Deep reinforcement learning based personalized reward allocation method and system

Abstract

The application relates to a personalized reward allocation method and system based on deep reinforcement learning, which comprises the following steps: in response to a request of a user for participating in a reward allocation activity through a client, collecting feature data of the user and reward pool information through a distributed data collection framework, and performing standardization processing to generate structured input data; processing the structured input data by using a deep reinforcement learning model to generate a personalized reward allocation probability, and ensuring that an allocation cost meets preset budget constraints by using a constraint optimization algorithm; and according to the personalized reward allocation probability, executing the reward allocation requested by the user by using a randomization algorithm and returning an allocation result. The application realizes personalized reward allocation through deep reinforcement learning, significantly improves user participation and satisfaction, ensures real-time performance and dynamic adaptability through distributed data processing and an online learning mechanism, and effectively controls the allocation cost through a constraint optimization algorithm, and is suitable for large-scale application.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a personalized reward distribution method and system based on deep reinforcement learning. Background Art

[0002] In user marketing activities, reward distribution systems are widely used to increase user engagement and brand loyalty. Traditional reward distribution methods mostly use fixed probability or random distribution strategies, which makes it difficult to dynamically adjust the distribution plan according to the user's personalized characteristics, resulting in poor user experience and waste of resources. In recent years, data-driven personalization technologies have gradually emerged, such as recommendation systems based on user profiles, but their application in reward distribution still has limitations, such as the lack of real-time optimization mechanisms and cost control capabilities. In existing technologies, some methods attempt to introduce machine learning, but they mostly rely on static models and have difficulty adapting to dynamic changes in user behavior. In addition, when processing multi-dimensional user data such as behavior sequences and preference information, traditional methods have insufficient data processing efficiency and model accuracy, and lack an effective budget constraint mechanism.

[0003] To address the above issues, there is an urgent need for a technical solution that can integrate multi-dimensional user data, dynamically generate personalized allocation probabilities, and achieve real-time optimization under budget constraints to improve the accuracy and efficiency of reward allocation and meet the needs of modern marketing scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a personalized reward distribution method and system based on deep reinforcement learning to solve the problem that the existing technology model has poor longitude and cannot control the budget.

[0005] To achieve one of the above-mentioned objectives, an embodiment of the present invention provides a personalized reward distribution method based on deep reinforcement learning, the method comprising:

[0006] In response to a user's request to participate in a reward distribution activity through the client, the user's feature data and reward pool information are collected through a distributed data collection framework, and standardized to generate structured input data;

[0007] Use deep reinforcement learning models to process structured input data, generate personalized reward allocation probabilities, and use constrained optimization algorithms to ensure that the allocation costs meet the preset budget constraints;

[0008] Based on the personalized reward allocation probability, a randomized algorithm is used to execute the reward allocation requested by the user and return the allocation result.

[0009] As a further improvement of an embodiment of the present invention, the method further includes that collecting the user's feature data and reward pool information through the distributed data collection framework includes:

[0010] The distributed data collection framework collects user feature data and reward pool information in real time through stream data processing technology, wherein:

[0011] The user feature data includes demographic characteristics, device information, behavior sequence data, consumption records and preference information;

[0012] The reward pool information includes the value of the reward item, the distribution heat and the user's preference information for the reward item;

[0013] The data collection also includes data cleaning of the user feature data and the reward pool information, and the data cleaning removes invalid data through an anomaly detection algorithm.

[0014] As a further improvement of an embodiment of the application, the method further comprises that the standardization processing to generate structured input data includes,

[0015] The standardization processing includes applying a normalization algorithm to convert the user feature data and the reward pool information after data cleaning into a unified dimension;

[0016] The non-numerical data in the user feature data and the reward pool information converted into a unified dimension are encoded into numerical vectors using embedding technology;

[0017] The numerical vectors are batch processed and format converted using a distributed data processing framework to generate a structured feature matrix suitable for deep neural network input.

[0018] As a further improvement of an embodiment of the application, the method further comprises that the processing of the structured input data using a deep reinforcement learning model to generate a personalized reward allocation probability includes,

[0019] The structured input data is encoded into a state vector using embedding technology;

[0020] The state vector is processed through a multi-layer deep neural network to generate a non-normalized weight of each reward item j ;

[0021] The allocation probability of each reward item is calculated according to the non-normalized weight and the reward value, and the formula is:

[0022]

[0023] wherein, is the allocation probability of the jth reward, is the target expected cost, is the price of the kth prize, k is the reward index, and n is the total number of rewards.

[0024] As a further improvement of an embodiment of the present application, the method further comprises that the ensuring, by the constraint optimization algorithm, that the allocation cost satisfies the preset budget constraint comprises,

[0025] applying a constraint loss function ensuring that the total expected cost and the single item probability do not exceed a preset upper limit, the constraint loss function being expressed as:

[0026]

[0027] wherein, is the total expected cost of all reward items, is the maximum allowed allocation probability upper limit of a single reward item, is a constraint strength coefficient.

[0028] As a further improvement of an embodiment of the present application, the method further comprises that the performing, by the randomization algorithm, the reward allocation requested by the user and returning the allocation result comprises,

[0029] generating a random number by a pseudo-random number generation algorithm ;

[0030] calculating the total winning probability of all reward items based on the allocation probability of each reward item ;

[0031] determining whether the user wins and obtains which reward item by comparing the random number with the probability distribution interval divided based on the allocation probability of each reward item , specifically:

[0032] if , it is determined that the user does not obtain a reward;

[0033] if , it is determined that the user obtains the jth reward item;

[0034] storing the allocation result to a distributed database through an application programming interface.

[0035] As a further improvement of an embodiment of the present application, the method further comprises collecting feedback data of the user on the allocation result, and optimizing the deep reinforcement learning model through an online learning algorithm to update the reward allocation probability generation strategy;

[0036] the optimizing, by the online learning algorithm, the deep reinforcement learning model comprises,

[0037] ​Collecting feedback data of the user through a client interface; the feedback data of the user includes a satisfaction score, a text evaluation, and interaction behavior data;

[0038] Performing sentiment analysis on the text evaluation by using a natural language processing algorithm, and generating an instant reward in combination with the satisfaction score as a reward return signal;

[0039] Updating the parameters of the deep neural network based on a comprehensive loss function using an online learning algorithm; the comprehensive loss function is represented as:

[0040]

[0041] wherein, is a weight hyperparameter, is a value loss function, and is represented as:

[0042]

[0043] wherein, is an expectation, is a loss function, is a target value, is a value function estimate in state s, and is represented as:

[0044]

[0045] wherein, i is a time step index, is a discount factor, is an instant reward at time t+i, is a state at time t.

[0046] To achieve one of the above-mentioned purposes, an embodiment of the present application further provides a personalized reward allocation system based on deep reinforcement learning, which comprises a data collection module, a probability generation module, and a reward allocation module.

[0047] The data collection module is configured to collect feature data of the user and reward pool information through a distributed data collection framework in response to a request of the user for participating in a reward allocation activity through a client, and to perform standardization processing to generate structured input data.

[0048] The probability generation module is configured to process the structured input data by using a deep reinforcement learning model, to generate a personalized reward allocation probability, and to ensure that an allocation cost meets a preset budget constraint by using a constraint optimization algorithm.

[0049] The reward allocation module is configured to execute the reward allocation requested by the user according to the personalized reward allocation probability by using a randomization algorithm, and to return an allocation result.

[0050] To achieve the above-mentioned one of the purposes of the application, an embodiment of the application further provides an electronic device, comprising a memory and a processor, characterized in that the memory stores a computer program executable on the processor, and the processor implements the steps in the personalized reward allocation method based on deep reinforcement learning as described above when executing the program.

[0051] To achieve the above-mentioned one of the purposes of the application, an embodiment of the application further provides a storage medium, characterized in that the storage medium stores a computer program, and the computer program implements the steps in the personalized reward allocation method based on deep reinforcement learning as described above when executed by a processor.

[0052] Compared with the prior art, the personalized reward allocation method and system based on deep reinforcement learning provided by the application realize personalized reward allocation through deep reinforcement learning, significantly improve user participation and satisfaction, distributed data processing and online learning mechanism ensure real-time and dynamic adaptability, constraint optimization algorithm effectively controls allocation cost, is suitable for large-scale application, and closed-loop process integrates data collection, probability generation, allocation execution and model optimization, enhances system efficiency and marketing effect. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is the overall flowchart of the personalized reward allocation method based on deep reinforcement learning described in the application.

[0054] Figure 2 is the architecture schematic diagram of the personalized reward allocation system based on deep reinforcement learning described in the application. DETAILED DESCRIPTION

[0055] The application will be described in detail below with reference to the specific embodiments shown in the drawings. However, these embodiments do not limit the application, and the changes in structure, method or function made by those skilled in the art based on these embodiments are included in the protection scope of the application.

[0056] The embodiments of the application will be described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the application, and cannot be understood as limiting the application.

[0057] In the first embodiment of the application, the application provides a personalized reward allocation method based on deep reinforcement learning, as shown in Figure 1 , comprising,

[0058] S1: In response to a user's request to participate in the reward allocation activity through the client, collect the user's feature data and reward pool information through the distributed data collection framework, and perform standardization processing to generate structured input data;

[0059] S2: Process the structured input data using a deep reinforcement learning model to generate a personalized reward allocation probability, and ensure that the allocation cost meets the preset budget constraint through a constraint optimization algorithm;

[0060] S3: According to the personalized reward allocation probability, use a randomization algorithm to execute the reward allocation requested by the user and return the allocation result.

[0061] In one specific embodiment of the present application, the user's feature data and reward pool information are collected through a distributed data collection framework, specifically,

[0062] The distributed data collection framework collects user feature data and reward pool information in real time through stream processing technology, wherein:

[0063] The user feature data includes demographic characteristics, device information, behavior sequence data, consumption records, and preference information;

[0064] The reward pool information includes the value of the reward item, the allocation heat, and the user's preference information for the reward item;

[0065] The data collection also includes data cleaning of the user feature data and the reward pool information, and the data cleaning removes invalid data through an anomaly detection algorithm.

[0066] It should be noted that the distributed data collection framework uses stream processing technology to achieve real-time collection of user feature data and reward pool information through a high-throughput, low-latency message queue. Specifically, the framework is deployed on a distributed server cluster and contains multiple data collection nodes. Each node interacts with the client (such as a mobile application or a web page) through an application programming interface to capture user behavior and preference data in real time. The collection process responds to the user's request to participate in the reward allocation activity through the client, such as clicking the "participate in the activity" button or entering the activity page.

[0067] Further, the user feature data includes the following dimensions: Demographic characteristics: including user age, gender, geographic location, occupation, etc. information, obtained through user registration information or third-party authorized data. Device information: including device model, operating system version, screen resolution, etc. collected through the client SDK. Behavior sequence data: including user click frequency, page dwell time, sharing behavior, etc. record user interaction sequence through event tracking technology (such as point-in-time). Consumption records: including historical order amount, purchase category, browsing record, etc. extracted from the e-commerce platform database. Preference information: including user's preference for reward types (such as physical rewards, discount coupons) and time sensitivity, obtained through user ratings or selection behavior on the client.

[0068] Further, the reward pool information includes the following content: Value of reward items: refers to the monetary value or equivalent cost of each reward item, such as the face value of a cash coupon or the market price of a physical reward, extracted from the reward pool database. Allocation heat: refers to the popularity of the reward item, calculated by counting the frequency of user clicks or selections on the reward item. User preference information for reward items: including user ratings or rankings of reward items, such as user priority selection of rewards of equal value collected through the client interface.

[0069] Further, data collection also includes data cleaning of the user feature data and the reward pool information, which is implemented as follows: Anomaly detection algorithm: uses statistical-based anomaly detection methods and machine learning algorithms to remove invalid data. For example, remove behavior data with abnormally high click frequency (above a pre-set threshold), or remove invalid coordinate values (such as beyond a reasonable range) in geographic location data. Data deduplication and integrity check: through a distributed data processing framework, the collected data is processed for deduplication, and the data integrity is checked to ensure that key fields (such as user ID, reward item ID) are not empty. The data after data cleaning is transmitted to the data storage module through a distributed message queue, providing high-quality input for subsequent standardized processing and deep reinforcement learning models.

[0070] In one specific embodiment of the present application, standardized processing is performed to generate structured input data, specifically,

[0071] The standardized processing includes applying normalization algorithms to convert the user feature data and the reward pool information after data cleaning into a unified dimension;

[0072] Non-numeric data in the user feature data and the reward pool information converted into a unified dimension are encoded into numerical vectors using embedding technology;

[0073] The numerical vectors are batched and formatted using a distributed data processing framework to generate a structured feature matrix suitable for deep neural network input.

[0074] It should be noted that the standardization process aims to convert the data-cleansed user feature data and reward pool information into a unified format suitable for deep neural network input, ensuring data consistency and model calculation efficiency. The process is executed on a distributed server cluster through a distributed data processing framework, including three sub-steps: normalization, embedding encoding, and batch processing, to generate a structured feature matrix.

[0075] Further, a normalization algorithm is applied to convert the data-cleansed user feature data and reward pool information into a unified dimension to eliminate dimension differences of different data dimensions. Preferably, the minimum-maximum normalization or Z-score normalization algorithm is used to process numerical data (such as the number of clicks in user behavior sequences and the value of reward items) to ensure that data values are distributed within a unified range, eliminating the influence of dimension differences on deep neural network training.

[0076] Further, for non-numerical data in the user feature data and the reward pool information (such as gender in demographic characteristics and category of reward items), embedding technology is used for encoding to convert it into a numerical vector. Specifically, a pre-trained embedding model or a custom embedding layer is used to map categorical variables to a high-dimensional vector space, generating a low-dimensional dense vector representation to preserve the semantic relationship of the data and adapt to the input requirements of the neural network.

[0077] Further, the numerical vectors are batched and formatted using a distributed data processing framework. Specifically, the distributed data processing framework performs batch aggregation on the numerical vectors through parallel computing, unifies the data format, organizes the data according to the input requirements of the deep neural network, and generates a structured feature matrix. The structured feature matrix contains the fusion representation of user features and reward pool information, arranged in fixed dimensions to ensure compatibility with the input layer of the deep reinforcement learning model.

[0078] In one specific embodiment of the present application, a deep reinforcement learning model is used to process structured input data to generate personalized reward allocation probabilities, specifically,

[0079] The structured input data is encoded into a state vector using embedding technology;

[0080] The state vector is processed through a multi-layer deep neural network to generate non-normalized weights for each reward item j ;

[0081] The allocation probability of each reward item is calculated according to the non-normalized weight and reward value, with the formula being:

[0082]

[0083] wherein, is the allocation probability of the jth prize, is the target expected cost, is the price of the kth prize, k is the prize index, and n is the total number of prizes.

[0084] It should be noted that the structured input data is encoded into a state vector by using embedding technology, specifically, user feature data and prize pool information are mapped into a high-dimensional continuous vector space by a pre-trained embedding model to capture the semantic relationship and feature correlation between data. The embedding process is implemented through a distributed computing framework to ensure the processing efficiency of high-dimensional data.

[0085] Further, the state vector is processed by a multi-layer deep neural network to generate non-normalized weights of each prize item, wherein the deep neural network adopts a multi-layer perceptron or recurrent neural network architecture, including an input layer, multiple hidden layers and an output layer, and processes non-linear relationships through an activation function to output a weight vector corresponding to the number of prize items.

[0086] Further, according to the non-normalized weights and prize values, the allocation probability of each prize item is calculated by the formula The probability calculation is performed by a distributed computing node to ensure real-time and accuracy, and a probability list is generated and stored in a memory database for subsequent prize allocation module calling.

[0087] In one specific embodiment of the present application, the allocation cost is ensured to meet the preset budget constraint by a constraint optimization algorithm, specifically,

[0088] The constraint loss function ensures that the total expected cost and the single item probability do not exceed the preset upper limit, and the constraint loss function is represented as:

[0089]

[0090] wherein, is the total expected cost of all prize items, is the maximum allowed allocation probability upper limit of a single prize item, is a constraint strength coefficient.

[0091] It should be noted that the constraint loss function is applied to optimize the deep reinforcement learning model to ensure that the total expected cost is equal to the preset budget, and the single item allocation probability does not exceed the maximum allowed probability upper limit. The first term of the constraint loss function penalizes the deviation of the total expected cost from the budget by a square error, and the second term A non-negative penalty is imposed on cases exceeding the upper probability limit by a rectified linear unit function.

[0092] Further, the optimization process is implemented by a distributed computing framework, and a gradient descent algorithm is used to adjust the parameters of the deep neural network to minimize the constrained loss function. The optimized probability distribution is transmitted to the reward allocation module through the in-memory database, ensuring accurate control of the allocation cost and effectiveness of the probability.

[0093] In one specific embodiment of the present application, a randomization algorithm is used to perform the user's request for reward allocation and return the allocation result, specifically,

[0094] A pseudo-random number generation algorithm is used to generate a random number ;

[0095] Based on the allocation probability of each reward item , the total probability of winning all reward items is calculated ;

[0096] By comparing the random number with the probability distribution interval divided based on the allocation probability of each reward item , it is determined whether the user wins and what kind of reward item the user wins, specifically:

[0097] If , it is determined that the user does not win a reward;

[0098] If , it is determined that the user wins the jth reward item;

[0099] The allocation result is stored in the distributed database through the application programming interface.

[0100] It should be noted that a high-performance pseudo-random number generator is used to generate uniformly distributed random numbers in a distributed computing environment, ensuring the fairness and randomness of the allocation process.

[0101] Further, the total probability of all reward items is calculated, and the calculation is completed by reading the pre-generated probability list from the in-memory database. By comparing the random number with the cumulative probability interval, it is determined whether the user wins a reward and the specific reward item. The comparison process is implemented by an efficient interval query algorithm, ensuring real-time performance. The allocation result (including user ID, reward item ID, allocation timestamp, etc.) is stored in the distributed database through the application programming interface to support subsequent feedback data collection and system audit, and the allocation result is returned to the user through the client interface.

[0102] In one specific embodiment of the present application, feedback data of users on the allocation result is collected, and a deep reinforcement learning model is optimized through an online learning algorithm to update the reward allocation probability generation strategy;

[0103] Specifically, the optimization of the deep reinforcement learning model through the online learning algorithm comprises,

[0104] The feedback data of the users is collected through a client interface; the feedback data of the users comprises satisfaction scores, text evaluations, and interaction behavior data;

[0105] The sentiment analysis of the text evaluations is performed through a natural language processing algorithm, and instant rewards are generated in combination with the satisfaction scores as reward return signals;

[0106] The online learning algorithm is used to update the deep neural network parameters based on a comprehensive loss function The comprehensive loss function is expressed as:

[0107]

[0108] wherein, is a weight hyperparameter, is a value loss function, and is expressed as:

[0109]

[0110] wherein, is an expectation, is a loss function, is a target value, is a value function estimate in state s, and is expressed as:

[0111]

[0112] wherein, i is a time step index, is a discount coefficient, is an instant reward at time t+i, is a state at time t.

[0113] It should be noted that the client interface acquires feedback data of users on the reward allocation result from a client (such as a mobile application or a webpage) through a real-time data transmission protocol, and the feedback data of the users comprises satisfaction scores (quantifying the satisfaction degree of the users in the form of numerical values), text evaluations (text feedback of the users on the reward items), and interaction behavior data (such as page staying time and sharing behavior).

[0114] Further, the natural language processing algorithm is used to perform sentiment analysis on the text evaluation, extract sentiment tendency features, and generate instant rewards by a weighted fusion algorithm combining the satisfaction score, as the reward return signal of the deep reinforcement learning model.

[0115] Further, the online learning is implemented through a distributed deep learning framework, and the neural network parameters are updated in real time by using a gradient descent algorithm, and the optimized model parameters are stored in a model server for subsequent probability generation.

[0116] In the second embodiment of the present application, the present application provides a personalized reward allocation system based on deep reinforcement learning, as shown in Figure 2 The system includes a data acquisition module 1, a probability generation module 2 and a reward allocation module 3.

[0117] The data acquisition module 1 is used to collect feature data and reward pool information of the user through a distributed data acquisition framework in response to the user's request for participating in the reward allocation activity through the client, and to generate structured input data by standardizing processing.

[0118] The probability generation module 2 is used to process the structured input data by using a deep reinforcement learning model to generate a personalized reward allocation probability, and to ensure that the allocation cost meets the preset budget constraint by using a constraint optimization algorithm.

[0119] The reward allocation module 3 is used to execute the reward allocation requested by the user according to the personalized reward allocation probability by using a randomization algorithm and return the allocation result.

[0120] In the third embodiment of the present application, the present application provides an electronic device, including a memory and a processor, characterized in that the memory stores a computer program executable on the processor, and the processor executes the program to realize the steps of the personalized reward allocation method based on deep reinforcement learning as described above.

[0121] In the fourth embodiment of the present application, the present application provides a storage medium, which stores a computer program, characterized in that the computer program is executed by a processor to realize the steps of the personalized reward allocation method based on deep reinforcement learning as described above.

[0122] In summary, the application provides a personalized reward allocation method and system based on deep reinforcement learning, which realizes personalized reward allocation through deep reinforcement learning, significantly improves user engagement and satisfaction, and ensures real-time and dynamic adaptability through distributed data processing and online learning mechanism. The constraint optimization algorithm effectively controls the allocation cost and is suitable for large-scale applications. The closed-loop process integrates data collection, probability generation, allocation execution and model optimization, enhancing system efficiency and marketing effectiveness.

[0123] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described modules can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0124] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, i.e., they can be located in one place or distributed to multiple network modules. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs.

[0125] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software function module.

[0126] The integrated module realized in the form of software function module can be stored in a computer readable storage medium. The software function module is stored in a storage medium, including a plurality of instructions for causing a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to execute part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk, and various program code storage media.

[0127] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A personalized reward distribution method based on deep reinforcement learning, characterized by: include, In response to a user's request to participate in a reward distribution activity through the client, the user's feature data and reward pool information are collected through a distributed data collection framework, and standardized to generate structured input data; The characteristic data includes demographic characteristics, device information, behavioral sequence data, consumption records and preference information; Use deep reinforcement learning models to process structured input data, generate personalized reward allocation probabilities, and use constrained optimization algorithms to ensure that the allocation costs meet the preset budget constraints; The method of using a deep reinforcement learning model to process structured input data and generate personalized reward distribution probabilities includes: Encoding the structured input data into a state vector using an embedding technique; The state vector is processed by a multi-layer deep neural network to generate a non-normalized weight for each reward item j ; The probability of allocating each reward item is calculated based on the non-normalized weight and reward value, and the formula is: ； in, is the probability of assigning the jth reward, is the target expected cost, is the price of the kth prize, k is the prize index, and n is the total number of prizes; The method of ensuring that the allocation cost meets the preset budget constraint by using the constraint optimization algorithm includes: Applying a constrained loss function To ensure that the total expected cost and the probability of a single item do not exceed the preset upper limit, the constraint loss function is expressed as: ； in, is the total expected cost of all reward items, The maximum allowable allocation probability for a single reward project, is the constraint strength coefficient; Based on the personalized reward distribution probability, a randomized algorithm is used to execute the reward distribution requested by the user and return the distribution result; The method of using a randomized algorithm to execute the reward distribution requested by the user and returning the distribution result includes: Generate random numbers using a pseudo-random number generation algorithm ; Based on the distribution probability of each reward item , calculate the sum of the winning probabilities of all reward items ; By dividing the random number Based on the probability of distribution of each reward item The probability distribution intervals are divided and compared to determine whether the user wins the prize and what kind of reward item he or she receives, specifically: like , it is determined that the user has not received the reward; like , it is determined that the user obtains the jth reward item; The allocation results are stored in the distributed database through the application program interface.

2. The personalized reward distribution method based on deep reinforcement learning according to claim 1, characterized in that: The user's feature data and reward pool information collected through the distributed data collection framework include: The distributed data collection framework collects user feature data and reward pool information in real time through streaming data processing technology, where: The reward pool information includes the value of the reward items, the distribution popularity and the user's preference information for the reward items; The data collection also includes data cleaning of the user feature data and the reward pool information, wherein the data cleaning removes invalid data through an anomaly detection algorithm.

3. The personalized reward distribution method based on deep reinforcement learning according to claim 2, characterized in that: The standardization process to generate structured input data includes: The standardization process includes applying a normalization algorithm to convert the user feature data and the reward pool information after data cleaning into a unified dimension; Using embedding technology to encode the user feature data converted to a unified dimension and the non-numerical data in the reward pool information into a numerical vector; The numerical vectors described in the distributed data processing framework are batch processed and format converted to generate a structured feature matrix suitable for deep neural network input.

4. The personalized reward distribution method based on deep reinforcement learning according to claim 1, characterized in that: It also includes collecting user feedback data on the allocation results and optimizing the deep reinforcement learning model through online learning algorithms to update the reward allocation probability generation strategy; The optimization of the deep reinforcement learning model by the online learning algorithm includes: Collect user feedback data through the client interface; The user feedback data includes satisfaction scores, text evaluations and interactive behavior data; Use natural language processing algorithms to perform sentiment analysis on the text reviews and generate instant rewards based on the satisfaction scores , as a reward return signal; Use online learning algorithm based on comprehensive loss function Update the deep neural network parameters; the comprehensive loss function is expressed as: ； in, is the weight hyperparameter, is the value loss function, expressed as: ； in, For expectations, is the loss function, is the target value, is the value function estimate in state s, expressed as: ； Where i is the time step index, is the discount factor, is the immediate reward at time t+i, is the state at time t.

5. A personalized reward distribution system based on deep reinforcement learning, applied to the personalized reward distribution method based on deep reinforcement learning as claimed in claim 1, characterized in that: Includes data acquisition module, probability generation module and reward distribution module; The data collection module is used to respond to a user's request to participate in a reward distribution activity through a client, collect the user's feature data and reward pool information through a distributed data collection framework, and perform standardization processing to generate structured input data; The probability generation module is used to process structured input data using a deep reinforcement learning model to generate personalized reward allocation probabilities and ensure that the allocation cost meets the preset budget constraints through a constrained optimization algorithm; The reward allocation module is used to execute the reward allocation requested by the user using a randomized algorithm according to the personalized reward allocation probability and return the allocation result.

6. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the personalized reward distribution method based on deep reinforcement learning as described in any one of claims 1 to 4 are implemented.

7. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the personalized reward distribution method based on deep reinforcement learning as described in any one of claims 1 to 4 are implemented.

Images