Recommendation method and device for migration resettlement housing resources, electronic equipment and storage medium

Abstract

The application discloses a kind of migration resettlement housing recommendation method, device, electronic equipment and storage medium.It relates to the field of migration resettlement housing management.The application is recommended by the housing recommendation model, and multi-dimensional data of user family information, pre-migration housing information and user demand information are fused, which significantly improves the accuracy of initial recommendation.And through real scene browsing interface, so that the user can more intuitively understand the first recommended housing.Collect first feedback information and second feedback information, wherein the first feedback information is obtained from voice information, to improve the user's satisfaction with subsequent recommended housing.A new second user demand information is generated by a large language model, and a second recommended housing is output by a housing recommendation model, so that the second recommended housing is more in line with the user's needs, significantly improving the degree of personalization and accuracy of the recommendation.It can improve the user satisfaction of resettlement housing recommendation and improve the overall efficiency of migration resettlement work.

Description

Technical Field

[0001] This application relates to the field of resettlement housing management, and in particular to a method, apparatus, electronic device and storage medium for recommending resettlement housing resources. Background Technology

[0002] With the acceleration of urbanization, the number of demolition and resettlement projects is increasing. How to efficiently and accurately recommend rebuilt housing to relocated residents has become a crucial issue in smart city and public services. Traditional housing recommendation methods typically rely on human experience or simple rule matching, such as simply allocating housing types based on the number of family members or making recommendations based on rigid conditions specified by the user. Such methods fail to provide resettlement housing that meets user needs, resulting in poor recommendation effectiveness. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method, apparatus, electronic device, and storage medium for recommending resettlement housing, which can automatically recommend resettlement housing to users and significantly improve the personalization and accuracy of the recommendations, thus facilitating the progress of resettlement projects.

[0004] The method, apparatus, electronic device, and storage medium for recommending relocation housing resources according to the first aspect of this application include: Receive user resettlement request message, the user resettlement request message including user identity information, user family information, housing information before relocation, and first user demand information; Based on the user's identity information, a first set of selectable properties is determined from a preset property database; Based on the first user demand information, the first set of available properties is filtered to obtain a second set of available properties; Determine the candidate housing vector corresponding to each housing in the second set of optional housing from the preset housing vector library; The housing information before migration, the first user demand information, the user family information and the candidate housing vector are input into a pre-trained housing recommendation model to obtain the first recommended housing. The first recommended property and its surrounding amenities are displayed through a real-view browsing interface. Acquire user voice information during the display of the real-view browsing interface, and convert the user voice information into text to obtain first feedback information; Obtain the user's second feedback information regarding the first recommended property; The first feedback information and the second feedback information are input into a preset large language model to obtain the second user demand information; the first user demand information and the second user demand information are concatenated to obtain the third user demand information. The housing information before migration, the third user's demand information, the user's family information, and the candidate housing vector are input into the housing recommendation model to obtain the second recommended housing.

[0005] The method for recommending resettlement housing according to the embodiments of this application has at least the following beneficial effects: After obtaining user identity information, user family information, pre-relocation housing information, and first user demand information, a first set of optional housing is determined based on the user identity information. This first set of optional housing is then filtered to obtain a second set of optional housing. This filters out housing that does not meet the user's resettlement needs from the source, ensuring the suitability of the recommended housing and reducing subsequent computational overhead. Then, candidate housing vectors corresponding to each housing in the second set of optional housing are determined from a pre-set housing vector library. A first recommended housing is obtained through a pre-trained housing recommendation model. This method integrates multi-dimensional data such as user family information, pre-relocation housing information, and user demand information for recommendation. Compared to traditional single-condition filtering methods, this method comprehensively covers the resettlement user's family housing needs, original living habits, and other information, deeply aligning with the livelihood attributes of resettlement and significantly improving the accuracy of the initial recommendation. Furthermore, the first recommended housing is displayed to the user through a real-view browsing interface, allowing the user to understand the first recommended housing more intuitively. The system collects first and second feedback from users regarding the primary recommended housing. The first feedback is obtained through user voice recordings during the real-world browsing interface. This voice recording typically reflects the user's most natural opinions about the primary recommended housing. A large language model can be used to uncover hidden user needs based on this first feedback, thereby improving user satisfaction with subsequent housing recommendations. The large language model then generates new second user demand information, which is then used to output secondary recommended housing through the housing recommendation model. This ensures that the secondary recommended housing better meets user needs, significantly improving the personalization and accuracy of the recommendations. This enhances user satisfaction with resettlement housing recommendations, effectively reduces communication and coordination costs in relocation and resettlement work, and improves the overall efficiency of the relocation and resettlement process.

[0006] According to some embodiments of the first aspect of this application, the housing recommendation model includes a first embedding layer, a second embedding layer, a third embedding layer, a first splicing layer, a second splicing layer, a third splicing layer, a first feature fusion layer, and a dot product layer; The pre-migration housing information, the first user demand information, the user family information, and the candidate housing vector are input into a pre-trained housing recommendation model to obtain the first recommended housing, including: Each field of the pre-migration housing information is input into the first embedding layer to obtain multiple first features, and the multiple first features are spliced ​​together through the first splicing layer to obtain a first spliced ​​feature; Each field of the first user demand information is input into the second embedding layer to obtain multiple second features, and the multiple second features are concatenated through the second concatenation layer to obtain a second concatenated feature; Each field of the user's family information is input into the third embedding layer to obtain multiple third features, and the multiple third features are then concatenated through the third splicing layer to obtain a third spliced ​​feature; The first splicing feature, the second splicing feature, and the third splicing feature are input into the first feature fusion layer to obtain the user demand feature; The matching score between the user demand feature and each candidate property vector is calculated using the dot product layer. The property corresponding to the candidate property vector with the highest matching score is selected as the first recommended property.

[0007] According to some embodiments of the first aspect of this application, the property recommendation model further includes a fourth embedding layer, a second feature fusion layer, and a classifier, and the training steps of the property recommendation model include: Acquire user demand training information, pre-migration housing training information, user family training information, real property tags, and description training information for multiple available properties; Each field of the pre-migration housing training information is input into the first embedding layer to obtain multiple first training features, and the multiple first training features are spliced ​​together through the first splicing layer to obtain the first training spliced ​​features. Each field of the user demand training information is input into the second embedding layer to obtain multiple second training features, and the multiple second training features are concatenated through the second concatenation layer to obtain second training concatenated features; Each field of the user's family training information is input into the third embedding layer to obtain multiple third training features, and the multiple third training features are spliced ​​together through the third splicing layer to obtain third training spliced ​​features; The first training splicing feature, the second training splicing feature, and the third training splicing feature are input into the first feature fusion layer to obtain the user demand training feature; Each field of the description training information of each of the optional housing units is input into the fourth embedding layer to obtain multiple fourth training features; The fourth training feature corresponding to the same available property is input into the second feature fusion layer to obtain the property training vector corresponding to the available property. The matching training score between the user demand training features and each of the housing listing training vectors is calculated using the dot product layer. The matching training scores are input into the classifier to obtain the recommended housing listings. Based on the recommended properties, the actual labels of the properties, and the preset loss function, the loss value is calculated. The property recommendation model is iteratively optimized based on the loss value to obtain the trained property recommendation model.

[0008] According to some embodiments of the first aspect of this application, the candidate housing vector is obtained through the following steps: Obtain the description information of each property from the property database; Each field of the description information is input into the fourth embedding layer of the trained housing recommendation model to obtain multiple fourth features; Each of the fourth features is input into the second feature fusion layer of the trained housing recommendation model to obtain the candidate housing vector.

[0009] According to some embodiments of the first aspect of this application, the processing steps of the first feature fusion layer or the second feature fusion layer include: Perform dimension mapping operations on multiple input features to obtain multiple mapped features with the same dimensions; Each of the mapping features is input into the attention unit of a multilayer perceptron with shared parameters to obtain the attention score corresponding to each mapping feature; Based on each attention score, weight normalization is performed to obtain the attention weight of each mapping feature; The weighted summation feature is obtained by performing a weighted summation process on each of the mapping features and the corresponding attention weights. The weighted summation features are input into a fully connected layer to obtain fused features, which are then used as the output features of the feature fusion layer.

[0010] According to some embodiments of the first aspect of this application, the first feedback information and the second feedback information are input into a preset large language model to obtain second user demand information, including: The first feedback information and the second feedback information are concatenated based on preset prompt words to obtain concatenated text; the prompt words are used to limit the output content and format of the large language model. The concatenated text is input into the large language model to obtain the second user requirement information.

[0011] According to some embodiments of the first aspect of this application, obtaining second feedback information from the user regarding the first recommended property includes: Obtain negative descriptions of the first recommended property; Each negative field of the negative description information is displayed to the user via a pop-up window; In response to detecting an operation command for the pop-up window, a target negative field is determined from each of the negative fields based on the operation command, and each of the target negative fields is used as the second feedback information.

[0012] A second aspect of this application provides a relocation housing recommendation device, comprising: The receiving module is used to receive user resettlement request messages, which include user identity information, user family information, pre-migration housing information, and first user demand information; The first determining module is used to determine a first set of selectable properties from a preset property database based on the user's identity information; The second determining module is used to filter the first set of available housing based on the first user demand information to obtain a second set of available housing. The third determining module is used to determine the candidate housing vector corresponding to each housing in the second optional housing set from the preset housing vector library; The first input module is used to input the pre-migration housing information, the first user demand information, the user family information and the candidate housing vector into a pre-trained housing recommendation model to obtain the first recommended housing; The display module is used to display the first recommended property and its surrounding facilities through a real-view browsing interface; The first acquisition module is used to acquire user voice information during the display of the real-view browsing interface, and convert the user voice information into text to obtain first feedback information. The second acquisition module is used to acquire second feedback information from the user regarding the first recommended property; The second input module is used to input the first feedback information and the second feedback information into a preset large language model to obtain the second user demand information; and to concatenate the first user demand information and the second user demand information to obtain the third user demand information. The third input module is used to input the pre-migration housing information, the third user demand information, the user family information, and the candidate housing vector into the housing recommendation model to obtain the second recommended housing.

[0013] A third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method for recommending relocation housing resources as described in any one of the first aspects of the embodiment.

[0014] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for recommending relocation housing resources as described in any one of the first aspects of the embodiment.

[0015] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0016] The present application will be further described below with reference to the accompanying drawings and embodiments, wherein: Figure 1 This is a flowchart illustrating the steps of the method for recommending relocation and resettlement housing in an embodiment of this application. Figure 2 This is a schematic diagram of the housing recommendation model in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of the device for recommending relocation and resettlement housing resources according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0018] In the description of this application, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0019] In the description of this application, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0020] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0021] In the description of this application, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0023] The first aspect of this application provides a method for recommending resettlement housing. This method can be deployed and executed on a terminal, on a server, or as software running on either the terminal or the server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the method for recommending resettlement housing, but is not limited to the above forms.

[0024] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0025] Reference Figure 1 , Figure 1 This is a flowchart illustrating the steps of a method for recommending relocation housing resources according to an embodiment of this application. The method for recommending relocation housing resources according to an embodiment of this application may include, but is not limited to, steps S110 to S200.

[0026] Step S110: Receive a user resettlement request message. The user resettlement request message includes user identity information, user family information, housing information before relocation, and first user demand information. It is worth noting that the method for recommending relocation housing resources in this application embodiment is applied to a service terminal. The service terminal is connected to a user terminal. The user sends a user relocation request message to the service terminal through the user terminal, so that the service terminal can receive the user relocation request message.

[0027] It is worth noting that user family information includes population size, generational information, and the gender and age of each family member. User family information is closely related to property recommendations. For example, population size and generational information are directly related to the floor plan of the recommended property; the age of family members is related to the geographical location of the recommended property. For instance, if a family's average age is 7 years old, the recommended property should be located near a primary school or kindergarten. It should be noted that the above user family information is only an example, and those skilled in the art can add or remove information according to actual circumstances. For example, user family information may also include the workplace of each family member.

[0028] It is worth noting that pre-migration housing information may include the area of ​​the original housing, the apartment layout, the geographical location of the original housing, the distance of the original housing from the hospital, the distance of the original housing from the school, and the distance of the original housing from the city center. The above-mentioned pre-migration housing information is only an example, and those skilled in the art can add or reduce information according to the actual situation.

[0029] Step S120: Determine the first set of available properties from the preset property database based on the user's identity information; It is worth noting that the preset housing database records the mapping relationship between user identity information and multiple housing listings. The housing listings corresponding to user identity information are the housing listings that the user can choose. Therefore, after obtaining user identity information, the corresponding first set of available housing listings can be determined, so that the recommended housing listings obtained later are housing listings that the user can choose, thus ensuring the compliance of the recommended housing listings.

[0030] Step S130: Based on the first user demand information, filter the first set of available properties to obtain the second set of available properties; It is worth noting that the first user demand information includes both soft demand information and hard demand information. For example, hard demand information can be specific restrictions on apartment type, area, or geographical location. In step S130, the first set of optional properties is filtered based on the hard demand information to obtain a second set of optional properties. All properties in the second set of optional properties meet the hard demand information. Step S130 of this embodiment uses hard demand information to filter properties, ensuring that the subsequently recommended properties necessarily meet the user's hard demand information, thus improving recommendation accuracy. Furthermore, by filtering properties, the number of optional properties is reduced, thereby reducing the consumption of computational resources in the subsequent model and improving the model's recommendation efficiency.

[0031] It is worth noting that step S130 of this application can obtain the user placement request message through a window, for example, by displaying a window on the user terminal, showing multiple soft requirement description words and multiple hard requirement description words in the window, and the user can determine the soft requirement description words and the hard requirement description words selected by the user in response to the operation instructions generated for the operation of the window. Each selected soft requirement description word constitutes soft requirement information, and each selected hard requirement description word constitutes hard requirement information.

[0032] It is worth noting that soft requirement descriptions are used to characterize the soft requirements for a property; hard requirement descriptions characterize the hard requirements for a property. For example, soft requirement descriptions could be "close to a hospital," "close to a school," or "larger area"; hard requirement descriptions could be "less than 3km from a school," "less than 5km from a hospital," or "three or more bedrooms." The examples of soft and hard requirement descriptions above are merely illustrative and should not be construed as limitations on the terms. Those skilled in the art can freely define soft and hard requirement descriptions according to actual circumstances.

[0033] Step S140: Determine the candidate housing vector corresponding to each housing in the second optional housing set from the preset housing vector library; It is worth noting that a property vector is constructed based on each property, and these property vectors are stored in a property vector library. The property vector corresponding to each property in the second set of optional properties is selected from the property vector library as a candidate property vector. The property recommendation model of this embodiment serves multiple users, meaning the method of this embodiment needs to be executed multiple times. Therefore, by pre-constructing and storing property vectors, and directly extracting the corresponding candidate property vector from the property vector library when needed, instead of repeatedly constructing property vectors each time the method is executed, computational overhead is reduced, and the processing efficiency of the property recommendation model is improved, thereby increasing the execution efficiency of the method of this embodiment.

[0034] Step S150: Input the housing information before migration, the first user demand information, the user family information and the candidate housing vector into the pre-trained housing recommendation model to obtain the first recommended housing. Step S160: Display the first recommended property and its surrounding amenities through a real-view browsing interface; It's worth noting that the user-facing virtual reality browsing interface displays rendered 3D models, including each property and its surrounding amenities. Furthermore, the top-recommended property is highlighted in the virtual reality browsing interface for easy understanding by the user.

[0035] For example, the method of this application embodiment is applied to a demolition and relocation project. This project includes a construction plan, which comprises spatial planning data, architectural design data, and infrastructure network data. The spatial planning data may include land boundaries, coordinates, and elevation design information for planning elements such as resettlement sites, roads, green spaces, and public facilities. The architectural design data may include building information models (BIM) or detailed design drawings of the geometric dimensions, structural types, floors, and unit types of resettlement housing and public buildings. The infrastructure network data may include the paths, burial depths, pipe diameters, and connection relationships of water supply, drainage, electricity, and communication pipelines. The three-dimensional model is a high-precision, dynamically analytic three-dimensional visualization model based on the aforementioned construction plan. This application embodiment does not limit the specific construction process of the three-dimensional model; those skilled in the art can use existing methods to construct it, and during the construction process, record the coordinates of each model element in the three-dimensional model. It should be noted that the three-dimensional model refers to a three-dimensional model of the entire city, including not only the model of the area indicated by the construction plan but also models of other areas besides those indicated by the construction plan.

[0036] Step S170: Obtain user voice information during the real-view browsing interface display, and convert the user voice information into text to obtain the first feedback information; It's worth noting that during the initial display of the recommended properties through the virtual reality browsing interface, a pop-up window requests recording permission from the user. After the user clicks the pop-up to grant recording permission, the client records the user's voice, which is then converted into text to obtain initial feedback. Since this voice information is obtained while the user is viewing the virtual reality browsing interface, and users are likely discussing this with family members during this time, their deeper needs regarding the housing options may be revealed. Therefore, converting the user's voice into text to obtain initial feedback allows for the uncovering of these deeper needs, ensuring that the subsequent second recommended properties meet these needs and improve user satisfaction.

[0037] Step S180: Obtain the user's second feedback information regarding the first recommended property; It should be noted that a pop-up window can be used to ask users if they are satisfied with the first recommended property. If the user is satisfied with the first recommended property, no further steps are needed, and the second recommended property will be set as the user's target property. If the user indicates through the pop-up window that they are not satisfied with the first recommended property, a text box will be displayed so that the user can enter their second feedback information.

[0038] Step S190: Input the first feedback information and the second feedback information into the preset large language model to obtain the second user demand information; concatenate the first user demand information and the second user demand information to obtain the third user demand information. Step S200: Input the housing information before migration, the third user's demand information, the user's family information and the candidate housing vector into the housing recommendation model to obtain the second recommended housing.

[0039] In some embodiments, after obtaining the second recommended property, it is displayed through a real-view browsing interface to allow users to intuitively understand the second recommended property and its surrounding amenities. A pop-up window then inquires whether the user is satisfied with the second recommended property. If the user is satisfied, a binding relationship is established between the second recommended property and the user's identity information, and this binding relationship is uploaded to a pre-set database.

[0040] It is worth noting that the method for recommending resettlement housing in the first aspect of this application, through steps S110 to S200, after obtaining user identity information, user family information, pre-relocation housing information, and first user demand information, determines a first set of optional housing based on the user identity information, filters the first set of optional housing to obtain a second set of optional housing, and filters out housing that does not meet the user's resettlement needs from the source, which not only ensures the suitability of the recommended housing but also reduces subsequent computational overhead. Then, candidate housing vectors corresponding to each housing in the second set of optional housing are determined from a preset housing vector library; and the first recommended housing is obtained through a pre-trained housing recommendation model. By integrating multi-dimensional data such as user family information, pre-relocation housing information, and user demand information for recommendation, compared with the traditional single-condition filtering recommendation method, it can comprehensively cover the family housing needs and original living habits of resettled users, deeply align with the livelihood attributes of resettlement, and significantly improve the accuracy of the initial recommendation. The first recommended housing is also displayed to the user through a real-view browsing interface, allowing the user to understand the first recommended housing more intuitively. The system collects first and second feedback from users regarding the primary recommended housing. The first feedback is obtained through user voice recordings during the real-world browsing interface. This voice recording typically reflects the user's most natural opinions about the primary recommended housing. A large language model can be used to uncover hidden user needs based on this first feedback, thereby improving user satisfaction with subsequent housing recommendations. The large language model then generates new second user demand information, which is then used to output secondary recommended housing through the housing recommendation model. This ensures that the secondary recommended housing better meets user needs, significantly improving the personalization and accuracy of the recommendations. This enhances user satisfaction with resettlement housing recommendations, effectively reduces communication and coordination costs in relocation and resettlement work, and improves the overall efficiency of the relocation and resettlement process.

[0041] Understandably, referring to Figure 2 , Figure 2 This is a schematic diagram of the housing recommendation model according to an embodiment of this application. The housing recommendation model includes a first embedding layer, a second embedding layer, a third embedding layer, a first splicing layer, a second splicing layer, a third splicing layer, a first feature fusion layer, and a dot product layer. The housing recommendation model also includes a fourth embedding layer, a second feature fusion layer, and a classifier. In practical applications, the housing recommendation model does not need to use a classifier; the classifier is used to assist in training the housing recommendation model.

[0042] It is understood that step S150 may include, but is not limited to, steps S210 to S250.

[0043] Step S210 inputs each field of the housing information before migration into the first embedding layer to obtain multiple first features, and then concatenates the multiple first features through the first concatenation layer to obtain the first concatenated feature; Step S220: Input each field of the first user demand information into the second embedding layer to obtain multiple second features, and then concatenate the multiple second features through the second concatenation layer to obtain the second concatenated features; Step S230: Input each field of the user's family information into the third embedding layer to obtain multiple third features, and then concatenate the multiple third features through the third concatenation layer to obtain the third concatenated features; It is worth noting that the first, second, third, and fourth embedding layers are used to map the text into vectors, avoiding combinatorial explosion and interference between fields. The model can learn the semantics of each field independently and concatenate multiple field vectors of the same information to completely preserve all the information without premature dimensionality reduction causing information loss.

[0044] In some embodiments, the first embedding layer, the second embedding layer, the third embedding layer, and the fourth embedding layer can use ALBERT as the embedding layer. ALBERT (A Lite BERT) is a lightweight variant of BERT used to generate embedded representations of the input data. The embedding layers perform feature extraction on the concatenated text to obtain a fixed-size embedding vector.

[0045] Step S240: Input the first splicing feature, the second splicing feature, and the third splicing feature into the first feature fusion layer to obtain the user demand features; It is worth noting that in step S240 of this application, the first spliced ​​feature, the second spliced ​​feature, and the third spliced ​​feature are fused through the first feature fusion layer, so that the resulting user demand feature carries information on the first user demand, user family information, and housing information before migration. This enables the model to learn the interaction weights between different information, making the user demand feature more discriminative.

[0046] Step S250: Calculate the matching score between the user demand features and each candidate housing vector using a dot product layer, and select the housing corresponding to the candidate housing vector with the highest matching score as the first recommended housing.

[0047] It is worth noting that in step S250, the dot product layer performs a dot product operation on the user demand feature vector and the candidate property vector. The result is used as a matching score, which characterizes the similarity between the user demand feature vector and the candidate property vector. Furthermore, the matching score quantitatively measures the degree of fit between the current user and each candidate property; a higher score indicates that the property better meets the user's needs. In step S250, the property corresponding to the candidate property vector with the highest similarity to the user demand features can be selected as the first recommended property.

[0048] This embodiment of the application, through steps S210 to S250, firstly performs feature extraction through steps S210 to S230, setting independent embedding and splicing layers for pre-migration housing information, user demand information, and user family information respectively. This achieves independent feature extraction of different categories of information and integration of similar features, effectively avoiding feature interference between different dimensions and types of data, and ensuring the feature integrity and representation accuracy of various types of information. Then, it performs fusion through the first feature fusion layer, which can comprehensively capture the inherent relationship between the user's hard policy requirements for resettlement, core housing needs, family carrying capacity, and original housing habits. The output user demand features can more accurately and comprehensively represent the user's full-dimensional resettlement demands. Finally, it calculates the matching score between the user demand features and the candidate housing vector through the dot product layer. The calculation logic is simple and efficient, the matching results are highly interpretable, and it can quickly identify the housing with the highest matching degree with the user's needs. At the same time, it significantly reduces the inference calculation overhead of the model, improves the response speed of housing recommendation, and optimizes the user interaction experience.

[0049] It should be noted that those skilled in the art should understand that the processing procedure of the housing recommendation model in step S200 is the same as that in steps S210 to S250, and this application will not elaborate on it.

[0050] It is understandable that the processing steps of the first feature fusion layer and the second feature fusion layer are the same. The processing of the feature fusion layer includes steps S310 to S350.

[0051] Step S310: Perform dimension mapping operation on the multiple input features respectively to obtain multiple mapped features with the same dimensions; It is worth noting that dimension mapping operation refers to a mathematical operation that transforms input features of different dimensions to the same specified dimension. Furthermore, this application does not specifically limit the specified dimension; those skilled in the art can set the specified dimension according to the actual situation.

[0052] For example, in step S240, if the input features are a first concatenation feature, a second concatenation feature, and a third concatenation feature, then it is necessary to first perform dimension mapping operations on the first concatenation feature, the second concatenation feature, and the third concatenation feature to obtain multiple mapped features. Each mapped feature is a first concatenation dimension feature, a second concatenation dimension feature, and a third concatenation dimension feature, and the dimensions of the first concatenation dimension feature, the second concatenation dimension feature, and the third concatenation dimension feature are the same.

[0053] Step S320: Input each mapping feature into the attention unit of the multilayer perceptron with shared parameters to obtain the attention score corresponding to each mapping feature; It is worth noting that the attention unit of a multilayer perceptron refers to the MLP. An MLP can consist of two fully connected layers and a non-linear activation function. By processing each mapped feature separately through the MLP, the attention score of each mapped feature can be obtained.

[0054] Step S330: Perform weight normalization based on each attention score to obtain the attention weight of each mapped feature; For example, the attention scores of each mapping feature can be normalized using Softmax to obtain the attention weight of each mapping feature.

[0055] Step S340: Perform weighted summation based on each mapping feature and its corresponding attention weight to obtain the weighted summation feature; For example, the weighted summation process can be represented as: ; Where F is the weighted summation feature, and n is the number of mapped features, for example, n is 3 in the first feature fusion module. Xi is the i-th mapped feature, and Yi is the attention weight of the i-th mapped feature.

[0056] Step S350: Input the weighted summation features into the fully connected layer to obtain the fused features, which are then used as the output features of the feature fusion layer.

[0057] It is worth noting that weighted summation features are essentially linear combinations of multiple mapping features based on their weights, and can only express first-order interaction relationships. By introducing fully connected layers with non-linear activation capabilities, the model can learn the complex and high-order cross-combination relationships between the dimensions within the weighted summation features.

[0058] In this embodiment, steps S310 to S350 first unify the dimensions of multiple input features through dimensional mapping. Then, a multilayer perceptron attention unit with shared parameters adaptively learns the attention weights of each feature, followed by weighted summation and fully connected layer processing. This allows the fusion process to automatically focus on key information and suppress irrelevant features, effectively preserving important interactions in multi-source heterogeneous information and enhancing the discriminative ability of the fused features. Applying this fusion structure to user demand feature fusion and housing feature fusion can comprehensively improve the expressive power and recommendation accuracy of the recommendation model.

[0059] It is understood that the training steps of the housing recommendation model in this application embodiment include steps S410 to S500.

[0060] Step S410: Obtain user demand training information, pre-migration housing training information, user family training information, real property tags, and description training information of multiple available properties; It is worth noting that historical data can be obtained from historical migration projects. This historical data includes user demand training information, housing training information before migration, user family training information, real property tags of historical users, and descriptive training information of available properties.

[0061] For example, follow-up visits are conducted with users from historical relocation projects to identify those satisfied with their new homes. Data is collected from these users, including their needs training information, pre-relocation housing training information, family training information, real property tags from historical users, and descriptive training information for available properties. Using data from satisfied users as training samples can improve the accuracy of the resulting property recommendation model.

[0062] For example, the descriptive training information of the available housing units can refer to the attributes of the available housing units, such as the location of the housing unit, the type of housing unit, the area of ​​the housing unit, the distance of the housing unit from the hospital, the distance of the housing unit from the school, etc.

[0063] Step S420: Each field of the pre-migration housing training information is input into the first embedding layer to obtain multiple first training features, and the multiple first training features are concatenated through the first concatenation layer to obtain the first training concatenated features. Step S430: Input each field of the user demand training information into the second embedding layer to obtain multiple second training features, and then concatenate the multiple second training features through the second concatenation layer to obtain the second training concatenation features. Step S440: Input each field of the user's family training information into the third embedding layer to obtain multiple third training features, and then concatenate the multiple third training features through the third concatenation layer to obtain the third training concatenation features. Step S450: Input the first training splicing feature, the second training splicing feature, and the third training splicing feature into the first feature fusion layer to obtain the user demand training feature; Step S460: Input each field of the description training information of each available property into the fourth embedding layer to obtain multiple fourth training features; Step S470: Input the fourth training feature corresponding to the same available property into the second feature fusion layer to obtain the property training vector corresponding to the available property. It should be noted that the processing procedures for the first feature fusion layer and the second feature fusion layer are the same. The specific processing procedure for the second feature fusion layer is described in steps S310 to S350, and will not be repeated here.

[0064] Step S480: Calculate the matching training score between the user demand training features and the training vector of each property listing through dot product layers. Step S490: Input multiple matching training scores into the classifier to obtain the training recommended properties; It's worth noting that while the property recommendation model doesn't require a classifier during actual inference, it's necessary to set one during training to facilitate subsequent loss calculations. The classifier used is Softmax, which converts the matching training scores into a probability distribution.

[0065] Step S500: Calculate the loss value based on the trained recommended listings, the real labels of the listings, and the preset loss function; Step S510: Iteratively optimize the housing recommendation model based on the loss value to obtain the trained housing recommendation model.

[0066] It is worth noting that in steps S500 and S510, the cross-entropy loss function is used to calculate the loss value, and the housing recommendation model is iteratively optimized based on the loss value until the number of iterations reaches a preset threshold, or the loss value is less than the preset loss threshold, then the optimization stops and the trained housing recommendation model is obtained.

[0067] It is worth noting that the parameters of the second feature fusion layer are optimized and updated during training. The second feature fusion layer is used to fuse multiple fourth training features of the same available property. Its fusion process is the same as that of the first feature fusion layer, which enables the model to capture the interaction relationships between multiple fourth training features of the same available property. This effectively preserves the important interaction relationships in multi-source heterogeneous information, making the features output by the second feature fusion layer more expressive.

[0068] Understandably, the candidate housing vector is obtained through steps S610 to S630.

[0069] Step S610: Obtain the description information of each property from the property database; Step S620: Input each field of the description information into the fourth embedding layer of the trained housing recommendation model to obtain multiple fourth features; Step S630: Input each fourth feature into the second feature fusion layer of the trained housing recommendation model to obtain the candidate housing vector.

[0070] It is worth noting that the housing database contains descriptive information for each property. This descriptive information can be set by those skilled in the art. For example, the descriptive information of a property may include its area, floor plan, distance from a hospital, distance from a school, distance from the city center, etc.

[0071] In this embodiment, steps S610 to S630 are used to calculate the property listing vectors offline. The candidate property listing vectors are stored in a property listing vector database. When the property recommendation model performs inference, it can directly retrieve the candidate property listing vectors without recalculation, thus improving the inference efficiency of the model. Furthermore, the processing of the second feature fusion layer is described in steps S310 to 350. Through the processing of the second feature fusion layer, the candidate property listing vectors gain stronger expressive power.

[0072] It is understood that step S190 may include, but is not limited to, steps S710 to S720.

[0073] Step S710: The first feedback information and the second feedback information are concatenated based on preset prompt words to obtain concatenated text; the prompt words are used to limit the output content and format of the large language model. Step S720: Input the concatenated text into the large language model to obtain the second user requirement information.

[0074] It is worth noting that, through steps S710 to S720, this embodiment of the application extracts the user's hidden needs from the first and second feedback information, thereby improving the accuracy of the subsequently obtained second recommended properties and increasing user satisfaction with the second recommended properties. By concatenating the first and second feedback information into a concatenated text based on preset prompts, and then inputting it into a large language model to obtain the second user need information, the prompts strictly limit the output content and format, guiding the large language model to accurately extract structured and usable need information from scattered voice and operation feedback, avoiding vague or non-standard output results, facilitating concatenation with the first user need information, thereby improving the accuracy, stability, and automation level of feedback parsing and need updating.

[0075] For example, the prompt could be: "You are a resettlement housing demand analysis assistant. Please summarize the housing conditions that the user cares about based on the user's voice feedback and the selected negative tags for the housing, outputting them in JSON format, with each demand as a separate field." For instance, the first feedback might be, "This house is okay in other aspects, but the floor is too high, and waiting for the elevator every day is inconvenient, plus there don't seem to be any hospitals nearby," and the second feedback might be, "Too far from the school" and "Too small." Then the output of the large language model would be: Requirement 1: The property must be on a low floor. Requirement 2: The property must be close to the school. Requirement 3: The property needs to be large. Requirement 4: The property must be close to the hospital. }

[0076] It should be noted that the embodiments of this application do not limit the specific large language model, and those skilled in the art can select a large language model from the prior art according to the actual situation.

[0077] It is understood that, in some other embodiments, the second feedback information can be obtained through the following steps: Step S810: Obtain negative description information of the first recommended property; It is worth noting that the negative descriptions are pre-set by those skilled in the art, who pre-define negative descriptions for each property. For example, negative descriptions for a property could include phrases such as "far from hospital," "far from school," or "small size."

[0078] Step S820: Display each negative field of the negative description information to the user via a pop-up window; Step S830: In response to detecting an operation command for the pop-up window, the selected target negative field is determined from each negative field based on the operation command, and each target negative field is used as the second feedback information.

[0079] It is worth noting that, through steps S810 to S830, the feedback operation of users' dissatisfaction with the first recommended housing is simplified from a complex manual description to a point-and-click operation, which significantly improves the collection efficiency and targeting of negative feedback. At the same time, the structured and field-based negative information can be directly used as input to the large language model to help generate more accurate second user needs, thereby making the adjusted recommendation results more effective in avoiding housing defects that users do not care about, and further improving user satisfaction with the placement recommendation.

[0080] A second aspect of this application provides a device for recommending relocation housing resources. (Refer to...) Figure 3 , Figure 3 This is a schematic diagram of the relocation housing recommendation device according to an embodiment of this application. The relocation housing recommendation device includes: The receiving module 310 is used to receive user resettlement request messages, which include user identity information, user family information, housing information before relocation, and first user demand information. The first determining module 320 is used to determine a first set of selectable housing resources from a preset housing resource database based on user identity information; The second determining module 330 is used to filter the first set of available housing based on the first user demand information to obtain a second set of available housing. The third determining module 340 is used to determine the candidate housing vector corresponding to each housing in the second optional housing set from the preset housing vector library; The first input module 350 is used to input the housing information before migration, the first user demand information, the user family information and the candidate housing vector into the pre-trained housing recommendation model to obtain the first recommended housing; The 360° display module is used to showcase the top recommended property and its surrounding amenities through a real-view browsing interface. The first acquisition module 370 is used to acquire user voice information during the display of the real-view browsing interface, and convert the user voice information into text to obtain the first feedback information. The second acquisition module 380 is used to acquire the user's second feedback information regarding the first recommended property; The second input module 390 is used to input the first feedback information and the second feedback information into a preset large language model to obtain the second user demand information; and to concatenate the first user demand information and the second user demand information to obtain the third user demand information. The third input module 400 is used to input the housing information before migration, the third user's demand information, the user's family information and the candidate housing vector into the housing recommendation model to obtain the second recommended housing.

[0081] The relocation housing recommendation device of this application embodiment is used to execute the relocation housing recommendation method of the first aspect of this application. When executing the method, it first receives a user relocation request message to obtain user identity information, user family information, pre-relocation housing information, and first user demand information. Based on the user identity information, it determines a first set of optional housing resources. The first set of optional housing resources is then filtered to obtain a second set of optional housing resources. This filters out housing resources that do not meet the user's relocation needs from the source, ensuring the suitability of the recommended housing resources and reducing subsequent computational overhead. Then, it determines candidate housing resource vectors corresponding to each housing resource in the second set of optional housing resources from a preset housing resource vector library. A first recommended housing resource is obtained through a pre-trained housing resource recommendation model. This method integrates multi-dimensional data such as user family information, pre-relocation housing information, and user demand information for recommendation. Compared to traditional single-condition filtering recommendation methods, it comprehensively covers the family housing needs and original living habits of relocated users, deeply aligning with the livelihood attributes of relocation and significantly improving the accuracy of initial recommendations. Finally, it displays the first recommended housing resource to the user through a real-view browsing interface, allowing the user to more intuitively understand the first recommended housing resource. The system collects first and second feedback from users regarding the primary recommended housing. The first feedback is obtained through user voice recordings during the real-world browsing interface. This voice recording typically reflects the user's most natural opinions about the primary recommended housing. A large language model can be used to uncover hidden user needs based on this first feedback, thereby improving user satisfaction with subsequent housing recommendations. The large language model then generates new second user demand information, which is then used to output secondary recommended housing through the housing recommendation model. This ensures that the secondary recommended housing better meets user needs, significantly improving the personalization and accuracy of the recommendations. This enhances user satisfaction with resettlement housing recommendations, effectively reduces communication and coordination costs in relocation and resettlement work, and improves the overall efficiency of the relocation and resettlement process.

[0082] It should be noted that the specific implementation of the relocation housing recommendation device is basically the same as the specific embodiment of the relocation housing recommendation method described above, and will not be repeated here. Subject to meeting the requirements of the embodiments of this application, the relocation housing recommendation device may also be equipped with other functional units to implement the relocation housing recommendation method in the above embodiments.

[0083] A third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method for recommending relocation housing resources according to any one of the first aspects of the embodiment. This electronic device can be any smart terminal, including a tablet computer, an in-vehicle computer, or similar device.

[0084] Reference Figure 4 , Figure 4This is a schematic diagram of the structure of an electronic device according to one embodiment. The electronic device includes: The processor 401 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 402 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 402 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 402 and is called and executed by the processor 401 to execute the recommended method for relocation and resettlement housing resources in the embodiments of this application. Input / output interface 403 is used to implement information input and output; The communication interface 404 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 405 transmits information between various components of the device (e.g., processor 401, memory 402, input / output interface 403, and communication interface 404); The processor 401, memory 402, input / output interface 403 and communication interface 404 are connected to each other within the device via bus 405.

[0085] According to a fourth aspect of this application, a computer-readable storage medium stores a computer program that, when executed by a processor, implements the method for recommending relocation housing resources according to any one of the first aspects of this application.

[0086] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0087] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0088] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0089] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0091] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0092] It should be understood that in this application, "at least one (item)" means one or more, and "more than one" means two or more. "And / or" is used to describe the mapping relationship between the mapped objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following mapped objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0093] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0094] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0096] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for recommending resettlement housing, characterized in that, include: Receive user resettlement request message, the user resettlement request message including user identity information, user family information, housing information before relocation, and first user demand information; Based on the user's identity information, a first set of selectable properties is determined from a preset property database; Based on the first user demand information, the first set of available properties is filtered to obtain a second set of available properties; Determine the candidate housing vector corresponding to each housing in the second set of optional housing from the preset housing vector library; The housing information before migration, the first user demand information, the user family information and the candidate housing vector are input into a pre-trained housing recommendation model to obtain the first recommended housing. The first recommended property and its surrounding amenities are displayed through a real-view browsing interface. Acquire user voice information during the display of the real-view browsing interface, and convert the user voice information into text to obtain first feedback information; Obtain the user's second feedback information regarding the first recommended property; The first feedback information and the second feedback information are input into a preset large language model to obtain the second user demand information; the first user demand information and the second user demand information are concatenated to obtain the third user demand information. The housing information before migration, the third user's demand information, the user's family information, and the candidate housing vector are input into the housing recommendation model to obtain the second recommended housing. The property recommendation model includes a first embedding layer, a second embedding layer, a third embedding layer, a first splicing layer, a second splicing layer, a third splicing layer, a first feature fusion layer, and a dot product layer; The pre-migration housing information, the first user demand information, the user family information, and the candidate housing vector are input into a pre-trained housing recommendation model to obtain the first recommended housing, including: Each field of the pre-migration housing information is input into the first embedding layer to obtain multiple first features, and the multiple first features are spliced ​​together through the first splicing layer to obtain a first spliced ​​feature; Each field of the first user demand information is input into the second embedding layer to obtain multiple second features, and the multiple second features are concatenated through the second concatenation layer to obtain a second concatenated feature; Each field of the user's family information is input into the third embedding layer to obtain multiple third features, and the multiple third features are then concatenated through the third splicing layer to obtain a third spliced ​​feature; The first splicing feature, the second splicing feature, and the third splicing feature are input into the first feature fusion layer to obtain the user demand feature; The matching score between the user demand feature and each candidate housing vector is calculated through the dot product layer, and the housing corresponding to the candidate housing vector with the largest matching score is taken as the first recommended housing. The property recommendation model further includes a fourth embedding layer, a second feature fusion layer, and a classifier. The training steps of the property recommendation model include: Acquire user demand training information, pre-migration housing training information, user family training information, real property tags, and description training information for multiple available properties; Each field of the pre-migration housing training information is input into the first embedding layer to obtain multiple first training features, and the multiple first training features are spliced ​​together through the first splicing layer to obtain first training spliced ​​features; Each field of the user demand training information is input into the second embedding layer to obtain multiple second training features, and the multiple second training features are concatenated through the second concatenation layer to obtain second training concatenation features; Each field of the user's family training information is input into the third embedding layer to obtain multiple third training features, and the multiple third training features are concatenated through the third concatenation layer to obtain third training concatenation features; The first training splicing feature, the second training splicing feature, and the third training splicing feature are input into the first feature fusion layer to obtain the user demand training feature; Each field of the description training information of each of the optional housing units is input into the fourth embedding layer to obtain multiple fourth training features; The fourth training feature corresponding to the same available property is input into the second feature fusion layer to obtain the property training vector corresponding to the available property. The matching training score between the user demand training features and each of the housing listing training vectors is calculated using the dot product layer. The matching training scores are input into the classifier to obtain the recommended housing listings. Based on the recommended properties, the actual labels of the properties, and the preset loss function, the loss value is calculated. The property recommendation model is iteratively optimized based on the loss value to obtain the trained property recommendation model.

2. The method for recommending resettlement housing according to claim 1, characterized in that, The candidate housing vector is obtained through the following steps: Obtain the description information of each property from the property database; Each field of the description information is input into the fourth embedding layer of the trained housing recommendation model to obtain multiple fourth features; Each of the fourth features is input into the second feature fusion layer of the trained housing recommendation model to obtain the candidate housing vector.

3. The method for recommending resettlement housing according to claim 1, characterized in that, The processing steps of the first feature fusion layer or the second feature fusion layer include: Perform dimension mapping operations on multiple input features to obtain multiple mapped features with the same dimensions; Each of the mapping features is input into the attention unit of a multilayer perceptron with shared parameters to obtain the attention score corresponding to each mapping feature; Based on each attention score, weight normalization is performed to obtain the attention weight of each mapping feature; The weighted summation feature is obtained by performing a weighted summation process on each of the mapping features and the corresponding attention weights. The weighted summation features are input into a fully connected layer to obtain fused features, which are then used as the output features of the feature fusion layer.

4. The method for recommending resettlement housing according to claim 1, characterized in that, The first feedback information and the second feedback information are input into a preset large language model to obtain the second user requirement information, including: The first feedback information and the second feedback information are concatenated based on preset prompt words to obtain concatenated text; the prompt words are used to limit the output content and format of the large language model. The concatenated text is input into the large language model to obtain the second user requirement information.

5. The method for recommending resettlement housing according to claim 1, characterized in that, Obtain second feedback information from the user regarding the first recommended property, including: Obtain negative descriptions of the first recommended property; Each negative field of the negative description information is displayed to the user via a pop-up window; In response to detecting an operation command for the pop-up window, a target negative field is determined from each of the negative fields based on the operation command, and each of the target negative fields is used as the second feedback information.

6. A device for recommending resettlement housing, characterized in that, Recommended method for implementing the relocation and resettlement housing resources as described in any one of claims 1 to 5; include: The receiving module is used to receive user resettlement request messages, which include user identity information, user family information, pre-migration housing information, and first user demand information; The first determining module is used to determine a first set of selectable properties from a preset property database based on the user's identity information; The second determining module is used to filter the first set of available housing based on the first user demand information to obtain a second set of available housing. The third determining module is used to determine the candidate housing vector corresponding to each housing in the second optional housing set from the preset housing vector library; The first input module is used to input the pre-migration housing information, the first user demand information, the user family information and the candidate housing vector into a pre-trained housing recommendation model to obtain the first recommended housing; The display module is used to display the first recommended property and its surrounding facilities through a real-view browsing interface; The first acquisition module is used to acquire user voice information during the display of the real-view browsing interface, and convert the user voice information into text to obtain first feedback information. The second acquisition module is used to acquire second feedback information from the user regarding the first recommended property; The second input module is used to input the first feedback information and the second feedback information into a preset large language model to obtain the second user demand information; and to concatenate the first user demand information and the second user demand information to obtain the third user demand information. The third input module is used to input the pre-migration housing information, the third user demand information, the user family information, and the candidate housing vector into the housing recommendation model to obtain the second recommended housing.

7. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the recommended method for relocation housing resources as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the recommended method for relocation and resettlement housing as described in any one of claims 1 to 5.

Images