Outbound flow rate control method and device

Abstract

The invention provides an outbound flow rate control method and device, and the method comprises the steps: receiving initial sensitivity and a target outbound flow rate, and the initial sensitivity is used for representing the initial change rate of the outbound flow rate; acquiring a current call completing rate and a resource occupancy rate; a slow start process is executed based on the initial sensitivity, in the slow start process, the outbound flow rate is gradually increased according to the current sensitivity until the target outbound flow rate or resource full load is reached, the current sensitivity represents the real-time change rate of the outbound flow rate, and the initial value of the current sensitivity is set as the initial sensitivity; adjusting the current sensitivity based on the resource occupancy rate; and when the outbound flow rate does not reach the target outbound flow rate and the current sensitivity is adjusted to zero, triggering resource capacity expansion, and when the outbound flow rate reaches the target outbound flow rate and the resource occupancy rate is lower than a preset threshold value, triggering resource capacity reduction.

Description

Technical Field

[0001] The present application relates to the technical field of telephone outbound calls, and in particular to an outbound call flow rate control method and device. Background Art

[0002] In the process of implementing this application, the inventor discovered that the existing outbound telephone calling technology has the following problems:

[0003] First, there are two basic characteristics of current outbound calls: the resources required to connect an outbound call are greater than those required to connect an outbound call; the connection flow rate cannot be directly controlled, as it depends on whether the called party answers the call and how long the connection takes.

[0004] Secondly, in the prior art, the connection flow rate control mainly relies on the experience of the operation and maintenance personnel to indirectly control the outbound call flow rate by controlling the outbound call flow rate. Specifically, the operation and maintenance personnel set the outbound call flow rate in advance based on their experience before the outbound call, and generally no longer revise it during the outbound call. Only in the event of a long-term overload will the outbound call be suspended and the outbound call flow rate be revised. For example, when the resources are fully loaded and can support a maximum connection flow rate of 100 calls / minute, the operation and maintenance personnel estimate the connection rate to be 10% based on experience, and the theoretical outbound call flow rate can be 1000 calls / minute, but because the estimated value is not accurate, the outbound call flow rate is usually set to 600 calls / minute or even lower to prevent resource overload.

[0005] This control method that relies on manual experience has the following problems:

[0006] Waste of resources: Since the outbound call flow rate is set by relying on manual experience to estimate the connection rate and thus indirectly set the connection flow rate, it is generally inclined to reserve a large margin of resources to prevent the need to suspend outbound calls and re-set the flow rate due to resource overload. This situation often leads to light-loaded resources, resulting in resource waste.

[0007] Difficulty in estimating the connection rate: Since the connection rate varies greatly in different environments (such as date, time period, outbound number segment, and outbound line provider), the error in manual estimation based on experience is extremely large.

[0008] Difficulty in scaling resources: When a high connection rate causes resource overload, it is basically impossible to achieve real-time expansion of resources because the outbound call flow rate cannot be accurately estimated. Summary of the invention

[0009] In view of this, the present application provides an outbound call flow rate control method and device, which are used to solve the problems in the prior art of relying on manual experience to control the outbound call flow rate, resulting in resource waste and difficulty in flow rate estimation.

[0010] The embodiment of the present application provides an outbound call flow rate control method, including:

[0011] receiving an initial sensitivity and a target outbound call flow rate, wherein the initial sensitivity is used to represent an initial rate of change of the outbound call flow rate;

[0012] Get the current connection rate and resource occupancy rate;

[0013] Based on the initial sensitivity, a slow start process is performed, during which the outbound call flow rate is gradually increased according to the current sensitivity until the target outbound call flow rate or resource full load is reached, wherein the current sensitivity represents the real-time change rate of the outbound call flow rate, and the initial value of the current sensitivity is set to the initial sensitivity;

[0014] Adjusting the current sensitivity based on the resource occupancy rate;

[0015] Furthermore, when the outbound call flow rate does not reach the target outbound call flow rate and the current sensitivity is adjusted to zero, resource expansion is triggered; when the outbound call flow rate reaches the target outbound call flow rate and the resource occupancy rate is lower than a preset threshold, resource reduction is triggered.

[0016] Optionally, performing the soft start process based on the initial sensitivity includes:

[0017] Setting an initial outbound call flow rate, wherein the initial outbound call flow rate is lower than the estimated full-load outbound call flow rate;

[0018] Based on the initial sensitivity, the current outbound call flow rate is increased according to the following formula every minute:

[0019] Current outbound call flow rate = outbound call flow rate in the last minute + current sensitivity × time interval.

[0020] Optionally, the adjusting the current sensitivity based on the resource occupancy rate includes:

[0021] When it is detected that the difference between the outbound call flow rate and the estimated full-load outbound call flow rate is less than the current sensitivity, the current sensitivity is reduced by using a binary approximation method;

[0022] The binary approximation method is: current sensitivity = (estimated full-load outbound call flow rate - current outbound call flow rate) / 2 / time interval.

[0023] Optionally, the triggering resource expansion includes:

[0024] When it is detected that the current sensitivity is reduced to zero and the target outbound call flow rate is not reached, increasing the resource quota;

[0025] The current sensitivity is reset to an initial value, and the outbound call flow rate is continued to be improved.

[0026] Optionally, the method further comprises:

[0027] Monitor whether resource expansion or reduction is successful;

[0028] When it is detected that resource expansion or reduction fails continuously for more than a preset time, an alarm message is sent.

[0029] Optionally, the method further comprises:

[0030] Based on historical outbound call data, a genetic algorithm is used to obtain the outbound call flow rate thresholds during peak and trough periods;

[0031] Wherein, the method of using a genetic algorithm to obtain the outbound call flow rate threshold value during the peak and trough periods includes:

[0032] Encode the outgoing call flow rate as a chromosome;

[0033] The connection rate is used as the fitness function;

[0034] Perform selection, crossover and mutation operations to obtain the optimal outbound call flow rate threshold during peak and trough periods.

[0035] Optionally, the method further comprises:

[0036] Dynamically adjust the target outbound call flow rate according to the outbound call flow rate threshold during the peak and trough periods;

[0037] Wherein, during the peak period, the target outbound call flow rate is increased to the outbound call flow rate threshold during the peak period;

[0038] During the trough period, the target outbound call flow rate is reduced to the outbound call flow rate threshold of the trough period.

[0039] Optionally, the method further comprises:

[0040] A meta-heuristic algorithm is used to generate multiple candidate solutions, each of which contains an outbound call flow rate adjustment strategy;

[0041] Evaluate the performance indicators of each candidate solution, including resource utilization and connection rate;

[0042] An optimal solution is selected from the candidate solutions as the optimal outbound call flow rate control strategy.

[0043] Optionally, the meta-heuristic algorithm is an ant colony algorithm, and the selecting the optimal solution from the candidate solutions as the outbound call flow rate control strategy includes:

[0044] Mapping the outbound call flow rate adjustment strategy to the path selection of ants;

[0045] Calculating pheromone concentration according to the resource utilization rate and the connection rate;

[0046] The pheromone concentration is iteratively updated until it converges to the optimal outgoing call flow rate control strategy.

[0047] The embodiment of the present application also provides an outbound call flow rate control device, comprising:

[0048] A receiving unit, used for receiving an initial sensitivity and a target outbound call flow rate, wherein the initial sensitivity is used for indicating an initial rate of change of the outbound call flow rate;

[0049] An acquisition unit, used to acquire the current connection rate and resource occupancy rate;

[0050] a control unit, configured to perform a slow start process based on the initial sensitivity, during which the outbound call flow rate is gradually increased according to the current sensitivity until the target outbound call flow rate or the resource is fully loaded, wherein the current sensitivity represents a real-time rate of change of the outbound call flow rate, and an initial value of the current sensitivity is set to the initial sensitivity;

[0051] An adjusting unit, configured to adjust the current sensitivity based on the resource occupancy rate;

[0052] The resource management unit is used to trigger resource expansion when the outbound call flow rate does not reach the target outbound call flow rate and the current sensitivity is adjusted to zero; and to trigger resource reduction when the outbound call flow rate reaches the target outbound call flow rate and the resource occupancy rate is lower than a preset threshold.

[0053] This application has the following beneficial effects:

[0054] 1. This application avoids the inaccuracy of manual experience by adopting a sensitivity-based slow start and automatic adjustment mechanism, so that the outbound call flow rate can be automatically adjusted to the optimal state, significantly improving resource utilization.

[0055] 2. This application monitors resource utilization in real time and automatically triggers the expansion and contraction mechanism, so that the system can adaptively adjust resource allocation, which not only avoids resource waste but also responds to high-load scenarios in a timely manner.

[0056] 3. This application introduces an intelligent optimization mechanism based on genetic algorithm and ant colony algorithm, which can better cope with complex scenarios such as peaks and troughs, and further improve the intelligence level of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0058] Figure 1 A flow chart of the outbound call flow rate control method provided in an embodiment of the present application;

[0059] Figure 2 A schematic diagram of outbound call flow rate adjustment based on slow start provided in an embodiment of the present application;

[0060] Figure 3 A schematic diagram of outbound call flow rate control based on sensitivity adjustment provided in an embodiment of the present application;

[0061] Figure 4 A schematic diagram of the resource expansion and contraction process provided in an embodiment of the present application;

[0062] Figure 5 This is a structural block diagram of the outbound call flow rate control device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to enable those skilled in the art to better understand the inventive concept, working principle and advantages of the embodiments of this specification, the schemes in the embodiments of this specification are described in detail below.

[0064] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] 1. Explanation of relevant terms

[0066] Before describing the technical solution of the present application in detail, the main technical terms involved in the present application are first explained:

[0067] Full Load (FL): refers to the consumption of computing power resources, memory resources, and bandwidth resources being approximately equal to the rated maximum operating conditions. Under FL conditions, resources are most effectively utilized.

[0068] Light Load (LL): refers to the situation where the consumption of computing power resources, memory resources, and bandwidth resources is far lower than the rated maximum operating conditions. Under LL conditions, some resources are not utilized, resulting in waste.

[0069] Overload (OL): refers to computing power resources, memory resources, and bandwidth resources that are greater than the rated maximum operating conditions. Under OL conditions, due to resource exhaustion, the system becomes unstable or even unable to complete the task normally.

[0070] Sensitivity: It is used to indicate the rate of change of the outgoing call flow rate, and the unit is flow / minute2. The higher the sensitivity, the faster the flow rate changes, and the lower the sensitivity, the slower the change. Among them:

[0071] A high sensitivity will quickly approach the full load condition, but may miss the optimal full load condition (coarse adjustment).

[0072] Low sensitivity will slowly approach the full load condition, but it is easier to find the optimal full load condition (fine adjustment).

[0073] Slow start:

[0074] The outbound call task starts with a lower outbound call flow rate and increases the outbound call flow rate with a certain sensitivity.

[0075] The initial outbound call flow rate is a loose fixed value, which generally meets the following two conditions:

[0076] Much lower than the theoretical full-load outbound call flow rate (i.e. one order of magnitude or more to prevent overload at startup)

[0077] Meet the concept of "large sample" in statistics, the sample size ≥ 30 (that is, greater than or equal to 30C / M, to ensure the first batch of data

[0078] The connection rate and other data are statistically significant)

[0079] Generally, the full-load outbound call flow rate that a single minimum resource can support is above 600C / M, so this value is taken as about 5-10% of the theoretical full-load outbound call flow rate of the resource.

[0080] Outbound call flow rate estimation:

[0081] During the soft start, parameters such as connection rate and resource occupancy rate will be collected from the monitoring and resource allocation services to estimate the next round of output.

[0082] Sensitivity estimation: During outbound calls, if the resource occupancy rate collected from the monitoring and resource allocation service is close to full load when the outbound call flow rate increases, the sensitivity is reduced (entering the fine-tuning stage) to prevent resource overload.

[0083] The difference between the current outbound call flow rate and the estimated maximum outbound call flow rate (based on the connection rate, connection time, and resource estimation) is 1 sensitivity (that is, if the sensitivity is not adjusted, the estimated maximum flow will be reached in 1 minute), and the sensitivity reduction is triggered. The sensitivity reduction adopts the "binary approximation method" to keep the outbound call flow rate stable. For example, the estimated maximum outbound call flow rate is 1000C / M, and the sensitivity is 200C / M2. At this time, the actual outbound call flow rate has reached 800C / M, and the sensitivity reduction is triggered. According to the "binary approximation" algorithm, the sensitivity should be reduced to (1000C / M-800C / M) / 2 / m=100C / M2. When the actual outbound call flow rate reaches 900C / M, the sensitivity is reduced again to (1000C / M-900C / M) / 2 / m=50C / M2. Gradually approach the maximum value of the outbound call flow rate.

[0084] Resource expansion and contraction: During outbound calls, the capacity will be automatically expanded or contracted based on the resource occupancy rate and sensitivity collected in the monitoring and resource allocation service. For example, if the outbound call flow rate has not reached the required flow rate and the sensitivity has dropped to 0C / M2, resource expansion will be triggered to prevent resource overload and task abnormality. On the contrary, if the outbound call flow rate has reached the required flow rate and the resource occupancy rate is low, resource contraction will be triggered to prevent resource waste.

[0085] Failure of resource expansion and contraction: If any abnormality causes the expansion and contraction to fail, it will only cause the outbound call flow rate to fail to reach the expected outbound call flow rate for a long time (for example, more than 10 minutes). After the system detects it, it will notify the technical staff by SMS, email and other means to investigate the cause of the failure.

[0086] 2. Method Embodiment

[0087] See also Figure 1 , the embodiment of the present application provides an outbound call flow rate control method, comprising the following steps:

[0088] S101. Receive initial sensitivity and target outbound call flow rate.

[0089] The initial sensitivity is used to indicate the initial rate of change of the outbound call flow rate. For example, the initial sensitivity can be set to 100 calls / minute2, indicating that the outbound call flow rate increases by 100 calls / minute every minute. The target outbound call flow rate indicates the outbound call rate that is expected to be achieved, for example, it can be set to 2000 calls / minute.

[0090] S102: Obtain the current connection rate and resource occupancy rate.

[0091] The connection rate indicates the ratio of the number of connected calls to the total number of dialed calls, and the resource utilization rate indicates the ratio of the currently used computing resources, storage resources, etc. to the total system resources. These data can be obtained in real time through the system monitoring module.

[0092] S103: Execute a soft start process based on the initial sensitivity.

[0093] During the slow start process, the initial value of the current sensitivity is first set to the received initial sensitivity, and then the outbound call flow rate is gradually increased according to the current sensitivity until the target outbound call flow rate is reached or the resources are fully loaded.

[0094] See also Figure 2 During the slow start process, the system gradually increases the outbound call flow rate according to the current sensitivity until the target outbound call flow rate is reached or the resources are fully loaded. The specific steps include:

[0095] 1. Set initial parameters

[0096] (1) Set the initial value of the current sensitivity to the received initial sensitivity;

[0097] (2) Set the initial outbound call flow rate, which must meet the following conditions:

[0098] The outbound call flow rate should be at least one order of magnitude lower than the estimated full-load flow rate to prevent overload during startup;

[0099] Greater than or equal to 30 calls per minute to meet the concept of "large sample" in statistics and ensure that indicators such as the connection rate of the first batch of data are statistically significant.

[0100] For example, when the full-load outbound call flow rate that a single minimum resource can support is 600 calls / minute, the initial outbound call flow rate can be set to 60 calls / minute.

[0101] 2. Calculation promotion process

[0102] (1) Calculate the flow rate increase value for each time interval based on the current sensitivity:

[0103] Flow rate increase value = current sensitivity × time interval

[0104] For example: when the current sensitivity is 100 ch / min2 and the time interval is 1 minute, the flow rate increase value is 100 ch / min.

[0105] (2) Update the current outbound call flow rate:

[0106] Current outbound call flow rate = outbound call flow rate in the previous time interval + flow rate increase value

[0107] 3. Slow start example

[0108] Assume the following system parameters:

[0109] - Initial sensitivity: 100 passes / minute2;

[0110] - Initial outgoing flow rate: 60 passes / minute;

[0111] - Target outgoing flow rate: 2000 passes / minute;

[0112] -Time interval: 1 minute

[0113] The flow rate changes during the slow start process are as follows:

[0114] -T0 time: 60 passes / minute (initial value);

[0115] -T1 time: 160 passes / minute (60+100×1);

[0116] -T2 time: 260 passes / minute (160+100×1);

[0117] -T3 time: 360 passes / minute (260+100×1);

[0118] ...and so on

[0119] 4. Termination condition judgment

[0120] After each increase in the outbound call flow rate, the system will determine whether any of the following termination conditions are met:

[0121] (1) Achieve target outgoing flow rate

[0122] At this point, the slow start process is completed and the system enters a stable operating state.

[0123] (2) Reaching full resource load

[0124] The specific manifestations are:

[0125] - Resource utilization is close to 100%

[0126] -CPU utilization, memory usage and other indicators reach the preset threshold

[0127] At this time, the sensitivity adjustment mechanism needs to be triggered (see S104 for details).

[0128] 5. Notes

[0129] (1) During the slow start process, the increase curve of the outgoing call flow rate increases linearly, which is due to the use of a fixed current sensitivity.

[0130] (2) The theoretical duration of slow start can be calculated by the following formula:

[0131] Theoretical duration = (target outgoing call flow rate - initial outgoing call flow rate) / current sensitivity

[0132] For example: target 2000 channels / minute, initial 60 channels / minute, current sensitivity 100 channels / minute 2

[0133] Theoretical duration = (2000-60) / 100 = 19.4 minutes

[0134] (3) The actual slow start process may be terminated early due to resource full load, or executed in segments due to resource expansion.

[0135] (4) The slow start process will not adjust the rising curve according to the real-time connection rate. This is mainly based on the following considerations:

[0136] - Keep control strategies simple and predictable;

[0137] - Fluctuations in the connection rate will be reflected in the resource utilization rate, which can be indirectly handled through the sensitivity adjustment mechanism;

[0138] -Avoid system shock due to overreaction.

[0139] Through the above-mentioned slow start mechanism, the system can smoothly increase the outbound call flow rate, avoiding the resource overload problem that may be caused by a sudden and substantial increase in the outbound call flow rate. At the same time, due to the use of a linear increase strategy based on sensitivity, the entire process is controllable and easy to adjust.

[0140] In one of the embodiments, when an outbound call task is started, in order to avoid sudden overload of system resources due to too fast increase in outbound call flow rate, this embodiment realizes smooth increase of outbound call flow rate through a slow start mechanism. Specifically, the system first sets the initial value of the current sensitivity to the received initial sensitivity, and then gradually increases the outbound call flow rate according to the current sensitivity until the target outbound call flow rate or resource full load state is reached.

[0141] First, the system needs to set a reasonable initial outbound call flow rate. The setting of the initial outbound call flow rate needs to meet two conditions at the same time: on the one hand, it must be lower than the estimated full-load outbound call flow rate by more than one order of magnitude to prevent overload when the system is just started; on the other hand, it must be greater than or equal to 30 calls / minute to meet the "large sample" requirement in statistics and ensure that the statistical indicators of the first batch of data have reference significance. Taking the full-load outbound call flow rate supported by a single minimum resource as 600 calls / minute as an example, the initial outbound call flow rate can be set to 60 calls / minute at this time, which not only meets the requirement of preventing overload, but also ensures the statistical significance of the data.

[0142] During the slow start process, the system will calculate the flow rate increase value for each time interval based on the current sensitivity. For example, if the current sensitivity is 100 cpm2 and the time interval is 1 minute, the flow rate increase value for the current time interval is 100 cpm. By adding the outbound call flow rate of the previous time interval to the flow rate increase value, the new outbound call flow rate value can be obtained. This linear increase method makes the whole process smoother and more controllable.

[0143] To understand the slow start process more intuitively, let's look at a specific example. Assume that the initial parameters of the system are set as follows: the initial sensitivity is 100 calls / minute2, the initial outbound call flow rate is 60 calls / minute, the target outbound call flow rate is 2000 calls / minute, and the time interval is 1 minute. Under this configuration, the system will gradually increase the outbound call flow rate according to a linear curve: at T0, the initial value is 60 calls / minute, at T1, it increases to 160 calls / minute (60+100×1), at T2, it reaches 260 calls / minute (160+100×1), at T3, it increases to 360 calls / minute (260+100×1), and so on.

[0144] After each increase in the outbound call flow rate, the system will check whether the slow start process needs to be terminated. The termination conditions include two situations: one is that the target outbound call flow rate is reached, which indicates that the slow start process has been completed and the system can enter a stable operation state; the other is that the resource is fully loaded, which is manifested by the resource occupancy rate approaching 100% or the CPU utilization rate, memory usage rate and other indicators reaching the preset threshold. At this time, the sensitivity adjustment mechanism needs to be triggered to prevent the system from overloading.

[0145] It should be noted that the theoretical duration of the slow start process can be calculated by dividing the target flow rate increment by the current sensitivity. For example, in the above example, the flow rate increment that the system needs to increase is 1940 ch / min (2000-60), and the current sensitivity is 100 ch / min2, so the theoretical duration is 19.4 minutes. However, in actual operation, due to resource constraints, the slow start process may be terminated early, or it may need to be completed in multiple stages due to resource expansion.

[0146] In addition, the soft start mechanism of this embodiment intentionally avoids adjusting the rising curve according to the real-time connection rate when designing. This is mainly based on the following considerations: first, this can maintain the simplicity and predictability of the control strategy; second, the fluctuation of the connection rate will be indirectly reflected by the change of resource occupancy rate, which can be handled by the subsequent sensitivity adjustment mechanism; finally, it avoids the system oscillation caused by over-reacting to the change of the connection rate.

[0147] By adopting this sensitivity-based slow start mechanism, the system can achieve a smooth increase in outbound call flow rate, effectively avoiding the resource overload problem caused by a sudden increase in flow rate. At the same time, since the entire process adopts a linear increase strategy, the control process is simple and clear, which facilitates parameter adjustment and status monitoring. This mechanism cooperates with the subsequent sensitivity adjustment and resource expansion and contraction mechanism to jointly ensure the stable operation of the outbound call system.

[0148] S104: Adjust the current sensitivity based on the resource occupancy rate.

[0149] like Figure 3 As shown in the figure, when the difference between the detected outbound call flow rate and the estimated full-load outbound call flow rate is less than the current sensitivity, the binary approximation method is used to reduce the current sensitivity:

[0150] Current sensitivity = (estimated full load outbound call flow rate - current outbound call flow rate) / 2 / time interval

[0151] For example, suppose:

[0152] -Estimated full load outbound call flow rate is 1000 calls / minute

[0153] -Current outbound call flow rate is 800 calls / minute

[0154] - The time interval is 1 minute

[0155] The current sensitivity after adjustment is: (1000-800) / 2 / 1=100 passes / minute2

[0156] In order to stabilize the outbound call flow rate at the optimal working point and avoid system overload due to excessive increase in flow rate, this embodiment designs a sensitivity adjustment mechanism based on resource occupancy rate. This mechanism achieves precise control of the growth of outbound call flow rate by dynamically adjusting the current sensitivity.

[0157] During the outbound call process, the system monitors the difference between the outbound call flow rate and the estimated full-load outbound call flow rate in real time. When this difference is less than the current sensitivity, it means that if the outbound call flow rate continues to increase according to the current sensitivity, it is likely to exceed the system's carrying capacity in the next time interval. To avoid this situation, the system will trigger the sensitivity adjustment mechanism to achieve fine-grained control of the outbound call flow rate by reducing the current sensitivity.

[0158] This embodiment uses the binary approximation method to adjust the current sensitivity. Specifically, the new current sensitivity value is equal to half of the difference between the estimated full-load outbound call flow rate and the current outbound call flow rate divided by the time interval. For example, when the estimated full-load outbound call flow rate is 1000 calls / minute, the current outbound call flow rate is 800 calls / minute, and the time interval is 1 minute, the adjusted current sensitivity will be reduced to 100 calls / minute2 (i.e. (1000-800) / 2 / 1). This adjustment method not only ensures that the outbound call flow rate can continue to increase, but also significantly reduces the risk of overload.

[0159] The choice of the bisection approximation method was carefully considered. First, the method is simple to implement, has low computational overhead, and is suitable for online real-time adjustment. Second, the strategy of reducing the sensitivity by half each time has good convergence, which can help the system quickly find a suitable working point. Finally, this gradual adjustment method can effectively avoid system oscillation and instability.

[0160] It is worth noting that the sensitivity adjustment process is a dynamic process. As the outbound call flow rate gradually approaches the estimated full-load outbound call flow rate, the current sensitivity will be adjusted multiple times, showing a gradually decreasing trend. This decreasing trend corresponds to the system's transition from "coarse adjustment" to "fine adjustment". In the early stages, a larger sensitivity value helps the system quickly increase the outbound call flow rate; when approaching full load, a smaller sensitivity value ensures that the system can smoothly approach the optimal working point.

[0161] Of course, the sensitivity adjustment mechanism may also trigger the handling of some special situations. For example, if the sensitivity is adjusted to close to zero or zero, this usually means that the current resources can no longer support a higher outbound call flow rate. At this time, if the target outbound call flow rate has not been reached, the system will trigger the resource expansion mechanism. On the contrary, if it is detected that the resource utilization rate continues to be lower than a certain threshold, the system may trigger the resource reduction mechanism. This coordination with the resource expansion and reduction mechanism enables the entire system to maintain optimal resource utilization efficiency under different load conditions.

[0162] In addition, the sensitivity adjustment mechanism also takes into account the issue of system stability. In order to prevent unnecessary adjustments due to instantaneous fluctuations, the system will appropriately smooth the resource utilization data. At the same time, in order to cope with sudden load changes, the system also retains the ability to adjust the sensitivity to a negative value, so that the outbound call flow rate can be actively reduced when necessary to maintain the stable operation of the system.

[0163] Through this sensitivity adjustment mechanism based on resource occupancy, the system can achieve precise control of the outbound call flow rate. This mechanism, together with the slow start process and resource expansion and contraction mechanism, constitutes a complete closed-loop control system. Its ultimate goal is to maximize resource utilization and provide the best outbound call service quality while ensuring stable system operation. In practical applications, this mechanism has shown good adaptability and robustness, and can effectively cope with various complex business scenarios.

[0164] S105: Execute resource expansion and contraction according to the system status.

[0165] When the outbound call flow rate does not reach the target outbound call flow rate and the current sensitivity is adjusted to zero, it means that the existing resources have reached full load, and resource expansion is triggered. Specifically, it includes:

[0166] 1) Increase resource quota

[0167] 2) Reset the initial value of the current sensitivity

[0168] 3) Continue to improve outbound call flow rate

[0169] When the outbound call flow rate reaches the target outbound call flow rate and the resource utilization rate is lower than the preset threshold, it indicates that there is resource waste, and resource reduction is triggered.

[0170] like Figure 4 As shown, the system will also monitor whether the resource expansion and contraction is successful. When continuous failures are detected exceeding a preset time (such as 10 minutes), an alarm message is sent to the operation and maintenance personnel.

[0171] In order to achieve the optimal configuration of system resources, this embodiment designs an automatic resource expansion and contraction mechanism. This mechanism monitors the system status in real time and triggers resource expansion or contraction operations at appropriate times, thereby ensuring service quality while avoiding resource waste.

[0172] During the operation of the outbound call system, there may be two typical scenarios where resource allocation needs to be adjusted. The first is when the outbound call flow rate has not yet reached the target value, but the current sensitivity has been reduced to zero. This indicates that the existing resources have reached full load and cannot support a higher outbound call flow rate. At this time, the system needs to obtain more resources by expanding capacity. The second is when the outbound call flow rate has reached the target value, but the resource utilization rate continues to be lower than the preset threshold. This means that there is resource redundancy in the system and resource allocation needs to be optimized by reducing capacity.

[0173] Taking resource expansion as an example, when the system detects that the current sensitivity has dropped to zero and the outbound call flow rate has not yet reached the target value, the expansion process will be initiated. First, the system will calculate the amount of resources that need to be expanded based on the difference between the current outbound call flow rate and the target flow rate, as well as resource utilization. Then, the system will initiate an expansion request to the resource management platform to apply for the allocation of new resources. After the resource allocation is completed, the system will reset the current sensitivity to a suitable initial value, allowing the outbound call flow rate to continue to increase. This process may be repeated multiple times until the target outbound call flow rate is finally reached.

[0174] In contrast, the resource scaling process is relatively simple. When the system detects that the resource utilization rate has been below the preset threshold (for example, 50%) for a period of time, the scaling operation will be triggered. When scaling, the system will comprehensively consider the current business load and historical data and release excess resources cautiously. In order to avoid frequent expansion and contraction of resources, the system usually retains a certain buffer margin when performing scaling.

[0175] In actual operation, resource expansion and contraction may fail for various reasons. For example, there may not be enough free resources in the resource pool for expansion, or some resources may not be released smoothly due to system reasons. In order to deal with these situations, this embodiment implements a complete exception handling mechanism. Specifically, when the system detects that resource expansion or contraction operations fail continuously and the duration exceeds a preset threshold (for example, 10 minutes), an alarm message will be automatically generated and the operation and maintenance personnel will be notified by SMS, email, etc. This enables relevant personnel to detect problems in a timely manner and take necessary intervention measures.

[0176] It should be noted that the failure of the resource expansion and contraction mechanism will not lead to the termination of the outbound call task. In the event of a capacity expansion failure, the system will continue to run at the maximum flow rate that the existing resources can support. This design ensures business continuity, but it may not be able to achieve the expected target flow rate. Similarly, in the event of a capacity reduction failure, the system will maintain the current resource configuration and continue to run, but it may cause a certain degree of resource waste.

[0177] In order to improve the accuracy of expansion and contraction, this embodiment not only considers the current system status when making resource adjustment decisions, but also refers to historical data. For example, the system will analyze the resource utilization of similar time periods in history, predict the resource demand during business peak periods, and make resource reserves in advance. This data-based predictive adjustment can significantly improve the system's response speed and reduce the lag of resource adjustment.

[0178] In addition, the resource expansion and contraction mechanism also works closely with the sensitivity adjustment mechanism. After performing the expansion operation, the system will re-evaluate the current sensitivity and select an appropriate value to restart the slow start process. This ensures that the newly added resources can be used smoothly and avoids new overload problems caused by too fast an increase in flow rate. Similarly, before performing the contraction operation, the system will also adjust the sensitivity appropriately to ensure that the outbound call flow rate can transition smoothly during the resource release process.

[0179] Through this complete resource expansion and contraction mechanism, the system can automatically adjust resource configuration according to changes in business load and achieve dynamic optimization of resource utilization. This mechanism, together with the slow start and sensitivity adjustment mechanism introduced earlier, forms an adaptive closed-loop control system that can effectively cope with complex and changing business scenarios and ensure the stable operation of the outbound call system and efficient resource utilization.

[0180] 3. Improved Embodiment

[0181] Based on the above basic method, the present application also provides two improved embodiments:

[0182] 1. Peak and trough control based on genetic algorithm

[0183] In order to better cope with the call characteristics in different time periods, this embodiment introduces a genetic algorithm to obtain the outbound call flow rate thresholds in the peak and trough time periods:

[0184] 1) Encode the outbound flow rate as chromosomes, and each gene represents the flow rate value in a certain period of time

[0185] 2) The connection rate is used as the fitness function. The higher the connection rate, the better the flow rate setting of the group.

[0186] 3) Optimize chromosomes through selection, crossover and mutation operations, and finally obtain the optimal outbound call flow rate threshold during peak and trough periods

[0187] Based on the obtained threshold, the system can dynamically adjust the target outbound call flow rate:

[0188] -During the peak period, increase the target outbound call flow rate to the peak threshold;

[0189] -During the trough period, reduce the target outbound flow rate to the trough threshold.

[0190] In the actual operation of the outbound call system, the outbound call effect often shows obvious peak and trough characteristics as the time period changes. For example, the outbound call connection rate is usually at its peak from 9:00 to 11:00 a.m. on weekdays, while it may be at its low point from 12:00 to 2:00 p.m. In order to better adapt to this periodic change, this embodiment introduces a peak and trough control mechanism based on a genetic algorithm to optimize the outbound call flow rate at different time periods in an intelligent way.

[0191] In the specific implementation, the outbound call flow rate needs to be encoded first. The system divides 24 hours a day into 48 time segments in half-hour units. The outbound call flow rate of each time segment is encoded as a gene, thus forming a chromosome containing 48 genes. The value range of each gene is set between the minimum outbound call flow rate and the maximum outbound call flow rate of the system, for example, it can be 0 to 2000 calls / minute. This encoding method not only ensures the continuity of time, but also provides a good foundation for the optimization operation of the genetic algorithm.

[0192] Next, the system will build a fitness function based on historical outbound call data. In this embodiment, the connection rate is selected as the main fitness indicator, and the resource utilization rate is combined as a constraint. Specifically, the fitness function can be expressed as: F = α × connection rate + β × resource utilization rate, where α and β are weight coefficients used to balance the importance of these two indicators. A higher connection rate and a moderate resource utilization rate will result in a higher fitness score, making it more likely to be retained during the evolution process.

[0193] In the iterative process of the genetic algorithm, the system optimizes chromosomes through three basic operations: selection, crossover, and mutation. The selection operation uses the roulette method, and the chromosome with higher fitness has a greater probability of being selected. The crossover operation randomly selects two parent chromosomes, exchanges genes at a certain point in time, and generates new daughter chromosomes. The mutation operation randomly changes the values ​​of certain genes with a small probability, which can help the algorithm jump out of the local optimal solution and explore a larger solution space.

[0194] In order to ensure the convergence of the algorithm and the practicality of the results, this embodiment also introduces some constraints. First, the change in the outbound call flow rate between adjacent time segments cannot be too large, so as to avoid drastic fluctuations in the system load. Second, the outbound call flow rate in certain special time periods (such as late at night) will be limited to a lower level. In addition, the system will also set corresponding penalty factors based on the resource costs of different time periods, so as to consider economic factors in the optimization process.

[0195] After multiple iterations of optimization, the genetic algorithm will eventually output a set of outbound call flow rate thresholds for peak and trough periods. These thresholds reflect the optimal outbound call strategy for different periods of time, and the system will dynamically adjust the target outbound call flow rate accordingly. For example, during the identified peak period (such as 9:00 a.m. to 11:00 a.m.), the system will increase the target outbound call flow rate to the peak threshold; and during the trough period (such as the lunch break), it will reduce the target outbound call flow rate to the trough threshold accordingly.

[0196] This genetic algorithm-based control method has strong adaptive capabilities. The system will re-run the genetic algorithm regularly (for example, every week) and update the peak and trough thresholds based on the latest historical data. This can not only adapt to seasonal changes, but also respond to changes in user behavior patterns in a timely manner. At the same time, since the algorithm itself is random, the method of taking the average value of multiple runs can also improve the stability and reliability of the results.

[0197] In order to verify the optimization effect, this embodiment also designs a complete effect evaluation mechanism. The system will record the changes in various indicators before and after optimization, including average connection rate, resource utilization, cost-effectiveness, etc. Practice shows that the use of this peak and valley control mechanism based on genetic algorithm can increase the overall connection rate of the system by more than 15%, while significantly reducing resource waste, and has good application value.

[0198] It should be emphasized that this peak and trough control mechanism does not operate independently, but is closely coordinated with the mechanisms introduced above, such as slow start, sensitivity adjustment, and resource expansion and contraction. In practical applications, the thresholds obtained by genetic algorithm optimization will serve as a reference for these basic mechanisms to assist the system in making better control decisions. This multi-level optimization strategy ultimately realizes the intelligent adjustment of the outbound call system in different time periods and scenarios, providing users with more efficient and stable services.

[0199] 2. Strategy optimization based on ant colony algorithm

[0200] In order to further optimize the control strategy, this embodiment introduces the ant colony algorithm:

[0201] 1) Mapping the outbound call flow rate adjustment strategy to the path selection of ants;

[0202] 2) Calculate pheromone concentration based on resource utilization and connection rate;

[0203] 3) By iteratively updating the pheromone concentration, it eventually converges to the optimal outbound call flow rate control strategy.

[0204] In complex and ever-changing outbound call scenarios, it is often difficult to adapt to various emergencies and special scenarios by relying solely on fixed control strategies. In order to further improve the intelligence level of the system, this embodiment introduces a strategy optimization mechanism based on the ant colony algorithm. This mechanism simulates the principle of ant foraging behavior, continuously explores and optimizes the control strategy of the outbound call flow rate, so that the system can respond to various business scenarios more flexibly.

[0205] In this embodiment, the control strategy of the outbound call flow rate is abstracted into the path selection problem of ants. Each possible control strategy corresponds to a feasible path for the ant from the starting point to the end point, where each node of the path represents a different combination of control parameters. For example, a complete control strategy may include multiple parameters such as initial sensitivity, sensitivity adjustment threshold, resource expansion and contraction trigger conditions, etc. This abstract method transforms complex multi-dimensional decision-making problems into intuitive path optimization problems.

[0206] In order to evaluate the advantages and disadvantages of different control strategies, the system has designed a comprehensive evaluation mechanism based on multiple indicators. The first is resource utilization, which indicates the efficiency of using system resources; the second is connection rate, which is a key indicator reflecting the effect of outbound calls; in addition, it also includes auxiliary indicators such as system stability and response time. The weighted combination of these indicators constitutes the "pheromone" on the path. The higher the concentration of pheromone, the better the effect of the control strategy.

[0207] During the operation of the ant colony algorithm, the system simulates the behavior of multiple ants at the same time. When each ant selects the next node, it will comprehensively consider the pheromone concentration and heuristic information on the path. Heuristic information is usually set based on business experience. For example, a parameter combination with low resource utilization will obtain a higher heuristic value. This dual guidance mechanism ensures that the algorithm can use historical experience without being completely constrained by historical data.

[0208] The update of pheromones is the core of the ant colony algorithm. After each round of iteration, the system will update the pheromones on all paths. On the one hand, the pheromones on all paths will gradually decay over time, which reflects the timeliness of historical experience; on the other hand, paths with better performance will receive pheromone enhancements, making these excellent control strategies more likely to be selected by subsequent ants. The specific formula for pheromone update is: τ(t+1)=(1-ρ)×τ(t)+Δτ, where ρ is the pheromone decay coefficient and Δτ is the pheromone increment.

[0209] In order to prevent the algorithm from falling into a local optimal solution, this embodiment also introduces an adaptive parameter adjustment mechanism. For example, when the system detects that most ants are choosing similar paths, it will appropriately increase the probability of random selection to encourage the algorithm to explore new solution spaces. At the same time, the system will also regularly inject some random control strategies, which may not perform well temporarily, but can provide more exploration directions for the algorithm.

[0210] In actual applications, the ant colony algorithm does not run all the time, but uses a periodic optimization method. The system collects complete operating data once in each optimization cycle (for example, every day), and then runs the ant colony algorithm to find a better control strategy. Once a strategy with better performance is found, the system will smoothly switch to the new strategy to avoid affecting system stability due to drastic changes in strategy.

[0211] It is particularly important to point out that, considering the real-time requirements of the outbound call system, this embodiment adopts some special optimization measures when deploying the ant colony algorithm. The first is parallel computing. The system will simulate the behavior of multiple groups of ants at the same time, making full use of the computing power of multi-core processors. The second is incremental optimization. The new round of optimization will be carried out based on the optimal results of the previous round, which can speed up the convergence of the algorithm.

[0212] Practical tests have shown that this strategy optimization mechanism based on the ant colony algorithm can effectively improve system performance. Under the same hardware conditions, the average connection rate of the optimized system has increased by about 12%, resource utilization has increased by about 20%, and the stability of the system has also been significantly improved. Especially in the face of sudden business peaks, the optimized system has shown stronger adaptability.

[0213] Finally, this strategy optimization mechanism based on the ant colony algorithm complements the aforementioned genetic algorithm optimization mechanism. The genetic algorithm mainly solves the problem of optimizing the flow rate threshold in different time periods, while the ant colony algorithm is committed to optimizing the control strategy itself. The combination of the two algorithms greatly improves the intelligence level of the entire outbound call system. In actual deployment, the system will flexibly select or combine these two optimization mechanisms according to the characteristics of the specific scenario to achieve the best control effect.

[0214] 4. Device Example

[0215] See also Figure 5 The present application also provides an outbound call flow rate control device, comprising:

[0216] A receiving unit, used for receiving an initial sensitivity and a target outgoing call flow rate;

[0217] An acquisition unit, used to acquire the current connection rate and resource occupancy rate;

[0218] A control unit, for executing a sensitivity-based soft start process;

[0219] An adjustment unit, used for adjusting the current sensitivity based on resource occupancy;

[0220] The resource management unit is used to perform expansion and reduction of resources.

[0221] The specific functions of each unit of the device correspond to the above method embodiments and will not be repeated here.

[0222] Those skilled in the art will appreciate that each unit in the apparatus of the above embodiments may be implemented by hardware or by running a software module on a processor. The present application is not limited to any specific combination of hardware and software.

[0223] 5. Application Examples

[0224] The following is a specific application example to illustrate the working process of this application:

[0225] Assume that there is an outbound call task, the required outbound call flow rate is 2000 calls / minute, the initial sensitivity is set to 100 calls / minute2, and the outbound call flow rate is 200 calls / minute at the beginning. Theoretically, the outbound call flow rate should reach the maximum flow rate after (2000-200) / 100=18 minutes.

[0226] However, due to resource limitations, when the flow rate increased to 1200 channels / minute, the negative feedback detected that the resources were close to full load. At this time, the sensitivity began to gradually decrease until it reached 0 channels / minute2, triggering resource expansion and ultimately meeting the demand of 2000 channels / minute.

[0227] After a period of time, the system monitors that the connection rate begins to decline (such as entering the lunch break period), and negative feedback detects that resource usage has dropped to a light load state, triggering resource reduction to reduce resource waste.

[0228] After the lunch break, the connection rate increased and the resources could not support the current flow rate. Negative feedback detected that the resources were about to be overloaded, and the sensitivity dropped to a negative value (such as -100 calls / minute2) until the outbound call flow rate stabilized at 1200 calls / minute, and the sensitivity returned to 0 calls / minute2. Because the required 2000 calls / minute was not reached, the system triggered resource expansion again.

[0229] It can be seen from the above process that the present application can automatically adjust the outbound call flow rate and resource allocation according to the actual situation, which not only ensures the stable operation of the task but also avoids the waste of resources.

[0230] All technical features in the embodiments may be implemented individually or in any combination. Those skilled in the art should understand that the specific embodiments cited herein are provided to illustrate the principles and features of the present application, but the present application is not limited to the specific embodiments.

[0231] In the description of this application, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a mechanical connection or an electrical connection, or it can be a detachable connection, or it can be integrated; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0232] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0233] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0234] Although the embodiments of this specification are disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, so the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A method for controlling outbound call flow rate, characterized in that: The method comprises: receiving an initial sensitivity and a target outbound call flow rate, wherein the initial sensitivity is used to represent an initial rate of change of the outbound call flow rate; Get the current connection rate and resource occupancy rate; Based on the initial sensitivity, a slow start process is performed, during which the outbound call flow rate is gradually increased according to the current sensitivity until the target outbound call flow rate or resource full load is reached, wherein the current sensitivity represents the real-time change rate of the outbound call flow rate, and the initial value of the current sensitivity is set to the initial sensitivity; Adjusting the current sensitivity based on the resource occupancy rate; as well as, When the outbound call flow rate does not reach the target outbound call flow rate and the current sensitivity is adjusted to zero, resource expansion is triggered; when the outbound call flow rate reaches the target outbound call flow rate and the resource occupancy rate is lower than the preset threshold, resource reduction is triggered.

2. The method according to claim 1, characterized in that: The performing the soft start process based on the initial sensitivity includes: Setting an initial outbound call flow rate, wherein the initial outbound call flow rate is lower than the estimated full-load outbound call flow rate; Based on the initial sensitivity, the current outbound call flow rate is increased according to the following formula every minute: Current outbound call flow rate = outbound call flow rate in the last minute + current sensitivity × time interval.

3. The method according to claim 1, characterized in that The adjusting the current sensitivity based on the resource occupancy rate includes: When it is detected that the difference between the outbound call flow rate and the estimated full-load outbound call flow rate is less than the current sensitivity, the current sensitivity is reduced by using a binary approximation method; The binary approximation method is: current sensitivity = (estimated full-load outbound call flow rate - current outbound call flow rate) / 2 / time interval.

4. The method according to claim 1, characterized in that: The triggering resource expansion includes: When it is detected that the current sensitivity is reduced to zero and the target outbound call flow rate is not reached, increasing the resource quota; The current sensitivity is reset to an initial value, and the outbound call flow rate is continued to be improved.

5. The method according to claim 1, characterized in that The method further comprises: Monitor whether resource expansion or reduction is successful; When it is detected that resource expansion or reduction fails continuously for more than a preset time, an alarm message is sent.

6. The method according to claim 1, characterized in that The method further comprises: Based on historical outbound call data, a genetic algorithm is used to obtain the outbound call flow rate thresholds during peak and trough periods; Wherein, the method of using a genetic algorithm to obtain the outbound call flow rate threshold value during the peak and trough periods includes: Encode the outgoing call flow rate as a chromosome; The connection rate is used as the fitness function; Perform selection, crossover and mutation operations to obtain the optimal outbound call flow rate threshold during peak and trough periods.

7. The method according to claim 6, characterized in that The method further comprises: Dynamically adjust the target outbound call flow rate according to the outbound call flow rate threshold during the peak and trough periods; Wherein, during the peak period, the target outbound call flow rate is increased to the outbound call flow rate threshold during the peak period; During the trough period, the target outbound call flow rate is reduced to the outbound call flow rate threshold of the trough period.

8. The method according to claim 1, characterized in that: The method further comprises: A meta-heuristic algorithm is used to generate multiple candidate solutions, each of which contains an outbound call flow rate adjustment strategy; Evaluate the performance indicators of each candidate solution, including resource utilization and connection rate; An optimal solution is selected from the candidate solutions as the optimal outbound call flow rate control strategy.

9. The method according to claim 8, characterized in that The meta-heuristic algorithm is an ant colony algorithm, and the step of selecting the optimal solution from the candidate solutions as the outbound call flow rate control strategy includes: Mapping the outbound call flow rate adjustment strategy to the path selection of ants; Calculating pheromone concentration according to the resource utilization rate and the connection rate; The pheromone concentration is iteratively updated until it converges to the optimal outgoing call flow rate control strategy.

10. An outbound call flow rate control device, characterized in that: The device comprises: A receiving unit, used for receiving an initial sensitivity and a target outbound call flow rate, wherein the initial sensitivity is used for indicating an initial rate of change of the outbound call flow rate; An acquisition unit, used to acquire the current connection rate and resource occupancy rate; a control unit, configured to perform a slow start process based on the initial sensitivity, during which the outbound call flow rate is gradually increased according to the current sensitivity until the target outbound call flow rate or the resource is fully loaded, wherein the current sensitivity represents a real-time rate of change of the outbound call flow rate, and an initial value of the current sensitivity is set to the initial sensitivity; an adjusting unit, configured to adjust the current sensitivity based on the resource occupancy rate; The resource management unit is used to trigger resource expansion when the outbound call flow rate does not reach the target outbound call flow rate and the current sensitivity is adjusted to zero; and to trigger resource reduction when the outbound call flow rate reaches the target outbound call flow rate and the resource occupancy rate is lower than a preset threshold.

Images