Product recommendation method and device, equipment and storage medium
Through the product recommendation model based on recurrent neural networks, personalized product recommendations are made using the historical operation data of the target account, which solves the problems of low efficiency and accuracy in traditional recommendation methods, realizes efficient and accurate personalized recommendations, and improves the user experience.
Patent Information
- Application Number
- CN202510979063.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-16
- Publication Date
- 2025-10-28
Smart Images
Figure CN120852014A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer technology, and in particular to a product recommendation method, apparatus, device, and storage medium. Background Technology
[0002] With the rapid development of financial technology, banking products are becoming increasingly diverse, and account needs are showing a trend of diversification. Recommending personalized products for individual accounts has become a key factor affecting user experience and is of significant research importance.
[0003] Currently, traditional product recommendation methods mainly rely on manual analysis or simple rule engines to recommend products to accounts. However, traditional methods are inefficient and inaccurate in recommending products to individual users, failing to provide personalized product recommendations and resulting in a poor user experience. Summary of the Invention
[0004] This invention provides a product recommendation method, apparatus, device, and medium to achieve efficient product recommendations to users, improve the efficiency and accuracy of product recommendations, thereby meeting the needs of personalized product recommendations and enhancing the user's service experience.
[0005] According to one aspect of the present invention, a product recommendation method is provided, the method comprising:
[0006] Based on the historical operation data of the target account, determine the target operation sequence corresponding to the target account;
[0007] The target operation sequence is input into a pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features.
[0008] Based on the output of the product recommendation model, the product recommendation result corresponding to the target account is determined.
[0009] According to another aspect of the present invention, a product recommendation device is provided, the device comprising:
[0010] The operation sequence determination module is used to determine the target operation sequence corresponding to the target account based on the historical operation data of the target account;
[0011] The product recommendation module is used to input the target operation sequence into a pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features.
[0012] The recommendation result determination module is used to determine the product recommendation result corresponding to the target account based on the output of the product recommendation model.
[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:
[0014] At least one processor; and
[0015] A memory communicatively connected to the at least one processor; wherein,
[0016] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the product recommendation method according to any embodiment of the present invention.
[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the product recommendation method described in any embodiment of the present invention.
[0018] The technical solution of this invention determines the target operation sequence corresponding to the target account based on the historical operation data of the target account, thereby preserving the dynamic evolution pattern of user behavior and providing input for subsequent models. The target operation sequence is input into a pre-trained product recommendation model for product recommendation. This product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer converts the target operation sequence into multiple target operation vectors. The hidden layer determines the product interaction features of the target account based on these vectors. The output layer recommends products to the target account based on these interaction features. This combines the static features of products with the dynamic behavior patterns of customers, thereby improving the accuracy and effectiveness of recommendations. Based on the output of the product recommendation model, the product recommendation result corresponding to the target account is determined, improving recommendation efficiency and accuracy, and providing personalized product recommendations for users. This invention captures the temporal dependencies between user operations through a product recommendation model, directly reflecting the intensity of user interests, achieving efficient product recommendations, improving the efficiency and accuracy of product recommendations, and thus satisfying personalized product recommendations for users and enhancing the user service experience.
[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a flowchart of a product recommendation method provided according to Embodiment 1 of the present invention;
[0022] Figure 2 This is a flowchart of a product recommendation method provided according to Embodiment 2 of the present invention;
[0023] Figure 3 This is a schematic diagram of a product recommendation device according to Embodiment 3 of the present invention;
[0024] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the product recommendation method of this invention. Detailed Implementation
[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0026] It should be noted that the terms "first," "second," "target," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0027] Example 1
[0028] Figure 1 This is a flowchart illustrating a product recommendation method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where product recommendations are made to users. The method can be executed by a product recommendation device, which can be implemented in hardware and / or software. This product recommendation device can be configured in an electronic device. Figure 1 As shown, the method includes:
[0029] S110. Based on the historical operation data of the target account, determine the target operation sequence corresponding to the target account.
[0030] The target account can refer to the account from which the product recommendation is to be made. Historical operation data can refer to the target account's behavior records on the application platform. For example, historical operation data can include browsing, clicking, adding to cart, purchasing, and favoriting data. The target operation sequence can refer to historical operation data sorted in chronological order and a fixed format.
[0031] Specifically, historical operation data of the target account on the corresponding operating platform can be obtained, and the obtained historical operation data can be preprocessed. For example, the preprocessing operation can be at least one of data cleaning and format processing, thereby obtaining the target operation sequence corresponding to the target account. This can preserve the dynamic evolution pattern of user behavior and provide input for subsequent time series models.
[0032] For example, S110 may include: determining the historical operation sequence corresponding to the target account based on the historical operation data of the target account; and determining the target operation sequence corresponding to the target account based on the historical operation sequence and a preset sequence length.
[0033] Here, the historical operation sequence can refer to the operation sequence obtained by converting historical operation data at different points in time according to a preset format. The preset sequence length can refer to the maximum length of the target operation sequence set in advance.
[0034] Specifically, the historical operation data of the target account is divided according to the time point (time stamp) of the operation, and the divided operation data is formatted according to a preset format to generate a historical operation sequence composed of operation data from multiple operation time points. The historical operation sequence is then filled or truncated according to a preset sequence length to generate the target operation sequence corresponding to the target account. This avoids information redundancy or insufficiency and provides input for subsequent time series models.
[0035] For example, the historical operation data of the target account can be cleaned and format converted. The resulting historical operation sequence is in the following format: [{1, 2020-01-01 12:00:00, Account ID, Bank Product, Product Risk Level, Amount},...{n, 2025-03-07 14:20:00, Account ID, Bank Product, Product Risk Level, Amount}]. Here, n can be the sequence length.
[0036] For example, based on the historical operation data of the target account, determining the historical operation sequence corresponding to the target account includes: dividing the historical operation data of the target account in chronological order to obtain multiple time-step operation data, wherein each time-step operation data includes the target operation characteristics of the target account at that point in time, and the target operation characteristics include: operation time, account identifier, product name, product risk level and product value information; sorting the multiple time-step operation data in chronological order to obtain the historical operation sequence corresponding to the target account.
[0037] In this context, time-step operation data refers to individual operation events derived from dividing the historical operation records of the target account in chronological order. Each event corresponds to an independent time step and includes all relevant details of the operation. Target operation features refer to key attributes extracted from the time-step operation data, used to describe the core characteristics and contextual information of user operations, and serve as the primary basis for model input.
[0038] Specifically, each operation time can be treated as an independent time step. Based on this time step, the historical operation data of the target account is divided sequentially from chronologically forward, generating multiple time-step operation data entries. Each time-step operation data entry corresponds to one operation record, avoiding information loss caused by fixed time windows. Each time-step operation data entry includes the target account's target operation characteristics at that point in time, including: operation time (timestamp), account identifier, product name (goods or service name), product risk level (predefined risk label), and product value information (such as price). By sorting multiple time-step operation data entries in chronological order, a historical operation sequence corresponding to the target account is obtained. This allows for prioritizing the most recent operations when processing the sequence, better aligning with the dynamic changes in user interests.
[0039] For example, determining the target operation sequence corresponding to the target account based on the historical operation sequence and a preset sequence length includes: in response to the sequence length corresponding to the historical operation sequence being less than a preset minimum sequence length, performing time-decay weighted interpolation on the historical operation sequence to determine the target operation sequence corresponding to the target account; in response to the sequence length corresponding to the historical operation sequence being greater than the preset minimum sequence length and less than the preset maximum sequence length, padding the historical operation sequence with null values to determine the target operation sequence corresponding to the target account; in response to the sequence length corresponding to the historical operation sequence being greater than the preset maximum sequence length, truncating the historical operation sequence according to the preset maximum sequence length in chronological order to determine the target operation sequence corresponding to the target account; and in response to the sequence length corresponding to the historical operation sequence being equal to the preset maximum sequence length, determining the historical operation sequence as the target operation sequence corresponding to the target account.
[0040] The preset minimum sequence length can refer to a pre-set threshold used to determine whether to perform data padding. The preset maximum sequence length can refer to the pre-set length of the target operation sequence.
[0041] Specifically, if the sequence length corresponding to a historical operation sequence is less than the preset minimum sequence length, decreasing weights are assigned based on the interval between the operation time and the current time. This performs time-decay weighted interpolation on the historical operation sequence to determine the target operation sequence corresponding to the target account, avoiding underfitting due to excessively short sequences. If the sequence length corresponding to a historical operation sequence is greater than the preset minimum sequence length but less than the preset maximum sequence length, null values are padded to the end of the historical operation sequence. This helps the model learn the correct temporal dependencies. If the sequence length corresponding to a historical operation sequence is greater than the preset maximum sequence length, the historical operation sequence is truncated chronologically from the latest to the earliest based on the preset maximum sequence length to determine the target operation sequence corresponding to the target account, better reflecting the user's current interests. If the sequence length corresponding to a historical operation sequence is equal to the preset maximum sequence length, the historical operation sequence is determined as the target operation sequence corresponding to the target account. By dynamically adjusting the sequence length, data sparsity and computational cost can be balanced.
[0042] S120. Input the target operation sequence into the pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features.
[0043] The product recommendation model can refer to a model built using a recurrent neural network (RNN) for recommending products to accounts, comprising an input layer, hidden layers, and an output layer. The input layer, which can be the first layer of the model, converts the target operation sequence into a model-readable vector representation. The hidden layers, which can be intermediate layers, extract temporal dependencies and user interests from the operation sequence. The output layer, which can be the last layer of the model, maps product interaction features to recommendation results. The target operation vector can be the vector representation obtained by reducing the dimensionality of the target operation sequence. Product interaction features can be user behavior preferences extracted from the target operation sequence by the hidden layer, reflecting the dynamic interaction between users and products.
[0044] Specifically, the target operation sequence is input into a pre-trained product recommendation model. The input layer of the product recommendation model receives the target operation sequence and converts it into multiple target operation vectors. The hidden layer of the product recommendation model analyzes the product preferences of the target account in the time series based on the multiple target operation vectors to determine the product interaction characteristics of the target account. The output layer of the product recommendation model calculates the recommendation probability of different products based on the product interaction characteristics, thereby realizing personalized product recommendations for the target account and improving the accuracy and effectiveness of the recommendations.
[0045] S130. Based on the output of the product recommendation model, determine the product recommendation results corresponding to the target account.
[0046] The product recommendation result can refer to the final recommendation list or probability distribution output by the model.
[0047] Specifically, based on the output of the product recommendation model, the product type or product name output by the model can be directly determined as the product recommendation result corresponding to the target account, thereby achieving automated and personalized product recommendations for the target account, improving recommendation efficiency, and enhancing user satisfaction.
[0048] For example, S130 may include: determining the product recommendation probability corresponding to different products based on the output of the product recommendation model, and determining the product with the highest product recommendation probability as the product recommendation result corresponding to the target account based on the product recommendation probability corresponding to each product.
[0049] Among them, the product recommendation probability can refer to the prediction value of the product recommendation model on the likelihood of the target account choosing a specific product. It is usually expressed in numerical form (between 0 and 1). The higher the value, the higher the model believes that the user has a higher interest or acceptance of the product.
[0050] Specifically, based on the output of the product recommendation model, the product recommendation probability corresponding to different products is determined according to the probability distribution of recommended products. The product recommendation probabilities corresponding to each product are then sorted, and the product with the highest recommendation probability is determined as the product recommendation result corresponding to the target account. This ensures that the recommendation result best matches the user's current behavior pattern and achieves personalized product recommendations for the target account.
[0051] In this embodiment, by using historical operation data of the target account, the target operation sequence corresponding to the target account is determined, thereby preserving the dynamic evolution pattern of user behavior and providing input for subsequent models. The target operation sequence is input into a pre-trained product recommendation model for product recommendation. This model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer converts the target operation sequence into multiple target operation vectors. The hidden layer determines the product interaction features of the target account based on these vectors. The output layer recommends products to the target account based on these interaction features. This combines the static features of products with the dynamic behavior patterns of customers, thereby improving the accuracy and effectiveness of recommendations. Based on the output of the product recommendation model, the product recommendation result corresponding to the target account is determined, improving recommendation efficiency and accuracy, and providing personalized product recommendations for users. This invention captures the temporal dependencies between user operations through a product recommendation model, directly reflecting the intensity of user interests, achieving efficient product recommendations, improving the efficiency and accuracy of product recommendations, and thus satisfying personalized product recommendations for users and enhancing the user service experience.
[0052] It should be noted that in this embodiment, when the target account corresponds to a bank user, it is important to consider that bank customers have different risk preferences for bank products due to factors such as age, occupation, region, income level, asset status, and debt situation. Furthermore, compared to general products, the risk level of bank products has a greater impact on customer behavior. Therefore, the user's risk preference characteristics and the risk level of the bank product can be added to the weight matrix of the product recommendation model to influence the recommendation result at the next time step. Specifically, the user's risk preference characteristics are the average risk level of the products purchased by the user, and this average can be used as a bias term input into the weight matrix of the hidden layer. This bias can be multiplied by the product risk level of the current behavior, amplifying the impact of the current behavior's product risk characteristics on the product.
[0053] Example 2
[0054] Figure 2 This is a flowchart of a product recommendation method provided in Embodiment 2 of the present invention. Based on the above embodiments, this embodiment optimizes the step of "inputting the target operation sequence into a pre-trained product recommendation model for product recommendation". Explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.
[0055] See Figure 2 Another product recommendation method provided in this embodiment specifically includes the following steps:
[0056] S210. Based on the historical operation data of the target account, determine the target operation sequence corresponding to the target account.
[0057] S220. Input the target operation sequence into the input layer for feature fusion encoding to generate multiple target operation vectors.
[0058] Specifically, the input layer maps discrete features such as operation type (e.g., click, add to cart) and product name for each row in the target operation sequence into low-dimensional dense vectors according to time sequence. It then normalizes continuous features such as timestamps and browsing durations for each row. Finally, it concatenates the low-dimensional dense vectors corresponding to the discrete features with the normalized continuous features to obtain a multi-dimensional operation vector, i.e., the target operation vector. By concatenating features such as operation type, product attributes, and time, the contextual information of the target account's behavior can be comprehensively represented, which is helpful for subsequent analysis of user product preference characteristics.
[0059] For example, the input layer includes a data transformation unit and a data augmentation unit; S220 may include: inputting the target operation sequence into the data transformation unit for feature fusion encoding to generate multiple candidate operation vectors; inputting the multiple candidate operation vectors into the data augmentation unit in chronological order to add a time decay factor to generate multiple target operation vectors.
[0060] Here, the candidate operation vector refers to the intermediate representation generated by the data transformation unit after feature fusion encoding of the target operation sequence. It is a multi-dimensional numerical vector that integrates various feature information of the user operation. The time decay factor refers to a weight coefficient calculated based on the interval between the operation time and the current time. It is used to adjust the importance of operations at different times in the candidate operation vector, with more recent operations having higher weights and older operations having lower weights.
[0061] Specifically, the target operation sequence is input into the data transformation unit. The data transformation unit compresses the high-dimensional sparse discrete features (such as product names) in the target operation sequence into low-dimensional dense vectors, while preserving semantic relationships (such as similar products having closer embedding vectors). It also generates multiple candidate operation vectors by concatenating features such as operation type, product attributes, and time, comprehensively representing the contextual information of user behavior. These candidate operation vectors are then input into the data augmentation unit in chronological order. The data augmentation unit calculates a time decay factor based on the interval between the timestamp of each candidate operation vector and the current time, and multiplies the candidate operation vector by the corresponding time decay factor to generate the target operation vector. This reduces the weight of older operations (such as browsing behavior from 7 days ago) and highlights the impact of recent behaviors (such as adding items to the cart yesterday) on product recommendations.
[0062] S230. Input multiple target operation vectors sequentially into the hidden layer in chronological order to extract the temporal dependency of account operations and generate product interaction features of the target account.
[0063] Specifically, multiple target operation vectors are input into the hidden layer in chronological order. The hidden layer processes the vector sequence composed of multiple target operation vectors using a Long Short-Term Memory (LSTM) network to determine the final hidden state corresponding to the target account, i.e., the product interaction features. This can achieve automatic learning of the abstract representation of user interests and reflect the temporal dependence of the entire target operation sequence.
[0064] It should be noted that, considering the vanishing and exploding gradient problems of recurrent neural networks, Long Short-Term Memory (LSTM) networks can be improved using gated recurrent units. LSTM controls the flow of information through input, forget, and output gates, while gated recurrent units simplify the structure of LSTM through update and reset gates.
[0065] S240. Input the product interaction features into the output layer to make product recommendations and generate product recommendation probability distribution information.
[0066] Specifically, product interaction features are input into the output layer, which performs a fully connected mapping on the product interaction features and generates recommendation probability distributions for different products based on the mapping results. This generates product recommendation probability distribution information, which directly reflects the user's preference strength for each product.
[0067] S250. Based on the output of the product recommendation model, determine the product recommendation results corresponding to the target account.
[0068] The technical solution of this embodiment is successful. This invention generates multiple target operation vectors by inputting the target operation sequence into the input layer for feature fusion encoding. This reduces the dimensionality of the target operation sequence while preserving semantic relationships, achieving multimodal data fusion and comprehensively representing the contextual information of user behavior. The multiple target operation vectors are then sequentially input into the hidden layer in chronological order to extract the temporal dependencies of account operations, generating product interaction features for the target account. This allows for the automatic learning of abstract representations of user interests and the differentiation of user product preferences based on temporal sequence. The product interaction features are then input into the output layer for product recommendation, generating product recommendation probability distribution information that directly reflects the user's preference strength for each product. This embodiment, through the collaborative mechanism of the input layer, hidden layer, and output layer, achieves the capture of the dynamic evolution of user behavior over time using a recurrent neural network-like structure, enabling personalized product recommendations and improving user satisfaction.
[0069] Example 3
[0070] Figure 3 This is a schematic diagram of a product recommendation device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: an operation sequence determination module 310, a product recommendation module 320, and a recommendation result determination module 330;
[0071] The operation sequence determination module 310 is used to determine the target operation sequence corresponding to the target account based on the historical operation data of the target account.
[0072] Product recommendation module 320 is used to input the target operation sequence into a pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features.
[0073] The recommendation result determination module 330 is used to determine the product recommendation result corresponding to the target account based on the output of the product recommendation model.
[0074] In this embodiment, by using historical operation data of the target account, the target operation sequence corresponding to the target account is determined, thereby preserving the dynamic evolution pattern of user behavior and providing input for subsequent models. The target operation sequence is input into a pre-trained product recommendation model for product recommendation. This product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer converts the target operation sequence into multiple target operation vectors. The hidden layer determines the product interaction features of the target account based on these vectors. The output layer recommends products to the target account based on these interaction features. This combines the static features of products with the dynamic behavior patterns of customers, thereby improving the accuracy and effectiveness of recommendations. Based on the output of the product recommendation model, the product recommendation result corresponding to the target account is determined, improving recommendation efficiency and accuracy, and providing personalized product recommendations for users. This invention captures the temporal dependencies between user operations through a product recommendation model, directly reflecting the intensity of user interests, achieving efficient product recommendations, improving the efficiency and accuracy of product recommendations, thus satisfying personalized product recommendations for users and enhancing the user service experience.
[0075] Optionally, the operation sequence determination module 310 includes:
[0076] The historical sequence determination unit is used to determine the historical operation sequence corresponding to the target account based on the historical operation data of the target account;
[0077] The target sequence determination unit is used to determine the target operation sequence corresponding to the target account based on the historical operation sequence and the preset sequence length.
[0078] Optionally, the historical sequence determination unit is specifically used to: divide the historical operation data of the target account in chronological order to obtain multiple time-step operation data, wherein each time-step operation data includes the target operation characteristics of the target account at that point in time, and the target operation characteristics include: operation time, account identifier, product name, product risk level, and product value information; and sort the multiple time-step operation data in chronological order to obtain the historical operation sequence corresponding to the target account.
[0079] Optionally, the target sequence determination unit is specifically configured to: in response to the sequence length corresponding to the historical operation sequence being less than a preset minimum sequence length, perform time-decay weighted interpolation on the historical operation sequence to determine the target operation sequence corresponding to the target account; in response to the sequence length corresponding to the historical operation sequence being greater than the preset minimum sequence length and less than the preset maximum sequence length, fill the historical operation sequence with null values to determine the target operation sequence corresponding to the target account; in response to the sequence length corresponding to the historical operation sequence being greater than the preset maximum sequence length, truncate the historical operation sequence according to the preset maximum sequence length in chronological order to determine the target operation sequence corresponding to the target account; and in response to the sequence length corresponding to the historical operation sequence being equal to the preset maximum sequence length, determine the historical operation sequence as the target operation sequence corresponding to the target account.
[0080] Optionally, the product recommendation module 320 includes:
[0081] An input unit is used to input the target operation sequence into the input layer for feature fusion encoding to generate multiple target operation vectors;
[0082] The hidden unit is used to input multiple target operation vectors into the hidden layer in chronological order to extract the temporal dependency of account operations and generate product interaction features of the target account.
[0083] The output unit is used to input the product interaction features into the output layer for product recommendation and generate product recommendation probability distribution information.
[0084] Optionally, the input layer includes a data conversion unit and a data augmentation unit; the input unit is specifically used to: input the target operation sequence into the data conversion unit for feature fusion encoding to generate multiple candidate operation vectors; and input the multiple candidate operation vectors into the data augmentation unit in chronological order to add a time decay factor to generate multiple target operation vectors.
[0085] Optionally, the recommendation result determination module is specifically used to: determine the product recommendation probability corresponding to different products based on the output of the product recommendation model, and determine the product with the highest product recommendation probability as the product recommendation result corresponding to the target account based on the product recommendation probability corresponding to each product.
[0086] The above-described apparatus can execute the product recommendation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the product recommendation method.
[0087] Example 4
[0088] Figure 4 This is a schematic diagram of the structure of an electronic device implementing the product recommendation method of an embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0089] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0090] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0091] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as product recommendation methods.
[0092] In some embodiments, the product recommendation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the product recommendation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the product recommendation method by any other suitable means (e.g., by means of firmware).
[0093] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the product recommendation method of the embodiments of the present invention.
[0094] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0095] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.
[0096] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0097] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0098] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0099] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0100] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0101] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A product recommendation method, characterized in that, include: Based on the historical operation data of the target account, determine the target operation sequence corresponding to the target account; The target operation sequence is input into a pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features. Based on the output of the product recommendation model, the product recommendation result corresponding to the target account is determined.
2. The method according to claim 1, characterized in that, The step of determining the target operation sequence corresponding to the target account based on the target account's historical operation data includes: Based on the historical operation data of the target account, determine the historical operation sequence corresponding to the target account; Based on the historical operation sequence and the preset sequence length, the target operation sequence corresponding to the target account is determined.
3. The method according to claim 2, characterized in that, The step of determining the historical operation sequence corresponding to the target account based on the target account's historical operation data includes: The historical operation data of the target account is divided into multiple time step operation data in chronological order. Each time step operation data includes the target operation characteristics of the target account at that point in time. The target operation characteristics include: operation time, account identifier, product name, product risk level, and product value information. The multiple time-step operation data are sorted in reverse chronological order to obtain the historical operation sequence corresponding to the target account.
4. The method according to claim 2, characterized in that, The step of determining the target operation sequence corresponding to the target account based on the historical operation sequence and the preset sequence length includes: In response to the fact that the sequence length corresponding to the historical operation sequence is less than the preset minimum sequence length, time decay weighted interpolation is performed on the historical operation sequence to determine the target operation sequence corresponding to the target account; In response to the fact that the sequence length corresponding to the historical operation sequence is greater than a preset minimum sequence length and less than a preset maximum sequence length, the historical operation sequence is padded with null values to determine the target operation sequence corresponding to the target account. In response to the fact that the sequence length corresponding to the historical operation sequence is greater than the preset maximum sequence length, the historical operation sequence is truncated according to the preset maximum sequence length in chronological order from back to front to determine the target operation sequence corresponding to the target account; In response to the historical operation sequence having a sequence length equal to a preset maximum sequence length, the historical operation sequence is determined as the target operation sequence corresponding to the target account.
5. The method according to claim 1, characterized in that, The step of inputting the target operation sequence into a pre-trained product recommendation model for product recommendation includes: The target operation sequence is input into the input layer for feature fusion encoding to generate multiple target operation vectors; Multiple target operation vectors are sequentially input into the hidden layer in chronological order to extract the temporal dependency of account operations and generate product interaction features of the target account. The product interaction features are input into the output layer for product recommendation, generating product recommendation probability distribution information.
6. The method according to claim 5, characterized in that, The input layer includes: a data conversion unit and a data enhancement unit; The step of inputting the target operation sequence into the input layer for feature fusion encoding to generate multiple target operation vectors includes: The target operation sequence is input into the data conversion unit for feature fusion encoding to generate multiple candidate operation vectors; Multiple candidate operation vectors are input into the data augmentation unit in chronological order from front to back, and a time decay factor is added to generate multiple target operation vectors.
7. The method according to claim 1, characterized in that, The step of determining the product recommendation result corresponding to the target account based on the output of the product recommendation model includes: Based on the output of the product recommendation model, the product recommendation probability corresponding to different products is determined, and according to the product recommendation probability corresponding to each product, the product with the highest product recommendation probability is determined as the product recommendation result corresponding to the target account.
8. A product recommendation device, characterized in that, include: The operation sequence determination module is used to determine the target operation sequence corresponding to the target account based on the historical operation data of the target account; The product recommendation module is used to input the target operation sequence into a pre-trained product recommendation model for product recommendation. The product recommendation model is trained based on a recurrent neural network and includes an input layer, a hidden layer, and an output layer. The input layer is used to convert the target operation sequence into multiple target operation vectors. The hidden layer is used to determine the product interaction features of the target account based on the multiple target operation vectors. The output layer is used to recommend products to the target account based on the product interaction features. The recommendation result determination module is used to determine the product recommendation result corresponding to the target account based on the output of the product recommendation model.
9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the product recommendation method according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the product recommendation method according to any one of claims 1-7.