A method for generating target customer characteristics of bank time deposits based on quantum assistance

Abstract

The present invention discloses a method for generating target customer features of bank time deposits based on quantum assistance, and relates to the technical field of data generation based on quantum computing. The present invention adopts a brand-new computing mode based on the basic principles of quantum mechanics, namely quantum computing, and quantum computers can generate more data due to their huge Hilbert space, and the randomness of classical computing is not completely random, while quantum computers have completely random characteristics when measuring based on entanglement and superposition characteristics. Based on these characteristics, the quality of the bank time deposit data features generated by the present invention is high, which helps to improve the success rate of bank time deposit marketing activities and greatly reduce marketing costs.

Description

Background Art

[0002] Fixed deposits are the main source of income for banks. Fixed deposits are cash investments held at financial institutions. Your money is invested at an agreed interest rate for a fixed time or period. In order to market fixed deposits to customers, banks have a variety of promotional programs such as email marketing, advertising marketing, telemarketing, digital marketing, etc. Among them, telemarketing campaigns are still one of the most effective ways to reach people. However, they require huge investments because banks need to hire large call centers to actually carry out the activities of marketing fixed deposits to customers. Therefore, it is crucial to identify the customers who are most likely to convert in advance so that they can be specifically targeted by phone, thereby avoiding blindly using large call centers to ask a large number of customers whether they have the intention of fixed deposits. Through this targeted marketing with a roughly targeted audience, the success rate of marketing activities can be greatly improved. At present, people mainly use classical machine learning to process bank fixed deposit data, and quantum-assisted generative adversarial tasks combined with quantum computing have not yet been involved in processing bank fixed deposit data.

[0003] At present, people mainly use classical machine learning to process bank time deposit data. Classical computers perform calculations based on classical bits, which is quite different from quantum computers that perform information processing based on quantum bits. Under the same bit situation, quantum computers can process and generate more data due to their huge Hilbert space, and the randomness of classical calculations is not completely random, while quantum computers have completely random characteristics when measuring based on their entanglement and superposition characteristics. Based on these characteristics, the generation tasks assisted by quantum computing can produce effects that cannot be achieved by pure classical neural network models. Summary of the invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and specifically provide a method for generating target customer characteristics of bank time deposits based on quantum assistance, which is as follows:

[0005] 1) In the first aspect, the present invention provides a method for generating target customer characteristics of bank time deposits based on quantum assistance, and the specific technical scheme is as follows:

[0006] S1. Based on the bank time deposit data set, quantum random noise is obtained;

[0007] S2. Input the quantum random noise into the classical generative neural network to generate fake bank time deposit data features, where the fake bank time deposit data features have the same dimension as the real bank time deposit data features;

[0008] S3, inputting the fake bank time deposit data features into the classical discriminant neural network to obtain a first prediction result, and inputting the real bank time deposit data features into the classical discriminant neural network to obtain a second prediction result;

[0009] S4, combining the first prediction result and the second prediction result into a first loss function for training a classical discriminant neural network, and updating parameters of the classical discriminant neural network according to the first loss function;

[0010] S5. When the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, the fake bank time deposit data features are input into the current classical discriminant neural network to obtain a prediction result, and according to the prediction result, a second loss function for training the classical generative neural network is constructed, and the parameters of the classical generative neural network are updated according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data features as true;

[0011] S6, return to execute S1, until the newly obtained classical discriminant neural network cannot judge whether it is a false bank time deposit data feature;

[0012] S7. Use the latest classical generative neural network to generate bank time deposit data features.

[0013] The beneficial effects of the method for generating target customer characteristics of bank time deposits based on quantum assistance provided by the present invention are as follows:

[0014] A new computing model based on the basic principles of quantum mechanics, namely quantum computing, is adopted. Quantum computers can generate more data due to their huge Hilbert space. The randomness of classical computing is not completely random, while quantum computers have completely random characteristics when measuring based on entanglement and superposition characteristics. Based on these characteristics, the quality of the bank time deposit data features generated by the present invention is relatively high, which helps to improve the success rate of bank time deposit marketing activities and greatly reduce marketing costs.

[0015] On the basis of the above scheme, the method for generating target customer characteristics of bank time deposits based on quantum assistance of the present invention can also be improved as follows.

[0016] Furthermore, S1 includes:

[0017] Preprocess the bank time deposit data set to obtain feature data used to characterize real bank time deposit data;

[0018] Construct a quantum circuit based on the characteristic data representing real bank time deposit data;

[0019] All quantum bits in the quantum circuit are measured to obtain measurement results, which are used as quantum random noise.

[0020] Furthermore, a quantum circuit is constructed based on characteristic data representing real bank time deposit data, including: determining the number of quantum bits based on the size of the characteristic data representing real bank time deposit data, and constructing a quantum circuit based on the number of quantum bits.

[0021] Furthermore, before updating the parameters of the classical generative neural network according to the second loss function, the method also includes: fixing the parameters of the current classical discriminative neural network.

[0022] 2) In the second aspect, the present invention also provides a bank time deposit target customer feature generation system based on quantum assistance, and the specific technical solution is as follows:

[0023] It includes a quantum random noise acquisition module, a training module and a data feature generation module;

[0024] The quantum random noise acquisition module is used to: obtain quantum random noise based on the bank time deposit data set;

[0025] The training module is used to: input quantum random noise into a classical generative neural network to generate fake bank time deposit data features, where the fake bank time deposit data features have the same dimension as the real bank time deposit data features; input the fake bank time deposit data features into a classical discriminant neural network to obtain a first prediction result, and input the real bank time deposit data features into a classical discriminant neural network to obtain a second prediction result; combine the first prediction result and the second prediction result into a first loss function for training the classical discriminant neural network, and update the parameters of the classical discriminant neural network according to the first loss function; when the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, input the fake bank time deposit data features into the current classical discriminant neural network to obtain a prediction result, and construct a second loss function for training the classical generative neural network according to the prediction result, and update the parameters of the classical generative neural network according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data features as true; return to call the quantum random noise acquisition module until the newly obtained classical discriminant neural network cannot determine whether it is a fake bank time deposit data feature;

[0026] The data feature generation module is used to generate bank time deposit data features using the latest classical generative neural network.

[0027] On the basis of the above scheme, the target customer feature generation system for bank time deposit based on quantum assistance of the present invention can also be improved as follows.

[0028] Furthermore, the quantum random noise acquisition module is specifically used for:

[0029] Preprocess the bank time deposit data set to obtain feature data used to characterize real bank time deposit data;

[0030] Construct a quantum circuit based on the characteristic data representing real bank time deposit data;

[0031] All quantum bits in the quantum circuit are measured to obtain measurement results, which are used as quantum random noise.

[0032] Furthermore, the quantum random noise acquisition module is also specifically used to: determine the number of quantum bits according to the size of characteristic data representing real bank time deposit data, and construct a quantum circuit according to the number of quantum bits.

[0033] Furthermore, the training module is also specifically used to: fix the parameters of the current classical discriminant neural network before updating the parameters of the classical generative neural network according to the second loss function.

[0034] 3) In a third aspect, the present invention further provides a computer device, comprising a processor, the processor being coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so that the computer device implements any of the above-mentioned methods for generating target customer characteristics for bank time deposits based on quantum assistance.

[0035] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium, in which at least one computer program is stored, and the at least one computer program is loaded and executed by a processor so that the computer implements any of the above-mentioned methods for generating target customer characteristics for bank time deposits based on quantum assistance.

[0036] It should be noted that the beneficial effects achieved by the technical solutions of the second to fourth aspects of the present invention and the corresponding possible implementation methods can be found in the above-mentioned technical effects of the first aspect and its corresponding possible implementation methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0038] Figure 1 A schematic diagram of a flow chart of a method for generating characteristics of target customers for bank time deposits based on quantum assistance according to an embodiment of the present invention;

[0039] Figure 2 A quantum circuit constructed based on characteristic data representing real bank time deposit data;

[0040] Figure 3The overall architecture of quantum-assisted generative adversarial neural networks;

[0041] Figure 4 Generate the network structure of a classic neural network;

[0042] Figure 5 It is the network structure of the classic discriminant neural network;

[0043] Figure 6 The figure is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, a method for generating target customer characteristics of bank time deposits based on quantum assistance in an embodiment of the present invention includes the following steps:

[0046] S1. Based on the bank time deposit data set, quantum random noise is obtained, including:

[0047] S10, preprocessing the bank time deposit data set to obtain feature data for characterizing real bank time deposit data;

[0048] S11, constructing a quantum circuit according to the characteristic data representing the real bank time deposit data, specifically, determining the number of quantum bits according to the size of the characteristic data representing the real bank time deposit data, and constructing the quantum circuit according to the number of quantum bits;

[0049] S12. Measure all quantum bits in the quantum circuit to obtain measurement results, and use the measurement results as quantum random noise.

[0050] The process of preprocessing the bank time deposit data set is as follows:

[0051] The non-numeric categories of some features in the downloaded bank time deposit data set are represented by different numbers, and then only the customer information who agrees to the time deposit in the bank time deposit data set is taken as the feature information for characterizing the real bank time deposit data in the present invention.

[0052] The process of constructing the quantum circuit is as follows: the dimension of the characteristic information representing the real bank time deposit data of each customer is 16 dimensions. Then based on R y (θ) and R x The (θ) rotation quantum gate and the CZ quantum gate are used to construct a quantum circuit for generating quantum random noise. In this embodiment, 10 quantum bits are used. The specific quantum circuit is as follows: Figure 2 shown.

[0053] S2. Input quantum random noise into the classical generative neural network to generate fake bank time deposit data features. The fake bank time deposit data features have the same dimension as the real bank time deposit data features.

[0054] S3, inputting the fake bank time deposit data features into the classical discriminant neural network to obtain a first prediction result, and inputting the real bank time deposit data features into the classical discriminant neural network to obtain a second prediction result;

[0055] Among them, the fake bank time deposit data characteristics refer to: the characteristic data representing the fake bank time deposit data generated by inputting quantum random noise into the classical generative neural network, and the real bank time deposit data characteristics refer to: the characteristic data representing the real bank time deposit data obtained through the bank time deposit data set.

[0056] The quantum-assisted generative adversarial neural network includes the classical generative neural network and the classical discriminative neural network. The classical generative neural network is also called the generator, and the classical discriminative neural network is also called the discriminator. The overall architecture is as follows: Figure 3 As shown, Figure 3 In the figure, DNN represents the classical discriminant neural network, GNN represents the classical generative neural network, the real data represents: the real bank time deposit data features, quantum noise refers to quantum random noise, and the 16-dimensional fake bank time deposit data features generated by inputting quantum random noise into the classical generative neural network are input into the classical discriminant neural network to obtain the first prediction result, and the real bank time deposit data features obtained through the bank time deposit data set are input into the classical discriminant neural network to obtain the second prediction result.

[0057] Among them, the classic generative neural network, i.e., the generator, includes a multi-layer classic fully connected neural network, such as Figure 4 As shown, the number of neural network nodes in each layer of the classical fully connected neural network is [10, 100, 50, 16], the dimension of the input layer is 10, which is used to input quantum random noise with a data dimension of 10 dimensions, and the data dimension of the output layer is 16 dimensions, which is used to output fake bank time deposit data features with a data dimension of 16 dimensions.

[0058] The structure of the classic discriminant neural network, i.e. the discriminator, is as follows Figure 5As shown, it includes multiple layers of classic fully connected neural networks, and the number of neural network nodes of each layer of the classic fully connected neural network is [16, 50, 100, 10, 1]. The dimension of the input layer is 16, which is used to input fake bank time deposit data features and real bank time deposit data features with 16 dimensions. The dimension of the last hidden layer is 10, which is used to correspond to the input layer dimension of the classic generative neural network, i.e., the generator. The dimension of the output layer is 1. The output layer performs a sigmoid function on the fake bank time deposit data features and the real bank time deposit data features input by the input layer to obtain a prediction result of the input fake bank time deposit data features, i.e., a first prediction result, and a prediction result of the input real bank time deposit data features, i.e., a second prediction result. The prediction result is used to judge whether the input feature data is a fake bank time deposit data feature or a real bank time deposit data feature. The prediction result of 1 indicates that the discriminator judges that the input is a real bank time deposit data feature, and the prediction result of 0 indicates that the discriminator judges that the input is a fake bank time deposit data feature.

[0059] S4. Combining the first prediction result and the second prediction result into a first loss function for training a classical discriminant neural network, and updating the parameters of the classical discriminant neural network according to the first loss function.

[0060] Among them, the first loss function LD is: LD = -E x [log(D(x))]-E z [log(1-D(G(z)))],E x [log(D(x))] represents: the average of the second prediction results D(x) of the classic discriminant neural network, i.e., the discriminator predicts each real bank time deposit data feature x in the training data set batch, E z [log(1-D(G(z)))] means: the average of the first prediction results D(G(z)) of the classic discriminant neural network, i.e., the discriminator, for each fake bank time deposit data feature z in the training data set batch, with a batch size of 100, which can also be set according to actual conditions. Then, the parameters of the classic discriminant neural network are updated based on the first loss function.

[0061] S5. When the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, the fake bank time deposit data features are input into the current classical discriminant neural network to obtain a prediction result, and according to the prediction result, a second loss function for training the classical generative neural network is constructed, and the parameters of the classical generative neural network are updated according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data features as true;

[0062] S6. Return to execute S1 until the newly obtained classical discriminant neural network cannot determine whether it is a false bank time deposit data feature.

[0063] Among them, the preset number threshold can be set according to actual conditions.

[0064] Before updating the parameters of the classical generative neural network according to the second loss function, the method also includes: fixing the parameters of the current classical discriminative neural network.

[0065] Among them, the second loss function is: LG = -E z [log(D(G(z)))], according to the second loss function, the parameters of the classical generative neural network are updated so that the current classical discriminant neural network can identify the fake bank time deposit data features as true. The generator, i.e. the classical generative neural network, and the discriminator, i.e. the classical discriminant neural network, are in an adversarial relationship. The purpose of the classical discriminant neural network is to distinguish whether the input is real bank time deposit data features or fake bank time deposit data features. The purpose of the classical generative neural network is to allow the classical discriminant neural network to recognize the bank time deposit data generated by it, i.e. the fake bank time deposit data features, as true. In other words, the latest classical discriminant neural network cannot distinguish the fake bank time deposit data features generated by the latest classical generative neural network as false.

[0066] The distribution of quantum random noise generated by the quantum circuit is represented by q θ Indicates that the distribution of the data output by the last hidden layer of the classic discriminant neural network is p, and q is obtained θ The cross entropy loss function of p and p is used to update the parameters of the quantum circuit based on the cross entropy loss function. Then, the back propagation algorithm is used to update the parameters of the fully connected neural network in the classical generative neural network and the classical discriminative neural network. Specifically:

[0067] The quantum circuit can generate 10-dimensional quantum random noise data, and its corresponding distribution is q θ Indicates that the last hidden layer of the discriminator can also output 10-dimensional feature data, and its corresponding distribution is represented by p, and then q can be obtained θ The cross entropy loss function of p is L(θ) = -E x [p log max(q θ ,ε)], where E x[] represents the cross entropy loss function and the average of the training data sample x, ε represents a small value, here 1e-14, and then the SPSA (Simultaneous perturbation stochastic approximation) approximation algorithm is used to update the constructed cross entropy loss function Figure 2 Parameters of the quantum circuit used to generate quantum random noise are shown.

[0068] Then, the back propagation algorithm is used to update the parameters of the fully connected neural network in the generator and the discriminator. In other words, the back propagation algorithm is used to update the parameters of the fully connected neural network in the classic generative neural network and the classic discriminative neural network. Finally, the discriminator and the generator are trained against each other for multiple times and stopped after convergence.

[0069] S7. Use the latest classical generative neural network to generate bank time deposit data features.

[0070] The present invention adopts a self-designed quantum-assisted generative adversarial neural network. First, a bank time deposit data set is downloaded and preprocessed. Then, only the customer information of the customer who agrees to the time deposit in the data set is taken as the feature information characterizing the real bank time deposit data. Then, based on the size of the feature information, the number of quantum bits to be used is selected. Then, a quantum circuit is constructed and all quantum bits are finally measured. The measurement result is used as quantum random noise. Then, the quantum random noise is input into a classical generative neural network to generate fake bank time deposit data features with the same dimension as the real bank time deposit data. Then, the bank time deposit data feature data generated by the generative neural network and the real bank time deposit data feature data are respectively input into the constructed classical discriminant neural network. Then, the classical discriminant neural network is output to obtain the prediction results of the real data and the data generated by the generative neural network. The discrimination results of the real data and the generated data are combined into a loss function for training the classical discriminant neural network. The parameters of the classical discriminant neural network are updated according to the obtained loss function corresponding to the classical discriminant neural network so that the discriminator can well identify whether the input is the real bank time deposit data or the bank time deposit data features generated by the generator. After the parameters of the classic discriminant neural network are updated for a set number of times, the bank time deposit data feature data generated by the generative neural network is input into the classic discriminant neural network, and the loss function for training the generative neural network is constructed according to the prediction results of the generated bank time deposit data feature data output by the classic discriminant neural network. Then, the parameters of the discriminator are fixed according to this loss function, and the parameters of the generator are updated so that the discriminator can judge the generated bank time deposit data as true. Through the continuous confrontation between the generator and the discriminator, the discriminator will not be able to distinguish whether the input is the generated bank time deposit data or the real bank time deposit data. Finally, the trained generator can be used to generate new bank time deposit data features that are different from and similar to the real bank time deposit data. Banks can target customers who agree to time deposits based on this feature.

[0071] The beneficial effects of the present invention are as follows:

[0072] 1) Currently, there is no specific research on the task of generating bank time deposit data in combination with quantum computing, and the quality of the data features of bank-approved time deposit customers generated by quantum-assisted generative adversarial neural networks is expected to exceed that of pure classical machine learning algorithms. It can realize the generation of data features of target customers of bank time deposits, improve the success rate of bank time deposit marketing activities, and greatly reduce marketing costs.

[0073] 2) Build an efficient quantum-assisted generative adversarial neural network model to generate target audience data for bank time deposits, improve the success rate of bank time deposit marketing activities, and greatly reduce marketing costs.

[0074] 3) The quantum-assisted generative adversarial neural network proposed in the present invention adopts a new computing model based on the basic principles of quantum mechanics, namely quantum computing. Quantum computers can generate more data due to their huge Hilbert space, and the randomness of classical computing is not completely random, while quantum computers have completely random characteristics when measuring based on their entanglement and superposition characteristics. Based on these characteristics, the bank time deposit generation task assisted by quantum computing can produce effects that cannot be achieved by pure classical neural network models.

[0075] 4) The present invention uses a classical neural network to perform the discriminator and generator tasks, and only generates quantum random noise through a quantum system. Therefore, the present invention can use the currently developed real quantum computers to generate bank time deposit data and other real-world fields.

[0076] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0077] A quantum-assisted bank time deposit target customer feature generation system according to an embodiment of the present invention comprises a quantum random noise acquisition module, a training module and a data feature generation module;

[0078] The quantum random noise acquisition module is used to: obtain quantum random noise based on the bank time deposit data set;

[0079] The training module is used to: input quantum random noise into a classical generative neural network to generate fake bank time deposit data features, where the fake bank time deposit data features have the same dimension as the real bank time deposit data features; input the fake bank time deposit data features into a classical discriminant neural network to obtain a first prediction result, and input the real bank time deposit data features into a classical discriminant neural network to obtain a second prediction result; combine the first prediction result and the second prediction result into a first loss function for training the classical discriminant neural network, and update the parameters of the classical discriminant neural network according to the first loss function; when the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, input the fake bank time deposit data features into the current classical discriminant neural network to obtain a prediction result, and construct a second loss function for training the classical generative neural network according to the prediction result, and update the parameters of the classical generative neural network according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data features as true; return to call the quantum random noise acquisition module until the newly obtained classical discriminant neural network cannot determine whether it is a fake bank time deposit data feature;

[0080] The data feature generation module is used to generate bank time deposit data features using the latest classical generative neural network.

[0081] Optionally, in the above technical solution, the quantum random noise acquisition module is specifically used for:

[0082] Preprocess the bank time deposit data set to obtain feature data used to characterize real bank time deposit data;

[0083] Construct a quantum circuit based on the characteristic data representing real bank time deposit data;

[0084] All quantum bits in the quantum circuit are measured to obtain measurement results, which are used as quantum random noise.

[0085] Optionally, in the above technical solution, the quantum random noise acquisition module is also specifically used to: determine the number of quantum bits according to the size of characteristic data representing real bank time deposit data, and construct a quantum circuit according to the number of quantum bits.

[0086] Optionally, in the above technical solution, the training module is further specifically used to: fix the parameters of the current classical discriminant neural network before updating the parameters of the classical generative neural network according to the second loss function.

[0087] It should be noted that the beneficial effects of the target customer feature generation system for bank time deposits based on quantum assistance provided in the above embodiment are the same as the beneficial effects of the target customer feature generation method for bank time deposits based on quantum assistance, which will not be repeated here. In addition, when the system provided in the above embodiment realizes its functions, it only takes the division of the above functional modules as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0088] In another embodiment, it includes a real bank time deposit data preprocessing and random noise quantum circuit building module, a generator and discriminator building module, a training discriminator module, a training generator module, and a model parameter updating module.

[0089] The real bank time deposit data preprocessing and random noise quantum circuit construction module is used to: represent the non-digital categories of some features in the downloaded bank time deposit data set with different numbers, and then only take the customer information of those who agree to time deposit in the data set as the feature information characterizing the real bank time deposit data in the present invention. Then, according to the size of the processed bank time deposit data feature dimension, select the appropriate number of quantum bits, construct a quantum circuit to generate quantum random noise, so as to generate fake bank customer data features that agree to time deposit through the generator designed later.

[0090] The generator and discriminator construction modules are used to: construct a generator and a discriminator based on the characteristics of the real data set and the quantum random noise generated by the previous module to generate the characteristics of fake bank time deposit data and distinguish between real bank time deposit data and generated bank time deposit data.

[0091] The training discriminator module is used to: combine the discriminator's discrimination results of the real bank time deposit data and the generated bank time deposit data input into a loss function, and then update the discriminator's parameters based on the loss function, so that the discriminator can distinguish whether the input is real bank time deposit data or bank time deposit data generated by the generator.

[0092] The training generator module is used to: input the bank time deposit data feature data generated by the generator into the classic discriminant neural network to obtain the prediction result, construct the loss function based on this result, and then fix the parameters of the discriminator. Based on this loss function, the parameters of the generator are updated so that the discriminator can judge the bank time deposit data generated by the generator as true.

[0093] The model parameter updating module is used for: updating the parameters of the quantum circuit for generating quantum random noise based on the SPSA approximation algorithm; the parameters of the generator and the discriminator can be updated through the back propagation algorithm of the machine learning framework; and finally, the discriminator and the generator are adversarially trained for multiple times so that the method for generating target customer features of bank time deposits based on quantum-assisted generative adversarial neural networks proposed in the present invention stops after convergence.

[0094] This is illustrated by the following examples.

[0095] Real bank time deposit data preprocessing and random noise quantum circuit building module:

[0096] The non-numeric categories of some features in the downloaded bank time deposit data set are represented by different numbers, and then only the customer information that agrees to time deposit in the data set is taken as the feature information that characterizes the real bank time deposit data in the present invention. Each customer contains 16 feature dimensions. Then based on R y (θ) and R x(θ) Rotation quantum gate and CZ quantum gate construct a quantum circuit for generating quantum random noise. Here, 10 quantum bits are used. The specific quantum circuit is as follows: Figure 2 shown.

[0097] Generator and discriminator building blocks:

[0098] The overall architecture of the quantum-assisted generative adversarial neural network designed by the present invention is as follows: Figure 3 As shown in the figure, DNN represents the classic discriminant neural network, GNN represents the classic generative neural network, and the 16-dimensional real bank time deposit data feature vector obtained based on the real bank time deposit data set in the previous module will be obtained through Figure 2 The quantum random noise generated by the quantum circuit is input into the generator to generate a 16-dimensional fake bank agreed time deposit data feature vector, and then the generated bank time deposit data and the real bank time deposit data are respectively input into the discriminator to obtain the discriminator's prediction results for the input data.

[0099] The generator is implemented using a fully connected neural network. Figure 4 As shown in the figure, it is composed of a multi-layer classic fully connected neural network. The number of neural network nodes in each layer is [10, 100, 50, 16], and the dimension of the input layer is 10, which is used to input 10-dimensional data from Figure 2 The quantum random noise after the quantum circuit is measured is shown in the figure. The dimension of the output layer is 16, which is used to generate fake bank-approved time deposit data.

[0100] After obtaining the fake bank time deposit data generated by the generative neural network, the generated fake bank time deposit data and the feature data of the real bank time deposit data can be input into the discriminator. The structure of the discriminator is as follows: Figure 5 As shown in the figure, it is composed of a multi-layer classic fully connected neural network. The number of neural network nodes in each layer is [16, 50, 100, 10, 1]. The dimension of the input layer is 16, which is used to input 16-dimensional real and generated bank time deposit data features. The dimension of the last hidden layer is 10, which corresponds to the input layer dimension of the generator. The dimension of the output layer is 1. By executing a sigmoid function on the output as the prediction result of the input bank time deposit data feature data, it is judged whether the input bank time deposit data is real bank time deposit data or fake bank time deposit data. The prediction result of 1 indicates that the input is a real bank time deposit data feature, and the result of 0 indicates that the input is a fake bank time deposit data feature.

[0101] Train the discriminator module:

[0102] Input the 16-dimensional real bank time deposit data features and the bank time deposit data features generated by the generator into Figure 5In the discriminator shown, the prediction results of the discriminator for the real bank time deposit data input sample and the generated bank time deposit data feature data sample are obtained respectively, and then the two prediction results are combined into a loss function, that is, the first loss function, as shown in the following formula:

[0103] LD=-E x [log(D(x))]-E z [log(1-D(G(z)))]

[0104] The symbol E in the figure x / z [ ] represents the average of the training data set batches for the discriminator's prediction results for the real or generated bank time deposit data feature data. Here, the batch size is 100, where x represents the real bank time deposit data feature, z represents the quantum random noise input to the generator through the quantum circuit, and G(z) represents the bank time deposit data feature data generated after the generator operates on the quantum random noise. D(x) represents the discriminator's prediction results for the real bank time deposit data sample, and D(G(z)) represents the discriminator's prediction results for the generated bank time deposit data sample. Then, based on the above-constructed loss function for the discriminator, the parameters of the discriminator are updated so that the discriminator can distinguish whether the input is real bank time deposit data or bank time deposit data generated by the generator.

[0105] Training generator module: Based on the real bank time deposit data preprocessing and random noise quantum circuit construction module, the 10-dimensional quantum random noise is input into the generator, and then the generated bank time deposit data feature data is input into the discriminator to obtain the prediction result of the discriminator to construct the loss function of the generator, that is, the second loss function, as shown in the following formula:

[0106] LG=-E z [log(D(G(z)))]

[0107] The parameters of the fixed discriminator are updated based on this loss function. The purpose is to make the discriminator identify the bank time deposit data generated by the generator as true. The generator and the discriminator are in an adversarial relationship. The purpose of the discriminator is to distinguish whether the input is true bank time deposit data or false bank time deposit data. The purpose of the generator is to make the discriminator recognize the bank time deposit data it generates as true, that is, to make the discriminator unable to distinguish that the bank time deposit data feature data input by the generator into the discriminator is false.

[0108] Update model parameter module: Figure 2 The quantum circuit can generate 10-dimensional quantum random noise data, and its corresponding distribution is q θIndicates that, similarly, the last hidden layer of the discriminator can also output 10-dimensional feature data, and its corresponding distribution is represented by p, and then q can be obtained θ The cross entropy loss function of and p is shown as follows:

[0109] L(θ)=-E x [p log max(q θ ,ε)]

[0110] In the formula, E x [] represents the cross entropy loss function obtained for the training data sample x and the average, ε represents a smaller value, here it is 1e-14, and then the SPSA approximation algorithm is used to update the constructed cross entropy loss function Figure 2 Parameters of the quantum circuit for generating quantum random noise are shown.

[0111] The back propagation algorithm can be used to update the parameters of the fully connected neural network in the generator and the discriminator. Finally, the discriminator and the generator are trained adversarially for multiple times, so that the bank time deposit target customer feature generation method based on the quantum-assisted generation adversarial neural network proposed in the present invention stops after convergence, so that the generator trained by the present invention can be used to generate new bank time deposit customer data features different from the real bank time deposit customer data features based on the quantum random noise generated by the quantum circuit, so that the bank can target customers who agree to time deposits based on this feature.

[0112] like Figure 6 As shown, a computer device 300 of an embodiment of the present invention includes a processor 320, the processor 320 is coupled to a memory 310, and the memory 310 stores at least one computer program 330, and the at least one computer program 330 is loaded and executed by the processor 320, so that the computer device 300 implements any of the above-mentioned methods for generating target customer characteristics of bank time deposits based on quantum assistance, specifically:

[0113] The computer device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors 320 (Central Processing Units, CPU) and one or more memories 310, wherein the one or more memories 310 store at least one computer program 330, and the at least one computer program 330 is loaded and executed by the one or more processors 320, so that the computer device 300 implements any of the methods for generating target customer characteristics of bank time deposits based on quantum assistance provided in the above embodiments. Of course, the computer device 300 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output, and the computer device 300 may also include other components for implementing device functions, which will not be repeated here.

[0114] A computer-readable storage medium according to an embodiment of the present invention stores at least one computer program, and the at least one computer program is loaded and executed by a processor so that a computer implements any of the above-mentioned methods for generating target customer characteristics for bank time deposits based on quantum assistance.

[0115] Optionally, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0116] In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes any of the above-mentioned methods for generating target customer characteristics of bank time deposits based on quantum assistance.

[0117] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects and represent the definition of a specific order or sequence. The order of use of similar objects can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0118] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0119] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0120] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for generating target customer characteristics of bank time deposits based on quantum assistance, characterized in that: include: S1. Based on the bank time deposit data set, quantum random noise is obtained; S2. Inputting the quantum random noise into a classical generative neural network to generate fake bank time deposit data features, wherein the fake bank time deposit data features have the same dimension as the real bank time deposit data features; S3, inputting the fake bank time deposit data features into the classical discriminant neural network to obtain a first prediction result, and inputting the real bank time deposit data features into the classical discriminant neural network to obtain a second prediction result; S4, combining the first prediction result and the second prediction result into a first loss function for training the classical discriminant neural network, and updating the parameters of the classical discriminant neural network according to the first loss function; S5. When the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, the fake bank time deposit data feature is input into the current classical discriminant neural network to obtain a prediction result, and according to the prediction result, a second loss function for training the classical generative neural network is constructed, and the parameters of the classical generative neural network are updated according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data feature as true; S6, return to execute S1, until the newly obtained classical discriminant neural network cannot judge whether it is a false bank time deposit data feature; S7. Generate bank time deposit data features using the latest classical generative neural network; S1 includes: Preprocess the bank time deposit data set to obtain feature data used to characterize real bank time deposit data; Construct a quantum circuit based on the characteristic data representing real bank time deposit data; Measuring all quantum bits in the quantum circuit to obtain measurement results, and using the measurement results as the quantum random noise; Constructing a quantum circuit according to characteristic data representing real bank time deposit data, including: determining the number of quantum bits according to the size of the characteristic data representing real bank time deposit data, and constructing a quantum circuit according to the number of quantum bits; Among them, the dimension of the characteristic information of each customer representing the real bank time deposit data is 16 dimensions, and then based on R y (θ) and R x (θ) Rotation quantum gate and CZ quantum gate are used to construct quantum circuits for generating quantum random noise; Using the SPSA approximation algorithm, the parameters of the quantum circuit used to generate quantum random noise are updated based on the constructed cross-entropy loss function.

2. According to claim 1, a method for generating target customer characteristics of bank time deposits based on quantum assistance is characterized in that: Before updating the parameters of the classical generative neural network according to the second loss function, the method also includes: fixing the parameters of the current classical discriminative neural network.

3. A bank time deposit target customer feature generation system based on quantum assistance, characterized in that: It includes a quantum random noise acquisition module, a training module and a data feature generation module; The quantum random noise acquisition module is used to: obtain quantum random noise according to a bank time deposit data set; The training module is used to: input the quantum random noise into a classical generative neural network to generate a fake bank time deposit data feature, wherein the fake bank time deposit data feature has the same dimension as the real bank time deposit data feature; input the fake bank time deposit data feature into a classical discriminant neural network to obtain a first prediction result, and input the real bank time deposit data feature into a classical discriminant neural network to obtain a second prediction result; combine the first prediction result and the second prediction result into a first loss function for training the classical discriminant neural network, and update the parameters of the classical discriminant neural network according to the first loss function; when the number of times the parameters of the classical discriminant neural network are updated reaches a preset number threshold, input the fake bank time deposit data feature into the current classical discriminant neural network to obtain a prediction result, construct a second loss function for training the classical generative neural network according to the prediction result, and update the parameters of the classical generative neural network according to the second loss function, so that the current classical discriminant neural network discriminates the fake bank time deposit data feature as true; Return to calling the quantum random noise acquisition module until the newly obtained classical discriminant neural network cannot determine whether it is a false bank time deposit data feature; The data feature generation module is used to generate bank time deposit data features using the latest classical generative neural network; The quantum random noise acquisition module is specifically used for: Preprocess the bank time deposit data set to obtain feature data used to characterize real bank time deposit data; Construct a quantum circuit based on the characteristic data representing real bank time deposit data; Measuring all quantum bits in the quantum circuit to obtain measurement results, and using the measurement results as the quantum random noise; The quantum random noise acquisition module is also specifically used to: determine the number of quantum bits according to the size of the characteristic data representing the real bank time deposit data, and construct a quantum circuit according to the number of quantum bits; Among them, the dimension of the characteristic information of each customer representing the real bank time deposit data is 16 dimensions, and then based on R y (θ) and R x (θ) Rotation quantum gate and CZ quantum gate are used to construct quantum circuits for generating quantum random noise; Using the SPSA approximation algorithm, the parameters of the quantum circuit used to generate quantum random noise are updated based on the constructed cross-entropy loss function.

4. According to claim 3, a system for generating target customer characteristics of bank time deposits based on quantum assistance is characterized in that: The training module is also specifically used to: fix the parameters of the current classical discriminant neural network before updating the parameters of the classical generative neural network according to the second loss function.

5. A computer device, characterized in that: The computer device includes a processor, the processor is coupled to a memory, at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor so that the computer device implements the method for generating target customer characteristics of bank time deposits based on quantum assistance as described in claim 1 or 2.

6. A computer-readable storage medium, characterized in that: At least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by the processor so that the computer implements the method for generating target customer characteristics of bank time deposits based on quantum assistance as described in claim 1 or 2.

Images