Pension demand prediction method and device, equipment and storage medium
By collecting market feedback information and using the improved Gray Wolf algorithm, the long-term and short-term memory network model is optimized, and the problems of high complexity of super parameter search and insufficient balance ability of Gray Wolf algorithm in the existing technology are solved, and efficient customer pension needs are achieved Improved forecasting and marketing conversion rates.
Patent Information
- Application Number
- CN202411644333.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-18
- Publication Date
- 2025-02-25
AI Technical Summary
When facing high-dimensional and multivariate hyperparameter search space, the existing technology has high computational complexity and slow convergence speed, which is easy to fall into local optimality. The gray wolf algorithm has a balanced and multi-modal problem in global exploration capabilities and local development. There are shortcomings in the ability to cope with dynamic changes in search space.
The market feedback score is determined by collecting market feedback information, the preset hyperparameter search space generates gray wolf individuals, and evaluates them based on the market feedback score and fitness function. The target gray wolf individual is determined using memory mechanisms and individual variation mechanisms, and a second evaluation is conducted, and the long-term and short-term memory network model is finally optimized to predict customer pension needs.
The memory mechanism and individual variation mechanism enhance search diversity and avoid local optimal traps. The improved Gray Wolf algorithm optimizes the long-term and short memory network model, improves prediction accuracy, generalization ability and training speed, and significantly improves marketing conversion rate.
Smart Images
Figure CN119514795A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a pension demand forecasting method, device, equipment and storage medium. Background Art
[0002] Among traditional optimization methods, grid search and random search are common techniques for selecting neural network hyperparameters, but these methods have high computational complexity, slow convergence speed, and are prone to falling into local optimality when faced with high-dimensional, multi-variable hyperparameter search spaces. In addition, the existing gray wolf algorithm still has problems such as insufficient balance between global exploration and local development, and lack of ability to cope with multi-peak problems and dynamic changes in search space.
[0003] From the above, it can be seen that how to achieve efficient optimization of LSTM (Long Short-Term Memory) hyperparameters to accurately identify customers' pension needs is an urgent problem that needs to be solved. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a pension demand prediction method, device, equipment and storage medium, which can achieve efficient optimization of LSTM hyperparameters to accurately identify customer pension needs. The specific scheme is as follows:
[0005] In a first aspect, the present application provides a pension demand forecasting method, comprising:
[0006] Determine a corresponding market feedback score through the collected market feedback information; the market feedback information is information related to customer behavior generated in pension marketing activities;
[0007] Generating each gray wolf individual through a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model;
[0008] Determine a first target gray wolf individual and a second target gray wolf individual in each of the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and perform a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result;
[0009] The long short-term memory network model is optimized based on the first evaluation result and the second evaluation result, and the customer's pension demand is predicted through the optimization result.
[0010] Optionally, determining a corresponding market feedback score through the collected market feedback information includes:
[0011] Market feedback information is obtained by collecting information related to customer behavior generated in pension marketing activities, and the market feedback information is quantified to obtain a market feedback score.
[0012] Optionally, generating each gray wolf individual by using a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result, includes:
[0013] Generating each gray wolf individual through a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function;
[0014] The fitness of each gray wolf individual is ranked based on the first evaluation result, and the gray wolf individuals that meet the preset interval conditions are eliminated.
[0015] Optionally, generating each gray wolf individual by using a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result, includes:
[0016] By presetting the hyperparameter search space, gray wolf individuals representing different hyperparameter configurations of the long short-term memory network model are generated, and a first evaluation is performed on the fitness of each gray wolf individual based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result.
[0017] Optionally, the first evaluation of the fitness of each of the gray wolf individuals based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result includes:
[0018] If the market feedback meets the preset key conditions, the market feedback weight corresponding to the market feedback score is increased;
[0019] If the market feedback does not meet the preset key conditions, the market feedback weight corresponding to the market feedback score is reduced.
[0020] Optionally, the determining a first target gray wolf individual and a second target gray wolf individual among the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and performing a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result, includes:
[0021] A first target gray wolf individual that meets a preset optimal condition is obtained from among the gray wolf individuals by using a memory mechanism in the gray wolf algorithm, and a second target gray wolf individual that varies from among the gray wolf individuals is obtained based on an individual variation mechanism in the gray wolf algorithm;
[0022] A summed gray wolf individual is obtained by summing the first target gray wolf individual and the second target gray wolf individual, and the summed gray wolf individual is evaluated based on the market feedback score and a preset fitness function to obtain a second evaluation result.
[0023] Optionally, after optimizing the long short-term memory network model based on the first evaluation result and the second evaluation result, the method further includes:
[0024] The process of optimizing the long short-term memory network model is divided into several feedback cycles, and the long short-term memory network model in the next feedback cycle is optimized through the feedback record of the current feedback cycle.
[0025] In a second aspect, the present application provides a pension demand prediction device, comprising:
[0026] A score collection module, used to determine a corresponding market feedback score through the collected market feedback information; the market feedback information is information related to customer behavior generated in pension marketing activities;
[0027] A first evaluation module is used to generate each gray wolf individual through a preset hyperparameter search space, and perform a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model;
[0028] A second evaluation module is used to determine a first target gray wolf individual and a second target gray wolf individual among the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and to perform a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result;
[0029] A demand prediction module is used to optimize the long short-term memory network model based on the first evaluation result and the second evaluation result, and predict the customer's pension demand through the optimization result.
[0030] In a third aspect, the present application provides an electronic device, including:
[0031] Memory, used to store computer programs;
[0032] A processor is used to execute the computer program to implement the aforementioned pension demand forecasting method.
[0033] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned pension demand forecasting method when executed by a processor.
[0034] The present application determines a corresponding market feedback score through the collected market feedback information; the market feedback information is information related to customer behavior generated in pension marketing activities; each gray wolf individual is generated through a preset hyperparameter search space, and a first evaluation is performed on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model; a first target gray wolf individual and a second target gray wolf individual in each gray wolf individual are determined by using a memory mechanism and an individual variation mechanism respectively, and a second evaluation is performed on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result; the long short-term memory network model is optimized based on the first evaluation result and the second evaluation result, and customer pension needs are predicted through the optimization result.
[0035] As can be seen from the above, the present application performs a first evaluation on the fitness of each gray wolf individual generated through the market feedback score determined by the market feedback information and the preset fitness function, and uses the memory mechanism and the individual variation mechanism to obtain the first target gray wolf individual and the second target gray wolf individual of each gray wolf individual, and then performs a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and the preset fitness function. The long-short memory network model is optimized through the two evaluation results of different gray wolf individuals to achieve the prediction of customer pension needs based on the optimization results. In this way, the search diversity is enhanced through the memory mechanism and the individual variation mechanism, and the local optimal trap is effectively avoided; the long-short memory network model is optimized by using the improved gray wolf algorithm, so that the long-short memory network model shows higher prediction accuracy, better generalization ability and faster training speed, so as to accurately identify customer needs and significantly improve marketing conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0037] Figure 1 A flowchart of a pension demand forecasting method disclosed in this application;
[0038] Figure 2 A flowchart of a specific pension demand forecasting method disclosed in this application;
[0039] Figure 3 This is a schematic diagram of the structure of a pension demand prediction device disclosed in this application;
[0040] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION
[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0042] At present, grid search and random search are common techniques for selecting neural network hyperparameters, but these methods have high computational complexity and slow convergence speed when facing high-dimensional, multi-variable hyperparameter search spaces, and are prone to falling into local optimality. To this end, the present application provides a pension demand forecasting method, which optimizes the long short-term memory network model using an improved gray wolf algorithm, so that the long short-term memory network model exhibits higher prediction accuracy, better generalization ability and faster training speed, thereby accurately identifying customer needs and significantly improving marketing conversion rates.
[0043] See also Figure 1 As shown, an embodiment of the present invention discloses a pension demand prediction method, comprising:
[0044] Step S11, determining a corresponding market feedback score through the collected market feedback information; the market feedback information is information related to customer behavior generated in pension marketing activities.
[0045] In this embodiment, in the pension marketing activities, the market feedback information related to the customer behavior generated in the pension marketing activities is collected, and the market feedback information is quantified to obtain a corresponding market feedback score. Specifically, the corresponding market feedback score is determined by the collected market feedback information, including: obtaining market feedback information by collecting information related to the customer behavior generated in the pension marketing activities, and quantifying the market feedback information to obtain a market feedback score.
[0046] It is understandable that the market feedback information may include pension product purchase rate, customer conversion rate, user satisfaction, etc. The market feedback information is collected in real time, converted into digital data, and then analyzed and processed to obtain the corresponding market feedback score. In a specific implementation, a marketing activity with a higher conversion rate corresponds to a higher market feedback score.
[0047] Step S12, generating each gray wolf individual through a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model.
[0048] In this embodiment, each gray wolf individual is generated in the preset hyperparameter search space, and a preset fitness function and the market feedback score are used to perform a first evaluation on the fitness of each gray wolf individual to obtain a corresponding first evaluation result. Specifically, the generation of each gray wolf individual through the preset hyperparameter search space, and the first evaluation of the fitness of each gray wolf individual based on the market feedback score and the preset fitness function to obtain the first evaluation result includes: generating each gray wolf individual through the preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and the preset fitness function; ranking the fitness of each gray wolf individual based on the first evaluation result, and eliminating the gray wolf individuals that meet the preset interval conditions.
[0049] Step S13, respectively using the memory mechanism and the individual variation mechanism to determine the first target gray wolf individual and the second target gray wolf individual among the gray wolf individuals, and performing a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and the preset fitness function to obtain a second evaluation result.
[0050] In this embodiment, the memory mechanism and the individual variation mechanism are introduced into the gray wolf algorithm to obtain the first target gray wolf individual and the mutated second target gray wolf individual that meet the preset priority conditions, and the first target gray wolf individual and the second target gray wolf individual are second evaluated using the preset fitness function to obtain the corresponding second evaluation result. Specifically, the memory mechanism and the individual variation mechanism are respectively used to determine the first target gray wolf individual and the second target gray wolf individual in each of the gray wolf individuals, and the first target gray wolf individual and the second target gray wolf individual are second evaluated based on the market feedback score and the preset fitness function to obtain the second evaluation result, including: using the memory mechanism in the gray wolf algorithm to obtain the first target gray wolf individual that meets the preset optimal condition in each of the gray wolf individuals, and obtaining the mutated second target gray wolf individual in each of the gray wolf individuals based on the individual variation mechanism in the gray wolf algorithm; obtaining the summed gray wolf individual by summing the first target gray wolf individual and the second target gray wolf individual, and evaluating the summed gray wolf individual based on the market feedback score and the preset fitness function to obtain the second evaluation result.
[0051] It can be understood that by introducing the memory mechanism to obtain the first target gray wolf individual that meets the preset priority conditions, the gray wolf algorithm can be prevented from falling into the local optimum; at the same time, the introduction of the individual variation mechanism allows some of the gray wolf individuals to randomly mutate some hyperparameters, which can increase the group's positive and global search capabilities. Among them, the calculation formulas corresponding to the memory mechanism and the individual variation mechanism are:
[0052] ;
[0053] in, To meet the preset priority conditions, the first target gray wolf individual, The second target of mutation is the gray wolf individual.
[0054] Step S14: Optimize the long short-term memory network model based on the first evaluation result and the second evaluation result, and predict the customer's pension demand through the optimization result.
[0055] In this embodiment, the process of optimizing the long short-term memory network model is an iterative process. After each optimization process result, the market performance of the optimized long short-term memory network model will be re-evaluated, and the optimization target and the preset hyperparameter search space will be adjusted to improve the actual application effect of the optimized long short-term memory network model. Specifically, after optimizing the long short-term memory network model based on the first evaluation result and the second evaluation result, it also includes: dividing the process of optimizing the long short-term memory network model into several feedback cycles, and optimizing the long short-term memory network model in the next feedback cycle through the feedback record of the current feedback cycle. It should be pointed out that the market feedback of each optimization process is recorded and used in the next optimization process, thereby forming a closed-loop result, that is, the optimization strategy and market performance drive each other to ensure the continuous optimization of the long short-term memory network model.
[0056] As can be seen from the above, the present application performs a first evaluation on the fitness of each gray wolf individual generated through the market feedback score determined by the market feedback information and the preset fitness function, and uses the memory mechanism and the individual variation mechanism to obtain the first target gray wolf individual and the second target gray wolf individual of each gray wolf individual, and then performs a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and the preset fitness function. The long-short memory network model is optimized through the two evaluation results of different gray wolf individuals to achieve the prediction of customer pension needs based on the optimization results. In this way, the search diversity is enhanced through the memory mechanism and the individual variation mechanism, and the local optimal trap is effectively avoided; the long-short memory network model is optimized by using the improved gray wolf algorithm, so that the long-short memory network model shows higher prediction accuracy, better generalization ability and faster training speed, so as to accurately identify customer needs and significantly improve marketing conversion rate.
[0057] It can be seen from the above embodiments that the present application optimizes the long short-term memory network model based on the improved gray wolf algorithm. Therefore, the process of first evaluating the fitness of each gray wolf individual based on the market feedback score and the preset fitness function is described.
[0058] See also Figure 2 As shown, the embodiment of the present invention discloses a specific pension demand forecasting method, including:
[0059] Step S21, determining a corresponding market feedback score through the collected market feedback information; the market feedback information is information related to customer behavior generated in pension marketing activities.
[0060] Step S22, generating individual gray wolves representing different hyperparameter configurations of the long short-term memory network model through a preset hyperparameter search space, and performing a first evaluation on the fitness of each of the individual gray wolves based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result.
[0061] In this embodiment, it can be understood that a group of gray wolf individuals are randomly generated in the preset hyperparameter search space, and each gray wolf individual represents a different hyperparameter configuration of the LSTM model, such as learning rate, batch size, number of hidden layer units, activation function, regularization system, etc.; and a preset fitness function is used to evaluate the hyperparameter configuration effect of each gray wolf individual, and the function comprehensively considers the prediction error on the validation set (such as mean square error, cross entropy), model training time and generalization ability. The definition of the first evaluation of the fitness of each gray wolf individual using the preset fitness function is as follows:
[0062] ;
[0063] in, is the prediction error on the validation set, For training time, For generalization ability, , and To adjust the weight, Provide a feedback score for the market. The weight coefficient corresponding to the market feedback score is used for weighted comprehensive evaluation together with other indicators (such as validation set error, training time, and generalization ability).
[0064] It is understood that the weight coefficient corresponding to the market feedback score is adjusted according to the importance and real-time changes of the market feedback. Specifically, the fitness of each of the gray wolf individuals is first evaluated based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result, including: if the market feedback meets the preset key conditions, then the market feedback weight corresponding to the market feedback score is increased; if the market feedback does not meet the preset key conditions, then the market feedback weight corresponding to the market feedback score is reduced.
[0065] In a specific implementation, if a significant decrease in the conversion rate is detected, the weight corresponding to the market feedback score is automatically increased, that is, the market feedback meets the preset key conditions, so as to guide the LSTM model to optimize in a more market-adaptive direction. It should be noted that the preset key conditions can be adjusted accordingly according to actual conditions and are not specifically limited here.
[0066] Furthermore, the search characteristics of the individual gray wolf are adjusted according to different stages of the optimization process. In the early stage of the optimization process, the global search is emphasized, and the step size is large to avoid falling into the local optimum; in the middle and late stages, the step size is gradually reduced to enhance the local development ability. The weight adjustment formula is:
[0067] ;
[0068] in, For the The search weight corresponding to the gray wolf individual in the sub-optimization process, is the maximum number of optimizations, is the initial weight value, is the final weight value.
[0069] Step S23, respectively using the memory mechanism and the individual variation mechanism to determine the first target gray wolf individual and the second target gray wolf individual among the gray wolf individuals, and performing a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and the preset fitness function to obtain a second evaluation result.
[0070] Step S24: Optimize the long short-term memory network model based on the first evaluation result and the second evaluation result, and predict the customer's pension demand through the optimization result.
[0071] Among them, the specific implementation process of the above steps S21, S23 and step S24 can refer to the corresponding content in the previous embodiment, and will not be repeated here.
[0072] As can be seen from the above, after obtaining the market feedback score, the fitness of each gray wolf individual is first evaluated through the market feedback score, the training time of the long short-term memory network model, the prediction error on the validation set, and the weight coefficients corresponding to the generalization ability. In the process of the first evaluation, it is determined whether the market feedback meets the preset key conditions to adjust the weight corresponding to the market feedback score. In this way, the LSTM model is repeatedly optimized by the improved gray wolf algorithm, so that the optimized LSTM model shows higher prediction accuracy, better generalization ability and faster training speed. In pension marketing applications, the model can accurately identify customer needs, significantly improve marketing conversion rate, and optimize the resource utilization efficiency of the system.
[0073] Accordingly, see Figure 3 As shown, the present application also provides a pension demand prediction device, comprising:
[0074] The score collection module 11 is used to determine the corresponding market feedback score through the collected market feedback information; the market feedback information is the information related to the customer behavior generated in the pension marketing activities;
[0075] A first evaluation module 12 is used to generate each gray wolf individual through a preset hyperparameter search space, and perform a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model;
[0076] A second evaluation module 13 is used to determine a first target gray wolf individual and a second target gray wolf individual among the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and to perform a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result;
[0077] The demand prediction module 14 is used to optimize the long short-term memory network model based on the first evaluation result and the second evaluation result, and predict the customer's pension demand through the optimization result.
[0078] As can be seen from the above, the present application performs a first evaluation on the fitness of each gray wolf individual generated through the market feedback score determined by the market feedback information and the preset fitness function, and uses the memory mechanism and the individual variation mechanism to obtain the first target gray wolf individual and the second target gray wolf individual of each gray wolf individual, and then performs a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and the preset fitness function. The long-short memory network model is optimized through the two evaluation results of different gray wolf individuals to achieve the prediction of customer pension needs based on the optimization results. In this way, the search diversity is enhanced through the memory mechanism and the individual variation mechanism, and the local optimal trap is effectively avoided; the long-short memory network model is optimized by using the improved gray wolf algorithm, so that the long-short memory network model shows higher prediction accuracy, better generalization ability and faster training speed, so as to accurately identify customer needs and significantly improve marketing conversion rate.
[0079] In some specific implementations, the score collection module 11 may specifically include:
[0080] The feedback information quantification unit is used to obtain market feedback information by collecting information related to customer behavior generated in pension marketing activities, and quantify the market feedback information to obtain a market feedback score.
[0081] In some specific implementations, the first evaluation module 12 may specifically include:
[0082] A first evaluation completion unit, configured to generate each gray wolf individual through a preset hyperparameter search space, and perform a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function;
[0083] The fitness ranking unit is used to rank the fitness of each gray wolf individual based on the first evaluation result, and eliminate the gray wolf individuals that meet the preset interval conditions.
[0084] In some specific implementations, the first evaluation module 12 may specifically include:
[0085] The first evaluation result determination unit is used to generate gray wolf individuals with different hyperparameter configurations representing the long short-term memory network model through a preset hyperparameter search space, and perform a first evaluation on the fitness of each gray wolf individual based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result.
[0086] In some specific implementations, the first evaluation module 12 may specifically include:
[0087] A weight increasing unit, configured to increase the market feedback weight corresponding to the market feedback score if the market feedback meets a preset key condition;
[0088] The weight reducing unit is used to reduce the market feedback weight corresponding to the market feedback score if the market feedback does not meet the preset key condition.
[0089] In some specific implementations, the second evaluation module 13 may specifically include:
[0090] A target individual determination unit is used to obtain a first target gray wolf individual that meets a preset optimal condition among the gray wolf individuals by using a memory mechanism in the gray wolf algorithm, and to obtain a second target gray wolf individual that varies among the gray wolf individuals based on an individual variation mechanism in the gray wolf algorithm;
[0091] The second evaluation completion unit is used to obtain a summed gray wolf individual by summing the first target gray wolf individual and the second target gray wolf individual, and evaluate the summed gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result.
[0092] In some specific implementations, the pension demand prediction device may further include:
[0093] The model optimization unit is used to divide the process of optimizing the long short-term memory network model into several feedback cycles, and optimize the long short-term memory network model in the next feedback cycle through the feedback record of the current feedback cycle.
[0094] Furthermore, the present application also discloses an electronic device. Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment, and the content in the figure cannot be regarded as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the pension demand prediction method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0095] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.
[0096] In addition, the memory 22 as a carrier for resource storage may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.
[0097] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, and can be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the pension demand forecasting method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include computer programs that can be used to complete other specific tasks.
[0098] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the pension demand forecasting method disclosed above. The specific steps of the method can refer to the corresponding contents disclosed in the above embodiments, and will not be repeated here.
[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0100] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0101] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0102] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0103] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A pension demand forecasting method, characterized in that: include: Determine the corresponding market feedback score through the collected market feedback information; The market feedback information is information related to customer behavior generated during pension marketing activities; Generating each gray wolf individual through a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model; Determine a first target gray wolf individual and a second target gray wolf individual in each of the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and perform a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result; The long short-term memory network model is optimized based on the first evaluation result and the second evaluation result, and the customer's pension demand is predicted through the optimization result.
2. The pension demand forecasting method according to claim 1, characterized in that: Determining the corresponding market feedback score through the collected market feedback information includes: Market feedback information is obtained by collecting information related to customer behavior generated in pension marketing activities, and the market feedback information is quantified to obtain a market feedback score.
3. The pension demand forecasting method according to claim 1, characterized in that: The step of generating each individual gray wolf by using a preset hyperparameter search space, and performing a first evaluation on the fitness of each individual gray wolf based on the market feedback score and a preset fitness function to obtain a first evaluation result, includes: Generating each gray wolf individual through a preset hyperparameter search space, and performing a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function; The fitness of each gray wolf individual is ranked based on the first evaluation result, and the gray wolf individuals that meet the preset interval conditions are eliminated.
4. The pension demand forecasting method according to claim 1, characterized in that: The step of generating each individual gray wolf by using a preset hyperparameter search space, and performing a first evaluation on the fitness of each individual gray wolf based on the market feedback score and a preset fitness function to obtain a first evaluation result, includes: By presetting the hyperparameter search space, gray wolf individuals representing different hyperparameter configurations of the long short-term memory network model are generated, and a first evaluation is performed on the fitness of each gray wolf individual based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients to obtain a first evaluation result.
5. The pension demand forecasting method according to claim 4, characterized in that: The first evaluation of the fitness of each of the gray wolf individuals based on the training time of the long short-term memory network model, the prediction error on the validation set, the generalization ability, the market feedback score and the corresponding weight coefficients is performed to obtain a first evaluation result, including: If the market feedback meets the preset key conditions, the market feedback weight corresponding to the market feedback score is increased; If the market feedback does not meet the preset key conditions, the market feedback weight corresponding to the market feedback score is reduced.
6. The pension demand forecasting method according to claim 1, characterized in that: The method comprises: determining a first target gray wolf individual and a second target gray wolf individual among the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and performing a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result, including: A first target gray wolf individual that meets a preset optimal condition is obtained from among the gray wolf individuals by using a memory mechanism in the gray wolf algorithm, and a second target gray wolf individual that varies from among the gray wolf individuals is obtained based on an individual variation mechanism in the gray wolf algorithm; A summed gray wolf individual is obtained by summing the first target gray wolf individual and the second target gray wolf individual, and the summed gray wolf individual is evaluated based on the market feedback score and a preset fitness function to obtain a second evaluation result.
7. The pension demand forecasting method according to any one of claims 1 to 6, characterized in that: After optimizing the long short-term memory network model based on the first evaluation result and the second evaluation result, the method further includes: The process of optimizing the long short-term memory network model is divided into several feedback cycles, and the long short-term memory network model in the next feedback cycle is optimized through the feedback record of the current feedback cycle.
8. A pension demand prediction device, characterized in that: include: A score collection module, used to determine the corresponding market feedback score through the collected market feedback information; The market feedback information is information related to customer behavior generated during pension marketing activities; A first evaluation module is used to generate each gray wolf individual through a preset hyperparameter search space, and perform a first evaluation on the fitness of each gray wolf individual based on the market feedback score and a preset fitness function to obtain a first evaluation result; the gray wolf individuals represent different hyperparameter configurations of the long short-term memory network model; A second evaluation module is used to determine a first target gray wolf individual and a second target gray wolf individual among the gray wolf individuals by using a memory mechanism and an individual variation mechanism respectively, and to perform a second evaluation on the first target gray wolf individual and the second target gray wolf individual based on the market feedback score and a preset fitness function to obtain a second evaluation result; A demand prediction module is used to optimize the long short-term memory network model based on the first evaluation result and the second evaluation result, and predict the customer's pension demand through the optimization result.
9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the pension demand forecasting method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the pension demand forecasting method according to any one of claims 1 to 7 is implemented.