Commodity recommendation method and device, electronic equipment and storage medium

Abstract

The application discloses a commodity recommendation method and device, electronic equipment and storage medium. The method comprises the following steps: determining a user historical behavior sequence according to collected purchase behaviors of a buried point in a preset time period; matching historical commodity attribute information corresponding to the user historical behavior sequence according to a static attribute set of a commodity, and forming a commodity historical purchase sequence based on the historical commodity attribute information; determining a preliminary candidate set based on the commodity historical purchase sequence; inputting the commodity historical purchase sequence into a trained consumption capacity model to predict a consumption level of the user in a category; and filtering the preliminary candidate set based on the consumption level to obtain a recommended candidate set. Through the technical scheme of the application, the user's consumption level portrait can be accurately obtained by fully and effectively utilizing the user's purchase behavior according to the user's behavior sequence, and a commodity recommendation decision can be made in combination with an actual scene and a purchase behavior, so that the accuracy and effectiveness of commodity recommendation are improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence recommendation, and more particularly to a method, apparatus, electronic device, and storage medium for recommending products. Background Technology

[0002] With the rapid development of e-commerce and recommendation technology, many researchers have gradually focused their attention on how to fully and effectively utilize purchase behavior to generate recommendation candidate sets, because purchase behavior, compared with other behaviors, can truly and effectively represent the user's intent.

[0003] In existing technologies, product recommendations are typically made to users by profiling their spending habits based on their purchasing behavior or by generating a set of recommended products based on their purchasing behavior.

[0004] In the process of realizing this invention, the inventors discovered the following problems with the prior art: the prior art cannot fully and effectively utilize users' purchasing behavior to obtain a more accurate profile of users' consumption levels, and it cannot make product recommendation decisions by combining actual scenarios and purchasing behavior. Summary of the Invention

[0005] This invention provides a product recommendation method, apparatus, electronic device, and storage medium, which can solve the problems in the prior art that cannot fully and effectively utilize users' purchasing behavior to obtain a more accurate profile of users' consumption levels, and cannot make product recommendation decisions based on actual scenarios and purchasing behavior.

[0006] In a first aspect, embodiments of the present invention provide a method for recommending goods, the method comprising:

[0007] The user's historical behavior sequence is determined based on the purchase behavior data collected within a preset time period.

[0008] The system matches historical product attribute information corresponding to the user's historical behavior sequence with the static attribute set of the product, and forms a historical purchase sequence of the product based on the historical product attribute information; wherein each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price;

[0009] A preliminary candidate set is determined based on the product's historical purchase sequence; the product's historical purchase sequence is then input into a trained consumption capacity model to predict the user's consumption level in the category;

[0010] The preliminary candidate set is filtered based on the consumption level to obtain the recommended candidate set.

[0011] Secondly, embodiments of the present invention provide a product recommendation device, the device comprising:

[0012] The behavior sequence determination module is used to determine the user's historical behavior sequence based on the purchased behavior data collected within a preset time period.

[0013] The purchase sequence determination module is used to match historical product attribute information corresponding to the user's historical behavior sequence according to the static attribute set of the product, and form a historical purchase sequence of the product based on the historical product attribute information; wherein, each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price;

[0014] A preliminary candidate set generation module is used to determine a preliminary candidate set based on the historical purchase sequence of the goods.

[0015] The consumption level prediction module is used to input the historical purchase sequence of the goods into the trained consumption capacity model to predict the user's consumption level in the category.

[0016] The recommended candidate set generation module is used to filter the preliminary candidate set based on the consumption level to obtain the recommended candidate set.

[0017] Thirdly, embodiments of the present invention provide an electronic device, the electronic device comprising:

[0018] At least one processor; and

[0019] A memory communicatively connected to the at least one processor; wherein,

[0020] 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 a method for recommending a product according to any embodiment of the present invention.

[0021] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the product recommendation method described in any embodiment of the present invention.

[0022] The technical solution of this invention, by determining the user's historical behavior sequence, forms a historical purchase sequence of goods based on historical product attribute information. A preliminary candidate set is obtained based on this historical purchase sequence. This sequence is then input into a trained consumption capacity model to predict the user's consumption level within a category. Finally, the preliminary candidate set is filtered based on the consumption level to obtain a recommended candidate set. Through this technical solution, the user's consumption level profile can be more accurately obtained by fully and effectively utilizing the user's purchase behavior based on the user's behavior sequence. Combining this with the actual scenario and purchase behavior, product recommendation decisions are made, improving the accuracy and effectiveness of product recommendations.

[0023] 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

[0024] 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.

[0025] Figure 1 This is a flowchart of a product recommendation method provided according to Embodiment 1 of the present invention;

[0026] Figure 2 This is a flowchart of a product recommendation method provided according to Embodiment 2 of the present invention;

[0027] Figure 3 This is a schematic diagram of the structure of a product recommendation device according to Embodiment 3 of the present invention;

[0028] 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

[0029] 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. 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.

[0030] It should be noted that the terms "first," "second," etc., 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 the 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.

[0031] With the rapid development of e-commerce and recommendation technology, many researchers are increasingly focusing on how to effectively utilize purchase behavior to generate recommendation candidate sets, as purchase behavior, compared to other behaviors, can truly and effectively represent user intent. Currently, the field of product recommendation often faces the following challenging problems:

[0032] 1. It is impossible to fully and effectively utilize users' purchasing behavior to obtain a more accurate profile of users' consumption levels.

[0033] Purchasing power is a crucial aspect of user profiling in e-commerce. Accurately identifying a user's purchasing power and recommending products within their budget that align with their preferences can significantly improve conversion rates. The most direct and effective way to assess a user's purchasing power is through their purchasing behavior. A common approach is to categorize users based on their purchase price using a certain method to calculate their spending level. However, this method has a drawback. If user A spends 6000 yuan on a mobile phone, while user B spends 2000 yuan on a mouse, user B's spending level might seem lower than user A's. However, common sense dictates that user A's 6000 yuan purchase is within a normal spending range, while user B's 2000 yuan purchase of a mouse indicates a higher spending level. Therefore, effectively utilizing user purchasing behavior to more accurately profile their spending power is of great significance.

[0034] 2. Unable to make product recommendation decisions based on actual scenarios and purchasing behavior.

[0035] In vertical e-commerce, after a user places an order, the recommendation system focuses on generating a candidate set of recommendations. Generally, there are two strategies: Strategy A: recommending similar products to the purchased item; Strategy B: recommending complementary products to the purchased item. For example, if user A buys a television, they are unlikely to buy another television in the short term, so we should recommend complementary products. If user B buys yogurt, and the probability of user B buying yogurt again in the short term is relatively high, we should recommend similar products to improve conversion rates. In practical e-commerce, making product recommendation decisions based on the actual scenario and purchasing behavior is a crucial step.

[0036] Example 1

[0037] 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 in the e-commerce field where product recommendations need to be made to users. This method can be executed by a product recommendation device, which can be implemented in hardware and / or software. The product recommendation device can be configured in a terminal or server with product recommendation functionality. Figure 1 As shown, the method includes:

[0038] S110. Determine the user's historical behavior sequence based on the purchased behavior data collected within the preset time period.

[0039] Wherein, the time period is a preset time length; in this embodiment, for example, the time period can be set to one year; furthermore, if the number of tracking purchase behaviors included within the time period is insufficient to determine the user's historical behavior sequence, then an appropriate adjustment is made to increase the time range of the time period.

[0040] The event tracking points can be defined as technologies and implementation processes for capturing, processing, and transmitting specific user behaviors or events. Furthermore, these event tracking points can be used to collect user behavior data. For example, to understand which buttons a user clicked, which pages they viewed, and what actions they took within an app, event tracking points can be used.

[0041] Furthermore, the aforementioned tracking purchase behavior can be: the user's purchase behavior collected by the system through tracking operations.

[0042] The user's historical behavior sequence includes: product identifiers of the products purchased by the user within a preset time period through tracking purchase behavior.

[0043] Optionally, the user's historical behavior sequence is determined based on the collected purchase behavior within a preset time period, including: sorting the product identification information of the user's purchase behavior in ascending order according to the user's purchase time; and outputting the sorted product identification information to obtain the user behavior sequence information.

[0044] Specifically, the ascending order is as follows: in the user's historical purchase sequence, the product identifier whose purchase time is furthest from the current system time is sorted at the beginning of the sequence, and the product identifier whose purchase time is closer to the current system time is sorted at the end of the sequence.

[0045] S120. Match historical product attribute information corresponding to the user's historical behavior sequence according to the static attribute set of the product, and form a historical purchase sequence of the product based on the historical product attribute information.

[0046] Each element in the product history purchase sequence includes the historical product identifier, historical product category, and historical product price.

[0047] The static attribute set of a product includes: the product identifier of all products in the system, as well as the product category and product price corresponding to the product identifier.

[0048] In an embodiment of the present invention, optionally, before matching historical product attribute information corresponding to the user's historical behavior sequence according to the static attribute set of the product, and forming a historical purchase sequence of the product based on the historical product attribute information, the method further includes: extracting historical product categories from the historical product attribute information based on the historical product attribute information corresponding to the user behavior sequence; finding the product information of all products included in the historical product categories to form a category information sequence; arranging the products in the category information sequence in descending order according to price, and dividing the sorted category information sequence to obtain n category buckets.

[0049] In this embodiment, specifically, after extracting the historical product categories from the historical product attribute information, all products with the same historical product category are formed into a category information sequence. Then, the product information corresponding to the products included in the category information sequence is sorted. Further, the products in the category information sequence are arranged in descending order of price, that is, the product with the highest price is placed at the beginning of the sequence. Finally, the sorted category information sequence is divided into n category bins. The division rules for the category bins can be preset manually. For example, if the division rule is set to set a bin every 200, then in the category information sequence, products with prices between 0 and 200 form the first category bin, products between 200 and 400 form the second bin, and so on, until the price of the product with the highest price in the category information sequence is also included in the bin range, at which point the division ends.

[0050] S130. Determine a preliminary candidate set based on the historical purchase sequence of the goods.

[0051] The preliminary candidate set includes: a set consisting of product identifiers of products that users are predicted to buy based on historical purchase sequences, product categories that match the product identifiers, and a mapping of product prices.

[0052] Optionally, obtaining a preliminary candidate set based on the product historical purchase sequence includes: converting the elements in the product historical purchase sequence into target vectors; inputting the target vectors into a preliminary candidate model, training the preliminary candidate model to obtain a trained preliminary candidate model; and inputting the user historical behavior sequence into the trained preliminary candidate model to obtain a preliminary candidate set of recommended products.

[0053] In this embodiment, the elements in the product history purchase sequence are processed by embedding to transform the elements into a target vector; wherein the target vector is a low-dimensional dense vector.

[0054] Furthermore, the embedding method can represent an object using a low-dimensional vector; furthermore, in this embodiment, the embedding method can represent an element in a product's historical purchase sequence using a low-dimensional vector.

[0055] Furthermore, the low-dimensional dense vector can be a dense matrix vector; for example, if the product historical purchase sequence includes 512 elements, then 512 fixed-dimensional vectors can be formed; furthermore, the low-dimensional dense vector can ensure that the target vector will not contain zero vectors or vectors with zero-dimensional values, thus ensuring the correctness of the calculation.

[0056] Optionally, the target vector is input into a preliminary candidate model, and the preliminary candidate model is trained to obtain a trained preliminary candidate model, including: selecting associated target vectors to form a target vector group; using a first part of the target vector group as the input of the preliminary candidate model, using a second part of the target vector group as the output of the preliminary candidate model, and training the preliminary candidate model to obtain a trained preliminary candidate model; wherein the purchase time corresponding to the first part of the vector is earlier than the purchase time corresponding to the second part of the vector.

[0057] Specifically, when determining the associated target vectors, the distance between the target vectors is calculated, target vectors that meet the preset distance requirements are selected as a group of associated target vectors, and a target vector group is formed.

[0058] In this embodiment, the vectors in the target vector group can be divided into a first part vector and a second part vector according to a preset rule. For example, the preset rule can be: starting from the first vector of the target vector group, setting the first 80% of the target vector group as the first part vector and the remaining 20% ​​as the second part vector, and training the preliminary candidate model to obtain the model parameters required by the preliminary candidate model, that is, obtaining the trained preliminary candidate model.

[0059] The initial candidate model can be a transformer model; further, the transformer model is a neural network that can learn context and thus meaning by tracking relationships in sequence data; further, using the transformer model effectively generates a recommendation candidate set based on the purchase sequence and product information, thereby improving the accuracy and effectiveness of product recommendations.

[0060] In this embodiment, specifically, the preliminary candidate model can predict the user's subsequent purchasing behavior by tracking the product's historical purchase sequence.

[0061] S140. Input the historical purchase sequence of the goods into the trained consumption capacity model to predict the user's consumption level in the category.

[0062] In this embodiment, optionally, before inputting the historical purchase sequence of goods into the trained consumption capacity model to predict the user's consumption level in the category, the method further includes: dividing each category into corresponding consumption levels based on price ranges; labeling the historical purchase sequences of goods in the training set with the consumption levels corresponding to the categories; inputting the labeled historical purchase sequences of goods into the consumption capacity model to train the consumption capacity model and obtain the trained consumption capacity model.

[0063] Specifically, a supervised training model is selected, and each category is divided into different consumption levels based on preset classification rules. The price range of the corresponding product in the product's historical purchase sequence is matched to determine the user's consumption level in that category. The consumption level is used as a label to annotate the information in the product's historical purchase sequence. Finally, the annotated product historical purchase sequence is input into the consumption capacity model to obtain the various model parameters in the supervised training model, thus completing the training of the consumption capacity model and obtaining the trained consumption capacity model.

[0064] It is easy to understand that users have different levels of spending power in different product categories. This embodiment uses a supervised training model to predict the price level of users' price preferences in different product categories.

[0065] In this embodiment of the invention, the historical purchase sequence of goods can be input into a trained consumption capacity model to output the user's consumption level in the category.

[0066] For example, if the user's historical purchase sequence includes purchases of mice for 500 yuan and 2000 yuan, then after inputting the historical purchase sequence into the trained consumption capacity model, the model can search for all price buckets under the mouse category according to the category buckets divided in step S120. If the model identifies the price of 500 yuan as falling into the price bucket of 0 to 1500 yuan, and the model identifies the price of 2000 yuan as falling into the price bucket of 1500 to 3000 yuan, then the predicted consumption level of the user in the mouse category is 0 to 3000.

[0067] S150. Filter the preliminary candidate set based on the consumption level to obtain the recommended candidate set.

[0068] The initial candidate set includes: a set consisting of product identifiers that users are predicted to buy based on the product's historical purchase sequence, product categories that match the product identifiers, and product prices; further, based on the product prices matched by the product identifiers and the consumption level, products that do not belong to the consumption level are eliminated, and only products that belong to the consumption level are retained, forming a recommendation candidate set.

[0069] The recommended candidate set includes: product identifiers, product prices, and product categories of products belonging to the aforementioned consumption level.

[0070] In an embodiment of the present invention, optionally, filtering the preliminary candidate set based on the consumption level to obtain a recommended candidate set includes: matching the consumption level with the n category buckets to obtain a category bucket corresponding to the consumption level, which is used as a target category bucket; extracting the category information sequence from the target category bucket, and filtering the preliminary candidate set based on the category information sequence to obtain a recommended candidate set.

[0071] In this embodiment, based on the product price matched by the product identifier, the category bucket corresponding to the consumption level is matched based on the consumption level, and the category information sequence in the target category bucket is extracted. Products that do not belong to the category information sequence are removed, and only products that belong to the category information sequence are retained to form a recommendation candidate set.

[0072] In this embodiment, predicting a user's spending level in a certain category through price sharding can more accurately assess the user's different spending levels in different categories. This avoids constructing a user's spending level based on a single price factor, effectively combining the user's spending level with the category, and preventing the use of a single price to judge the user's spending ability.

[0073] The technical solution of this invention, by determining the user's historical behavior sequence, forms a historical purchase sequence of goods based on historical product attribute information. A preliminary candidate set is obtained based on this historical purchase sequence. This sequence is then input into a trained consumption capacity model to predict the user's consumption level within a category. Finally, the preliminary candidate set is filtered based on the consumption level to obtain a recommended candidate set. Through this technical solution, the user's consumption level profile can be more accurately obtained by fully and effectively utilizing the user's purchase behavior based on the user's behavior sequence. Combining this with the actual scenario and purchase behavior, product recommendation decisions are made, improving the accuracy and effectiveness of product recommendations.

[0074] Example 2

[0075] Figure 2 This is a flowchart of a product recommendation method provided in Embodiment 2 of the present invention.

[0076] Correspondingly, such as Figure 2 As shown, the method includes:

[0077] S210. Determine the user's historical behavior sequence based on the purchased behavior data collected within the preset time period.

[0078] S220. Match historical product attribute information corresponding to the user's historical behavior sequence according to the static attribute set of the product, and form a historical purchase sequence of the product based on the historical product attribute information.

[0079] Each element in the product history purchase sequence includes the historical product identifier, historical product category, and historical product price.

[0080] S230. Transform the elements in the product history purchase sequence into a target vector.

[0081] S240. Input the target vector into the preliminary candidate model, train the preliminary candidate model, and obtain the trained preliminary candidate model.

[0082] S250. Input the user's historical behavior sequence into the trained preliminary candidate model to obtain a preliminary candidate set of recommended products.

[0083] S260. Input the historical purchase sequence of the goods into the trained consumption capacity model to predict the user's consumption level in the category.

[0084] S270. Filter the preliminary candidate set based on the consumption level to obtain the recommended candidate set.

[0085] The technical solution of this invention, by determining the user's historical behavior sequence, forms a historical purchase sequence of goods based on historical product attribute information. Elements in the historical purchase sequence are transformed into target vectors, which are then input into a preliminary candidate model to obtain a trained preliminary candidate model. After inputting the user's historical behavior sequence into the trained preliminary candidate model to obtain a preliminary candidate set of recommended goods, the historical purchase sequence is input into a trained consumption capacity model to predict the user's consumption level in a category. Finally, the preliminary candidate set is filtered based on the consumption level to obtain a recommended candidate set. Through this technical solution, the user's consumption level profile can be more accurately obtained by fully and effectively utilizing the user's purchase behavior based on the user's behavior sequence. Combining actual scenarios and purchase behavior, product recommendation decisions are made, improving the accuracy and effectiveness of product recommendations.

[0086] Specific application scenarios

[0087] To more clearly illustrate the technical solutions provided in the embodiments of the present invention, this embodiment provides a specific implementation method as follows:

[0088] The data preparation for implementing this specific embodiment is as follows: purchase behavior data collected within a preset time period, including user identifier, purchase time and purchased product identifier, material information, including product identifier, category identifier and price, and features for building the model, etc.

[0089] Furthermore, the feature-based method for constructing the model provided in this embodiment includes: the Transformer method.

[0090] The specific steps are as follows:

[0091] 1. First, filter the purchase behavior based on the user's preset time period and use the data within a certain time range for subsequent operations. For example, one year's worth of order data can be used for relevant calculations. If the amount of purchased data is insufficient to support the data, the time range can be appropriately increased.

[0092] 2. Calculate aggregated results based on user identifiers and sorted in ascending order by purchase time.

[0093] 3. Based on the item-category mapping relationship, generate a product purchase history sequence.

[0094] 4. Based on the generated product historical purchase sequence, use embedding technology to generate a low-dimensional dense vector, which is then used as input for the transformer technology.

[0095] 5. Techniques for training the transformer to obtain the transformer model parameters.

[0096] 6. Use the transformer model parameters to generate a certain number of preliminary candidate sets, which will be used as input for step 9.

[0097] 7. Using statistical methods, price ranges are established for each category, and corresponding price levels are defined.

[0098] 8. Based on the user's order behavior, construct a feature model. The label of the feature model is the price level generated in step 7. This model is used to predict the user's price level in each category, thereby obtaining the price range in that category.

[0099] 9. Take the preliminary recommendation candidate set obtained by the transformer technology and combine it with the user's consumption price range in the category, and perform post-processing logic. That is, assuming that the user's candidate set is product A, and its category is CateID, use the model obtained in step 8 to predict the maximum price level that the user is willing to pay in the category CateID. Find the price range according to the category-price level dictionary, and determine whether product A matches the obtained price range. If it matches, keep the recommendation result; otherwise, filter the recommendation result, and finally obtain the recommendation candidate set.

[0100] Example 3

[0101] Figure 3This is a schematic diagram of a product recommendation device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes:

[0102] The behavior sequence determination module 310 is used to determine the user's historical behavior sequence based on the purchased behavior within a preset time period.

[0103] The purchase sequence determination module 320 is used to match historical product attribute information corresponding to the user's historical behavior sequence according to the static attribute set of the product, and form a historical purchase sequence of the product based on the historical product attribute information; wherein, each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price;

[0104] The preliminary candidate set generation module 330 is used to determine a preliminary candidate set based on the historical purchase sequence of the goods.

[0105] The consumption level prediction module 340 is used to input the historical purchase sequence of the goods into the trained consumption capacity model to predict the user's consumption level in the category.

[0106] The recommended candidate set generation module 350 is used to filter the preliminary candidate set based on the consumption level to obtain the recommended candidate set.

[0107] The technical solution of this invention, by determining the user's historical behavior sequence, forms a historical purchase sequence of goods based on historical product attribute information. A preliminary candidate set is obtained based on this historical purchase sequence. This sequence is then input into a trained consumption capacity model to predict the user's consumption level within a category. Finally, the preliminary candidate set is filtered based on the consumption level to obtain a recommended candidate set. Through this technical solution, the user's consumption level profile can be more accurately obtained by fully and effectively utilizing the user's purchase behavior based on the user's behavior sequence. Combining this with the actual scenario and purchase behavior, product recommendation decisions are made, improving the accuracy and effectiveness of product recommendations.

[0108] Based on the above embodiments, the preliminary candidate set generation module 330 may include:

[0109] A vector transformation unit is used to transform elements in the product historical purchase sequence into a target vector.

[0110] The model training unit is used to input the target vector into the preliminary candidate model, train the preliminary candidate model, and obtain the trained preliminary candidate model.

[0111] The preliminary candidate set acquisition unit is used to input the user's historical behavior sequence into the trained preliminary candidate model to obtain a preliminary candidate set of recommended products.

[0112] Based on the above embodiments, the model training unit further includes:

[0113] The vector group generation unit is used to select associated target vectors to form a target vector group.

[0114] The preliminary candidate model training unit is used to train the preliminary candidate model by taking the first part of the target vector group as the input of the preliminary candidate model and the second part of the target vector group as the output of the preliminary candidate model, thereby obtaining the trained preliminary candidate model; wherein the purchase time corresponding to the first part of the vector is earlier than the purchase time corresponding to the second part of the vector.

[0115] Based on the above embodiments, the purchase sequence determination module 320 may include:

[0116] The product category extraction unit is used to extract the historical product category from the historical product attribute information based on the historical product attribute information corresponding to the user behavior sequence;

[0117] The category information sequence generation unit is used to find the product information of all products included in the historical product category and form a category information sequence;

[0118] The category binning unit is used to sort the products in the category information sequence in descending order according to price, and divide the sorted category information sequence into n category bins.

[0119] Based on the above embodiments, the consumption level prediction module 340 may include:

[0120] Consumer level classification unit, used to divide each category into corresponding consumer levels based on price range;

[0121] The annotation unit is used to annotate the historical purchase sequences of goods in the training set according to the consumption level corresponding to the category;

[0122] The consumption capacity model acquisition unit is used to input the labeled historical purchase sequence of goods into the consumption capacity model, train the consumption capacity model, and obtain the trained consumption capacity model.

[0123] Based on the above embodiments, the recommended candidate set generation module 350 may include:

[0124] The target category bucketing determination unit is used to match the consumption level with the n category buckets to obtain the category bucket corresponding to the consumption level, which is used as the target category bucket.

[0125] The recommended candidate set acquisition unit is used to extract the category information sequence from the target category bucket, and filter the preliminary candidate set based on the category information sequence to obtain the recommended candidate set.

[0126] Based on the above embodiments, the behavior sequence determination module may include:

[0127] The sorting unit is used to sort the product identification information of the user's purchase behavior in ascending order according to the user's purchase time.

[0128] The numbering information output unit is used to output the product identification information after ascending sorting to obtain user behavior sequence information.

[0129] The product recommendation device provided in this embodiment of the invention can execute the product recommendation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0130] Example 4

[0131] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device 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 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., 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.

[0132] 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.

[0133] 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.

[0134] 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.

[0135] Specifically, the method includes:

[0136] The user's historical behavior sequence is determined based on the purchase behavior data collected within a preset time period.

[0137] The system matches historical product attribute information corresponding to the user's historical behavior sequence with the static attribute set of the product, and forms a historical purchase sequence of the product based on the historical product attribute information; wherein each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price;

[0138] A preliminary candidate set is determined based on the historical purchase sequence of the aforementioned products;

[0139] The historical purchase sequence of the products is input into the trained consumption capacity model to predict the user's consumption level in the category;

[0140] The preliminary candidate set is filtered based on the consumption level to obtain the recommended candidate set.

[0141] 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).

[0142] 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.

[0143] Computer programs used to implement 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 executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0144] 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.

[0145] 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).

[0146] 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.

[0147] 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.

[0148] 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.

[0149] 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 method for recommending products, characterized in that, include: The user's historical behavior sequence is determined based on the purchase behavior recorded within a preset time period. The system matches historical product attribute information corresponding to the user's historical behavior sequence with the static attribute set of the product, and forms a historical purchase sequence of the product based on the historical product attribute information; wherein each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price; A preliminary candidate set is determined based on the historical purchase sequence of the aforementioned products; The historical purchase sequence of the products is input into the trained consumption capacity model to predict the user's consumption level in the category; The preliminary candidate set is filtered based on the consumption level to obtain the recommended candidate set; Before matching historical product attribute information corresponding to the user's historical behavior sequence based on the product's static attribute set, and forming a product historical purchase sequence based on the historical product attribute information, the process further includes: Based on the historical product attribute information corresponding to the user behavior sequence, extract the historical product category from the historical product attribute information; Find the product information of all products included in the historical product category to form a category information sequence; The products in the category information sequence are sorted in descending order of price, and the sorted category information sequence is divided into n category buckets. The preliminary candidate set is filtered based on the consumption level to obtain a recommended candidate set, including: The consumption level is matched with the n category buckets to obtain the category bucket corresponding to the consumption level, which is used as the target category bucket. Extract the category information sequence from the target category bucket, and filter the preliminary candidate set based on the category information sequence to obtain the recommended candidate set.

2. The method according to claim 1, characterized in that, A preliminary candidate set is obtained based on the historical purchase sequence of the goods, including: Transform the elements in the product's historical purchase sequence into a target vector; The target vector is input into the preliminary candidate model, and the preliminary candidate model is trained to obtain the trained preliminary candidate model. The user's historical behavior sequence is input into the trained preliminary candidate model to obtain a preliminary candidate set of recommended products.

3. The method according to claim 2, characterized in that, The target vector is input into the preliminary candidate model, and the preliminary candidate model is trained to obtain a trained preliminary candidate model, including: Select the associated target vectors to form a target vector group; The first part of the target vector group is used as the input of the preliminary candidate model, and the second part of the target vector group is used as the output of the preliminary candidate model. The preliminary candidate model is trained to obtain a trained preliminary candidate model. The purchase time corresponding to the first part of the vector is earlier than the purchase time corresponding to the second part of the vector.

4. The method according to claim 1, characterized in that, Before inputting the historical purchase sequence of the goods into the trained consumer spending power model to predict the user's spending level in the category, the following steps are also included: Each category is divided into corresponding consumption levels based on price range; The historical purchase sequences of goods in the training set are labeled with the consumption level corresponding to the category; The labeled product historical purchase sequence is input into the consumption capacity model, and the consumption capacity model is trained to obtain the trained consumption capacity model.

5. The method according to claim 1, characterized in that, Based on the collected purchase behavior data within a preset time period, the user's historical behavior sequence is determined, including: The product identification information of the user's purchase behavior is sorted in ascending order according to the user's purchase time; The sorted product identification information is output to obtain user behavior sequence information.

6. A product recommendation device, characterized in that, include: The behavior sequence determination module is used to determine the user's historical behavior sequence based on the purchased behavior data collected within a preset time period. The purchase sequence determination module is used to match historical product attribute information corresponding to the user's historical behavior sequence with the static attribute set of the product, and form a historical purchase sequence of the product based on the historical product attribute information; wherein, each element in the historical purchase sequence of the product includes a historical product identifier, a historical product category, and a historical product price; A preliminary candidate set generation module is used to determine a preliminary candidate set based on the historical purchase sequence of the goods. The consumption level prediction module is used to input the historical purchase sequence of the goods into the trained consumption capacity model to predict the user's consumption level in the category. The recommended candidate set generation module is used to filter the preliminary candidate set based on the consumption level to obtain the recommended candidate set; The purchase sequence determination module includes: The product category extraction unit is used to extract the historical product category from the historical product attribute information based on the historical product attribute information corresponding to the user behavior sequence; The category information sequence generation unit is used to find the product information of all products included in the historical product category and form a category information sequence; The category binning unit is used to sort the products in the category information sequence in descending order according to price, and to divide the sorted category information sequence into n category bins. The recommendation candidate set generation module includes: The target category bucketing determination unit is used to match the consumption level with the n category buckets to obtain the category bucket corresponding to the consumption level, which is used as the target category bucket. The recommended candidate set acquisition unit is used to extract the category information sequence from the target category bucket, and filter the preliminary candidate set based on the category information sequence to obtain the recommended candidate set.

7. 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 recommended method for the product according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the recommended method for the product of any one of claims 1-5.

Images