A digital home supply chain e-commerce category management recommendation system and method

Abstract

The present invention discloses a digital home supply chain e-commerce category management and recommendation system, including: a portrait construction unit, which is used to collect biological signals of different target objects in real time and establish a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity; a comprehensive management and control unit, which is connected to the portrait construction unit in communication, and the comprehensive management and control unit is used to receive each of the simulated portrait data chains, and analyze each of the simulated portrait data chains to evaluate the target demand. The present invention sets the recommendation system in the home commodity display area of ​​home sales, and draws user portraits by obtaining the user's biological signals, thereby analyzing the user's real needs through the user portrait, and making customized recommendations to the user to increase sales.

Description

Technical Field

[0001] The present invention relates to the field of e-commerce technology, and in particular to a digital home supply chain e-commerce product category management recommendation system and method. Background Art

[0002] When purchasing home furnishing products, users usually choose to go to physical showrooms to experience and feel them. Therefore, home furnishing products require a large physical space for display, which in turn restricts the product selection and purchasing model of home furnishing product e-commerce. It is impossible to grasp customer needs and preferences through online browsing, and therefore it is impossible to analyze and understand customers and then complete customized recommendations, resulting in customer loss and reduced sales.

[0003] In the prior art, CN201711258520.2 discloses a design method for an information-based home furnishing online mall intelligent recommendation system, which uses computer software technology to develop system programs, uses WEB technology to develop system servers, and uses database technology to store and manage system information data. The main functions of the system include display of furniture products, collection and analysis of target needs and market information, and recommendation of product information. Through the development and implementation of system functions, merchants can collect and analyze user preferences and industry dynamic data of the market through the Internet, and by discovering users' potential needs, they can increase the sales rate of products.

[0004] Although CN201711258520.2 can analyze the target's real needs based on the user's browsing history, it cannot satisfy the user's real experience, and therefore cannot accurately capture the user's preferences, and it is still difficult to attract customers and increase sales. Summary of the invention

[0005] The purpose of the present invention is to provide a digital home supply chain e-commerce category management and recommendation system to solve the technical problems in the prior art that the system cannot satisfy the user's real experience, and thus cannot accurately capture the user's preferences, and is still difficult to attract customers and increase sales.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0007] A digital home supply chain e-commerce category management recommendation system, including:

[0008] A portrait construction unit, used for collecting biological signals of different target objects in real time and establishing a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity;

[0009] A comprehensive management and control unit is communicatively connected to the portrait construction unit, and is used to receive each of the simulated portrait data chains and analyze each of the simulated portrait data chains to evaluate target requirements;

[0010] A demand model building unit, which is in communication connection with the overall management and control unit, and is used to receive the target demand evaluated by the overall management and control unit, and build a target commodity demand model according to the target demand;

[0011] The customized recommendation unit is communicatively connected with the demand model building unit, and is used to customize different real-life experience displays for different types of target objects based on the target commodity demand model.

[0012] As a preferred solution of the present invention, the portrait construction unit is arranged in the home goods display area, and the portrait construction unit includes a collection unit for collecting biological signals of the target object, and a building unit for building a simulated portrait data chain of the target object in communication connection with the data transmission end of the collection unit, and a storage unit for providing storage, update and reading functions for the simulated portrait data chain. The collection unit includes a stress sensor assembly laid on the ground of the home goods display area for collecting biological signals of the target object. The specific method of collecting biological signals of the target object to establish the simulated portrait data chain is:

[0013] The collection unit collects biological signals to mark the target object and records the behavior path of the target object in the home product display area;

[0014] The category keywords of the target product are extracted from the behavior path, and the category keywords are stored in layers to form a single intention layer. All intention layers are connected in the form of a linked list to maintain the radial depth of the behavior path to form a simulated portrait data chain.

[0015] As a preferred solution of the present invention, the specific method for the overall management and control unit to analyze each of the simulated portrait data chains to evaluate the target demand includes:

[0016] The intentionality of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the intentionality is:

[0017]

[0018] The relevance of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the relevance is:

[0019]

[0020] The attention degree of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the attention degree is:

[0021]

[0022] The global search algorithm is used to obtain the target demand of the target object on the objective function jointly established by the intention, relevance and attention. The calculation formula of the objective function is:

[0023]

[0024] Among them, a i , a j is the category keyword of the i,jth intention layer, is all pairs of a i The set of keywords of the intent layer with outgoing links, L(a j ) is a j The number of outgoing links in the intention layer, T ai for a i The number of intention layers, N is the total number of intention layers, [y1(a j )]、[y2(a j )]、[y3(a j )] is a correction function for minimizing intention, relevance and attention.

[0025] As a preferred solution of the present invention, the target commodity demand model is used to cluster all target objects with the same target demand into clusters, and the output result of the target commodity demand model is a plurality of target object clusters. The specific method of the demand model construction unit to construct the target commodity demand model according to the target demand includes:

[0026] Quantify the target demands of all target objects into a target commodity demand sample set, where one target object is represented as a target commodity demand sample, and the target demand is represented as a sample feature;

[0027] Performing feature dimension reduction on the target commodity demand sample set to obtain key feature quantities, and updating the target commodity demand sample set according to the key feature quantities;

[0028] Randomly extract 50%, 30% and 20% of the data from the target commodity demand sample set as a training set, a test set and a validation set;

[0029] The training set, test set and validation set are used to model multiple groups of target commodity demand model models on a variety of machine learning clustering algorithms, and the optimal target commodity demand model is selected from the multiple groups of target commodity demand models.

[0030] As a preferred solution of the present invention, the specific method for performing feature dimension reduction on the target commodity demand sample set to obtain key feature quantities is:

[0031] All features quantified from all target demands in the target commodity demand sample set are paired in pairs to obtain double feature pairs, and the feature values ​​of the double feature pairs are brought into the similarity calculation formula to obtain the similarity coefficient of the double feature pairs. The similarity calculation formula is:

[0032]

[0033] Among them, X is the feature set, x i , x j is the eigenvalue of the i-th, j-th feature in the feature set, p(x i ,x j ) is x i and x j The joint probability distribution function of i ) and p(x j ) are x i and x j The marginal probability distribution function of

[0034] Traversing and selecting the first double feature pair whose similarity coefficient is higher than the similarity threshold, and randomly selecting the first feature in the first double feature pair, and correspondingly deleting the first features in all double feature pairs until the traversal of the first double feature pair is completed;

[0035] The features obtained by pairing and splitting all double feature pairs are summarized to form the key feature quantity.

[0036] As a preferred solution of the present invention, the optimal target commodity demand model is selected from the target commodity demand model based on the clustering evaluation index. The specific method includes:

[0037] Draw cluster evaluation index curves of multiple groups of target commodity demand models in sequence;

[0038] Compare the values ​​of clustering evaluation index curves of multiple groups of target commodity demand models in turn, and select the one corresponding to the maximum value of the vertex of the clustering evaluation index curve as the optimal target commodity demand model;

[0039] Among them, clustering evaluation indicators include modularity density and normalized mutual information.

[0040] As a preferred solution of the present invention, the customized recommendation unit includes a plurality of real-scene model display units installed in the home display area, and the real-scene model generation unit is communicatively connected to the real-scene model display unit, the real-scene model display unit includes a real-scene green screen and a projection device assembly, and the real-scene display unit is also provided with a second acquisition unit for collecting biological signals of the target object, and the second acquisition unit is communicatively connected to the real-scene model generation unit. The specific method for constructing the customized recommendation unit includes:

[0041] The second collecting unit collects the biological signal of the target object entering the real scene model display unit, and feeds it back to the real scene model generation unit;

[0042] The real scene model generation unit calibrates the target object cluster according to the biological signal of the target object, and establishes a biological human model of the target object based on the biological signal, and uses the biological human model to customize and modify the target requirements of the target object cluster to construct an experience real scene model and synchronously feed back to the real scene model display unit;

[0043] The real-scene model display unit displays the real scene to the target object on the real-scene green screen according to the experience real-scene model.

[0044] As a preferred solution of the present invention, the biological signal includes the foot area and standing force of the target object collected by the collection unit, and the specific method of establishing the biological human body model includes:

[0045] The biological height data of the user is calculated based on the area of ​​both feet, and the biological weight data of the user is calculated based on the standing force;

[0046] The biological body model of the target object is constructed by combining the biological height data and the biological weight data using the human body modeling algorithm.

[0047] As a preferred solution of the present invention, the specific method of using the biological human body model to customize and modify the target needs to construct an experience reality model includes:

[0048] The sizes of the benchmark model of the target requirement and the benchmark model of the real application scenario of the target requirement are proportionally modified according to the biological human body model to obtain a customized model of the target requirement;

[0049] For customized models that meet target needs, we select real-life application scenario models of equal proportions and jointly build an experiential real-life model that matches the real home environment.

[0050] As a preferred solution of the present invention, the present invention provides a recommendation method according to the digital home supply chain e-commerce category management recommendation system, comprising the following steps:

[0051] Step S1, the portrait construction unit collects biological signals of different target objects in real time and establishes a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity;

[0052] Step S2, the overall management and control unit receives each of the simulated portrait data chains, and analyzes each of the simulated portrait data chains to evaluate target requirements;

[0053] Step S3: the demand model building unit receives the target demand evaluated by the overall management and control unit, and builds a target commodity demand model according to the target demand;

[0054] Step S4: The customized recommendation unit customizes different real-life experience displays for different types of target objects based on the target product demand model.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The present invention sets the recommendation system in the home furnishing commodity display area of ​​home furnishing sales, and draws a user portrait by acquiring the user's biological signal, thereby analyzing the user's real needs through the user portrait, and making customized recommendations to the user to increase sales. At the same time, different real-scene experience displays are customized for users through real-scene construction technology, which is intuitive and can quickly obtain the real scene of the home furnishing commodity in the user's daily use, so that the user can have a deeper perception of the commodity, which can not only enable the user to obtain a better shopping experience, but also enable the user and the commodity supplier to obtain the best benefits as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0058] Figure 1 A system structure block diagram provided for an embodiment of the present invention;

[0059] Figure 2 A flow chart of a method provided by an embodiment of the present invention.

[0060] The numbers in the figure represent the following:

[0061] 1-Portrait building unit; 2-Overall management and control unit; 3-Demand model building unit; 4-Customized recommendation unit. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] like Figure 1 As shown, a digital home supply chain e-commerce category management recommendation system includes:

[0064] A portrait construction unit, used for collecting biological signals of different target objects in real time and establishing a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity;

[0065] The portrait construction unit is arranged in the home goods display area, and the portrait construction unit includes a collection unit for collecting biological signals of the target object, a building unit for building a simulated portrait data chain of the target object in communication with the data transmission end of the collection unit, and a storage unit for providing storage, update and reading functions for the simulated portrait data chain. The collection unit includes a stress sensor assembly laid on the ground of the home goods display area for collecting biological signals of the target object. The specific method of collecting biological signals of the target object to establish the simulated portrait data chain is:

[0066] The collection unit collects biological signals to mark the target object and records the behavior path of the target object in the home product display area;

[0067] Among them, the target object is the shopping user who enters the home furnishing product display area. The target object entering the home furnishing product display area is marked by using unique biological signals to ensure the accuracy of the marking. The height, weight and other data of the target object can also be obtained through biological signals, and then a biological human body model is constructed as a customization benchmark for the customized experience scene, which will increase the shopping favorability of the target object and enable it to better feel the advantages and disadvantages of the product for itself.

[0068] The category keywords of the target product are extracted from the behavior path, and the category keywords are stored in layers to form a single intention layer. All intention layers are connected in the form of a linked list to maintain the radial depth of the behavior path to form a simulated portrait data chain.

[0069] The simulated portrait data chain established by the portrait construction unit has an update cycle, which is determined based on the actual application scenario and can be one day, one week, or other orders of magnitude, all within the protection scope of this embodiment.

[0070] The storage unit will save the simulation portrait data chain of all objects in the current cycle and the simulation portrait data chain of the previous cycle. If the simulation portrait data chain of the target object in the current cycle exists, the simulation portrait data chain of the current cycle will be saved. If the simulation portrait data chain of the target object in the current cycle does not exist, the simulation portrait data chain of the previous cycle will still be selected as the simulation portrait data chain of the current cycle, and the simulation portrait data chain of the current cycle will be transmitted to the overall management and control unit.

[0071] A comprehensive management and control unit is communicatively connected to the portrait construction unit, and is used to receive each of the simulated portrait data chains and analyze each of the simulated portrait data chains to evaluate target requirements;

[0072] The specific method by which the overall control unit analyzes each of the simulated image data chains to evaluate target requirements includes:

[0073] The intentionality of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the intentionality is:

[0074]

[0075] The relevance of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the relevance is:

[0076]

[0077] The attention degree of the target object is extracted and quantified on the simulated portrait data chain, and the calculation formula of the attention degree is:

[0078]

[0079] The global search algorithm is used to obtain the target demand of the target object on the objective function jointly established by the intention, relevance and attention. The calculation formula of the objective function is:

[0080]

[0081] Among them, a i , a j is the category keyword of the i,jth intention layer, is all pairs of a i The set of keywords of the intent layer with outgoing links, L(a j ) is a j The number of outgoing links in the intention layer, T ai for a i The number of intention layers, N is the total number of intention layers, [y1(a j )]、[y2(a j)]、[y3(a j )] is a correction function for minimizing intention, relevance and attention.

[0082] The problem of extracting the target needs of the target object in the simulation portrait data chain is converted into a multi-objective search problem. The intention, relevance and attention between the types of goods in the behavior path can be discovered from the simulation portrait data chain summarized from the disordered behavior path of the target object. The target needs that best meet the real needs of the target object can be searched by solving the objective function that represents the trade-off between intention, relevance and attention. Therefore, goods can be recommended to the target object based on the target needs, which effectively improves the recommendation success rate.

[0083] The acquisition of target requirements is synchronized and updated periodically along with the simulation portrait data link.

[0084] A demand model building unit, which is in communication connection with the overall management and control unit, and is used to receive the target demand evaluated by the overall management and control unit, and build a target commodity demand model according to the target demand;

[0085] The target commodity demand model is used to cluster all target objects with the same target demand into clusters. The output result of the target commodity demand model is multiple target object clusters, and a unified experience scene generation can be performed for multiple target objects in the same target object cluster at the same time. The specific method for the demand model construction unit to construct the target commodity demand model according to the target demand includes:

[0086] Quantify the target demands of all target objects into a target commodity demand sample set, where one target object is represented as a target commodity demand sample, and the target demand is represented as a sample feature;

[0087] Performing feature dimension reduction on the target commodity demand sample set to obtain key feature quantities, and updating the target commodity demand sample set according to the key feature quantities;

[0088] Randomly extract 50%, 30% and 20% of the data from the target commodity demand sample set as a training set, a test set and a validation set;

[0089] The training set, test set and validation set are used to model multiple groups of target commodity demand model models on a variety of machine learning clustering algorithms, and the optimal target commodity demand model is selected from the multiple groups of target commodity demand models.

[0090] The specific method for performing feature dimensionality reduction on the target commodity demand sample set to obtain key feature quantities is:

[0091] All features quantified from all target demands in the target commodity demand sample set are paired in pairs to obtain double feature pairs, and the feature values ​​of the double feature pairs are brought into the similarity calculation formula to obtain the similarity coefficient of the double feature pairs. The similarity calculation formula is:

[0092]

[0093] Among them, X is the feature set, x i , x j is the eigenvalue of the i-th, j-th feature in the feature set, p(x i ,x j ) is x i and x j The joint probability distribution function of i ) and p(x j ) are x i and x j The marginal probability distribution function of .

[0094] Traversing and selecting the first double feature pair whose similarity coefficient is higher than the similarity threshold, and randomly selecting the first feature in the first double feature pair, and correspondingly deleting the first features in all double feature pairs until the traversal of the first double feature pair is completed;

[0095] The features obtained by pairing and splitting all double feature pairs are summarized to form the key feature quantity.

[0096] The similarity coefficient refers to the degree of similarity between two features. The similarity coefficient is used to quantify the similarity between the first-order features. Then, multiple features with high similarity between the first-order features can be represented by only one feature, thereby achieving feature dimensionality reduction while avoiding deletion of valid features, reducing computational complexity and improving efficiency.

[0097] The optimal target commodity demand model is selected from the target commodity demand model based on the clustering evaluation index, and the specific method includes:

[0098] Draw cluster evaluation index curves of multiple groups of target commodity demand models in sequence;

[0099] Compare the values ​​of clustering evaluation index curves of multiple groups of target commodity demand models in turn, and select the one corresponding to the maximum value of the vertex of the clustering evaluation index curve as the optimal target commodity demand model;

[0100] Among them, clustering evaluation indicators include modularity density and normalized mutual information.

[0101] Specifically, assuming

[0102]

[0103]

[0104] Among them, Y k , Y j are single samples in the same target object cluster, Y s For Y k , Y j A single sample in different target object clusters, M is Y k , Y j The total number of samples in the target object cluster, N is Y s The total number of samples in the target object cluster, R is the total number of target object clusters, and T is the transposition operator.

[0105] The specific calculation formula for modularity density is: The specific calculation formula of mutual information is: Y is the positive sample Y k The true division.

[0106] The larger the D value is, the higher the compactness of the target objects within the target object cluster divided by the optimal target commodity demand model is, that is, the higher the similarity is, so that the distribution network devices divided into a target object cluster have consistent target requirements.

[0107] The larger the MNI is, the higher the similarity between the target demand of the target object in the target object cluster divided by the optimal target commodity demand model and the actual target demand is, and the higher the evaluation accuracy of the optimal target commodity demand model is.

[0108] The optimal target commodity demand model is updated periodically with the target demand

[0109] The customized recommendation unit is communicatively connected with the demand model building unit, and is used to customize different real-life experience displays for different types of target objects based on the target commodity demand model.

[0110] The customized recommendation unit includes a plurality of real-scene model display units installed in the home display area, and the real-scene model generation unit is communicatively connected to the real-scene model display unit, the real-scene model display unit includes a real-scene green screen and a projection device assembly, and the real-scene display unit is also provided with a second acquisition unit for collecting biological signals of the target object, and the second acquisition unit is communicatively connected to the real-scene model generation unit. The specific method for constructing the customized recommendation unit includes:

[0111] The second collecting unit collects the biological signal of the target object entering the real scene model display unit, and feeds it back to the real scene model generation unit;

[0112] The real scene model generation unit calibrates the target object cluster according to the biological signal of the target object, and establishes a biological human model of the target object based on the biological signal, and uses the biological human model to customize and modify the target requirements of the target object cluster to construct an experience real scene model and synchronously feed back to the real scene model display unit;

[0113] The real-scene model display unit displays the real scene to the target object on the real-scene green screen according to the experience real-scene model.

[0114] At the same time, all target objects acquired by all second acquisition models are clustered into clusters according to the target commodity demand model, so that the target demand benchmark model and the target demand real application scenario benchmark model are uniformly selected for multiple target objects in the same target object cluster, avoiding the one-by-one selection of the target demand benchmark model and the target demand real application scenario benchmark model for each target object, thereby improving the system operation efficiency.

[0115] The biological signal includes the area of ​​both feet and the standing force of the target object collected by the collection unit. The specific method of establishing the biological human body model includes:

[0116] The biological height data of the user is calculated based on the area of ​​both feet, and the biological weight data of the user is calculated based on the standing force;

[0117] The biological body model of the target object is constructed by combining the biological height data and the biological weight data using the human body modeling algorithm.

[0118] The specific method of using the biological human body model to customize and modify the target needs to construct an experience reality model includes:

[0119] The sizes of the benchmark model of the target requirement and the benchmark model of the real application scenario of the target requirement are proportionally modified according to the biological human body model to obtain a customized model of the target requirement;

[0120] Customized models that meet target needs are assembled and constructed into experiential real-life models that match the real home environment.

[0121] The sizes of the benchmark model of the target demand and the benchmark model of the real application scenario of the target demand are modified in proportion to the biological human body model. The target object can be set up with a real experience scene that belongs only to itself, so that the object can better experience the advantages and disadvantages of the product and make better product selection. At the same time, the object can also have a customized service experience, which helps home furnishing suppliers lock in customers and avoid customer loss.

[0122] Based on the above-mentioned digital home supply chain electronic product category management recommendation system, the present invention provides a recommendation method, including the following steps:

[0123] Step S1, the portrait construction unit collects biological signals of different target objects in real time and establishes a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity;

[0124] Step S2, the overall management and control unit receives each of the simulated portrait data chains, and analyzes each of the simulated portrait data chains to evaluate target requirements;

[0125] Step S3: the demand model building unit receives the target demand evaluated by the overall management and control unit, and builds a target commodity demand model according to the target demand;

[0126] Step S4: The customized recommendation unit customizes different real-life experience displays for different types of target objects based on the target product demand model.

[0127] The present invention sets the recommendation system in the home furnishing commodity display area of ​​home furnishing sales, and draws a user portrait by acquiring the user's biological signal, thereby analyzing the user's real needs through the user portrait, and making customized recommendations to the user to increase sales. At the same time, different real-scene experience displays are customized for users through real-scene construction technology, which is intuitive and can quickly obtain the real scene of the home furnishing commodity in the user's daily use, so that the user can have a deeper perception of the commodity, which can not only enable the user to obtain a better shopping experience, but also enable the user and the commodity supplier to obtain the best benefits as a whole.

[0128] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A digital home supply chain e-commerce category management and recommendation system, characterized in that: include; A portrait construction unit, used for collecting biological signals of different target objects in real time and establishing a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity; A comprehensive management and control unit is communicatively connected to the portrait construction unit, and is used to receive each of the simulated portrait data chains and analyze each of the simulated portrait data chains to evaluate target requirements; A demand model building unit, which is in communication connection with the overall management and control unit, and is used to receive the target demand evaluated by the overall management and control unit, and build a target commodity demand model according to the target demand; A customized recommendation unit, which is in communication connection with the demand model building unit, and is used to customize different real-life experience displays for different types of target objects according to the target commodity demand model; The specific methods by which the overall control unit analyzes each simulated image data link to assess target requirements include: The intentionality of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula of the intentionality is: The relevance of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula for the relevance is: The attention degree of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula of the attention degree is: The global search algorithm is used to obtain the target demand of the target object based on the objective function jointly established by intention, relevance and attention. The calculation formula of the objective function is: Among them, a i , a j is the category keyword of the i, jth intention layer, M ai is all pairs of a i The set of keywords of the intent layer with outgoing links, L(a j ) is a j The number of outgoing links in the intention layer, T ai for a i The number of intention layers, N is the total number of intention layers, [y1(a j )]、[y2(a j )]、[y3(a j )] is a correction function for minimizing intention, relevance and attention.

2. According to claim 1, a digital home supply chain e-commerce category management and recommendation system is characterized by: The portrait construction unit is arranged in the home goods display area, and the portrait construction unit includes a collection unit for collecting biological signals of the target object, a building unit for building a simulated portrait data chain of the target object in communication with the data transmission end of the collection unit, and a storage unit for providing storage, update and reading functions for the simulated portrait data chain. The collection unit includes a stress sensor assembly laid on the ground of the home goods display area for collecting biological signals of the target object. The specific method of collecting biological signals of the target object to establish the simulated portrait data chain is: The collection unit collects biological signals to mark the target object and records the behavior path of the target object in the home product display area; The category keywords of the target product are extracted from the behavior path, and the category keywords are stored in layers to form a single intention layer. All intention layers are connected in the form of a linked list to maintain the radial depth of the behavior path to form a simulated portrait data chain.

3. A digital home supply chain e-commerce category management and recommendation system according to claim 2, characterized in that: The target commodity demand model is used to cluster all target objects with the same target demand into clusters. The output result of the target commodity demand model is a plurality of target object clusters. The specific method of the demand model construction unit to construct the target commodity demand model according to the target demand includes: Quantify the target demands of all target objects into a target commodity demand sample set, where one target object is represented as a target commodity demand sample, and the target demand is represented as a sample feature; Performing feature dimension reduction on the target commodity demand sample set to obtain key feature quantities, and updating the target commodity demand sample set according to the key feature quantities; Randomly extract 50%, 30% and 20% of the data from the target commodity demand sample set as a training set, a test set and a validation set; The training set, test set and validation set are used to model multiple groups of target commodity demand model models on a variety of machine learning clustering algorithms, and the optimal target commodity demand model is selected from the multiple groups of target commodity demand models.

4. A digital home supply chain e-commerce category management and recommendation system according to claim 3, characterized in that: The specific method for performing feature dimensionality reduction on the target commodity demand sample set to obtain key feature quantities is: All features quantified from all target demands in the target commodity demand sample set are paired in pairs to obtain double feature pairs, and the feature values ​​of the double feature pairs are brought into the similarity calculation formula to obtain the similarity coefficient of the double feature pairs. The similarity calculation formula is: Among them, X is the feature set, x i , x j is the eigenvalue of the i-th, j-th feature in the feature set, p(x i ,x j ) is x i and x j The joint probability distribution function of i ) and p(x j ) are x i and x j The marginal probability distribution function of Traversing and selecting the first double feature pair whose similarity coefficient is higher than the similarity threshold, and randomly selecting the first feature in the first double feature pair, and correspondingly deleting the first features in all double feature pairs until the traversal of the first double feature pair is completed; The features obtained by pairing and splitting all double feature pairs are summarized to form the key feature quantity.

5. A digital home supply chain e-commerce category management and recommendation system according to claim 4, characterized in that: The optimal target commodity demand model is selected from the target commodity demand model based on the clustering evaluation index, and the specific method includes: Draw cluster evaluation index curves of multiple groups of target commodity demand models in sequence; Compare the values ​​of clustering evaluation index curves of multiple groups of target commodity demand models in turn, and select the one corresponding to the maximum value of the vertex of the clustering evaluation index curve as the optimal target commodity demand model; Among them, clustering evaluation indicators include modularity density and normalized mutual information.

6. A digital home supply chain e-commerce category management and recommendation system according to claim 5, characterized in that: The customized recommendation unit includes a plurality of real-scene model display units installed in the home display area, and a real-scene model generation unit that is communicatively connected to the real-scene model display unit. The real-scene model display unit includes a real-scene green screen and a projection device assembly. The real-scene model display unit is also provided with a second acquisition unit for acquiring biological signals of the target object, and the second acquisition unit is communicatively connected to the real-scene model generation unit. The specific method for constructing the customized recommendation unit includes: The second acquisition unit acquires the biological signal of the target object entering the real-scene model display unit, and feeds it back to the real-scene model generation unit; The real scene model generation unit calibrates the target object cluster according to the biological signal of the target object, and establishes a biological human model of the target object based on the biological signal, and uses the biological human model to customize and modify the target requirements of the target object cluster to construct an experience real scene model and synchronously feed back to the real scene model display unit; The real-scene model display unit displays the real scene to the target object on the real-scene green screen according to the experience real-scene model.

7. A digital home supply chain e-commerce category management and recommendation system according to claim 6, characterized in that: The biological signal includes the foot area and standing force of the target object collected by the collection unit. The specific method of establishing the biological human body model includes: The biological height data of the user is calculated based on the area of ​​both feet, and the biological weight data of the user is calculated based on the standing force; The biological body model of the target object is constructed by combining the biological height data and the biological weight data using the human body modeling algorithm.

8. A digital home supply chain e-commerce category management and recommendation system according to claim 7, characterized in that: The specific method of using the biological human body model to customize and modify the target needs to construct an experience reality model includes: The sizes of the benchmark model of the target requirement and the benchmark model of the real application scenario of the target requirement are proportionally modified according to the biological human body model to obtain a customized model of the target requirement; For customized models that meet target needs, we select real-life application scenario models of equal proportions and jointly build an experiential real-life model that matches the real home environment.

9. A recommendation method of the digital home supply chain e-commerce category management recommendation system according to any one of claims 1 to 8, characterized in that: The following steps are involved: Step S1, the portrait construction unit collects biological signals of different target objects in real time and establishes a simulated portrait data chain for recording and representing the target demand of each target object for the target commodity; Step S2, the overall management and control unit receives each of the simulated portrait data chains, and analyzes each of the simulated portrait data chains to evaluate target requirements; Step S3: the demand model building unit receives the target demand evaluated by the overall management and control unit, and builds a target commodity demand model according to the target demand; Step S4: the customized recommendation unit customizes different real-life experience displays for different types of target objects according to the target product demand model; The specific methods by which the overall control unit analyzes each simulated image data link to assess target requirements include: The intentionality of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula of the intentionality is: The relevance of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula for the relevance is: The attention degree of the target object is extracted and quantified on the simulated portrait data chain. The calculation formula of the attention degree is: The global search algorithm is used to obtain the target demand of the target object based on the objective function jointly established by intention, relevance and attention. The calculation formula of the objective function is: Among them, a i , a j is the category keyword of the i,jth intention layer, is all pairs of a i The set of keywords of the intent layer with outgoing links, L(a j ) is a j The number of outgoing links in the intention layer, T ai for a i The number of intention layers, N is the total number of intention layers, [y1(a j )]、[y2(a j )]、[y3(a j )] is a correction function for minimizing intention, relevance and attention.

Images