time.

[0043] The specific value of the first time selected can be 2s, 5s, 10s, 20s, 50s, etc., which is not specifically limited here. But generally speaking, the first time selected should not be too short. A first time that is too short will make the prediction result unsuitable, thereby increasing the number of predictions. The first time selected should not be too long either. A first time that is too long will make the prediction result suitable, and the prediction result corresponding to a time shorter than the first time may also be suitable, which will result in a late outbound call time and reduce the customer service efficiency.

[0044] In the specific process of outbound call prediction, you can choose a time from three different time lengths, namely, long time, medium time, and short time, to predict the outbound call. If the result of the predicted outbound call is that the outbound call can be made, then a shorter time is selected from the long time, medium time, and short time to predict the outbound call, until the result of the predicted outbound call is that the outbound call cannot be made. At this time, the time of the previous predicted outbound call is taken as the first time, and an outbound call is made for the current agent after the first time; and if the result of the predicted outbound call is that the outbound call cannot be made, then a longer time is selected from the long time, medium time, and short time to predict the outbound call judgment, until the result of the predicted outbound call is that the outbound call can be made. At this time, the current time of the predicted outbound call is taken as the first time, and an outbound call is made for the current agent after the first time.

[0045] For example: when making the first predictive outbound call, predict whether an outbound call can be made after 10 seconds. If yes, then continue to predict whether an outbound call can be made after 8 seconds. If not, then determine to make an outbound call for the current agent after 10 seconds. Another example: when making the first predictive outbound call, predict whether an outbound call can be made after 10 seconds. If not, then continue to predict whether an outbound call can be made after 12 seconds. If yes, then determine to make an outbound call for the current agent after 12 seconds.

[0046] After the first time is determined, the first probability of successfully calling the customer after the first time is determined, that is, the probability of being able to connect with the customer after the first time is predicted. Through the above first probability, the probability of connecting with the customer after the first time can be determined, and then combined with the probability of the current agent ending the call after the first time, it is determined whether to make an outbound call for the current agent after the first time.

[0047] Specifically, the first probability of successfully calling the customer after the first time can be determined by collecting statistics on historical outbound call data, estimating through specific services, etc. The specific determination method is not limited here.

[0048] S202: Determine a second probability that the current agent has not ended the call after the first time.

[0049] Since it takes a certain amount of time to wait from the time an outbound call is made to the time the call is answered, in order to improve the service efficiency of the agent, it is necessary to start making an outbound call for the next call while the agent is on the previous call. In this way, after the agent completes the previous call, the next call can be connected directly, which shortens the waiting time between the two calls of the agent, thereby improving the service efficiency of the agent.

[0050] At the same time, in order to avoid the situation where the next customer has answered the call while the current agent is still talking to the previous customer, in addition to determining the first probability of successfully calling the customer after the first time, it is also necessary to determine the second probability that the current agent has not ended the call after the first time. Comprehensively considering the probability that the customer answers the call after the first time and the probability that the current agent ends the current call can more accurately predict whether it is appropriate to make an outbound call for the current agent after the first time.

[0051] It should be noted that the outbound call prediction method provided in the embodiment of the present application is for outbound call prediction for a certain seat. That is to say, at the current moment, the outbound call prediction method provided in the embodiment of the present application can be used to predict whether each seat needs an outbound call after the first time, and then the number of seats that need outbound calls is used as the number of outbound calls.

[0052] Specifically, the second probability that the current agent has not ended the call after the first time can be determined by counting historical outbound call data, estimating through specific services, etc. The specific determination method is not limited here.

[0053] S203: Determine whether to make an outbound call for the current agent after the first time according to the first probability and the second probability.

[0054] Determining the first probability of successfully calling the customer after the first time is equivalent to knowing the probability that the customer answers the call after the first time. Determining the second probability that the current agent has not ended the call after the first time is equivalent to knowing the probability that the current agent cannot provide service to the customer who answers the call after the first time. Combining the first probability and the second probability, it is possible to know the probability that the customer has answered the call but the current agent cannot provide service, and thus determine whether to make an outbound call for the current agent after the first time based on the probability.

[0055] Specifically, the first probability may be multiplied by the second probability, and then based on the obtained product, it is determined whether to make an outbound call for the current agent after the first time. The first probability may be multiplied by the second probability, and then the product may be multiplied by a parameter, and then based on the obtained result, it is determined whether to make an outbound call for the current agent after the first time. Different weights may be assigned to the first probability and the second probability, respectively, and then the weighted first probability and the second probability may be multiplied, and then based on the obtained product, it is determined whether to make an outbound call for the current agent after the first time. The specific method for determining whether to make an outbound call for the current agent after the first time based on the first probability and the second probability is not limited here.

[0056] From the above content, it can be seen that the method for predicting outbound calls provided in the embodiment of the present application first determines the first probability of successfully making an outbound call to a customer after the first time; then, determines the second probability that the current agent has not ended the call after the first time, so that the probability that the customer answers the call after the first time and the probability that the current agent cannot provide services to the customer who answers the call are determined; finally, based on the first probability and the second probability, it is possible to know the probability that the customer has answered the call but the current agent cannot provide services, and thus, based on the probability, determine whether to make an outbound call for the current agent after the first time. In this way, it can be ensured to the greatest extent that an outbound call is made to the customer after the time, and after the customer answers the call, the current agent just ends the previous call, so that the current agent can seamlessly provide services to the next customer, reducing harassment to the customer, and then accurately predicting outbound calls for each agent, and improving the service efficiency of each agent.

[0057] Further, as Figure 2 As a refinement and extension of the method shown, an embodiment of the present application also provides a method for predicting outbound calls. Figure 3 The process diagram of the method for predicting outbound calls in the embodiment of the present application is as follows Figure 2 , see Figure 3 As shown, the method may include:

[0058] S301: Obtain the historical time consumed for successful outbound calls of multiple agents and the connection rate of outbound calls of multiple agents.

[0059] The so-called historical time consumed by multiple agents for successful outbound calls is the time consumed by each agent from the start of the outbound call to the completion of the call and the establishment of a call with the external line. In other words, in a historical outbound call process, the time from the start of the outbound call to the customer to the customer answering the call is one historical time consumed.

[0060] Specifically, the time X1 taken from when agent 1 starts making an outbound call until a call is completed and a call is established with the outside line is collected; the time X2 taken from when agent 2 starts making an outbound call until a call is completed and a call is established with the outside line is collected; ...; the time Xn taken from when agent n starts making an outbound call until a call is completed and a call is established with the outside line is collected.

[0061] The so-called connection rate of outbound calls by multiple agents is the ratio of the total number of outbound calls connected by multiple agents to the total number of calls by multiple agents.

[0062] S302: Determine a first probability of successfully making an outbound call to a customer after the first time according to the integration result of the historical time consumption for successful outbound calls of multiple agents at the first time and the connection rate of the outbound calls of multiple agents.

[0063] Since it is necessary to integrate the historical time consumed by multiple agents for successful outbound calls at the first time, it is necessary to first fit the first normal distribution curve based on the historical time consumed by multiple agents for successful outbound calls, so as to integrate the first normal distribution curve using the first time to obtain the integration result of the historical time consumed by multiple agents for successful outbound calls at the first time. Since the integration result is statistically calculated based on the normal distribution curve, it can accurately represent the probability of being able to connect with the customer after the first time under the premise that the customer must answer the phone. Then, combined with the connection rate of multiple agents' outbound calls, the first probability of successfully calling the customer after the first time can be accurately obtained.

[0064] However, the number of historical outbound calls is not just one, so when fitting the first normal distribution curve, it is necessary to use data of multiple outbound calls, that is, the historical time spent on successful outbound calls by multiple agents. The multiple agents here can be all agents that the call center can control, or they can be part of the agents that the call center can control. There is no limit on the specific number of multiple agents here.

[0065] Specifically, step S302 may include:

[0066] a1: Determine the first average value of the historical time consumed by multiple agents to successfully make outbound calls.

[0067] Specifically, it can be calculated by the following formula (1).

[0068]

[0069] Wherein, E1 represents the first average value, X represents the historical time consumption, and n represents the number of seats.

[0070] a2: determining a first variance of the historical consuming time for successful outbound calls of the multiple agents based on the historical consuming time for successful outbound calls of the multiple agents and the first average value.

[0071] Specifically, it can be calculated by the following formula (2).

[0072]

[0073] Wherein, D1 represents the first variance, E1 represents the first average value, X represents the historical time consumption, and n represents the number of seats.

[0074] a3: Determine the first time distribution of the successful outbound calls of the multiple agents according to the historical time consumed by the successful outbound calls of the multiple agents, the first average value and the first variance.

[0075] The first time distribution here is the first normal distribution curve.

[0076] The first normal distribution curve after fitting can be expressed by the following formula (3).

[0077]

[0078] Wherein, F1(t1) represents the first normal distribution curve, D1 represents the first variance, E1 represents the first average value, and t1 represents the consuming time.

[0079] a4: Determine a first initial probability of successfully calling the customer after the first time according to the integral result of the first time distribution.

[0080] Specifically, since the first time distribution can characterize the distribution of the historical time consumed for successful outbound calls of multiple agents, and is embodied in a normal distribution curve, that is, the first normal distribution curve is a normal distribution function with the horizontal axis representing the time consumed for successful outbound calls, and the first time is a specific time consumed for successful outbound calls, therefore, by integrating the first normal distribution curve on the horizontal axis from 0 to the first time, the probability of successful outbound calls after the first time can be accurately counted, that is, the obtained integration result can accurately characterize the probability of the customer answering the call after the first time.

[0081] a5: Determine a first probability of successfully calling the customer after the first time according to the first initial probability and the connection rates of the outbound calls of the multiple agents.

[0082] Since not every outbound call will be successful, there may be situations where the customer is on the phone, refuses the call, is out of service area, or the phone is turned off. Therefore, when determining the first probability of a successful outbound call after the first time, it is also necessary to consider whether the customer answers the call, that is, the connection rate.

[0083] Specifically, the connection rate can be calculated by the following formula (4).

[0084]

[0085] Of course, the connection rate can be calculated not only by the above formula (4), but also by other formulas. For example: As long as the connection rate is positively correlated with the ratio of the total number of connection times to the total number of outbound calls, the specific calculation method of the connection rate is not limited here.

[0086] Specifically, the first probability can be calculated by the following formula (5).

[0087]

[0088] Among them, F3(t1) represents the first probability of successful outbound call after a certain elapsed time, F1(t1) represents the first normal distribution curve, and t1 represents a certain elapsed time.

[0089] Since the above-mentioned integral result represents the probability of successfully calling the customer after the first time, assuming that the customer will definitely answer the call, and in fact, the customer may not necessarily answer the call during a call, therefore, by multiplying the above-mentioned integral result by the connection rate, the factor of whether the customer's call is connected can be taken into account, thereby accurately determining the first probability of successfully calling the customer after the first time.

[0090] S303: Obtain the historical call time of multiple agents and the call time of the current agent.

[0091] The so-called historical call time of multiple seats refers to the multiple times spent by each seat in the process of historical calls, from the time the call is established with the external line to the time the call ends.

[0092] Specifically, the time Y1 required for agent 1 to complete a call with an outside line is collected; the time Y2 required for agent 2 to complete a call with an outside line is collected; ...; the time Yn required for agent n to complete a call with an outside line is collected.

[0093] In the specific implementation process, after the agent establishes a call with the customer, the agent and the customer are disconnected due to equipment or network problems, or the agent is unable to end the call due to the customer's deliberate obstruction, resulting in an extremely long call time. These are not situations where the agent normally provides services to the customer. When fitting the second normal distribution curve of the historical call time of multiple agents, if these data are used, the accuracy of the second normal distribution curve in representing the distribution of agent call time will be reduced. Therefore, after collecting the initial historical call time of multiple agents, the historical call time with abnormal time tags in the initial historical call time of multiple agents is deleted to obtain the historical call time of multiple agents. The historical calls of multiple agents after deleting the abnormal time are fitted with the second normal distribution curve, and the fitted second normal distribution curve can more accurately represent the normal distribution of the historical call time of the agent.

[0094] The abnormal time label can be a label added by the agent to the call based on the actual call situation after the agent and the corresponding customer have finished talking. It can also be a label automatically added by the detection device to the call based on the call situation (including call content, signal status, etc.) between the agent and the corresponding customer during the call. The specific method of adding abnormal labels is not limited here.

[0095] The so-called current agent's call time refers to the time a certain agent has spent since the call with the outside line was established to the present.

[0096] S304: Determine a second probability that the current agent has not ended the call after the first time according to the sum of the first time and the call time of the current agent and the integration result of the historical call time of multiple agents.

[0097] Here, the sum of the first time and the call time of the current agent can be called the second time. Since the historical call time of multiple agents is integrated, the time used in the integration needs to be a complete call time, that is, the time from the start of the call to the end of the call. If the current agent is already in a call, then the first time is not a complete call time. Therefore, the second time obtained by adding the current agent's call time to the first time can represent the total time spent by the current agent in the current call process. Then, by integrating the historical call time of multiple agents using the second time, the second probability that the current agent has not ended the call after the first time can be accurately calculated.

[0098] Since the historical call times of multiple agents need to be integrated according to the second time, it is necessary to first fit a second normal distribution curve based on the historical call times of multiple agents, so as to integrate the second normal distribution curve using the second time to obtain the integration result of the historical call times of multiple agents at the second time. Since the integration result is statistically calculated based on the normal distribution curve, it can accurately represent the second probability that the current agent has not received a call after the first time.

[0099] However, the number of historical calls is not limited to one, so when performing curve fitting, it is necessary to use data of multiple calls, that is, the historical call time of multiple agents. The multiple agents here can be all agents that the call center can control, or they can be part of the agents that the call center can control. There is no limit on the specific number of multiple agents.

[0100] Specifically, step S304 may include:

[0101] b1: Determine the second average value of the historical talk time of multiple agents.

[0102] Specifically, it can be calculated by the following formula (6).

[0103]

[0104] Wherein, E2 represents the second average value, Y represents the historical call time, and n represents the number of seats.

[0105] b2: Determine a second variance of the historical call times of the multiple agents based on the historical call times of the multiple agents and the second average value.

[0106] Specifically, it can be calculated by the following formula (7).

[0107]

[0108] Wherein, D2 represents the second variance, E2 represents the second mean, Y represents the historical call time, and n represents the number of seats.

[0109] b3: Determine a second time distribution of the historical call times of the multiple agents according to the historical call times of the multiple agents, the second average value, and the second variance.

[0110] The second time distribution here is the second normal distribution curve.

[0111] The fitted second normal distribution curve can be expressed by the following formula (8).

[0112]

[0113] Wherein, F2(t2) represents the second normal distribution curve, D2 represents the second variance, E2 represents the second mean value, and t2 represents the call time.

[0114] In the specific implementation process, due to the different types of business handled by different seats and the different business capabilities of different seats, there will be differences between the second normal distribution curves corresponding to different seats, and there will be differences between the second normal distribution curve corresponding to a single seat and the second normal distribution curve corresponding to all seats. Although this difference is relatively small, it will still have a certain impact on the accuracy of outbound call prediction for each seat when the number of seats corresponding to the call center is large. Therefore, when fitting the second normal distribution curve, on the one hand, a second normal distribution curve can be fitted based on the historical call time of all seats, and this second normal distribution curve is used to determine the second probability that each seat has not ended the call after the first time. Since fitting a second normal distribution curve takes a short time, the efficiency of fitting the second normal distribution curve can be improved, thereby improving the efficiency of predicting outbound calls. On the other hand, the historical call time of each seat can be screened out from the historical call time of multiple seats, and the second normal distribution curve of the corresponding seat can be fitted based on the historical call time of each seat. Different seats correspond to different second normal distribution curves. Then, the second probability that the current agent has not ended the call after the first time can be determined based on the sum of the first time and the call time of the current agent, and the integration result of the second normal distribution curve of the agent. Since the second normal distribution curve is fitted based on the historical call time of the current agent, it can accurately represent the historical call situation of the current agent. By using the second normal distribution curve belonging to the current agent to determine the second probability that the current agent has not ended the call after the first time, the second probability that the current agent has not ended the call after the first time can be determined more accurately, thereby improving the accuracy of outbound call prediction.

[0115] b4: Determine a second probability that the current agent has not ended the call after the first time by integrating the second time distribution based on the sum of the first time and the call time of the current agent.

[0116] Specifically, the second probability can be calculated by the following formula (9).

[0117]

[0118] Among them, F4(t2) represents the second probability that the current agent does not end the call after a certain call time, F2(t2) represents the second normal distribution curve, T represents the longest time in the historical call time, and t2 represents a certain call time.

[0119] It should be noted that steps S301-S302 and steps S303-S304 may be executed simultaneously or one after the other. The execution order of steps S301-S302 and steps S303-S304 is not limited here.

[0120] S305: Determine, based on the product of the first probability and the second probability, a third probability that the current agent makes an outbound call after the first time to cause harassment to the customer.

[0121] Since the first probability can represent the probability that the customer answers the call after the first time, and the second probability can represent the probability that the current agent cannot provide service to the customer who answers the call after the first time, the product of the first probability and the second probability can represent the probability that the customer has answered the call after the first time, but the current agent cannot provide service to the customer, that is, the probability of harassing the customer by making outbound calls after the first time, that is, the third probability.

[0122] Specifically, the third probability can be calculated by the following formula (10).

[0123] F5(t)=F3(t1)×F4(t2) Formula (10)

[0124] Among them, F5(t) represents the third probability of causing harassment to customers by making an outbound call after a certain time, F3(t1) represents the first probability of successful outbound call after a certain time spent, and F4(t2) represents the second probability that the current agent has not ended the call after a certain call time. In fact, the above-mentioned certain time spent and certain call time both refer to a certain time thereafter. For example: the first time.

[0125] Of course, in addition to calculating the third probability using the above formula (10), a coefficient may be multiplied on the basis of the product of the first probability and the second probability to obtain the final third probability. The specific calculation method of the third probability is not limited here.

[0126] S306: Determine whether the third probability is less than a preset threshold; if so, execute S307; if not, execute S308.

[0127] S307: Determine to make an outbound call for the current agent after the first time.

[0128] S308: Determine not to make an outbound call for the current agent after the first time.

[0129] When the third probability is less than the preset threshold, it means that the customer has answered the call, and the probability that the current agent cannot provide service is low. Therefore, determining to make an outbound call for the current agent after the first time can not only improve the work efficiency of the current agent, but also reduce harassment to the customer. When the third probability is greater than or equal to the preset threshold, it means that the customer has answered the call, and the probability that the current agent cannot provide service is high. Therefore, determining not to make an outbound call for the current agent after the first time can greatly reduce harassment to the customer.

[0130] In actual applications, the preset threshold can be set according to actual business. The specific value of the preset threshold is not limited here.

[0131] In the process of determining whether to make an outbound call after the first time, the first time should not be too long. If a longer first time is chosen, the predicted result will most likely be able to make an outbound call, which will lose the meaning of the prediction itself. In view of this, the first time can be made shorter than the longest time in the historical call time of all agents.

[0132] Specifically, the selection of the first time can be restricted by the following formula (11).

[0133]

[0134] Among them, t represents the first time, E2 represents the second average value, D2 represents the second variance, and T represents the longest time in the historical call time. The corresponding probability in the second normal distribution curve is about 0.997. In other words, The second normal distribution curve corresponds to a maximum time. The value of the first time does not exceed There will not be a high probability that an outbound call can be made after the first time, so that the predictive outbound call method provided in the embodiment of the present application has practical predictive significance.

[0135] In actual applications, if it is preliminarily determined that an outbound call can be made for the current agent after the first time, and if the current agent is on a call under special circumstances, for example, the customer deliberately makes things difficult and the call cannot be ended normally, then if an outbound call is still made for the agent after the first time, the probability that the agent will end the current call after the first time is low. In this case, there will still be a situation of causing harassment to the customer. In view of this, step S307 can be further optimized.

[0136] c1: If the third probability is less than the preset threshold, a prompt message is output to the current agent.

[0137] The prompt information is used to prompt the current agent to make an outbound call after the first time.

[0138] c2: Get the feedback result of the current agent based on the prompt information.

[0139] After seeing the prompt information, the current agent will know that the predicted outbound call device will arrange an outbound call for him / her after the first time. The current agent can give feedback to the prompt information according to his / her actual situation, and the feedback result can indicate whether he / she can successfully end the current call after the first time.

[0140] c3: Determine whether the feedback result indicates that the current agent agrees to make an outbound call after the first time; if so, execute c4; if not, execute c5-c8.

[0141] c4: Determine to make an outbound call for the current agent after the first time.

[0142] c5: retrieve the call records of the current agent in the current call.

[0143] c6: Determine whether there is any abnormality in the content of the call record; if so, execute c7; if not, execute c8.

[0144] c7: Determine not to make an outbound call for the current agent after the first time.

[0145] c8: Determine to make an outbound call for the current agent after the first time.

[0146] If there are abnormalities in the content of the call record, for example: the customer deliberately makes things difficult, the customer's comprehension is low, etc., it means that the current agent is in a special situation during this call. It is not that he does not want to continue to serve other customers, but that he is restricted by objective reasons and cannot successfully end the current call in the first time. In this case, no outbound calls will be made for the agent in the first time to avoid the agent not ending the current call and harassing subsequent customers who have answered the call.

[0147] If there is nothing abnormal in the call record, it means that nothing special happened to the current agent during the call. It may be that the current agent no longer wants to continue serving other customers. In this case, continue to make outbound calls for the agent as soon as possible to improve the agent's service efficiency.

[0148] Based on the same inventive concept, as an implementation of the above method, the embodiment of the present application also provides a predictive outbound call device. Figure 4 The structure of the prediction device in the embodiment of the present application is shown in FIG. Figure 1 , see Figure 4 As shown, the device may include:

[0149] The first determination module 401 is used to determine a first probability of successfully calling a customer after a first time.

[0150] The second determining module 402 is used to determine a second probability that the current agent has not ended the call after the first time.

[0151] The outbound call prediction module 403 is used to determine whether to make an outbound call for the current agent after the first time according to the first probability and the second probability.

[0152] Further, as Figure 4As a refinement and expansion of the device shown, an embodiment of the present application also provides a predictive outbound call device. Figure 5 The structure of the device for predicting outbound calls in the embodiment of the present application is shown in FIG. Figure 2 , see Figure 5 As shown, the device may include:

[0153] The first determining module 501 includes:

[0154] The first acquisition unit 5011 is used to acquire the historical time consumed by multiple agents' successful outbound calls and the connection rate of the multiple agents' outbound calls.

[0155] The first determining unit 5012 is used to determine a first probability of successfully calling the customer after the first time according to the integration result of the historical consuming time at the first time and the connection rate.

[0156] The first time is shorter than the longest time among the historical call times of all agents.

[0157] The first determination unit 5012 is specifically used to: determine a first average value of the historical consumed time; determine a first variance of the historical consumed time based on the historical consumed time and the first average value; determine a first time distribution of successful outbound calls of the multiple agents based on the historical consumed time, the first average value and the first variance; determine a first initial probability of successful outbound calls to customers after the first time based on an integral result of the first time distribution over the first time; determine a first probability of successful outbound calls to customers after the first time based on the first initial probability and the connection rate.

[0158] The second determining module 502 includes:

[0159] The second acquisition unit 5021 is used to acquire the historical call time of multiple agents and the call time of the current agent.

[0160] The second acquisition unit 5021 is specifically used to: acquire the initial historical call time of the multiple seats; delete the historical call time carrying abnormal time tags in the initial historical call time, and obtain the historical call time of the multiple seats.

[0161] The second determining unit 5022 is used to determine a second probability that the current agent has not ended the call after the first time according to an integration result of the historical call time and the sum of the first time and the completed call time.

[0162] The second determination unit 5022 is specifically used to: determine a second average value of the historical call time; determine a second variance of the historical call time based on the historical call time and the second average value; determine a second time distribution of the historical calls of the multiple agents based on the historical call time, the second average value and the second variance; determine a second probability that the current agent has not ended the call after the first time based on an integration result of the second time distribution which is a sum of the first time and the call time.

[0163] The second determination unit 5022 is further specifically used to: filter out the historical call time of the current seat from the historical call time of the multiple seats; and determine a second probability that the current seat has not ended the call after the first time by integrating the historical call time of the current seat based on the sum of the first time and the call time.

[0164] The outbound call prediction module 503 includes:

[0165] The calculation unit 5031 is used to determine, based on the product of the first probability and the second probability, a third probability that an outbound call is made by the current agent after the first time to cause harassment to the customer.

[0166] The outbound call prediction unit 5032 is used to determine to make an outbound call for the current agent after the first time if the third probability is less than a preset threshold.

[0167] The predicted outbound call unit 5032 is specifically used to: if the third probability is less than the preset threshold, output a prompt message to the current agent, the prompt message is used to prompt to make an outbound call for the current agent after the first time; obtain the feedback result of the current agent based on the prompt message; if the feedback result indicates that the current agent agrees to make an outbound call after the first time, determine to make an outbound call for the current agent after the first time; if the feedback result indicates that the current agent does not agree to make an outbound call after the first time, retrieve the call record of the current agent in the current call, and if there is no abnormality in the content of the call record, determine to make an outbound call for the current agent after the first time.

[0168] It should be noted here that the description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the device embodiment of the present application, please refer to the description of the method embodiment of the present application for understanding.

[0169] Based on the same inventive concept, an embodiment of the present application also provides an electronic device. Figure 6This is a schematic diagram of the structure of the electronic device in the embodiment of the present application, see Figure 6 As shown, the electronic device may include: a processor 601, a memory 602, and a bus 603; wherein the processor 601 and the memory 602 communicate with each other via the bus 603; the processor 601 is used to call program instructions in the memory 602 to execute the methods in one or more of the above embodiments.

[0170] It should be noted that the description of the above electronic device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the electronic device embodiment of this application, please refer to the description of the method embodiment of this application for understanding.

[0171] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium, which may include: a stored program; wherein, when the program is running, the device where the storage medium is located is controlled to execute the method in one or more of the above embodiments.

[0172] It should be noted here that the description of the above storage medium embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the storage medium embodiment of the present application, please refer to the description of the method embodiment of the present application for understanding.

[0173] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting outbound calls, characterized in that: The method comprises: Determine the first probability of successfully calling the customer after the first time; Get the historical call time of multiple agents and the call time of the current agent; Determine a second probability that the current agent has not ended the call after the first time according to the sum of the first time and the completed call time and the integration result of the historical call time; Determine, according to the product of the first probability and the second probability, a third probability that an outbound call is made for the current agent after the first time to cause harassment to the customer; If the third probability is less than a preset threshold, it is determined to make an outbound call for the current agent after the first time.

2. The method according to claim 1, characterized in that Determining a first probability of successfully calling the customer after the first time includes: Obtain the historical time consumed for successful outbound calls of multiple agents and the connection rate of the outbound calls of the multiple agents; A first probability of successfully calling the customer after the first time is determined according to the integration result of the historical consuming time at the first time and the connection rate.

3. The method according to claim 2, characterized in that Determining a first probability of successfully calling the customer after the first time according to the integration result of the historical consuming time at the first time and the connection rate includes: Determining a first average of the historical elapsed times; Determine a first variance of the historical consuming time according to the historical consuming time and the first average value; Determine a first time distribution of successful outbound calls of the plurality of agents according to the historical consuming time, the first average value, and the first variance; Determining a first initial probability of successfully calling the customer after the first time according to an integral result of the first time distribution over the first time; A first probability of successfully calling the customer after a first time is determined according to the first initial probability and the connection rate.

4. The method according to claim 1, characterized in that: The determining, based on the sum of the first time and the completed call time and the integration result of the historical call time, a second probability that the current agent has not ended the call after the first time, includes: Determining a second average of the historical call durations; Determine a second variance of the historical call time according to the historical call time and the second average value; Determine a second time distribution of historical calls of the plurality of agents according to the historical call time, the second average value, and the second variance; A second probability that the current agent has not ended the call after the first time is determined based on an integration result of the second time distribution of the sum of the first time and the call time.

5. The method according to claim 1, characterized in that The acquisition of historical call times of multiple agents includes: Obtaining initial historical call times of the multiple agents; The historical call times carrying abnormal time tags in the initial historical call times are deleted to obtain the historical call times of the multiple agents.

6. The method according to claim 1, characterized in that The determining, based on the sum of the first time and the completed call time and the integration result of the historical call time, a second probability that the current agent has not ended the call after the first time, includes: Filtering the historical call time of the current agent from the historical call time of the multiple agents; A second probability that the current agent has not ended the call after the first time is determined according to the sum of the first time and the completed call time and the integration result of the historical call time of the current agent.

7. The method according to claim 1, characterized in that The first time is shorter than the longest time among the historical talk times of all agents.

8. The method according to claim 1, characterized in that If the third probability is less than a preset threshold, determining to make an outbound call for the current agent after the first time includes: If the third probability is less than the preset threshold, outputting a prompt message to the current agent, wherein the prompt message is used to prompt the current agent to make an outbound call after the first time; Obtaining the feedback result of the current agent according to the prompt information; If the feedback result indicates that the current agent agrees to make an outbound call after the first time, determining to make an outbound call for the current agent after the first time; If the feedback result indicates that the current agent does not agree to make an outbound call after the first time, the call record of the current agent in the current call is retrieved, and if there is no abnormality in the call record, it is determined to make an outbound call for the current agent after the first time.

9. A predictive outbound call device, characterized in that: The outbound call prediction device comprises: A first determination module, used to determine a first probability of successfully calling the customer after a first time; The second determination module is used to obtain the historical call time of multiple agents and the call time of the current agent; Determine a second probability that the current agent has not ended the call after the first time according to the sum of the first time and the completed call time and the integration result of the historical call time; An outbound call prediction module, configured to determine, based on the product of the first probability and the second probability, a third probability that an outbound call will be made for the current agent after the first time, causing harassment to the customer; If the third probability is less than a preset threshold, it is determined to make an outbound call for the current agent after the first time.

10. A predictive outbound call system, characterized in that: The system comprises: Business system, used to generate outbound call demand; A call center connected to the business system, used to obtain the outbound call demand and make outbound calls to customers based on the outbound call demand; A predictive outbound call device, connected to the call center, for controlling whether the call center makes an outbound call to the customer based on the outbound call demand after the first time; the predictive outbound call device is used to execute the method according to any one of claims 1 to 8; A client, connected to the call center, for receiving a call from the call center; An agent is connected to the call center and is used to connect to the customers connected to the call center.

11. An electronic device, characterized in that: The electronic device comprises: a processor, a memory, and a bus; wherein the processor and the memory communicate with each other via the bus; and the processor is used to call program instructions in the memory to execute the method as claimed in any one of claims 1 to 8.

12. A computer-readable storage medium, characterized in that: The storage medium comprises: a stored program; wherein, when the program is running, the device where the storage medium is located is controlled to execute the method as claimed in any one of claims 1 to 8.

Images