Home decoration product recommendation method, electronic equipment and computer readable storage medium

Abstract

The embodiment of the invention relates to a home decoration product recommendation method, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining the individual home decoration behavior data of a target user, carrying out the coding of the individual home decoration behavior data, and obtaining the individual behavior characteristics corresponding to the individual home decoration behavior data; acquiring a group home decoration behavior feature set, and determining a group feature vector associated with a first attribute of a home decoration product included in the individual home decoration behavior data from the group home decoration behavior feature set as a target group home decoration behavior feature; generating behavior fusion features based on the individual behavior features and the target group home decoration behavior features; inputting the behavior fusion features into a pre-trained home decoration product recommendation model, and obtaining a home decoration product recommendation result output by the home decoration product recommendation model; and determining a target home decoration product from the at least one recommended product based on the recommendation probability corresponding to each recommended product. By adopting the scheme, the accuracy of home decoration product recommendation can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of product recommendation technology, and in particular to a method for recommending home decoration products, an electronic device, and a computer-readable storage medium. Background Technology

[0002] In the home decoration industry, consumers face the dilemma of choosing suitable products from a vast array of options. With the development of deep learning technology, recommendation systems now exist that use users' historical behavioral data during the product selection process to recommend products for future consideration, thus alleviating the user's predicament in choosing the right items from a massive selection.

[0003] However, existing recommendation systems rely solely on users' historical behavior data to recommend home improvement products, making it difficult to guarantee the accuracy of the recommendations. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this disclosure provides a method for recommending home decoration products, an electronic device, and a computer-readable storage medium.

[0005] This disclosure provides a method for recommending home decoration products, the method comprising: Acquire individual home decoration behavior data of the target user, and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes the home decoration products that the target user browsed or purchased within a preset historical period. A set of group home decoration behavior features is obtained, and a group feature vector associated with the first attribute of the home decoration product included in the individual home decoration behavior data is determined from the set of group home decoration behavior features as the target group home decoration behavior features. The set of group home decoration behavior features includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical selection records of home decoration products of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes. Generate behavioral fusion features based on the individual behavioral characteristics and the target group's home decoration behavioral characteristics; The behavior fusion features are input into a pre-trained home decoration product recommendation model to obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product. A target home decoration product is determined from the at least one recommended product based on the recommendation probability corresponding to each recommended product, and the target home decoration product is used to recommend it to the target user.

[0006] This disclosure also provides a home decoration product recommendation device, the device comprising: The first processing module is used to acquire individual home decoration behavior data of the target user and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes the home decoration products that the target user browsed or purchased within a preset historical period. The second processing module is used to acquire a set of group home decoration behavior features and determine a group feature vector associated with the first attribute of the home decoration product included in the individual home decoration behavior data as the target group home decoration behavior feature from the set of group home decoration behavior features. The set of group home decoration behavior features includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical selection records of home decoration products of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes. The feature fusion module is used to generate behavior fusion features based on the individual behavior features and the home decoration behavior features of the target group; The result acquisition module is used to input the behavior fusion features into a pre-trained home decoration product recommendation model, and obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product. The product determination module is used to determine a target home decoration product from the at least one recommended product based on the recommendation probability corresponding to each recommended product, and the target home decoration product is used to recommend it to the target user.

[0007] This disclosure also provides an electronic device, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the home decoration product recommendation method provided in this disclosure.

[0008] This disclosure also provides a computer-readable storage medium storing a computer program for performing the home decoration product recommendation method provided in this disclosure.

[0009] The technical solution provided in this disclosure has the following advantages compared with the prior art: The home decoration product recommendation scheme provided in this disclosure acquires individual home decoration behavior data of target users, encodes the individual home decoration behavior data to obtain individual behavior features corresponding to the individual home decoration behavior data, the individual home decoration behavior data includes home decoration products browsed or purchased by the target user within a preset historical period; and acquires a group home decoration behavior feature set, and determines a group feature vector associated with the first attribute of the home decoration products included in the individual home decoration behavior data as the target group home decoration behavior feature from the group home decoration behavior feature set, the group home decoration behavior feature set includes features based on multiple... The historical home decoration product selection records of each user determine the group attribute selection features corresponding to different home decoration product attributes. The group attribute selection features include group feature vectors for different categories under the corresponding attributes. Behavioral fusion features are generated based on individual behavioral features and target group home decoration behavioral features. The behavioral fusion features are input into a pre-trained home decoration product recommendation model to obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product. Based on the recommendation probability corresponding to each recommended product, a target home decoration product is determined from the at least one recommended product. The target home decoration product is used to recommend to the target user. By adopting the above technical solution, individual home decoration behavior data and group home decoration behavior feature sets of target users are obtained. For home decoration products contained in the individual home decoration behavior data, the associated target group home decoration behavior features are determined from the group home decoration behavior feature set. These features are then fused with the individual behavior features determined based on the individual home decoration behavior data and input into the home decoration product recommendation model for home decoration product recommendation. This ensures that home decoration product recommendations consider both the individual home decoration preferences of target users and the general preferences for home decoration product matching, which helps to improve the satisfaction of target users with the recommended home decoration products and improve the accuracy of home decoration product recommendations. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0011] Figure 1 A flowchart illustrating a home decoration product recommendation method provided in this embodiment of the disclosure; Figure 2 A flowchart illustrating yet another home decoration product recommendation method provided in this disclosure embodiment; Figure 3 A flowchart illustrating another method for recommending home decoration products provided in this embodiment of the present disclosure; Figure 4A flowchart illustrating another home decoration product recommendation method provided in this embodiment of the disclosure; Figure 5 A schematic diagram of the architecture of a home decoration product recommendation scheme provided in a specific embodiment of this disclosure; Figure 6 This is a schematic diagram of the structure of a home decoration product recommendation device provided in an embodiment of the present disclosure; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0012] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0013] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0014] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0015] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0016] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0017] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0018] In the home decoration industry, consumers often face a "matching dilemma" when confronted with a vast array of home decoration products. They need to ensure stylistic consistency (e.g., a Scandinavian-style sofa paired with a solid wood coffee table), functional compatibility (e.g., furniture size suitable for small apartments), and construction feasibility (e.g., matching tiles with grout). However, current home decoration product recommendation systems typically rely solely on users' historical click data, failing to model the complex multidimensional relationships between products (e.g., style-space-craftsmanship). This often leads to recommendations that result in style conflicts, size mismatches, or construction inconsistencies. Furthermore, home decoration decisions exhibit strong sequential dependence (the user's selection path may be...). (The path is "floor → wall paint → lighting fixtures" or the reverse path). Traditional collaborative filtering algorithms struggle to dynamically adapt to this non-linear decision-making process. Furthermore, for newly listed products or new users, the lack of historical user behavior data or product interaction records makes it difficult for the recommendation system to accurately match user preferences or product characteristics, resulting in a serious cold start problem. In addition, the characteristics of high-priced, low-frequency consumption make users' tolerance for error extremely high. A single incorrect recommendation can lead to significant financial losses. This necessitates that the recommendation system must be interpretable (e.g., explaining that "this lighting fixture is recommended because its metal material complements the industrial-style coffee table you have already selected").

[0019] To address at least one of the shortcomings of traditional recommendation systems, this disclosure provides a home decoration product recommendation solution. It aims to solve the pain point for users in the home decoration industry who, faced with a vast array of product choices during the home decoration process, struggle to efficiently find related products that are style-consistent, functionally compatible, and within their budget. Traditional recommendation systems (such as collaborative filtering) rely solely on historical behavioral data for home decoration product recommendations, failing to effectively model complex matching rules, dynamic decision-making processes, and cold-start scenarios. This solution achieves scenario-based intelligent matching recommendations, resolving the core contradiction between the high complexity of decision-making in the home decoration field and the low level of user expertise.

[0020] The following describes the home decoration product recommendation method provided in this disclosure with reference to specific embodiments.

[0021] Figure 1 This is a flowchart illustrating a home decoration product recommendation method provided in an embodiment of this disclosure. This method can be executed by a home decoration product recommendation device provided in this embodiment, wherein the device can be implemented using software and / or hardware, and is generally integrated into an electronic device. Figure 1 As shown, the method includes: Step 101: Obtain individual home decoration behavior data of the target user and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes the home decoration products that the target user browsed or purchased within a preset historical period.

[0022] The target user can be any user who needs home improvement product recommendations. Individual home improvement behavior data for the target user includes at least the home improvement products browsed or purchased by the target user within a preset historical time period; that is, the target user's most recent home improvement behavior. A home improvement behavior can be understood as a series of home improvement products browsed consecutively on a single day, or a series of home improvement products purchased consecutively within a certain time period. It is understandable that the preset historical time period will vary depending on the time of the user's most recent home improvement behavior. For example, if user A browsed a series of home improvement products consecutively on a single day, then for user A, the preset historical time period is the day user A browsed home improvement products; if user B purchased home improvement products within a certain time period on a single day, then for user B, the preset historical time period is the time period within the day user B purchased home improvement products.

[0023] In this embodiment, the acquired individual home decoration behavior data can be encoded to transform fragmented user home decoration behaviors into quantifiable preference features, thereby obtaining individual behavior features corresponding to the individual home decoration behavior data. These individual behavior features include, at a minimum, encoded features corresponding to the attributes of the home decoration products included in the individual home decoration behavior data. These attributes include, but are not limited to, the categories of home decoration products (e.g., flooring, wall coverings, stools, dining tables, wardrobes, etc.) and styles (e.g., minimalist, retro, Chinese style).

[0024] As an example, an individual home decoration preference encoder can be pre-trained, and the individual home decoration behavior data of the target user can be input into the trained individual home decoration preference encoder for encoding, and the individual behavior features output by the individual home decoration preference encoder can be obtained.

[0025] Step 102: Obtain the group home decoration behavior feature set, and determine the group feature vector associated with the first attribute of the home decoration products included in the individual home decoration behavior data from the group home decoration behavior feature set as the target group home decoration behavior feature.

[0026] The group home decoration behavior feature set includes group attribute selection features corresponding to different home decoration product attributes, determined based on the historical home decoration product selection records of multiple users. These group attribute selection features include group feature vectors for different categories under each attribute, which represent the vector representation of an attribute category in the group data. Different home decoration product attributes include, but are not limited to, product categories and styles. There are multiple categories under each attribute; for example, for category attributes, the corresponding categories may include, but are not limited to, flooring, wall coverings, dining tables, wardrobes, and tiles. For style attributes, the corresponding categories may include, but are not limited to, Nordic style, light luxury style, Chinese style, and minimalist style.

[0027] In this embodiment, for the home decoration products included in the acquired individual home decoration behavior data, a group feature vector associated with the first attribute can be determined from the acquired group home decoration behavior feature set as the target group home decoration behavior feature based on the attributes of each home decoration product (referred to as the first attribute for ease of description and differentiation). It is understood that in this embodiment, the first attribute refers to the actual attribute category to which the home decoration product belongs, such as the dining table category under the category attribute.

[0028] Step 103: Generate behavioral fusion features based on individual behavioral characteristics and target group home decoration behavioral characteristics.

[0029] In this embodiment, after obtaining the individual behavioral features corresponding to individual home decoration behavior data and the group feature vector associated with the first attribute of the home decoration products included in the individual home decoration behavior data as the target group's home decoration behavior features, the individual behavioral features and the target group's home decoration behavior features can be fused to generate behavioral fusion features. Through fusion, the group tendencies of all users and the individual preferences of users can be combined, so as to improve the accuracy of home decoration product recommendations by combining the general public's matching tendencies and the individual's unique matching preferences.

[0030] In the feature fusion process, for the same home decoration product in the individual home decoration behavior data, all the encoding vectors corresponding to the attributes of the home decoration product in the individual behavior features are fused with the group feature vectors corresponding to the attributes in the target group home decoration behavior features, and then the remaining encoding vectors in the individual behavior features are fused to obtain the behavior fusion features.

[0031] It should be noted that this embodiment does not limit the specific fusion method. Fusion can be performed by vector concatenation, or by weighted summation, addition, averaging, etc.

[0032] Step 104: Input the behavior fusion features into the pre-trained home decoration product recommendation model to obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product.

[0033] The home decoration product recommendation model is pre-trained. It can use a multi-layer Transformer codec as the initial model and train it using collected training samples. When the model converges or the number of iterations reaches a preset number, the training is complete, and a trained home decoration product recommendation model is obtained. The trained home decoration product recommendation model can predict home decoration products based on the input data and output at least one recommended product and the recommendation probability corresponding to each recommended product. The higher the recommendation probability, the higher the matching degree between the corresponding recommended product and the user's previous home decoration behavior, and the higher the possibility of being adopted by the user.

[0034] In this embodiment, the obtained behavioral fusion features are input into a pre-trained home decoration product recommendation model. The home decoration product recommendation model predicts the next home decoration product based on the input behavioral fusion features and outputs a home decoration product recommendation result. The home decoration product recommendation result includes at least one recommended product and the recommendation probability corresponding to each recommended product.

[0035] Step 105: Determine the target home decoration product from at least one recommended product based on the recommendation probability corresponding to each recommended product. The target home decoration product is used to recommend to the target user.

[0036] In this embodiment, after obtaining the home decoration product recommendation results output by the home decoration product recommendation model, a target home decoration product can be determined from at least one recommended product based on the recommendation probability corresponding to each recommended product in the home decoration product recommendation results, and recommended to the target user.

[0037] As an example, based on the recommendation probability of each recommended product, the recommended product with the highest recommendation probability can be selected as the target recommended product and recommended to the target user.

[0038] The home decoration product recommendation scheme provided in this embodiment acquires individual home decoration behavior data of target users and encodes the individual home decoration behavior data to obtain individual behavior features corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes home decoration products browsed or purchased by the target user within a preset historical period. It also acquires a group home decoration behavior feature set and determines a group feature vector associated with a first attribute of the home decoration products included in the individual home decoration behavior data from the group home decoration behavior feature set as the target group home decoration behavior feature. The group home decoration behavior feature set includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical home decoration product selection records of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes. It generates behavior fusion features based on individual behavior features and target group home decoration behavior features. It inputs the behavior fusion features into a pre-trained home decoration product recommendation model to obtain the home decoration product recommendation result output by the home decoration product recommendation model. The home decoration product recommendation result includes at least one recommended product and the recommendation probability corresponding to each recommended product. Based on the recommendation probability corresponding to each recommended product, it determines a target home decoration product from the at least one recommended product, and the target home decoration product is used to recommend it to the target user. By adopting the above technical solution, individual home decoration behavior data and group home decoration behavior feature sets of target users are obtained. For home decoration products contained in the individual home decoration behavior data, the associated target group home decoration behavior features are determined from the group home decoration behavior feature set. These features are then fused with the individual behavior features determined based on the individual home decoration behavior data and input into the home decoration product recommendation model for home decoration product recommendation. This ensures that home decoration product recommendations consider both the individual home decoration preferences of target users and the general preferences for home decoration product matching, which helps to improve the satisfaction of target users with the recommended home decoration products and improve the accuracy of home decoration product recommendations.

[0039] In some embodiments, such as Figure 2 As shown, based on the aforementioned embodiments, step 102, which involves obtaining a set of group home decoration behavior features, may include the following sub-steps: Step 201: Obtain the historical selection records of home decoration products from multiple users.

[0040] The multiple users may include or exclude the target users mentioned above. Each user's home decoration product selection history may include one or more home decoration behavior data of that user, and the home decoration behavior data shall at least include the home decoration products selected by the user.

[0041] Step 202: Based on the attribute information of each home decoration product in the historical selection record of home decoration products, generate at least one attribute selection sequence diagram. The attribute selection sequence diagram is used to represent the order in which multiple users select home decoration products with corresponding attributes.

[0042] The attribute information of each home decoration product in the historical selection record can include the category attribute, style attribute, material attribute, spatial parameters (such as apartment size), construction process (such as waterproofing requirements), etc. of each home decoration product, as well as the time information of when a home decoration product was selected. Based on the attribute information of each home decoration product, an attribute selection time sequence diagram can be constructed for the key attributes (such as category, style, material) among these attributes to obtain at least one attribute selection time sequence diagram.

[0043] It should be noted that in this embodiment, the category attributes, style attributes, material attributes, etc. corresponding to each home decoration product are pre-set, and the correspondence between each home decoration product and the category attributes, material attributes, style attributes, etc. can be established in order to obtain the attributes of each home decoration product.

[0044] Taking the construction of an attribute selection sequence diagram corresponding to category attributes as an example, we can determine the order in which all users select home decoration product categories based on the historical selection records of home decoration products for multiple users, and obtain the category selection sequence diagram of the home decoration product categories selected by the group corresponding to these multiple users.

[0045] Taking the construction of a style attribute selection sequence diagram as an example, assuming the preset style levels include basic styles (k=1 level) and derived styles (k=2 level), the basic style is such as Nordic minimalism, and the derived style is such as Nordic eclectic light luxury. Based on the historical selection records of home decoration products by multiple users, the order in which all users selected home decoration styles is determined, resulting in a style selection sequence diagram for the style of home decoration products selected by the corresponding groups of these multiple users. The style selection sequence diagram includes at least one k=1 level node and a k=2 level node. It is understood that the inclusion of two levels of nodes in the style selection sequence diagram is only for example. In actual construction, more levels can be expanded according to the data complexity. For example, if it is necessary to refine the style classification, three levels of nodes can be added (such as "Nordic minimalism - natural wood color scheme" and "Nordic eclectic - light luxury metal elements").

[0046] It should be noted that, in this embodiment of the disclosure, time series graphs of different attribute dimensions can be extracted from the historical data of user groups' selection of home decoration products (i.e., the historical selection records of home decoration products by multiple users) according to specific business needs, so as to learn the overall tendency of user groups to select home decoration products from more perspectives.

[0047] In this embodiment, by generating at least one attribute selection time sequence graph based on the attribute information of each home decoration product in the historical selection record of home decoration products, heterogeneous data such as home decoration products (e.g., sofas, tiles), style tags (e.g., Nordic style, industrial style), spatial parameters, and construction techniques can be uniformly represented as a graph structure. The nodes in the generated time sequence graph not only contain unique product codes or identifiers, but also introduce virtual space nodes (e.g., living room walls, bathroom floors) and style anchor nodes (e.g., "light luxury metal elements"). The complex relationships between home decoration products are depicted through multi-dimensional edge relationships (style association edges, physical compatibility edges, and construction constraint edges). The attribute values ​​of the edges come from structured data (e.g., size, specifications), unstructured data (e.g., style semantics in text descriptions), and cross-modal data (e.g., spatial relationships in image parsing). The weights are dynamically adjusted by combining domain rules (e.g., construction standards, product compatibility) and user behavior data (e.g., browsing preferences). Finally, the association modeling and reasoning of heterogeneous data are realized through graph neural network and other technologies, solving the adaptation problem of traditional recommendation systems to complex scenarios.

[0048] Step 203: Perform feature extraction on at least one attribute selection time sequence graph to obtain the group attribute selection feature corresponding to each attribute selection time sequence graph. The group attribute selection feature includes the group feature vector corresponding to each node in the attribute selection time sequence graph.

[0049] In this embodiment, for at least one obtained attribute selection time sequence map, feature extraction can be performed on each attribute selection time sequence map to obtain the group attribute selection features corresponding to each attribute selection time sequence map.

[0050] As an example, a graph convolutional neural network can be used to extract features from the time sequence graph of each attribute selection to obtain the group attribute selection features of the corresponding attribute (e.g., category, style).

[0051] The group attribute selection features include the group feature vector corresponding to each node in the attribute selection time sequence graph. Each node in the attribute selection time sequence graph corresponds to an attribute category. By extracting features from the attribute selection time sequence graph, the group feature vector corresponding to each attribute category can be obtained. For example, assuming there are 1000 categories of all home decoration products, the obtained category selection time sequence graph contains 1000 nodes. The number of rows of group attribute selection features corresponding to the category attributes obtained through feature extraction is 1000, with each row corresponding to the group feature vector of one category attribute category.

[0052] Step 204: Based on the group attribute selection features corresponding to the time sequence diagram of each attribute selection, construct a group home decoration behavior feature set.

[0053] In this embodiment, after extracting features from the time sequence diagram of each attribute selection to obtain the group attribute selection features of the corresponding attribute, the group attribute selection features of these attributes constitute the group home decoration behavior feature set.

[0054] The home decoration product recommendation scheme provided in this disclosure generates at least one attribute selection time series graph based on the historical selection records of home decoration products of multiple users. Features are extracted from each of the at least one attribute selection time series graphs to obtain the group attribute selection features corresponding to each attribute selection time series graph. This leads to the construction of a group home decoration behavior feature set. As a result, the home decoration product matching preference features of the general public can be mined, providing group tendencies for home decoration product recommendations. This provides data support for subsequent recommendations of the next home decoration product by combining the matching tendencies of the general public and the unique matching preferences of individuals.

[0055] In some embodiments, such as Figure 3 As shown, based on the aforementioned embodiments, step 102, which involves determining the home decoration behavior characteristics of the target group, may include the following sub-steps: Step 301: Traverse the home decoration products in the individual home decoration behavior data and obtain the first attribute of the currently traversed home decoration product.

[0056] In this embodiment, the acquired individual home decoration behavior data includes at least one home decoration product. These home decoration products can be traversed sequentially, and the first attribute of the currently traversed home decoration product can be obtained.

[0057] Step 302: Determine the target group attribute selection feature corresponding to the first attribute from the group attribute selection features of the group home decoration behavior feature set.

[0058] In this embodiment, the group home decoration behavior feature set includes at least one group attribute selection feature corresponding to an attribute. After obtaining the first attribute of the currently traversed home decoration product, the group attribute selection feature corresponding to the first attribute can be determined from the group home decoration behavior feature set based on the first attribute. For ease of description and distinction, the group attribute selection feature corresponding to the first attribute is called the target group attribute selection feature.

[0059] It is understood that in this embodiment, the first attribute refers to the actual attribute category to which the home decoration product belongs, such as the dining table category under the category attribute. Based on the first attribute, it can be determined whether the first attribute belongs to the category attribute, the style attribute, or other attributes. Therefore, based on the attribute to which the first attribute category belongs, the corresponding target group attribute selection features can be determined from the group home decoration behavior feature set. For example, if the first attribute is a dining table, then it can be determined that the first attribute belongs to the category attribute, and the determined target group attribute selection features are the group attribute selection features corresponding to the category attribute.

[0060] Step 303: Determine the group feature vector corresponding to the attribute identifier of the first attribute from the target group attribute selection features.

[0061] In this embodiment, after determining the target group attribute selection feature corresponding to the first attribute, the group feature vector corresponding to the attribute identifier of the first attribute can be further determined from the target group attribute selection feature based on the attribute identifier of the first attribute.

[0062] It is understood that in this embodiment, the group attribute selection feature includes group feature vectors of different categories under the corresponding attribute, and each home decoration product attribute (referring to the specific category under the attribute) has its unique number (i.e. attribute identifier). Based on the attribute identifier, the corresponding group feature vector can be found from the target group attribute selection feature.

[0063] Step 304: After traversing all home decoration products, determine the group feature vector corresponding to the home decoration products as the home decoration behavior characteristics of the target group.

[0064] In this embodiment, each home decoration product in the individual home decoration behavior data is traversed, and the group feature vector corresponding to the first attribute of each home decoration product is obtained. After traversing all home decoration products, the group feature vectors corresponding to all home decoration products are determined as the home decoration behavior characteristics of the target group.

[0065] The home decoration product recommendation scheme provided in this embodiment traverses the home decoration products in individual home decoration behavior data, first filters out the corresponding target group attribute selection features from the group home decoration behavior feature set according to the first attribute of each home decoration product, and then determines the group feature vector corresponding to the attribute identifier from the target group attribute selection features according to the attribute identifier of the first attribute, thereby determining the target group home decoration behavior features. This can accurately find the group feature vector corresponding to each attribute of the home decoration product, providing effective data support for accurate home decoration product recommendation.

[0066] Typically, the selection of home decoration products may exhibit a strong temporal dependence. For example, if a user's first home decoration activity primarily involves selecting flooring products, their next activity might involve choosing wall decoration products. Therefore, in this embodiment, when encoding individual home decoration behavior data, the time of selecting each product can also be encoded. Thus, in some embodiments, when encoding individual home decoration behavior data to obtain corresponding individual behavioral characteristics, user information, product attribute information, and time information can be retrieved. User information, obtained with user authorization, includes, but is not limited to, the user's age, gender, spending power, and decoration experience. Attribute information includes, but is not limited to, the product category, style, size, and material. Time information represents the time each product was selected or viewed. Next, user information is encoded to obtain user characteristics, and the product attribute information and time information are encoded separately to obtain the product attribute characteristics and time characteristics. Each product has its own unique attribute characteristics and time characteristics.

[0067] For example, structured data such as users' age, gender, and spending power can be mapped into 64-dimensional vectors through a fully connected layer, and users' home decoration experience (levels 0-5) can be embedded in a segment embedding method to obtain user features. This not only enhances the richness of user features but also better portrays the user's home decoration background.

[0068] A user's renovation experience level can be determined in the following ways: (1) Based on renovation behavior data: Based on the complexity of the project in the user's historical renovation records (such as whether it involves water and electricity renovation, custom furniture, etc.), material selection (such as the proportion of environmentally friendly materials used), and process requirements (such as whether fine construction is required), the user is classified into levels through rules or models (such as level 0 being "newbie" and level 5 being "expert").

[0069] (2) User self-evaluation or survey: Users are asked to fill in their renovation experience through questionnaires, interviews and other means (such as "whether they have completed renovation independently" and "the degree of attention to construction details"). The results are then mapped to the corresponding renovation experience level.

[0070] (3) Combining consumption data: By analyzing the price range of home decoration products purchased by users (such as the proportion of high-end products) and brand preferences (whether they choose well-known brands), their experience level can be indirectly inferred.

[0071] By using experience level embedding to label a user's home decoration behavior, the aim is to transform the user's decoration experience into quantifiable numerical features, making it easier for the model to better understand user preferences. The higher the experience level of a user, the less likely their purchasing preferences may reflect the preferences of the user group. On the contrary, the lower the experience level of a user, the better it reflects the understanding and preferences of the vast majority of users regarding the selection of home decoration products, thus enabling more accurate recommendations.

[0072] For example, when encoding attribute information, the categories, styles, and other attributes of several home decoration products selected from a user's individual home decoration behavior data can be one-hot encoded separately to better learn the user's preferences for different attributes of home decoration products. When encoding time information, the Time2Vec algorithm can be used to encode the time information of the user's selection of home decoration products to better capture the temporal patterns of the user's selection of home decoration products. The Time2Vec algorithm is a method for vectorizing time information, which will not be explained in detail in this disclosure.

[0073] Finally, after obtaining user characteristic information, the attribute characteristics of each home decoration product, and the time characteristics, individual behavioral characteristics can be determined based on these characteristics. The determined individual behavioral characteristics include the user characteristics and the corresponding attribute and time characteristics for each home decoration product.

[0074] In this embodiment of the disclosure, by encoding the user information, the attribute information and time information corresponding to the home decoration products in the individual home decoration behavior data, user characteristics, attribute characteristics and time characteristics of each home decoration product are obtained, and then individual behavior characteristics are obtained. This can help the home decoration product recommendation model to better learn the preferences of different users for different attributes of home decoration products and capture the temporal characteristics of users' selection of home decoration products, thereby making better recommendations for home furnishing products.

[0075] Since individual home decoration behavior data may contain more than one home decoration product, and a single home decoration product may have more than one attribute, the determined individual behavior features contain attribute features corresponding to more than one home decoration product, and the target group home decoration behavior features contain group feature vectors corresponding to more than one attribute. Therefore, it is necessary to fuse the attribute features of the same home decoration product and the group feature vectors to achieve the fusion of group-level and individual-level features. Thus, in some embodiments, when generating behavior fusion features based on individual behavior features and target group home decoration behavior features, the target attribute features and target time features corresponding to each home decoration product in the individual home decoration behavior data can be obtained from the individual behavior features; and the group feature vector corresponding to each home decoration product can be obtained from the target group home decoration behavior features; then, the target attribute features, target time features, and group feature vectors corresponding to the same home decoration product are concatenated to obtain the attribute fusion features corresponding to each home decoration product; finally, the behavior fusion features are generated based on the user features in the individual behavior features and the attribute fusion features corresponding to each home decoration product.

[0076] In other words, for each home decoration product in the individual home decoration behavior data, the attribute features (referred to as target attribute features for ease of description and differentiation) and time features (referred to as target time features for ease of description and differentiation) of the home decoration product are obtained from the determined individual behavior characteristics, and the group feature vector corresponding to the home decoration product is obtained from the target group home decoration behavior characteristics. It can be understood that the group feature vector corresponding to a home decoration product includes all group feature vectors corresponding to the attributes of the home decoration product included in the target group home decoration behavior characteristics. Then, the target attribute features, target time features, and all group feature vectors of the home decoration product are concatenated (i.e., concatenated) to obtain the attribute fusion feature of the home decoration product. Finally, the attribute fusion features of all home decoration products are concatenated with the user features in the individual behavior characteristics to obtain the behavior fusion feature.

[0077] In this embodiment of the disclosure, by concatenating the attribute features, time features, and group feature vectors of the same home decoration product in the individual behavioral features and the home decoration behavioral features of the target group, the attribute fusion features of each home decoration product are obtained. This realizes the encoding of home decoration products at the group level and the individual level, and realizes the group-individual fusion encoding of a user's home decoration behavior, providing effective data support for accurate home decoration product recommendations.

[0078] Figure 4 This is a flowchart illustrating another home decoration product recommendation method provided in this embodiment of the present disclosure. This embodiment further optimizes the aforementioned home decoration product recommendation method based on the previous embodiment. Figure 4 As shown, the method includes: Step 401: Obtain individual home decoration behavior data of the target user and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes the home decoration products that the target user browsed or purchased within a preset historical period.

[0079] Step 402: Obtain the group home decoration behavior feature set, and determine the group feature vector associated with the first attribute of the home decoration products included in the individual home decoration behavior data as the target group home decoration behavior feature from the group home decoration behavior feature set.

[0080] Among them, the group home decoration behavior feature set includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical selection records of home decoration products of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes.

[0081] Step 403: Generate behavioral fusion features based on individual behavioral characteristics and target group home decoration behavioral characteristics.

[0082] Step 404: Input the behavior fusion features into the pre-trained home decoration product recommendation model to obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product.

[0083] It should be noted that, in this embodiment, the description of steps 401-404 can be found in the explanation of steps 101-104 in the previous embodiment, and their implementation principles are similar, so they will not be repeated here.

[0084] Step 405: Generate at least one attribute selection sequence diagram based on the historical selection records of home decoration products of multiple users.

[0085] It should be noted that, in this embodiment, the specific implementation method of generating at least one attribute selection sequence diagram based on the historical selection records of home decoration products of multiple users can be found in the explanation of generating at least one attribute selection sequence diagram in the aforementioned embodiment. The implementation principle is similar and will not be repeated here.

[0086] Step 406: Based on at least one attribute selection time sequence diagram, determine the matching tendency transfer attention matrix corresponding to at least one attribute selection time sequence diagram.

[0087] In this embodiment, by selecting a time sequence diagram for each attribute, a corresponding matching tendency transfer attention matrix can be determined. The matching tendency transfer attention matrix has a size of N*N, where N represents the number of attribute categories contained under an attribute, such as the number of category categories contained under a category attribute. The i-th row in the matching tendency transfer attention matrix represents the transition from attribute category p of home decoration products. iThe probability of moving to other attribute categories under this attribute.

[0088] In some embodiments, when determining the pairing tendency transfer attention matrix corresponding to each attribute selection time sequence map, at least one attribute selection time sequence map can be input into the pairing tendency encoder for encoding to obtain at least one attribute encoded feature map; then, the at least one attribute encoded feature map is input into the pairing tendency transfer attention network for inter-node transfer probability learning to obtain the pairing tendency transfer attention matrix corresponding to at least one attribute selection time sequence map.

[0089] Among them, the match preference encoder is used to learn the features of each node in the input attribute selection time sequence graph and re-encode them, outputting the corresponding attribute encoding feature map; the match preference transfer attention network is used to relearn the input attribute encoding feature map, learn the node-to-node transition probability in the input attribute encoding feature map, obtain an N*N match preference transfer attention matrix through feature transformation and output it.

[0090] Suppose an attribute selection time series graph is denoted as X, with dimensions N×C1. N represents the number of nodes in the graph, and C1 represents the initial feature size of each node. For example, if the feature size of a node in a type selection time series graph is defined as 4, then C1 is 4. X is input into a matching preference encoder for encoding, resulting in an attribute-encoded feature map of size N*C2. It's important to note that C2 differs from C1 because the node features of the attribute selection time series graph are re-encoded by the matching preference encoder. Next, the N*C2 attribute-encoded feature map is input into a matching preference transfer attention network for relearning. This network learns the node-to-node transition probabilities in the attribute-encoded feature map and outputs an N*N matching preference transfer attention matrix. Each row and column in this matrix represent a node. For example, in the matching preference transfer attention matrix corresponding to a style selection time series graph, row i and column j represent the probability that after selecting the style represented by row i, the user will subsequently select the home decoration product represented by column j. The matching preference transfer attention matrix (denoted as Φ) corresponding to the attribute selection time series graph X is determined as follows:

[0091]

[0092]

[0093] Using the above method, the matching tendency shift attention matrix corresponding to the timing diagram of each attribute selection can be determined.

[0094] Step 407: Based on the matching preference attention shift matrix, adjust the recommendation probability of each recommended product to obtain the adjusted probability of each recommended product.

[0095] In this embodiment, after obtaining at least one recommended product and the recommendation probability corresponding to each recommended product from the home decoration product recommendation model, the recommendation probability of each recommended product can be adjusted based on the matching tendency attention shift matrix to obtain the adjusted probability of each recommended product.

[0096] In some embodiments, when adjusting the recommendation probability corresponding to each recommended product, the matching tendency attention shift matrix can be queried based on the first attribute of the home decoration products in the individual home decoration behavior data and the attribute of each recommended product (referred to as the second attribute for ease of description and differentiation), to determine the shift probability corresponding to each recommended product (referred to as the target shift probability for ease of description and differentiation). The target shift probability represents the probability that a user will continue to choose a home decoration product with the second attribute after choosing a home decoration product with the first attribute. Then, the recommendation probability of the user is adjusted based on the target shift probability corresponding to each recommended product to obtain the adjusted probability corresponding to each recommended product.

[0097] In other words, for each recommended product output by the home decoration product recommendation model, the target transfer probability for each recommended product can be determined by querying the matching tendency transfer attention matrix corresponding to each attribute based on the second attribute of the recommended product (i.e., the category under each attribute) and the first attribute of the home decoration product in the individual home decoration behavior data. For example, for the category attribute, the matching tendency transfer attention matrix corresponding to the category attribute is queried. The transfer probability from the first attribute to the second attribute is found in the matching tendency transfer attention matrix corresponding to the category attribute. Assuming that the first attribute is located in the m-th row of the matching tendency transfer attention matrix and the second attribute is located in the n-th column of the matching tendency transfer attention matrix, then the element in the m-th row and j-th column of the matching tendency transfer attention matrix is ​​the determined target transfer probability.

[0098] When adjusting the recommendation probability of a product based on the target transition probability of each recommended product, the adjusted probability can be determined by multiplying the target transition probability of the same recommended product with its corresponding recommendation probability. Furthermore, since home decoration products may have multiple attributes and multiple determined target transition probabilities, the final transition probability can be determined by summing or averaging these multiple probability values. This final transition probability can then be used to adjust the recommendation probability of the recommended product to obtain the adjusted probability.

[0099] Step 408: Determine the target home decoration product based on the adjusted probability corresponding to each recommended product.

[0100] In this embodiment, after determining the adjusted probability of each recommended product, the target home decoration product can be determined based on the adjusted probability of each recommended product. For example, the recommended product with the highest adjusted probability can be determined as the target home decoration product.

[0101] The home decoration product recommendation scheme provided in this disclosure generates at least one attribute selection time sequence diagram based on the historical selection records of home decoration products by multiple users. Based on the at least one attribute selection time sequence diagram, a matching tendency transfer attention matrix corresponding to the at least one attribute selection time sequence diagram is determined. Then, the recommendation probability of each recommended product is adjusted based on the matching tendency transfer attention matrix to obtain the adjusted probability of each recommended product. The target home decoration product is determined based on the adjusted probability of each recommended product. Thus, the recommendation probability of the recommended products output by the home decoration product recommendation model is adjusted based on the transfer probability between attributes, so that the transfer probability between attributes is integrated in the home decoration product recommendation process, which can further improve the accuracy of home decoration product recommendation.

[0102] Figure 5 This is a schematic diagram of the architecture of a home decoration product recommendation scheme provided in a specific embodiment of this disclosure. Taking the constructed group home decoration behavior feature set, which includes group style selection features and group category selection features, as an example... Figure 5As shown, based on the historical home decoration product selection records of multiple users, a style selection time series graph and a category selection time series graph are first generated. For example, the style selection time series graph includes k=1 level nodes (basic styles such as "Nordic Minimalism") and k=2 level nodes (derived styles such as "Nordic Eclectic Luxury"). Then, the first graph's convolutional network is used to extract features from the style selection time series graph to obtain group style selection features, and the second graph's convolutional network is used to extract features from the category selection time series graph to obtain the corresponding group category selection features. When recommending home decoration products to users, individual home decoration behavior data of users is obtained, and an individual home decoration product matching preference encoder is used to encode and embed user information, home decoration product attribute information, and time information in the individual home decoration behavior data. The above information is converted into a specific format, such as binary vectors, through encoding, and then the specific format data is mapped to a low-dimensional space through embedding, thereby obtaining user features, attribute features, and time features, which in turn constitute individual behavior features. Next, the temporal multimodal information fusion module fuses individual behavioral features with the target group's home decoration behavioral features retrieved from the group's home decoration behavioral feature set, resulting in behavioral fusion features. These features are then input into the encoder / decoder, which includes a Transformer layer to encode and decode the input behavioral fusion features, outputting the user's next potentially interesting home decoration product, the product's attributes, and the time of selection. Furthermore, a matching preference shifting attention mechanism can be used to determine the matching preference shifting attention matrices corresponding to the style selection time series and category selection time series, respectively. These matrices are then used to adjust the next home decoration product output by the encoder / decoder, resulting in the final determined target home decoration product.

[0103] Figure 6 This is a schematic diagram of a home decoration product recommendation device provided in an embodiment of this disclosure. The device can be implemented by software and / or hardware, and is generally integrated into an electronic device for recommending home decoration products. Figure 6 As shown, the home decoration product recommendation device 50 includes: a first processing module 510, a second processing module 520, a feature fusion module 530, a result acquisition module 540, and a product determination module 550. Among them, The first processing module 510 is used to acquire individual home decoration behavior data of the target user and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes home decoration products that the target user browsed or purchased within a preset historical period. The second processing module 520 is used to obtain a group home decoration behavior feature set and determine the group feature vector associated with the first attribute of the home decoration product included in the individual home decoration behavior data as the target group home decoration behavior feature from the group home decoration behavior feature set. The group home decoration behavior feature set includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical selection records of home decoration products of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes. The feature fusion module 530 is used to generate behavior fusion features based on individual behavior features and target group home decoration behavior features; The result acquisition module 540 is used to input the behavior fusion features into the pre-trained home decoration product recommendation model and obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product. The product determination module 550 is used to determine the target home decoration product from at least one recommended product based on the recommendation probability corresponding to each recommended product. The target home decoration product is used to recommend to the target user.

[0104] Optionally, the second processing module 520 includes: The data acquisition unit is used to acquire the historical selection records of home decoration products from multiple users. The graph generation unit is used to generate at least one attribute selection sequence graph based on the attribute information of each home decoration product in the historical selection record of home decoration products. The attribute selection sequence graph is used to represent the order in which multiple users select home decoration products with corresponding attributes. The feature extraction unit is used to extract features from at least one attribute selection time sequence graph to obtain the group attribute selection features corresponding to each attribute selection time sequence graph. The group attribute selection features include the group feature vector corresponding to each node in the attribute selection time sequence graph. The feature set construction unit is used to select features based on the group attributes corresponding to the time series diagram for each attribute selection, and to construct a group home decoration behavior feature set.

[0105] Optionally, the second processing module 520 includes: The attribute acquisition unit is used to traverse the home decoration products in the individual home decoration behavior data and obtain the first attribute of the currently traversed home decoration product. The first determining unit is used to determine the target group attribute selection feature corresponding to the first attribute from the group attribute selection features of the group home decoration behavior feature set; The second determining unit is used to determine the group feature vector corresponding to the attribute identifier of the first attribute from the target group attribute selection features; The third determining unit is used to determine the group feature vector corresponding to the home decoration products as the home decoration behavior characteristics of the target group after all home decoration products have been traversed.

[0106] Optionally, the first processing module 510 is also used for: Obtain user information, attribute information of home decoration products, and time information from individual home decoration behavior data; User information is encoded to obtain user characteristics; The attribute information and time information corresponding to home decoration products are encoded to obtain the attribute characteristics and time characteristics of home decoration products. Individual behavioral characteristics are determined based on user characteristics, attribute characteristics, and time characteristics.

[0107] Further optionally, the feature fusion module 530 is also used for: Extract the target attribute features and target time features corresponding to each home decoration product from the individual home decoration behavior data; Obtain the group feature vector corresponding to each home decoration product from the home decoration behavior characteristics of the target group; By concatenating the target attribute features, target time features, and group feature vectors corresponding to the same home decoration product, the attribute fusion features corresponding to each home decoration product are obtained. Behavioral fusion features are generated based on user features in individual behavioral features and attribute fusion features corresponding to each home decoration product.

[0108] Optionally, the recommended home improvement product device 50 also includes: The graph generation module is used to generate at least one attribute selection sequence graph based on the home decoration product historical selection records of multiple users. The matrix determination module is used to determine the matching tendency transfer attention matrix corresponding to the selection time sequence diagram based on at least one attribute selection time sequence diagram. Product identification module 550 includes: The probability adjustment unit is used to adjust the recommendation probability of each recommended product based on the pairing preference attention shift matrix, so as to obtain the adjusted probability of each recommended product; The product determination unit is used to determine the target home decoration product based on the adjusted probability corresponding to each recommended product.

[0109] Further, optionally, the matrix determination module is also used for: Input the timing map of at least one attribute selection into the matching tendency encoder for encoding to obtain the feature map of at least one attribute encoding. Input at least one attribute-encoded feature map into the pairing tendency transfer attention network to learn the transfer probability between nodes, and obtain the pairing tendency transfer attention matrix corresponding to at least one attribute selection time sequence map.

[0110] Optionally, the probability adjustment unit is also used for: Based on the first attribute and the second attribute of each recommended product, query the matching tendency to transfer attention matrix and determine the target transfer probability corresponding to each recommended product; The recommendation probability of each recommended product is adjusted based on the target transition probability of each recommended product, resulting in the adjusted probability of each recommended product.

[0111] The home decoration product recommendation device provided in this disclosure can execute the home decoration product recommendation method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of executing the method.

[0112] To implement the above embodiments, this disclosure also proposes a computer program product, including a computer program / instructions, which, when executed by a processor, implements the home decoration product recommendation method in the above embodiments. Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure.

[0113] The following is a detailed reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing the electronic device 700 in the embodiments of this disclosure. The electronic device 700 in the embodiments of this disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0114] like Figure 7 As shown, the electronic device 700 may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device 700. The processing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0115] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic device 700 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 700 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0116] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product 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 device 709, or installed from storage device 708, or installed from ROM 702. When the computer program is executed by processing device 701, it performs the functions defined in the home improvement product recommendation method of embodiments of this disclosure.

[0117] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0118] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0119] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0120] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the home decoration product recommendation method provided in any embodiment of this disclosure.

[0121] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0122] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0123] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0124] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0125] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. 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 fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0126] According to one or more embodiments of this disclosure, this disclosure provides an electronic device, including: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the home improvement product recommendation method as provided in any of the present disclosures.

[0127] According to one or more embodiments of the present disclosure, the present disclosure provides a computer-readable storage medium storing a computer program for performing a home improvement product recommendation method as described in any of the present disclosure.

[0128] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0129] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0130] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for recommending home decoration products, characterized in that, include: Acquire individual home decoration behavior data of the target user, and encode the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data. The individual home decoration behavior data includes the home decoration products that the target user browsed or purchased within a preset historical period. A set of group home decoration behavior features is obtained, and a group feature vector associated with the first attribute of the home decoration product included in the individual home decoration behavior data is determined from the set of group home decoration behavior features as the target group home decoration behavior features. The set of group home decoration behavior features includes group attribute selection features corresponding to different home decoration product attributes determined based on the historical selection records of home decoration products of multiple users. The group attribute selection features include group feature vectors of different categories under the corresponding attributes. Generate behavioral fusion features based on the individual behavioral characteristics and the target group's home decoration behavioral characteristics; The behavior fusion features are input into a pre-trained home decoration product recommendation model to obtain the home decoration product recommendation results output by the home decoration product recommendation model. The home decoration product recommendation results include at least one recommended product and the recommendation probability corresponding to each recommended product. A target home decoration product is determined from the at least one recommended product based on the recommendation probability corresponding to each recommended product, and the target home decoration product is used to recommend it to the target user.

2. The method according to claim 1, characterized in that, The acquisition of the group's home decoration behavior feature set includes: Obtain the home decoration product selection history of multiple users; Based on the attribute information of each home decoration product in the historical selection record of the home decoration products, at least one attribute selection sequence diagram is generated. The attribute selection sequence diagram is used to represent the order in which the multiple users select home decoration products with corresponding attributes. Feature extraction is performed on the at least one attribute selection time sequence graph to obtain the group attribute selection feature corresponding to each attribute selection time sequence graph. The group attribute selection feature includes the group feature vector corresponding to each node in the attribute selection time sequence graph. Based on the group attribute selection features corresponding to each attribute selection time sequence diagram, the group home decoration behavior feature set is constructed.

3. The method according to claim 1, characterized in that, The step of determining the group feature vector associated with the first attribute of the home decoration products included in the individual home decoration behavior data from the group home decoration behavior feature set as the target group home decoration behavior feature includes: Traverse the home decoration products in the individual home decoration behavior data and obtain the first attribute of the currently traversed home decoration product; From the group attribute selection features of the group home decoration behavior feature set, determine the target group attribute selection feature corresponding to the first attribute; Determine the group feature vector corresponding to the attribute identifier of the first attribute from the target group attribute selection features; Once all the home decoration products have been traversed, the group feature vector corresponding to the home decoration products is determined as the home decoration behavior feature of the target group.

4. The method according to claim 1, characterized in that, The process of encoding the individual home decoration behavior data to obtain the individual behavior characteristics corresponding to the individual home decoration behavior data includes: Obtain user information, attribute information of home decoration products, and time information from the individual home decoration behavior data; The user information is encoded to obtain user characteristics; The attribute information and time information corresponding to the home decoration products are encoded respectively to obtain the attribute features and time features corresponding to the home decoration products; The individual behavioral characteristics are determined based on the user characteristics, the attribute characteristics, and the time characteristics.

5. The method according to claim 4, characterized in that, The generation of behavioral fusion features based on the individual behavioral characteristics and the target group's home decoration behavioral characteristics includes: From the individual behavioral characteristics, obtain the target attribute features and target time features corresponding to each home decoration product in the individual home decoration behavior data; Obtain the group feature vector corresponding to each home decoration product from the home decoration behavior characteristics of the target group; The target attribute features, target time features, and group feature vectors corresponding to the same home decoration product are concatenated to obtain the attribute fusion features corresponding to each home decoration product. The behavior fusion feature is generated based on the user features in the individual behavior features and the attribute fusion features corresponding to each home decoration product.

6. The method according to claim 1, characterized in that, The method further includes: Generate at least one attribute selection sequence diagram based on the home decoration product selection history records of multiple users; Based on the at least one attribute selection time sequence diagram, determine the matching tendency shift attention matrix corresponding to the at least one attribute selection time sequence diagram; The step of determining the target home decoration product from the at least one recommended product based on the recommendation probability corresponding to each recommended product includes: Based on the matching preference attention shift matrix, the recommendation probability of each recommended product is adjusted to obtain the adjusted probability of each recommended product; The target home decoration product is determined based on the adjusted probability corresponding to each recommended product.

7. The method according to claim 6, characterized in that, The step of determining the matching tendency shifting attention matrix corresponding to the at least one attribute selection time series graph based on the at least one attribute selection time series graph includes: The at least one attribute selection time sequence map is input into the matching tendency encoder for encoding to obtain at least one attribute encoded feature map; The at least one attribute encoding feature map is input into the pairing tendency transfer attention network to learn the inter-node transfer probability, thereby obtaining the pairing tendency transfer attention matrix corresponding to the at least one attribute selection time series map.

8. The method according to claim 6, characterized in that, The step of adjusting the recommendation probability of each recommended product based on the matching preference attention shift matrix to obtain the adjusted probability of each recommended product includes: Based on the first attribute and the second attribute of each recommended product, query the matching tendency attention shift matrix to determine the target shift probability corresponding to each recommended product; The recommendation probability of each recommended product is adjusted based on the target transition probability corresponding to each recommended product, thus obtaining the adjusted probability corresponding to each recommended product.

9. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the home decoration product recommendation method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for executing the home decoration product recommendation method according to any one of claims 1-8.

Images