Dynamic flow distribution method and system

Abstract

The invention provides a dynamic flow distribution method and system, and relates to the technical field of flow distribution. The method comprises the steps of obtaining a target order; a multi-dimensional weight model is adopted, and the weight of each dimension of the target order is adjusted according to the market supply and demand relationship; the dimensions of the target order comprise the order amount, the account number grade, the emergency degree, the historical completion rate and the special requirement; inputting the adjusted weight of each dimension of the target order into a long short-term memory artificial neural network to obtain a predicted value score of the target order; a dynamic flow distribution algorithm is adopted, and a corresponding target beater is matched for the target order according to the prediction value; the dynamic flow distribution algorithm comprises a dual-objective optimization model; and solving the dual-objective optimization model by adopting a multi-objective genetic algorithm and an exposure attenuation function. According to the invention, dynamic intelligent adjustment of the order exposure strategy can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic distribution, and more particularly, to a dynamic traffic distribution method and system. Background Art

[0002] Game leveling service platforms generally adopt an order distribution mechanism based on keyword matching or simple sorting rules, which has the following technical bottlenecks:

[0003] Defects of static allocation: Traditional systems mostly rely on fixed rules (such as price reverse order / release time sorting), ignoring the dynamic decay characteristics of order value and the real-time changes in the capabilities of game levelers. For example, high-value urgent orders may sink to the bottom due to late release time, while stale and inefficient orders occupy the exposure positions for a long time, resulting in a GMV loss of more than 30% for the platform (industry research data).

[0004] Single matching dimension: Existing technologies usually only use the order amount or the level of game levelers as the matching basis, lacking the ability of multi-dimensional collaborative analysis. According to the public data of a certain platform, the order timeout rate caused by ignoring the device performance of game levelers is as high as 25%, and more than 30% of the game leveler resources are idle due to the lack of consideration of time period preferences.

[0005] Ecological imbalance problem: Top game levelers are overexposed and overloaded with work (the average daily order volume is 5 - 8 times that of ordinary game levelers), while novice game levelers have difficulty in obtaining growth opportunities due to uneven traffic distribution, resulting in a monthly game leveler churn rate of 15% - 20%, significantly increasing the platform operation cost.

[0006] Weak risk control: Existing systems lack an intelligent downgrading mechanism, and risk orders such as fraudulent orders and user orders with a high dispute rate can still obtain normal exposure. A third-party evaluation shows that the proportion of platform complaints caused by such orders exceeds 35%.

[0007] In response to the above problems, the industry has tried to introduce a collaborative filtering recommendation algorithm, but it has two major limitations: (1) It is difficult to handle the time-sensitive requirements unique to the game leveling scenario, and the non-linear decay characteristics of order value over time are not modeled; (2) It cannot balance the multi-objective optimization requirements of maximizing platform revenue and maintaining a healthy game leveler ecosystem. Summary of the Invention

[0008] The purpose of the present invention is to provide a dynamic traffic distribution method and system to achieve dynamic and intelligent adjustment of the order exposure strategy for the deficiencies in the above-mentioned existing technologies.

[0009] To achieve the above purpose, the technical solutions adopted in the embodiments of the present application are as follows:

[0010] In a first aspect, an embodiment of the present application provides a dynamic traffic allocation method, including: obtaining a target order; adjusting the weights of each dimension of the target order according to the market supply and demand relationship by using a multi-dimensional weight model; the dimensions of the target order include order amount, account level, urgency, historical completion rate, and special requirements; inputting the adjusted weights of each dimension of the target order into a long short-term memory artificial neural network to obtain the predicted value score of the target order; using a dynamic traffic allocation algorithm to match the corresponding target hitman for the target order according to the predicted value score; the dynamic traffic allocation algorithm includes a bi-objective optimization model; the bi-objective optimization model is solved by using a multi-objective genetic algorithm and an exposure decay function.

[0011] In one implementation, the initial weights of each dimension of the target order include: the initial weight of the order amount is 50%, the initial weight of the account level is 20%, the initial weight of the urgency is 15%, the initial weight of the historical completion rate is 10%, and the initial weight of the special requirements is 5%.

[0012] In one implementation, the step of using the dynamic traffic allocation algorithm to match the corresponding target hitman for the target order according to the predicted value score includes: using the multi-objective genetic algorithm and the exposure decay function, based on the bi-objective optimization model, to match the corresponding target hitman for the target order according to the predicted value score.

[0013] In one implementation, the step of using the multi-objective genetic algorithm and the exposure decay function, based on the bi-objective optimization model, to match the corresponding target hitman for the target order according to the predicted value score includes: inputting the predicted value score into the bi-objective optimization model; the bi-objective optimization model is as follows:

[0014] Suppose there is an order set and a hitman set The bi-objective optimization model is as follows:

[0015] Objective 1: Maximize the platform revenue

[0016]

[0017] Objective 2: Maximize the hitman satisfaction

[0018]

[0019] Constraint conditions:

[0020]

[0021] Among them, V i represents order O iPrediction value score; M ij Indicates order O i And the fighter P j Degree of match; x ij Indicates a binary decision variable (1 indicates that order O is assigned i To the fighter P j , 0 otherwise); S j Indicates the satisfaction quantification value of the fighter P j ; W j Indicates the fighter P j Maximum working load capacity; workload j Indicates the fighter P j Current working load;

[0022] The multi-objective genetic algorithm is used to solve the bi-objective optimization model to obtain the target fighter corresponding to the target order; the exposure attenuation function is used to dynamically adjust the search space of the multi-objective genetic algorithm.

[0023] In one implementation, the target fighter exists in the database in the form of a fighter profile; the fighter profile is determined by a fighter profile modeling system.

[0024] In one implementation, before using the dynamic traffic allocation algorithm to match the target order with the corresponding target fighter according to the prediction value score, the method further includes: obtaining multi-dimensional data of the fighter; the multi-dimensions include win rate, response speed, device performance, time period preference, customer evaluation, and special skills; using a dynamic weight adjustment algorithm to adjust the weights of the multi-dimensional data; using an interest graph analyzer to analyze the adjusted multi-dimensional data; and performing modeling according to the analysis results to obtain a six-dimensional ability evaluation system.

[0025] In one implementation, using the interest graph analyzer to analyze the adjusted multi-dimensional data includes: constructing a heterogeneous network using Neo4j according to the adjusted multi-dimensional data; where the node types include fighters, game characters, equipment, and tactics; the edge relationships include fighter-character, character-equipment, and fighter-tactics; using the GraphSAGE algorithm to generate node vectors to obtain the deep association features of the heterogeneous network; and determining the interest degree of the fighter for each skill according to the skill association strength formula.

[0026] In one implementation, after obtaining the multi-dimensional data of the fighter, the method further includes: real-time collecting the operation heat map data of the fighter through the buried point technology to dynamically update the six-dimensional ability evaluation system; the operation heat map includes the operation event density per minute.

[0027] In one implementation, after inputting the adjusted dimensional weights of the target order into the long short-term memory artificial neural network to obtain the predicted value score of the target order, the method further includes: adjusting the traffic position of the target order in the three-level traffic control system according to the predicted value score; the three-level traffic control system includes a golden traffic position, a regular traffic position, and a downgraded traffic position; the golden traffic position includes the top 5% of the orders, the regular traffic position includes the middle 70% of the orders, and the downgraded traffic position includes the bottom 25% of the orders.

[0028] In a second aspect, an embodiment of the present application further provides a dynamic traffic allocation system, including: an order value score evaluation engine for inputting the adjusted dimensional weights of the target order into the long short-term memory artificial neural network to obtain the predicted value score of the target order; a fighter portrait modeling system for obtaining multi-dimensional data of the fighter; the multi-dimensions include win rate, response speed, device performance, time period preference, customer evaluation, and special skills; adjusting the weights of the multi-dimensional data using a dynamic weight adjustment algorithm; analyzing the adjusted multi-dimensional data using an interest graph analyzer; performing modeling according to the analysis results to obtain a six-dimensional ability evaluation system; a dynamic traffic allocation module for using a dynamic traffic allocation algorithm to match a corresponding target fighter for the target order according to the predicted value score; a visual operation console, including a three-dimensional traffic sand table, a drag-and-drop rule configurator, and a real-time data dashboard; the three-dimensional traffic sand table is used to display the order value score range on the X-axis, the fighter level distribution on the Y-axis, and the time period traffic heat on the Z-axis; the real-time data dashboard is used to monitor the exposure conversion rate, warn of downgraded orders, and analyze the traffic position efficiency.

[0029] In a third aspect, an embodiment of the present application provides a computer device, including: a processor, a storage medium, and a bus, where the storage medium stores program instructions executable by the processor. When the computer device runs, the processor communicates with the storage medium through the bus, and the processor executes the program instructions to perform the steps of any of the above methods.

[0030] In a fourth aspect, an embodiment of the present application provides a non-volatile computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it performs the steps of any of the above methods.

[0031] The beneficial effects of the present application are as follows: First, obtain the target order; Second, adopt a multi-dimensional weight model to adjust the weights of each dimension of the target order according to the market supply and demand relationship; The dimensions of the target order include order amount, account level, urgency, historical completion rate, and special requirements; Third, input the adjusted weights of each dimension of the target order into a long short-term memory artificial neural network to obtain the predicted value score of the target order; Finally, adopt a dynamic traffic allocation algorithm to match the corresponding target hitters for the target order according to the predicted value score; The dynamic traffic allocation algorithm includes a bi-objective optimization model; The bi-objective optimization model is solved by using a multi-objective genetic algorithm and an exposure decay function. By introducing a multi-objective optimization algorithm and a reinforcement learning mechanism, the present application realizes the dynamic intelligent adjustment of the order exposure strategy. Compared with the traditional fixed traffic allocation method, the actual measurement shows that the order matching efficiency is increased by 42%, the satisfaction of hitters receiving orders is increased by 28%, and the platform GMV is increased by 19%. The specially designed visual operation console enables operators to adjust the strategy parameters in real time, taking into account the balance between system automation and manual intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a flowchart of a dynamic traffic allocation method provided by an embodiment of the present application;

[0034] Figure 2 It is a flowchart of a dynamic traffic allocation method provided by an embodiment of the present application;

[0035] Figure 3 It is a flowchart of a dynamic traffic allocation method provided by an embodiment of the present application;

[0036] Figure 4 It is a flowchart of a dynamic traffic allocation method provided by an embodiment of the present application;

[0037] Figure 5 It is a structural diagram of a dynamic traffic allocation system provided by an embodiment of the present application;

[0038] Figure 6 It is a structural diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0040] Therefore, the detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts fall within the scope of protection of the present application.

[0041] In the description of the present application, it should be noted that if terms such as "upper", "lower", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this application is usually placed during use, it is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present application.

[0042] In addition, terms such as "first", "second", etc. in the description and claims of the present invention and the above accompanying drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0043] It should be noted that the features in the embodiments of the present application can be combined with each other without conflict.

[0044] Figure 1 A schematic flowchart of a dynamic traffic allocation method provided for the embodiments of the present application; as Figure 1 shown, the method includes:

[0045] Step 110, obtain a target order.

[0046] Among them, the target order, i.e., the leveling task, is the core data entity for system scheduling. Generally, it is a structured data object, including the full-dimensional description and dynamic status information of the leveling task. It can be saved in the JSON Schema format in the system. Among them, the full-dimensional description of the leveling task includes value density, spatio-temporal constraints, skill vectors, and risk tags; the dynamic status information includes the dynamic status. Specifically, the full-dimensional description and dynamic status information are shown in Table 1 below:

[0047]

[0048] Table 1

[0049] Step 120: Use a multi-dimensional weight model to adjust the weights of each dimension of the target order according to the market supply and demand relationship.

[0050] Among them, the dimensions of the target order include order amount, account level, urgency, historical completion rate, and special requirements.

[0051] Among them, the multi-dimensional weight model is the core algorithm engine for order value evaluation. It adopts a dynamic weighted fusion mechanism to quantify multiple characteristic dimensions of the order into a unified value score. The model realizes the autonomous evolution of weight parameters through a real-time feedback mechanism and machine learning optimization.

[0052] In actual operation, the initial weights of each dimension of the target order are as follows: the initial weight of the order amount is 50%, the initial weight of the account level is 20%, the initial weight of the urgency is 15%, the initial weight of the historical completion rate is 10%, and the initial weight of the special requirements is 5%. The above initial weights can be obtained through historical data regression analysis. The dynamic adjustment of the weights of each dimension of the subsequent target order can be adjusted based on the initial weights of each dimension of the above target order. The specific dynamic adjustment logic is shown in Table 2 below; in this way, the weight adjustment efficiency can be improved.

[0053] The specific information of each dimension of the above target order is shown in Table 2 below:

[0054]

[0055] Table 2

[0056] In actual operation, a real-time feedback mechanism can be adopted to update the weights about every 30 minutes; the weight update is shown in the following formula (1):

[0057]

[0058] Among them, η represents the learning rate, generally defaulting to 0.01; α and β represent the target balance coefficients;

[0059] The constraint conditions are:

[0060] The fluctuation range of the single - dimensional weight ≤ ±15%;

[0061] The total weight change in adjacent cycles ≤ ±5%;

[0062] The full - dimensional weight ≥ 40%.

[0063] Step 130: Input the weights of each dimension of the adjusted target order into the long - short - term memory artificial neural network to obtain the predicted value score of the target order.

[0064] Among them, the long - short - term memory artificial neural network can adopt the existing long - short - term memory artificial neural network; specifically, the long - short - term memory network (LSTM, Long Short - Term Memory) is a type of recurrent neural network in time, which is specially designed to solve the long - term dependence problem existing in general RNNs (recurrent neural networks). All RNNs have a chain - like form of repeating neural network modules. In a standard RNN, this repeating structural module has a very simple structure, such as a tanh layer; the main idea of the long - short - term memory network is: store information in individual memory cells, and the memory cells in different hidden layers form a conveyor belt (the red line in the figure) through a small amount of linear interaction to achieve the flow of information. At the same time, a "gate" structure is introduced to add or delete information in the memory cells and control the flow of information.

[0065] The long - short - term memory artificial neural network can capture the temporal change law of order features, and then respond to market supply - demand fluctuations in real - time (for example: the sudden increase in the weight of device requirements due to the launch of a new game).

[0066] Step 140: Adopt a dynamic traffic allocation algorithm to match the corresponding target hitters for the target order according to the predicted value score.

[0067] Among them, the dynamic traffic allocation algorithm includes a bi - objective optimization model; the bi - objective optimization model is solved using a multi - objective genetic algorithm and an exposure decay function.

[0068] Specifically, the above - mentioned step 140 includes the following steps:

[0069] Use a multi - objective genetic algorithm and an exposure decay function, based on the bi - objective optimization model, to match the corresponding target hitters for the target order according to the predicted value score.

[0070] Among them, both the exposure decay function and the bi - objective optimization model are newly developed.

[0071] Specifically, as Figure 2 shown, the above - mentioned steps can further include the following steps 210 and 220:

[0072] Step 210: Input the predicted value score into the bi-objective optimization model. The bi-objective optimization model is as follows:

[0073] Suppose there is an order set and a hitman set The bi-objective optimization model is shown by the following formula:

[0074] Objective 1: Maximize the platform revenue, as shown by formula (2) below:

[0075]

[0076] Objective 2: Maximize the hitman's satisfaction, as shown by formula (3) below:

[0077]

[0078] The constraint condition is shown by formula (4) below:

[0079]

[0080] Among them, V i represents the predicted value score of order O i ; M ij represents the matching degree between order O i and hitman P j ; x ij represents a binary decision variable (1 means allocating order O i to hitman P j , 0 otherwise); S j represents the quantified value of hitman P j 's satisfaction; W j represents the maximum working capacity of hitman P j ; workload j represents the current workload of hitman P j .

[0081] Furthermore, the calculation logic of V i is: V i = α · amount + β · account level...; M ij = cos(order demand vector, hitman skill vector); Among them, represents the total revenue of the matching orders, λ represents the historical order acceptance volume; W j = daily order acceptance limit · order difficulty coefficient; t is the number of allocated orders.

[0082] The dual-objective optimization model in this step can ensure that high-value orders are preferentially matched with high-quality beaters, prevent excessive exploitation of top beaters, maintain ecological balance, and find the optimal balance solution set through the Pareto frontier. Secondly, the constraints are dynamic constraints, including the workload feedback mechanism and the satisfaction decay function; finally, in the processing of mixed variables, the discrete variable x ij and the continuous variable S j can be co-optimized. The branch and bound method is used to handle integer constraints, and the sequential quadratic programming is combined to optimize continuous variables.

[0083] Step 220: Solve the dual-objective optimization model using the multi-objective genetic algorithm to obtain the target beaters corresponding to the target orders.

[0084] Among them, the exposure decay function is used to dynamically adjust the search space of the multi-objective genetic algorithm.

[0085] The multi-objective genetic algorithm can use the improved NSGA-II multi-objective genetic algorithm.

[0086] The exposure decay function is used to quantify the value decay of the order over time and control its exposure intensity at the beater end, as shown in the following formula (5):

[0087] f(t) = V base × e -λt × C compensation (5)

[0088] Among them, V base represents the order base value score, which is calculated by the multi-dimensional weight model; λ represents the decay rate coefficient; t represents the order survival time; C compensation represents the compensation factor, which is generally activated when the anti-starvation mechanism is triggered.

[0089] Furthermore, the dynamic adjustment rule of the decay rate coefficient λ is as follows:

[0090] During the low peak period (01:00 - 08:00), λ = 0.02;

[0091] During the flat peak period (08:00 - 18:00), λ = 0.05;

[0092] During the high peak period (18:00 - 24:00), λ = 0.08;

[0093] During the sudden traffic period (such as: flash sale activities, etc.), λ = 0.12.

[0094] The calculation rule of time t is as shown in the following formula (6):

[0095]

[0096] Among them, ti is the current timestamp; t0 is the order creation timestamp.

[0097] In actual operation, the timer pauses when the hitman actively checks the order; for orders that span multiple days, the t value is reset at 08:00 the next day, and t new = t - 24.

[0098] The starvation prevention compensation mechanism is as shown in the following formula (7):

[0099] When the direct current time of the order exceeds the time threshold T threshold it is activated:

[0100]

[0101] where T threshold can be set to 6 hours; γ = 0.05 (compensation rate per hour).

[0102] Furthermore, the target hitman exists in the database in the form of a hitman profile; the hitman profile is determined by the hitman profile modeling system.

[0103] Among them, the hitman profile modeling system aims to construct a dynamically updated hitman ability evaluation system through multi-dimensional data collection, analysis, and modeling, providing accurate decision-making basis for the intelligent matching algorithm. The system adopts a modular design, including a data collection layer, a feature engineering layer, a model calculation layer, and an application interface layer, supporting a hybrid computing architecture that combines real-time update and offline batch processing.

[0104] Specifically, as Figure 3 shown, before the above step 140, the dynamic traffic allocation method provided by the embodiments of the present application may further include the following steps 310 to 340:

[0105] Step 310, obtain multi-dimensional data of the hitman.

[0106] Among them, the multi-dimensions include win rate, response speed, device performance, time period preference, customer evaluation, and special skills.

[0107] Step 320, adjust the weights of the multi-dimensional data using the dynamic weight adjustment algorithm.

[0108] Among them, the dynamic weight adjustment algorithm is as shown in the following formula (8):

[0109]

[0110] where E is the matching efficiency objective function; α, β are learning rate parameters; is the platform policy adjustment amount.

[0111] Step 330: Analyze the adjusted multi-dimensional data using an interest graph analyzer.

[0112] Among them, the analysis process of the interest graph analyzer mainly includes: graph structure construction, graph embedding calculation, and interest degree calculation.

[0113] Specifically, as Figure 4 shown, the above Step 330 can further include the following Steps 410 to 430:

[0114] Step 410: Use Neo4j to construct a heterogeneous network based on the adjusted multi-dimensional data.

[0115] Among them, the node types include thugs, game characters, equipment, and tactics; the edge relationships include thug-character, character-equipment, and thug-tactics.

[0116] Step 420: Use the GraphSAGE algorithm to generate node vectors and obtain the deep association features of the heterogeneous network.

[0117] In actual operation, this step generally outputs 128-dimensional embedding vectors, which can capture deep association features.

[0118] Step 430: Determine the interest degree of the thug for each skill according to the skill association strength formula.

[0119] Among them, the interest degree I ab is calculated as shown in the following formula (9):

[0120]

[0121] Among them, χ ab represents the number of times skills a and b are used together; φ a represents the number of times skill a is used; represents the number of times skill b is used.

[0122] Step 340: Model according to the analysis results to obtain a six-dimensional ability evaluation system.

[0123] Among them, the obtained six-dimensional ability evaluation system is shown in Table 3 below:

[0124]

[0125] Table 3

[0126] In actual operation, the operation heat map data of the thug can also be collected in real time through the data burying technology to dynamically update the six-dimensional ability evaluation system; the operation heat map includes the operation event density per minute.

[0127] In actual operation, after the above step 130, the traffic position of the target order can also be adjusted in the three-level traffic control system according to the predicted value score. The three-level traffic control system includes a golden traffic position, a regular traffic position, and a downgraded traffic position. The golden traffic position includes the top 5% of orders, the regular traffic position includes the middle 70% of orders, and the downgraded traffic position includes the bottom 25% of orders.

[0128] After introducing the dynamic traffic allocation method of the exemplary embodiments of the present disclosure, next, reference is made to Figure 5 to describe the dynamic traffic allocation system 500 of the exemplary embodiments of the present disclosure.

[0129] Reference is made to Figure 5 , the dynamic traffic allocation system 500 includes: an order value score evaluation engine 510 for inputting the weights of each dimension of the adjusted target order into a long short-term memory artificial neural network to obtain the predicted value score of the target order; a striker profile modeling system 520 for obtaining multi-dimensional data of the striker. The multi-dimensions include win rate, response speed, device performance, time period preference, customer evaluation, and special skills. The weights of the multi-dimensional data are adjusted by using a dynamic weight adjustment algorithm; the adjusted multi-dimensional data is analyzed by using an interest graph analyzer; modeling is performed according to the analysis results to obtain a six-dimensional ability evaluation system; a dynamic traffic allocation module 530 for using a dynamic traffic allocation algorithm to match a corresponding target striker for the target order according to the predicted value score; a visual operation console 540 including a three-dimensional traffic sand table, a drag-and-drop rule configurator, and a real-time data dashboard. The three-dimensional traffic sand table is used to display the order value score range on the X-axis, the striker level distribution on the Y-axis, and the time period traffic heat on the Z-axis. The real-time data dashboard is used to monitor the exposure conversion rate, warn about downgraded orders, and analyze the traffic position effectiveness.

[0130] The above device is used to execute the method provided in the foregoing embodiment, and its implementation principle and technical effects are similar and will not be elaborated herein.

[0131] The above-mentioned modules may be one or more integrated circuits configured to implement the above methods. For example: one or more Application Specific Integrated Circuits (ASICs), or, one or more microprocessors, or, one or more Field Programmable Gate Arrays (FPGAs), etc. Again, when a certain module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a Central Processing Unit (CPU) or other processors that can call program code. Again, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0132] Figure 6 Schematic diagram of the computer device provided by the embodiment of the present application. This device may be integrated into a terminal device or a chip of a terminal device, and the terminal may be a computing device with data processing capabilities.

[0133] This device includes: a processor 601, a storage medium 602, and a bus 603.

[0134] The storage medium 602 stores program instructions executable by the processor 601. When the computer device 600 runs, the processor 601 communicates with the storage medium 602 through the bus 603, and the processor 601 executes the program instructions to execute the above method embodiments. The specific implementation manners and technical effects are similar and will not be elaborated here.

[0135] Optionally, the present invention further provides a program product, such as a computer-readable storage medium, including a program that is used to execute the above method embodiments when executed by a processor.

[0136] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the displayed or discussed coupling or direct coupling or communication connection between each other may be through some interfaces. The indirect coupling or communication connection of the device or unit may be in an electrical, mechanical or other form.

[0137] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0138] In addition, each functional unit in various embodiments of the present invention may be integrated in a processing unit, may exist separately physically for each unit, or two or more units may be integrated in one unit. The above integrated unit may be implemented in the form of hardware or in the form of a hardware plus software functional unit.

[0139] The above integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above software functional unit stored in a storage medium includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to execute some steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drive, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated as: ROM), random access memory (English: Random Access Memory, abbreviated as: RAM), magnetic disk or optical disc and other various media that can store program codes.

[0140] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A dynamic traffic allocation method, characterized in that, Including: Obtain a target order; Adopt a multi-dimensional weight model to adjust the weights of each dimension of the target order according to the market supply and demand relationship; the dimensions of the target order include order amount, account level, urgency, historical completion rate, and special requirements; Input the adjusted weights of each dimension of the target order into a long short-term memory artificial neural network to obtain the predicted value score of the target order; Adopt a dynamic traffic allocation algorithm to match the corresponding target hitman for the target order according to the predicted value score; the dynamic traffic allocation algorithm includes a bi-objective optimization model; the bi-objective optimization model is solved using a multi-objective genetic algorithm and an exposure decay function.

2. The method according to claim 1, wherein The initial weights of each dimension of the target order include: the initial weight of the order amount is 50%, the initial weight of the account level is 20%, the initial weight of the urgency is 15%, the initial weight of the historical completion rate is 10%, and the initial weight of the special requirements is 5%.

3. The method according to claim 1, characterized in that The step of adopting the dynamic traffic allocation algorithm to match the corresponding target hitman for the target order according to the predicted value score includes: Using the multi-objective genetic algorithm and the exposure decay function, based on the bi-objective optimization model, match the corresponding target hitman for the target order according to the predicted value score.

4. The method according to claim 3, wherein The step of using the multi-objective genetic algorithm and the exposure decay function, based on the bi-objective optimization model, to match the corresponding target hitman for the target order according to the predicted value score includes: Input the predicted value score into the bi-objective optimization model; the bi-objective optimization model is as follows: Suppose there exists an order set and a thug set The double-objective optimization model is as follows: Objective 1: Maximize platform revenue Objective 2: Maximize hitman satisfaction Constraints: Among them, V i represents the predicted value score of order O i ; M ij represents the matching degree of order O i and fighter P j ; x ij represents a binary decision variable (1 means allocating order O i to fighter P j , 0 otherwise); S j represents the quantified satisfaction value of fighter P j ; W j represents the maximum working bearing capacity of fighter P j ; workload j represents the current workload of fighter P j ; Use the multi-objective genetic algorithm to solve the bi-objective optimization model to obtain the target hitman corresponding to the target order; the exposure decay function is used to dynamically adjust the search space of the multi-objective genetic algorithm.

5. The method according to claim 1, characterized in that, The target hitman exists in the database in the form of a hitman profile; the hitman profile is determined by a hitman profile modeling system.

6. The method according to claim 1, characterized in that, Before the step of adopting the dynamic traffic allocation algorithm to match the corresponding target hitman for the target order according to the predicted value score, the method further includes: Obtain multi-dimensional data of hitmen; the multi-dimensions include win rate, response speed, device performance, time period preference, customer evaluation, and special skills; Adopt a dynamic weight adjustment algorithm to adjust the weights of the multi-dimensional data; Use an interest graph analyzer to analyze the adjusted multi-dimensional data; Build a model according to the analysis results to obtain a six-dimensional ability evaluation system.

7. The method according to claim 6, characterized in that The step of using the interest graph analyzer to analyze the adjusted multi-dimensional data includes: According to the adjusted multi-dimensional data, use Neo4j to construct a heterogeneous network; where the node types include hitmen, game characters, equipment, and tactics; the edge relationships include hitman-character, character-equipment, and hitman-tactics; Use the GraphSAGE algorithm to generate node vectors to obtain the deep association features of the heterogeneous network; Determine the interest degree of the hitman for each skill according to the skill association intensity formula.

8. The method according to claim 6, characterized in that After obtaining the multi-dimensional data of the hitman, the method further includes: Real-time collecting the operation heat map data of the hitman through the data embedding technology to dynamically update the six-dimensional ability evaluation system; the operation heat map includes the operation event density per minute.

9. The method according to claim 1, characterized in that, After inputting the adjusted weights of each dimension of the target order into the long short-term memory artificial neural network to obtain the predicted value score of the target order, the method further includes: Adjusting the traffic position of the target order in the three-level traffic control system according to the predicted value score; the three-level traffic control system includes the golden traffic position, the normal traffic position, and the downgraded traffic position; the golden traffic position includes the top 5% of the orders, the normal traffic position includes the middle 70% of the orders, and the downgraded traffic position includes the last 25% of the orders.

10. A dynamic traffic allocation system, characterized in that, Including: An order value score evaluation engine for inputting the adjusted weights of each dimension of the target order into the long short-term memory artificial neural network to obtain the predicted value score of the target order; A hitman portrait modeling system for obtaining the multi-dimensional data of the hitman; the multi-dimensions include the winning rate, response speed, device performance, time period preference, customer evaluation, and special skills; Adjusting the weights of the multi-dimensional data by using a dynamic weight adjustment algorithm; Analyzing the adjusted multi-dimensional data by using an interest graph analyzer; modeling according to the analysis results to obtain a six-dimensional ability evaluation system; A dynamic traffic allocation module for using a dynamic traffic allocation algorithm to match a corresponding target hitman for the target order according to the predicted value score; A visual operation console, including a three-dimensional traffic sand table, a drag-and-drop rule configurator, and a real-time data dashboard; the three-dimensional traffic sand table is used to display the order value score interval on the X-axis, the hitman level distribution on the Y-axis, and the time period traffic heat on the Z-axis; the real-time data dashboard is used to monitor the exposure conversion rate, warn about downgraded orders, and analyze the traffic position effectiveness.

Images