Virtual resource allocation method and device, storage medium and electronic equipment

Abstract

The invention discloses a virtual resource allocation method and device, a storage medium and electronic equipment. The method comprises the following steps: acquiring predicted demand quantities respectively corresponding to a plurality of initial objects in a preset time period; screening the plurality of initial objects according to the predicted demands corresponding to the plurality of initial objects, and determining a plurality of target objects; and based on the predicted total transaction amount of the plurality of initial objects in the predetermined time period and the cost-benefit ratio of the virtual resources input in the predetermined time period, determining the target allocation amount of the virtual resources allocated to the plurality of target objects, the cost-benefit ratio being used for representing the influence of the input of the virtual resources on the transaction increment. According to the method and the device, the technical problem of non-ideal virtual resource allocation efficiency in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of machine learning and intelligent decision-making, and specifically, to a virtual resource allocation method, device, storage medium and electronic device. Background Art

[0002] For virtual trading platforms, how to allocate resources to related activities to achieve the optimal solution for the interests of all parties when virtual resources are limited has become an urgent problem that major platforms need to solve. Related technologies statically allocate resources based on historical data or optimize resource allocation with a single goal. Both rely on historical data for resource allocation and lack the prediction of environmental trends, resulting in resource allocation decisions lagging behind environmental changes. In practical applications, faced with large-scale data sets, especially virtual trading environments with extremely large numbers of objects, the computational complexity and resource consumption of optimization algorithms have become the main bottlenecks limiting their application. At present, there are technical problems in related technologies such as unreasonable resource allocation mechanisms and the inability to take into account both the business objectives of virtual trading platforms and the efficient allocation of limited resources.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a virtual resource allocation method, device, storage medium and electronic device to at least solve the technical problem of unsatisfactory virtual resource allocation efficiency existing in the related art.

[0005] According to one aspect of an embodiment of the present application, a virtual resource allocation method is provided, including: obtaining predicted demand quantities corresponding to a plurality of initial objects respectively within a predetermined time period; screening the plurality of initial objects according to the predicted demand quantities corresponding to the plurality of initial objects respectively, and determining a plurality of target objects; based on the predicted total transaction volume of the plurality of initial objects within the predetermined time period, and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, determining a target allocation quantity of the virtual resources to be allocated to the plurality of target objects, wherein the cost-effectiveness ratio is used to represent the impact of the investment of the virtual resources on the transaction increment.

[0006] Optionally, the screening of the multiple initial objects according to the predicted demands respectively corresponding to the multiple initial objects to determine the multiple target objects includes: acquiring image information and text information respectively associated with the multiple initial objects; determining the similarities between the multiple initial objects based on the image information and text information respectively corresponding to the multiple initial objects; and determining the multiple target objects according to the similarities between the multiple initial objects and the predicted demands respectively corresponding to the multiple initial objects.

[0007] Optionally, the multiple initial objects are screened according to the similarities between the multiple initial objects to determine the multiple target objects, including: removing objects whose similarities are greater than a predetermined similarity threshold from the multiple initial objects to obtain multiple first objects; performing clustering processing according to the similarities between the multiple first objects to obtain clustering results; and selecting some objects from each type of object indicated in the clustering results according to the predicted demands respectively corresponding to the multiple initial objects to obtain the multiple target objects.

[0008] Optionally, obtaining the predicted demand quantities respectively corresponding to the multiple initial objects within a predetermined time period includes: obtaining current feature parameters respectively corresponding to the multiple initial objects; based on the current feature parameters, using a long short-term memory network for processing to determine the predicted demand quantities respectively corresponding to the multiple initial objects, wherein the long short-term memory network is trained based on historical feature parameters respectively corresponding to the multiple initial objects.

[0009] Optionally, the target allocation amount of the virtual resources to be allocated to the multiple target objects is determined based on the predicted total transaction volume of the multiple initial objects within the predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, including: determining an objective function based on the predicted total transaction volume and the cost-effectiveness ratio; using a population algorithm to iteratively process multiple individuals to determine fitness values ​​corresponding to the multiple individuals, wherein each of the multiple individuals is used to indicate a candidate allocation amount corresponding to the multiple target objects, and the fitness value is the value of the objective function obtained by using the corresponding individual; until a target individual with an optimal solution that balances the predicted total transaction volume and the cost-effectiveness ratio exists among the multiple individuals, and the target allocation amounts corresponding to the multiple target objects are determined based on the candidate allocation amounts corresponding to the multiple target objects indicated by the target individual.

[0010] Optionally, determining the objective function based on the predicted total transaction volume and the cost-effectiveness ratio includes: generating a first sub-function based on the predicted total transaction volume; generating a second sub-function based on the cost-effectiveness ratio; and generating the objective function based on the first sub-function and the second sub-function.

[0011] Optionally, the method also includes: for the first sub-function, determining a second object among the multiple target objects according to a first predetermined condition; obtaining a first predetermined individual based on performing allocation amount increase processing on the second object; for the second sub-function, determining a third object among the multiple target objects according to a second predetermined condition; obtaining a second predetermined individual based on performing allocation amount increase processing on the third object; obtaining the multiple individuals based on the first predetermined individual and the second predetermined individual.

[0012] According to another aspect of an embodiment of the present application, a virtual resource allocation device is provided, including: a demand acquisition module, used to acquire predicted demand quantities corresponding to multiple initial objects respectively within a predetermined time period; a target object determination module, used to screen the multiple initial objects according to the predicted demand quantities corresponding to the multiple initial objects respectively, and determine multiple target objects; an allocation quantity determination module, used to determine a target allocation quantity of the virtual resources allocated to the multiple target objects based on the predicted total transaction volume of the multiple initial objects within the predetermined time period, and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, wherein the cost-effectiveness ratio is used to represent the impact of the investment of the virtual resources on the transaction increment.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing any one of the virtual resource allocation methods.

[0014] According to another aspect of an embodiment of the present application, an electronic device is provided, comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the virtual resource allocation methods described.

[0015] In the embodiment of the present application, the predicted demand corresponding to multiple initial objects in a predetermined time period is obtained; the multiple initial objects are screened according to the predicted demand corresponding to the multiple initial objects to determine multiple target objects; based on the predicted total transaction volume of the multiple initial objects in the predetermined time period and the cost-effectiveness ratio of the virtual resources invested in the predetermined time period, the target allocation amount of the virtual resources allocated to the multiple target objects is determined, wherein the cost-effectiveness ratio is used to represent the impact of the investment of the virtual resources on the transaction increment. The purpose of fine and intelligent resource allocation is achieved, the technical effect of improving the efficiency of virtual resource allocation is achieved, and the technical problem of unsatisfactory virtual resource allocation efficiency existing in the related technology is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 is a flowchart of an optional virtual resource allocation method provided according to an embodiment of the present application;

[0018] Figure 2is a schematic diagram of an optional virtual resource allocation method provided according to an embodiment of the present application;

[0019] Figure 3 It is a schematic diagram of an optional virtual resource allocation device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution 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 only 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 should fall within the scope of protection of the present application.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0023] The Long Short-Term Memory (LSTM) model is a special recurrent neural network architecture. The LSTM model can effectively manage information flow, so that the network can remember important information and forget irrelevant information when processing sequence data, thereby making more accurate predictions of input data. The LSTM model can be used to predict the demand for an object. It analyzes the historical demand data of the object, user behavior and other multi-dimensional time series information, captures the long-term trends and short-term changes, and provides more accurate demand forecast results.

[0024] CLIP is a deep learning model that learns the association between images and text by pre-training on a large number of image-text pairs, maps images and text into the same vector space, and thus calculates the similarity between images and text.

[0025] The K-means algorithm iteratively assigns data points to K clusters to find the inherent structure and pattern of the data. The K-means algorithm is highly efficient and capable of processing large amounts of data. It can quickly process massive amounts of information, reduce the consumption of computing resources, and ensure the accuracy of clustering results.

[0026] NSGA-Ⅱ (Non-dominated Sorting Genetic Algorithm II) is an efficient multi-objective genetic algorithm used to solve multi-objective optimization problems. NSGA-Ⅱ can handle multiple conflicting objectives at the same time and find a trade-off solution that satisfies all objectives, that is, the Pareto optimal solution set.

[0027] According to an embodiment of the present application, a method embodiment of virtual resource allocation is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] Figure 1 is a flowchart of an optional virtual resource allocation method provided according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0029] Step S102, obtaining the predicted demand quantities corresponding to the multiple initial objects in a predetermined time period;

[0030] It can be understood that in the data preparation stage, based on the historical data of the object and the environmental trend analysis, the predicted demand corresponding to multiple initial objects in the predetermined time period is obtained, which ensures the foresight and accuracy of the virtual resource allocation decision.

[0031] In an optional embodiment, the predicted demand quantities corresponding to the multiple initial objects in a predetermined time period are obtained, including: obtaining current feature parameters corresponding to the multiple initial objects respectively; based on the current feature parameters, using a long short-term memory network for processing to determine the predicted demand quantities corresponding to the multiple initial objects respectively, wherein the long short-term memory network is trained based on the historical feature parameters corresponding to the multiple initial objects respectively.

[0032] It can be understood that, first, the current feature parameters of each initial object are collected. The feature parameters cover multi-dimensional data such as the historical demand, price fluctuations, attribute information, and user evaluation of the object. The pre-trained long short-term memory network (LSTM) model is used to deeply analyze and process the current feature parameters. The LSTM model can effectively capture the time series dependencies in the data, and accurately predict the predicted demand corresponding to each initial object based on the learning of historical feature parameters. The introduction of the LSTM model significantly enhances the dynamic adaptability and predictability of virtual resource allocation decisions, ensuring that in a complex and changing environment, the virtual resource allocation plan can be closer to actual needs, thereby improving resource utilization efficiency.

[0033] Step S104, screening the multiple initial objects according to the predicted demand quantities respectively corresponding to the multiple initial objects, and determining multiple target objects;

[0034] It can be understood that based on the predicted demand corresponding to each of the multiple initial objects, the multiple initial objects are screened to determine the final target object. The predicted demand is used to set a threshold or adopt a specific sorting algorithm to eliminate objects with low predicted demand, ensuring that resources can be concentrated on objects with higher demand potential. By focusing on high-demand predicted objects, the virtual trading platform can accurately invest limited resources to drive higher transaction volume and user engagement, thereby maximizing business goals.

[0035] In an optional embodiment, multiple initial objects are screened according to predicted demands respectively corresponding to the multiple initial objects to determine multiple target objects, including: acquiring image information and text information respectively associated with the multiple initial objects; determining similarities between the multiple initial objects based on the image information and text information respectively corresponding to the multiple initial objects; determining multiple target objects according to the similarities between the multiple initial objects and the predicted demands respectively corresponding to the multiple initial objects.

[0036] It can be understood that the image information and text information associated with each initial object within a predetermined time period are collected, and based on the image information and text information corresponding to multiple initial objects, objects with similar features are identified using relevant algorithms, the similarity between multiple initial objects is determined, and multiple target objects are determined based on the similarity and the predicted demand corresponding to the multiple initial objects. By calculating the similarity of the objects and the predicted demand, a large number of similar objects are effectively filtered out, the variable scale in the optimization process is reduced, and the solution time of the optimization algorithm is greatly shortened.

[0037] Optionally, the above-mentioned similarity calculation algorithm can be multiple, for example, the Contrastive Language-Image Pre-training model (CLIP). In the virtual trading scenario, the CLIP algorithm can efficiently analyze the image and description text of the object, accurately calculate the visual and semantic similarity between objects, and provide data support for the screening and classification of multiple objects. The core advantage of the CLIP algorithm is that it can process cross-modal data, that is, it can process images and texts at the same time. In the field of virtual trading, objects are basically defined by pictures and description texts. By using the CLIP algorithm, the similarity of objects can be evaluated more comprehensively.

[0038] In an optional embodiment, multiple initial objects are screened according to similarities between the multiple initial objects to determine multiple target objects, including: removing objects whose similarities are greater than a predetermined similarity threshold from the multiple initial objects to obtain multiple first objects; performing clustering processing according to the similarities between the multiple first objects to obtain clustering results; and selecting some objects from each type of object indicated in the clustering results according to the predicted demands respectively corresponding to the multiple initial objects to obtain multiple target objects.

[0039] It can be understood that by calculating the similarity matrix of image and text information, objects with similarities greater than a predetermined similarity threshold are identified and eliminated, and multiple first objects, i.e., a set of objects after preliminary screening, are obtained. Then, the multiple first objects are clustered to form a clustering result. Finally, based on the predicted demand corresponding to the multiple initial objects, some objects are selected from each type of object indicated in the clustering result as target objects to enter the final resource allocation optimization process. Through multi-stage screening and clustering, effective optimization and segmentation of multiple objects are achieved, and the accuracy and efficiency of resource allocation are improved. Removing objects with high similarity avoids repeated resource allocation for homogeneous objects and reduces resource waste.

[0040] Optionally, the above clustering methods can be multiple, for example, using the K-means algorithm. In the scenario of resource allocation for virtual trading activities, the K-means algorithm needs to determine the number of clusters K, and the K value is determined based on business needs and data analysis. The algorithm randomly selects K data points as the initial cluster centers and assigns all other data points to the nearest cluster center. Then, the algorithm recalculates the center of each cluster and assigns the data points again. This process is iterated repeatedly until the cluster center no longer changes significantly or reaches a predetermined number of iterations. Using the K-means algorithm for clustering, the objects that have been initially screened can be classified according to their demand forecasts, categories and other characteristics. Through K-means clustering, it can be ensured that the comprehensive characteristics of the object are taken into account when allocating resources, so as to achieve a more refined resource allocation strategy.

[0041] Step S106, based on the predicted total transaction volume of multiple initial objects within a predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, determine the target allocation amount of virtual resources to multiple target objects, wherein the cost-effectiveness ratio is used to represent the impact of the investment of virtual resources on the transaction increment.

[0042] It can be understood that the model is used to predict the total transaction volume of multiple initial objects within a predetermined time period. The total transaction volume is a quantitative indicator that reflects the potential of the virtual trading platform. Based on the predicted total transaction volume and the cost-effectiveness ratio of the virtual resources invested within the predetermined time, the target allocation of virtual resources to multiple target objects is determined. The cost-effectiveness ratio is the ratio of the impact of the investment of virtual resources on the transaction increment. This indicator comprehensively considers the input and output of resources and is a key parameter for evaluating the effectiveness of resource allocation strategies. Combining the multi-objective optimization of the predicted total transaction volume and the cost-effectiveness ratio, the virtual trading platform can make more scientific and effective resource allocation under limited resources.

[0043] Optionally, the multi-objective optimization model can be multiple. For example, the predicted total transaction volume (GMV, Gross Merchandise Volume) and cost-effectiveness ratio are used as objective functions to determine the target allocation amount and establish a multi-objective optimization model, in which resource limits and object quota limits are used as constraints. Through the multi-objective optimization solution of the NSGAⅡ algorithm, a set of virtual resource allocation schemes can be obtained, which finds a balance between GMV maximization and cost-effectiveness optimization, namely the so-called Pareto optimal solution. Finally, the system will select one or more optimal solutions as the target allocation amount and determine the specific resource allocation for each object.

[0044] A multi-objective optimization model of GMV (maxGMV, GMV maximization) and cost-effectiveness ratio (maxfee_rate, cost-effectiveness optimization) is established. The predicted demand is substituted into the GMV and cost-effectiveness ratio calculation formula to obtain the objective function. At the same time, constraints are constructed to meet the actual business restrictions, and the following optimization model is obtained:

[0045] maxGMV=f1(x_sale i ,w i ,price i )

[0046] maxfee_rate = f2(x_sale i ,w i ,price i )

[0047]

[0048] w i ≥0

[0049] where w i is a decision variable, representing the virtual resource quota of each object activity, i represents different object numbers; x_sale i represents the demand for object i, price i represents the price of object i, f1 and f2 represent the function expressions of GMV and cost-effectiveness, st represents satisfying the constraint conditions, n represents a total of n objects, g represents satisfying the overall virtual resource limit, r and q are fixed values ​​that the quotas of some objects must be greater than or less than, and s, k and p all represent the numbers of the object arrangement order.

[0050] In an optional embodiment, based on the predicted total transaction volume of multiple initial objects within a predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, the target allocation amount of virtual resources to multiple target objects is determined, including: determining the objective function based on the predicted total transaction volume and the cost-effectiveness ratio; using a population algorithm to iteratively process multiple individuals to determine the fitness values ​​corresponding to the multiple individuals, wherein each individual among the multiple individuals is used to indicate the candidate allocation amount corresponding to the multiple target objects, and the fitness value is the value of the objective function obtained by using the corresponding individual; until there is a target individual among the multiple individuals that has the optimal solution that balances the predicted total transaction volume and the cost-effectiveness ratio, the target allocation amounts corresponding to the multiple target objects are determined based on the candidate allocation amounts corresponding to the multiple target objects indicated by the target individual.

[0051] It can be understood that the total predicted transaction volume is obtained based on the prediction model, and the objective function is established in combination with the cost-effectiveness of the virtual resources invested in the time period. The population algorithm is used to generate an initial population containing multiple individuals, and the iterative process begins. Each individual is actually a candidate allocation amount, indicating how many virtual resources each target object will obtain. The fitness value of each individual is calculated, and the fitness value reflects the performance in pursuing the two goals of maximizing the total transaction volume and maintaining a high cost-effectiveness. The iterative process continues until the target individual that can balance the optimal solution of the predicted total transaction volume and the cost-effectiveness is found. According to the candidate allocation amount indicated by the target individual, the target allocation amount corresponding to multiple target objects is determined. The resource allocation strategy based on prediction and multi-objective optimization can more accurately predict the demand for the object, and reasonably allocate limited virtual resources accordingly to achieve the maximum transaction volume and the optimal cost-effectiveness.

[0052] In an optional embodiment, an objective function is determined based on the predicted total transaction volume and cost-effectiveness ratio, including: generating a first sub-function based on the predicted total transaction volume; generating a second sub-function based on the cost-effectiveness ratio; and generating an objective function based on the first sub-function and the second sub-function.

[0053] It can be understood that the first sub-function is generated based on the predicted total transaction volume. The second sub-function is generated based on the cost-effectiveness ratio, and the objective function is generated based on the first sub-function and the second sub-function. The function is used to find the best balance point between the total transaction volume and the cost-effectiveness ratio, which is the so-called Pareto optimal solution. By constructing a multi-objective function, while ensuring the growth of transaction volume, effective control of resources is achieved, so as to achieve the best effect under limited virtual resources.

[0054] In an optional embodiment, the method also includes: for the first sub-function, determining the second object among multiple target objects according to the first predetermined condition; obtaining the first predetermined individual based on performing allocation amount increase processing on the second object; for the second sub-function, determining the third object among multiple target objects according to the second predetermined condition; obtaining the second predetermined individual based on performing allocation amount increase processing on the third object; obtaining multiple individuals based on the first predetermined individual and the second predetermined individual.

[0055] It can be understood that for the first sub-function, according to the first predetermined condition, such as historical demand, environmental trend forecast, etc., the second object is selected from multiple target objects, and the second object is processed to increase the allocation amount to obtain the first predetermined individual; for the second sub-function, according to the second predetermined condition, the third object is determined from multiple target objects; the third object is processed to increase the allocation amount to obtain the second predetermined individual; based on the first predetermined individual and the second predetermined individual, multiple individuals are obtained. Determining individuals according to predetermined conditions ensures that the objects can maintain or improve benefits under resource allocation and avoid inefficient use of resources.

[0056] Through the above step S102, the predicted demand corresponding to multiple initial objects in the predetermined time period is obtained; in step S104, multiple initial objects are screened according to the predicted demand corresponding to the multiple initial objects, and multiple target objects are determined; in step S106, based on the predicted total transaction volume of multiple initial objects in the predetermined time period and the cost-effectiveness ratio of virtual resources invested in the predetermined time period, the target allocation amount of virtual resources allocated to multiple target objects is determined, wherein the cost-effectiveness ratio is used to represent the impact of virtual resource investment on transaction increment. The purpose of fine and intelligent resource allocation can be achieved, and the technical effect of improving the efficiency of virtual resource allocation is achieved, thereby solving the technical problem of unsatisfactory virtual resource allocation efficiency existing in related technologies.

[0057] Based on the above embodiments and optional embodiments, the present application proposes an optional implementation method. Based on the actual difficulties of virtual trading business, a machine learning model LSTM for demand prediction and a multi-objective virtual resource allocation optimization model are established to reasonably allocate activity resources while satisfying actual resource constraints, so as to maximize benefits and assist business staff in making decisions. Figure 2is a schematic diagram of an optional virtual resource allocation method provided according to an embodiment of the present application, such as Figure 2 As shown, the algorithm flow steps are shown.

[0058] Assuming that the virtual resources invested in an activity on a virtual trading platform is a yuan, there are m objects participating in the activity, and each object has n indicators reflecting its own historical situation, such as the lowest price in history, the number of visitors in the past 30 days, the number of clicks in the past 30 days, the number of transactions in the past 30 days, etc., a decision matrix reflecting the indicator information of all objects is constructed, and the decision matrix is ​​standardized. The demand prediction model LSTM is established according to the standardized matrix, and the GMV (f1) and cost-effectiveness (f2) multi-objective optimization models are established. The virtual resource allocation is optimized and solved by improving the NSGAⅡ optimization algorithm, and the virtual resource quota of each object is iteratively obtained.

[0059] Step S1, training the demand prediction model LSTM. First, all object information is processed with features, such as historical demand, third-level categories, brands, prices, etc., and input into the machine learning model for training. After parameter tuning, the feature information of all objects before the event starts is updated to obtain the predicted demand for the event;

[0060] The pre-trained long short-term memory network (LSTM) model is used to deeply analyze and process the current feature parameters. The LSTM model can effectively capture the time series dependencies in the data and accurately predict the forecast demand corresponding to each initial object based on the learning of historical feature parameters. The introduction of the LSTM model significantly enhances the dynamic adaptability and predictability of virtual resource allocation decisions, ensuring that in a complex and changing environment, the virtual resource allocation plan can be closer to actual needs, thereby improving resource utilization efficiency.

[0061] Step S2, construct a multi-objective optimization model. The predicted demand of all active objects is obtained from the first step. Based on this, it is substituted into the GMV and cost-effectiveness calculation formula to obtain the objective function. At the same time, constraints are constructed to meet the actual business restrictions, and the following optimization model is obtained:

[0062] maxGMV=f1(x_sale i ,w i ,price i )

[0063] maxfee_rate = f2(x_sale i ,w i ,price i )

[0064]

[0065] w i≥0

[0066] where w i is a decision variable, representing the virtual resource quota of each object activity, i represents different object numbers; x_sale i represents the demand for object i, price i represents the price of object i, f1 and f2 represent the function expressions of GMV and cost-effectiveness, st represents satisfying the constraint conditions, n represents a total of n objects, g represents satisfying the overall virtual resource limit, r and q are fixed values ​​that the quotas of some objects must be greater than or less than, and s, k and p all represent the numbers of the object arrangement order.

[0067] Step S3, optimization of variable space based on CLIP and k-means. First, use the CLIP algorithm to calculate the text and image similarity of all objects, filter out objects with high similarity, and then introduce the K-means algorithm to cluster the remaining objects, and select the appropriate number of objects in each category based on the predicted demand;

[0068] In virtual trading scenarios, the CLIP algorithm can efficiently analyze the images and description texts of objects, accurately calculate the visual and semantic similarities between objects, and provide data support for the screening and classification of multiple objects. The core advantage of the CLIP algorithm is that it can process cross-modal data, that is, it can process images and texts at the same time. In the field of virtual trading, objects are basically defined by images and description texts. By using the CLIP algorithm, the similarity of objects can be evaluated more comprehensively.

[0069] Clustering is performed using algorithms such as K-means. In the scenario of resource allocation for virtual trading activities, the K-means algorithm needs to determine the number of clusters K, and the K value is determined based on business needs and data analysis. The algorithm randomly selects K data points as the initial cluster centers and assigns all other data points to the nearest cluster center. The algorithm then recalculates the center of each cluster and assigns data points again. This process is repeated until the cluster center no longer changes significantly or the predetermined number of iterations is reached. Clustering using the K-means algorithm can classify the objects that have been initially screened according to their demand forecasts, categories, and other characteristics. Through K-means clustering, it can be ensured that the comprehensive characteristics of the object are taken into account when allocating resources, and a more refined resource allocation strategy can be implemented.

[0070] Step S4, the greedy strategy generates the initial population. Set the size of the population to N, and consider the two goals of actual GMV and cost-effectiveness. Give priority to evenly allocating g to objects with the largest GMV contribution and objects with small resource usage, generate 2 individuals, and allocate total virtual resources g according to the GMV contribution ratio and resource usage ratio, respectively, to obtain 2 individuals, and the remaining N-4 individuals are randomly generated to form the original population P;

[0071] The greedy strategy is to take the best or optimal choice in the current state in each step, hoping to lead to the best or optimal result in the world. In generating the initial population, the greedy strategy is used to quickly build the starting point of the algorithm, providing a good start for the subsequent search or optimization process.

[0072] Step S5, population replication. Calculate the objective function of the individuals in the population, and use the roulette calculation method to select a new population Q of size N from the initial population;

[0073] The roulette calculation method determines the probability of an individual being selected based on the relative size of the individual's fitness value. Each individual is assigned a weight proportional to its fitness value. The higher the fitness value of the individual, the greater the probability of being selected.

[0074] Step S6, selection, crossover, and mutation. Generate a new generation of population through selection, crossover, and mutation;

[0075] Step S7: rounding back. Considering that the amount of virtual resources actually released needs to be an integer, the w of all individuals is rounded back. i Round up and round down, and recalculate the objective function of the individuals in the population;

[0076] Step S8, fast non-dominated sorting. After merging the offspring and the parent generation, select individuals with high fitness, then calculate the crowding degree, select individuals with high crowding degree to the next generation, and together form a new population of size N;

[0077] Step S9, the iteration stops. Unless the optimal Pareto front is found, the iteration stops, otherwise, steps S6, S7, and S8 are repeated.

[0078] The virtual resource quota of each object is obtained based on the final optimization for staff's decision-making reference.

[0079] The optional implementation method mentioned above can at least achieve the following effects: by introducing the prediction model LSTM, NSGAⅡ algorithm, CLIP algorithm and K-means algorithm, uncertainty is converted into certainty, and the object of solution is filtered and calculated, so that the number of variables to be solved is greatly reduced, and the model variable, i.e., virtual resource quota, is efficiently searched and solved. At the same time, the initial individuals are greedily generated according to the actual business goals when the population is initialized, which can greatly improve the solution convergence efficiency in the actual application process and improve the optimization solution speed; the realistic constraints of the actual virtual resource allocation, the rounding needs of the actual activity virtual resource allocation and multiple business goals are taken into account, so that each goal reaches the Pareto optimality and the target benefit of limited resources is maximized. The two methods of rounding up and rounding down the solution generated by the population are combined to maximize the validity of the rounding solution, and the result is reasonable and scientific.

[0080] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0081] In this embodiment, a virtual resource allocation device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0082] According to an embodiment of the present application, a device embodiment for implementing a virtual resource allocation method is also provided. Figure 3 is a schematic diagram of a virtual resource allocation device according to an embodiment of the present application, such as Figure 3 As shown, the above-mentioned virtual resource allocation device includes a demand acquisition module 302, a target object determination module 304, and an allocation determination module 306. The device is described below.

[0083] The demand acquisition module 302 is used to acquire the predicted demand corresponding to the multiple initial objects in a predetermined time period;

[0084] The target object determination module 304 is connected to the demand acquisition module 302 and is used to screen the multiple initial objects according to the predicted demand quantities respectively corresponding to the multiple initial objects to determine the multiple target objects;

[0085] The allocation amount determination module 306 is connected to the target object determination module 304, and is used to determine the target allocation amount of virtual resources to multiple target objects based on the predicted total transaction volume of multiple initial objects within a predetermined time period, and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, wherein the cost-effectiveness ratio is used to represent the impact of the investment of virtual resources on the transaction increment.

[0086] In a virtual resource allocation device provided by an embodiment of the present application, a demand acquisition module 302 is provided to acquire the predicted demand corresponding to multiple initial objects in a predetermined time period; a target object determination module 304 is connected to the demand acquisition module 302, and is used to screen multiple initial objects according to the predicted demand corresponding to the multiple initial objects, and determine multiple target objects; an allocation determination module 306 is connected to the target object determination module 304, and is used to determine the target allocation of virtual resources to multiple target objects based on the predicted total transaction volume of multiple initial objects in a predetermined time period, and the cost-effectiveness ratio of virtual resources invested in the predetermined time period, wherein the cost-effectiveness ratio is used to represent the impact of the investment of virtual resources on the transaction increment. The purpose of fine and intelligent resource allocation is achieved, and the technical effect of improving the efficiency of virtual resource allocation is achieved, thereby solving the technical problem of unsatisfactory virtual resource allocation efficiency existing in the related art.

[0087] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0088] It should be noted that the above-mentioned module for obtaining demand quantity 302, module for determining target object 304, and module for determining allocation quantity 306 correspond to steps S102 to S106 in the embodiment, and the examples and application scenarios implemented by the above-mentioned modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment. It should be noted that the above-mentioned modules as part of the device can be run in a computer terminal.

[0089] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0090] The virtual resource allocation device may also include a processor and a memory. The demand acquisition module 302, the target object determination module 304, the allocation determination module 306, etc. are all stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions.

[0091] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0092] An embodiment of the present application provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a virtual resource allocation method is implemented.

[0093] The embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program stored in the memory and can be run on the processor. When the processor executes the program, the following steps are implemented: obtaining the predicted demand corresponding to multiple initial objects in a predetermined time period; screening the multiple initial objects according to the predicted demand corresponding to the multiple initial objects, and determining multiple target objects; based on the predicted total transaction volume of the multiple initial objects in the predetermined time period, and the cost-effectiveness ratio of the virtual resources invested in the predetermined time, determining the target allocation amount of virtual resources to the multiple target objects, wherein the cost-effectiveness ratio is used to represent the impact of the investment of virtual resources on the transaction increment. The device in this article can be a server, a PC, etc.

[0094] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining predicted demand quantities corresponding to multiple initial objects within a predetermined time period; screening the multiple initial objects according to the predicted demand quantities corresponding to the multiple initial objects, and determining multiple target objects; based on the predicted total transaction volume of the multiple initial objects within the predetermined time period, and the cost-effectiveness of the virtual resources invested within the predetermined time, determining the target allocation amount of virtual resources to the multiple target objects, wherein the cost-effectiveness is used to represent the impact of the investment of virtual resources on the transaction increment.

[0095] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0096] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0099] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0100] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0101] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0102] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0103] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0104] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A virtual resource allocation method, characterized in that: include: Obtaining the predicted demand quantities corresponding to the multiple initial objects in a predetermined time period; Screening the multiple initial objects according to the predicted demands respectively corresponding to the multiple initial objects to determine multiple target objects; Based on the predicted total transaction volume of the multiple initial objects within the predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, the target allocation amount of the virtual resources to the multiple target objects is determined, wherein the cost-effectiveness ratio is used to represent the impact of the investment of the virtual resources on the transaction increment.

2. The method according to claim 1, characterized in that The screening of the multiple initial objects according to the predicted demands respectively corresponding to the multiple initial objects to determine the multiple target objects includes: Acquire image information and text information respectively associated with the plurality of initial objects; Determining similarities between the multiple initial objects based on the image information and text information respectively corresponding to the multiple initial objects; The multiple target objects are determined according to the similarities between the multiple initial objects and the predicted demands respectively corresponding to the multiple initial objects.

3. The method according to claim 2, characterized in that The screening of the multiple initial objects according to the similarities between the multiple initial objects to determine the multiple target objects includes: Removing objects whose similarity is greater than a predetermined similarity threshold from the multiple initial objects to obtain multiple first objects; Performing clustering processing according to the similarities between the multiple first objects to obtain a clustering result; According to the predicted demands respectively corresponding to the multiple initial objects, some objects are selected from each type of objects indicated in the clustering result to obtain the multiple target objects.

4. The method according to claim 1, characterized in that: The obtaining of the predicted demand quantities respectively corresponding to the multiple initial objects within the predetermined time period includes: Obtaining current feature parameters corresponding to the multiple initial objects respectively; Based on the current characteristic parameters, a long short-term memory network is used for processing to determine the predicted demand quantities corresponding to the multiple initial objects respectively, wherein the long short-term memory network is trained based on the historical characteristic parameters corresponding to the multiple initial objects respectively.

5. The method according to any one of claims 1 to 4, characterized in that: The step of determining the target allocation amount of the virtual resources to be allocated to the multiple target objects based on the predicted total transaction amount of the multiple initial objects within the predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period includes: Determining an objective function based on the predicted total transaction volume and the cost-effectiveness ratio; Adopting a population algorithm, iteratively processing a plurality of individuals, and determining fitness values ​​corresponding to the plurality of individuals, wherein each of the plurality of individuals is used to indicate a candidate allocation amount corresponding to the plurality of target objects, and the fitness value is a value of the objective function obtained by adopting the corresponding individual; Until a target individual with the optimal solution of balancing the predicted total transaction volume and the cost-effectiveness ratio exists among the multiple individuals, the target allocation amounts corresponding to the multiple target objects respectively are determined based on the candidate allocation amounts corresponding to the multiple target objects respectively indicated by the target individual.

6. The method according to claim 5, characterized in that The determining of the objective function based on the predicted total transaction volume and the cost-effectiveness ratio includes: Based on the predicted total transaction volume, generating a first sub-function; Based on the cost-effectiveness ratio, generating a second sub-function; The objective function is generated based on the first sub-function and the second sub-function.

7. The method according to claim 6, characterized in that The method further comprises: For the first sub-function, determining a second object from among the plurality of target objects according to a first predetermined condition; Based on performing allocation amount increasing processing on the second object, a first predetermined individual is obtained; For the second sub-function, determining a third object from among the plurality of target objects according to a second predetermined condition; Based on performing allocation amount increasing processing on the third object, a second predetermined individual is obtained; The plurality of individuals are obtained based on the first predetermined individual and the second predetermined individual.

8. A virtual resource allocation device, characterized in that: include: A demand acquisition module is used to acquire the predicted demand corresponding to a plurality of initial objects in a predetermined time period; A target object determination module, configured to screen the multiple initial objects according to the predicted demand quantities respectively corresponding to the multiple initial objects, and determine multiple target objects; The allocation amount determination module is used to determine the target allocation amount of the virtual resources allocated to the multiple target objects based on the predicted total transaction volume of the multiple initial objects within the predetermined time period and the cost-effectiveness ratio of the virtual resources invested within the predetermined time period, wherein the cost-effectiveness ratio is used to represent the impact of the investment of the virtual resources on the transaction increment.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the virtual resource allocation method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the virtual resource allocation method described in any one of claims 1 to 7.

Images