Intelligent seat arrangement system collaborative decision-making method and device based on machine learning model, and storage medium
By using an intelligent seating arrangement system based on machine learning models, combining deep neural networks and graph neural networks, and dynamically calculating and optimizing algorithms, a reasonable seating allocation scheme is generated. This solves the shortcomings of existing technologies in scene understanding and constraint handling, and achieves efficient and personalized seating arrangements.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 浙江中易慧能科技有限公司
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-15
AI Technical Summary
Existing technologies cannot dynamically understand and quantify the importance or closeness of relationships of personnel based on the scenario, making it difficult to efficiently and accurately meet various complex constraints. They also lack self-learning and adaptive mechanisms, resulting in a single and rigid optimization objective.
An intelligent seating arrangement system based on machine learning models is adopted. By acquiring scene type and personnel attribute features, it uses deep neural networks and graph neural networks for dynamic calculation, and combines rule-based and metaheuristic optimization algorithms to generate a seating allocation scheme that meets multiple objectives.
It achieves a deep understanding of the differentiated needs of different scenarios, accurately assesses the importance of personnel, and efficiently searches for the best seating allocation scheme in multiple aspects, overcoming the shortcomings of traditional methods in terms of intelligence and constraint processing.
Smart Images
Figure CN122048299A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent seat allocation technology, and in particular to a collaborative decision-making method, system and device for intelligent seat allocation systems based on machine learning models. Background Technology
[0002] Seating arrangements are a fundamental and crucial preparatory step in various social and organizational activities, including meetings, banquets, celebrations, classroom sessions, and transportation. A well-arranged seating arrangement not only relates to etiquette and hierarchy, and the efficiency of resource allocation, but also directly impacts the effectiveness of communication, the meeting atmosphere, and even the achievement of event goals. Traditional seating arrangements rely heavily on the organizer's personal experience, subjective judgment, and manual coordination; this process is typically time-consuming, labor-intensive, and struggles to simultaneously consider multiple complex factors.
[0003] With the development of information technology, some auxiliary seating management tools or simple algorithms have emerged. For example, rule-based automated tools allow for the pre-setting of simple rules (such as sorting by job title or alphabetical order of last name) and then automatically filling in the seats. However, they cannot adapt to the differentiated needs of different scenarios, let alone handle complex and flexible constraints such as requiring personnel A and B to be adjacent or departments C and D to be spaced apart. Manual drag-and-drop graphical tools provide a visual interface where users can drag and drop seats. While flexible, they have low levels of intelligence and are essentially still manual seating arrangements, failing to reduce the cognitive burden and decision-making costs for users. They are inefficient and prone to errors when dealing with large numbers of people. A few studies have attempted to apply operations research or simple greedy algorithms to seat allocation to optimize a single objective. However, these methods often simplify interpersonal relationships to numerical values, lacking the ability to learn from data and adapt adaptively. They cannot understand deep social relationships and contextual situations, and are difficult to integrate multiple objectives.
[0004] In summary, existing technologies generally suffer from the following prominent shortcomings: They cannot dynamically understand and quantify the importance or closeness of relationships among individuals based on the scenario, indicating a low level of intelligence; they struggle to efficiently and accurately meet complex constraints of various types and levels, such as mandatory seating or prohibition of adjacent seating, especially lacking effective solutions when constraints conflict, indicating weak constraint handling capabilities; they cannot continuously learn and optimize from historical arrangements or user feedback, treating each seating arrangement as an isolated task, failing to accumulate experience, indicating a lack of self-learning and adaptive mechanisms; and they struggle to achieve an intelligent balance among multiple competing objectives such as weight balancing, social harmony, and spatial layout, indicating a singular and rigid optimization objective.
[0005] Therefore, there is an urgent need in this field for an intelligent seating arrangement method that can deeply integrate scene understanding, possess relation learning capabilities, and perform multi-objective collaborative optimization, in order to overcome the above-mentioned shortcomings and achieve efficient, reasonable, and personalized automatic seating arrangements. Summary of the Invention
[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a collaborative decision-making method for intelligent seating systems based on machine learning models, comprising the following steps: S1: Obtain the scenario type selected by the user, load the corresponding weight parameter configuration template according to the scenario type, and obtain the set of personnel attribute features, table layout attribute set, and personnel relationship constraint set including the personnel to be seated; S2: Based on the scenario type, the weight parameter configuration template, and the personnel attribute feature set, a machine learning model is used to dynamically calculate and output the comprehensive weight score of each person waiting to be seated. A sequence of ranked personnel is generated based on the comprehensive weight score. S3: Using the sorted personnel sequence as the initial allocation order, the table layout attribute set as the spatial constraint, and the personnel relationship constraint set as the constraint condition that must be satisfied, a hybrid allocation strategy combining rules and metaheuristic optimization algorithms is adopted to allocate seats and iteratively optimize the personnel to be seated. By optimizing the preset multi-objective function, the final seat allocation scheme is output.
[0007] Preferably, step S2 includes: Extract the personnel attribute feature vector corresponding to the personnel attribute features, and the scene feature vector corresponding to the scene type; The personnel attribute feature vector is concatenated with the scene feature vector to form a combined feature vector; The combined feature vectors are input into a pre-trained deep neural network model, which performs nonlinear mapping and regression prediction to output the comprehensive weight score for each person waiting to be seated.
[0008] Preferably, in step S1, the set of relationship constraints between personnel is obtained, including: Construct a personnel relationship diagram, in which nodes correspond to personnel to be seated and edges correspond to the relationships between personnel. The relationships include at least one of historical co-seat records, organizational affiliation, or communication collaboration frequency. The personnel relationship graph is input into the graph neural network model along with the current scene features selected by the current user to obtain the potential relationship strength score between each pair of personnel and the corresponding constraint type suggestion. Set a relationship strength score threshold, and present relationships with potential relationship strength scores exceeding the threshold to the user. After user confirmation, add them to the set of interpersonal relationship constraints.
[0009] Preferably, the pre-training process of the deep neural network model includes: Collect historical seating arrangement cases, including personnel attribute characteristics, scenario types, and corresponding effective seating allocation schemes; Based on the seating arrangement scheme corresponding to each historical seating case, the relative importance of each person in the scheme is deduced in reverse, and this relative importance is used as the weight score label for training the deep neural network model. Using the personnel attribute characteristics and scene type in historical seating cases as model input, and the corresponding weight score labels as supervision signals, the deep neural network model is trained end-to-end. During training, regularization and cross-validation techniques are used to optimize the deep neural network model.
[0010] Preferably, in step S3, a hybrid allocation strategy is used for seat allocation, including: For each person waiting to be seated, check each of their corresponding candidate seats one by one to determine whether the candidate seat meets all the hard constraints related to that person; If the person belongs to a group that requires them to sit together, then the candidate seat must be at the same table and adjacent to the person who has already been assigned a seat in that group. If the person belongs to a group that cannot be seated together, the candidate seat cannot be at the same table or adjacent to any person already assigned in that group; After filtering out all candidate seats that meet the above hard constraints, the target seat for the person is selected from them according to preset heuristic rules. The preset heuristic rules include prioritizing filling tables that are not full, prioritizing seats near the head or with a good view, and prioritizing making the average comprehensive weight of the people at the current table close to the overall average comprehensive weight of all people. If a person waiting to be seated has no candidate seats that meet the hard constraints, the backtracking adjustment mechanism will be activated to adjust the seats of people who have already been assigned, or the user will be directly prompted that there is a conflict of relationship constraints between people and the user needs to adjust the constraint rules.
[0011] Preferably, the metaheuristic optimization algorithm includes: Define a state space and neighborhood operations, where the state space corresponds to a seating arrangement scheme for each state, and the neighborhood operations are to randomly swap the seats of people at two different tables or randomly rotate the seating order of several consecutive people at the same table. Take the initial feasible allocation scheme as the current state S_curr, and calculate the objective function value E_curr of the current state S_curr, E_curr = F(S_curr); In each iteration, a new seat allocation state S_new is generated by applying a random neighborhood operation to the current state S_curr, ensuring that the new seat allocation state S_new satisfies all hard constraints. A new objective function value E_new is calculated, E_new=F(S_new). The target energy difference ΔE is calculated based on the objective function value and the new objective function value, ΔE = E_new-E_curr. If the target energy difference ΔE<0, the new seat allocation state S_new is accepted as the current state, i.e., S_curr=S_new and E_curr=E_new. If the target energy difference ΔE≥0, the new seat allocation state S_new is accepted with probability P=exp(-ΔE / T'), where T' is the current temperature parameter. During the iteration process, the temperature parameter T' is gradually reduced according to the predefined annealing schedule; When the temperature parameter T' drops to the preset minimum threshold, or when no new state S_new is received for several consecutive iterations, the iteration loop terminates. At this time, the allocation scheme corresponding to the current state S_curr is the optimized final seat allocation scheme.
[0012] Preferably, the multi-objective optimization function includes: Where S represents a seating arrangement scheme, and T is the total number of tables. Let be the variance of the combined weighted scores of all people at the t-th table, and let Adj be the set of all adjacent seat pairs. To evaluate the social harmony function of individuals a and b sitting next to each other, its value is given by a pre-trained affinity model, where SoftC is the set of all soft constraints. The penalty function is used to measure the degree to which scheme S violates the soft constraint c. α, β, and γ are adjustable weighting coefficients used to balance the importance of different optimization objectives.
[0013] Preferably, after step S3, step S4 is further included: The final seating allocation scheme is presented graphically on the user interface, with different weighted personnel marked with different colors or sizes, and constraints indicated by lines or icons. It provides drag-and-drop functionality, allowing users to manually adjust the seating of any person; Record the complete adjustment sequence from the user's automated plan to the final confirmed plan; The adjusted sequence is used as reinforcement learning feedback to update the parameters of the deep neural network model and the graph neural network model.
[0014] Based on the same concept, the present invention also provides a computer device, characterized in that it includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor causes the processor to perform the steps of a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model as described in any one of the embodiments.
[0015] Based on the same concept, the present invention also provides a storage medium storing computer-readable instructions, characterized in that, when the computer-readable instructions are executed by one or more processors, the one or more processors cause the one or more processors to perform the steps of a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model as described in any one of the embodiments.
[0016] Compared with the prior art, the beneficial effects of the present invention are: (1) By acquiring the scene type selected by the user and loading the corresponding weight parameter configuration template, and constructing a personnel relationship diagram and using graph neural network to intelligently predict constraints, the present invention enables the system to deeply understand the differentiated needs of different seating scenarios and automatically mine the implicit interpersonal relationships, thereby significantly improving the scene adaptability and intelligence level of seating arrangement, and overcoming the defects of rigid rules and reliance on human experience in traditional methods.
[0017] (2) This invention integrates personnel attributes with scene features and inputs them into a pre-trained deep neural network model for dynamic weight calculation, thereby achieving accurate and quantitative evaluation of personnel importance or priority. This evaluation can be dynamically adjusted according to the scene context, thus providing a data-driven and objective ranking basis for subsequent allocation, and solving the limitations of fixed weights and inability to adaptively learn in traditional methods.
[0018] (3) This invention uses a sorted sequence as the initial order, layout and hard constraints as boundary conditions, and adopts a hybrid allocation strategy that combines rules and metaheuristic optimization algorithms. It also optimizes a well-defined multi-objective function to efficiently search for a globally superior seat allocation scheme in terms of weight balance, social affinity and other aspects, under the premise of strictly satisfying all user constraints. It achieves a comprehensive optimization goal while ensuring practicality, and overcomes the shortcomings of simple algorithms that have a single optimization goal and are difficult to handle complex constraints. Attached Figure Description
[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.
[0020] Figure 1 This is a flowchart of a collaborative decision-making method for an intelligent seating system based on a machine learning model, according to the present invention.
[0021] Figure 2 This is a flowchart illustrating the collaborative decision-making method of an intelligent seating arrangement system based on a machine learning model, according to the present invention. Figure 3 This is a conference room layout diagram for a collaborative decision-making method of an intelligent seating arrangement system based on a machine learning model, according to the present invention. Figure 4 This is a scenario setting diagram for a collaborative decision-making method for an intelligent seating system based on a machine learning model, according to the present invention. Figure 5 This is a personnel management diagram for a collaborative decision-making method of an intelligent seating system based on a machine learning model, according to the present invention. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Obviously, the described embodiments are only some, not all, of the embodiments described in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without creative effort are within the scope of protection of this application.
[0023] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a” and “an” used herein, and “the”, may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.
[0024] First Embodiment Please see Figure 1 and Figure 2 As shown, this embodiment provides a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model, including the following steps: S1: Obtain the scenario type selected by the user, load the corresponding weight parameter configuration template according to the scenario type, and obtain the set of personnel attribute features, table layout attribute set, and inter-personal relationship constraint set, including the personnel to be seated. Specifically, in this embodiment, taking an important internal strategic seminar as an example, there are approximately 30 participants, using 1 main table (10 people) and 3 secondary tables (6-7 people per table). Please refer to [link / reference]. Figure 3 , Figure 4 and Figure 5As shown, the system provides a graphical user interface for scene selection, personnel and table information management, constraint setting, result visualization, and manual adjustment. After the user selects a business meeting scene in the system interface, a predefined business meeting weight parameter configuration template is automatically loaded. The weight parameter configuration template defines that the position weight coefficient is dominant (e.g., CEO=1000, Vice President=800, Director=600, Manager=400, Employee=200), and the department weight is adjusted as a secondary factor (with a slight increase for core business departments). The allocation strategy is balanced by default, and the weight balance coefficient α in the optimization objective is set relatively high.
[0025] Please see Figure 3 As shown, the list of attendees is synchronized from the company's HR system, and attributes such as name, position, department, and years of service are automatically extracted to form a set of personnel attribute features. The administrator then draws the meeting room layout in the layout editor, with one rectangular main table (10 seats, facing the podium) and three round side tables (capacities of 7, 7, and 6 respectively) to generate a set of table layout attributes.
[0026] Preferably, in step S1, the set of relationship constraints between personnel is obtained, including: Construct a personnel relationship diagram, where nodes correspond to personnel to be seated and edges correspond to the relationships between personnel. These relationships include at least one of the following: historical shared seating records, organizational affiliation, or communication collaboration frequency. Input the personnel relationship graph and the current scene features selected by the current user into the graph neural network model GNN to obtain the potential relationship strength score between each pair of personnel and the corresponding constraint type suggestion. A relationship strength score threshold is set. Relationship suggestions with potential strength scores exceeding the threshold are presented to the user. After user confirmation, these are added to the set of inter-personal relationship constraints. Specifically, in this embodiment, the user explicitly sets hard constraints: the CEO must sit at the head of the table, and the head of the finance department and the head of the audit department cannot sit next to each other. The graph neural network model (GNN) analyzes the organizational structure and project collaboration data and recommends soft constraints: it is suggested that Manager Wang of Project A sit next to Senior Engineer Li, and that sales and marketing personnel sit alternately to promote communication. The administrator adopted some of the suggestions.
[0027] S2: Based on the scenario type, weight parameter configuration template and personnel attribute feature set, the machine learning model is used to dynamically calculate and output the comprehensive weight score of each person to be seated. The sequence of ranked personnel is generated according to the comprehensive weight score. Specifically, in this embodiment, the first part of the sequence is usually the CEO, vice president, key director, etc.
[0028] Preferably, step S2 includes: Extract personnel attribute feature vectors (such as [job code, department code, years of service]) corresponding to personnel attribute features, and scene feature vectors corresponding to scene types (such as business meetings); The personnel attribute feature vector is concatenated with the scene feature vector to form a combined feature vector; The combined feature vectors are input into a pre-trained deep neural network model (DNN). The DNN performs nonlinear mapping and regression prediction to output a comprehensive weight score for each person to be seated (for example, a non-executive director responsible for core innovative businesses may receive a higher score than some vice presidents). Specifically, in this embodiment, the DNN is trained based on massive amounts of historical meeting seating data and can understand that in strategic seminars, directors with cross-departmental experience may need to participate in core discussions more than vice presidents in a single area.
[0029] Preferably, the pre-training process of the deep neural network model includes: Collect historical seating arrangement cases, including personnel attribute characteristics, scenario types, and corresponding effective seating allocation schemes; Based on the seating arrangement scheme corresponding to each historical seating case, the relative importance of each person in the scheme is deduced in reverse, and this relative importance is used as the weight score label for training the deep neural network model. Using the personnel attribute characteristics and scene type in historical seating cases as model input, and the corresponding weight score labels as supervision signals, the deep neural network model is trained end-to-end. During training, regularization and cross-validation techniques are used to optimize the deep neural network model.
[0030] S3: Using the initial allocation order of the sorted personnel sequence, the set of table layout attributes as spatial constraints, and the set of personnel relationship constraints as mandatory constraints, a hybrid allocation strategy combining rules and metaheuristic optimization algorithms is employed to allocate seats and iteratively optimize the personnel to be seated. By optimizing a preset multi-objective function, the final seat allocation scheme is output. Specifically, in this embodiment, the allocation engine allocates seats according to the sorted sequence, starting with the CEO. First, the CEO is placed at the head of the main table (satisfying the hard constraint), and then other senior executives are allocated seats at the main table in turn, while checking constraints such as non-adjacency. After the main table is full, a round-robin method is used to allocate subsequent personnel to each secondary table, and attempts are made to satisfy the soft adjacency suggestions recommended by GNN. At this stage, all hard constraints are ensured to be satisfied, generating a feasible solution. The final seating table is as follows: the main table is for senior executives; each secondary table has overlapping functions (e.g., each table has R&D, sales, and operations personnel); Manager Wang and Senior Engineer Li, who are recommended to be adjacent, are arranged in adjacent positions at the same table; the finance and audit heads are assigned to different tables.
[0031] Preferably, in step S3, a hybrid allocation strategy is used for seat allocation, including: For each person waiting to be seated, check each of their corresponding candidate seats one by one to determine whether the candidate seat meets all the hard constraints related to that person; If the person belongs to a group that requires them to sit together, then the candidate seat must be at the same table and adjacent to the person who has already been assigned a seat in that group. If the person belongs to a group that cannot be seated together, the candidate seat cannot be at the same table or adjacent to any person already assigned in that group; After filtering out all candidate seats that meet the above hard constraints, the target seat for the person is selected from them according to the preset heuristic rules. The preset heuristic rules include prioritizing filling tables that are not full, prioritizing seats near the head or with a good view, and prioritizing making the average comprehensive weight of the people at the current table close to the overall average comprehensive weight of all people. If a person waiting to be seated has no candidate seats that meet the hard constraints, the backtracking adjustment mechanism will be activated to adjust the seats of people who have already been assigned, or the user will be directly prompted that there is a conflict of relationship constraints between people and the user needs to adjust the constraint rules.
[0032] Preferably, the metaheuristic optimization algorithm includes: Define a state space and neighborhood operations, where the state space corresponds to a seating arrangement scheme for each state, and the neighborhood operations are to randomly swap the seats of people at two different tables or randomly rotate the seating order of several consecutive people at the same table. Take the initial feasible allocation scheme as the current state S_curr, and calculate the objective function value E_curr of the current state S_curr, E_curr = F(S_curr); In each iteration, a new seat allocation state S_new is generated by applying a random neighborhood operation to the current state S_curr, ensuring that the new seat allocation state S_new satisfies all hard constraints. The new objective function value E_new is calculated, E_new=F(S_new). The target energy difference ΔE is calculated based on the objective function value and the new objective function value, ΔE = E_new - E_curr. If the target energy difference ΔE<0, the new seat allocation state S_new is accepted as the current state, i.e., S_curr=S_new and E_curr=E_new. If the target energy difference ΔE ≥ 0, the new seat allocation state S_new is accepted with probability P=exp(-ΔE / T'), where T' is the current temperature parameter. During the iteration process, the temperature parameter T' is gradually reduced according to the predefined annealing schedule; When the temperature parameter T' drops to the preset minimum threshold, or when no new state S_new is accepted for several consecutive iterations, the iteration loop terminates. At this point, the allocation scheme corresponding to the current state S_curr is the optimized final seating allocation scheme. Specifically, in this embodiment, the algorithm randomly swaps two people at different tables (ensuring no violation of hard constraints) and calculates the objective function value of the new scheme. If the new scheme is better (F value decreases), it is accepted; if it is worse, it is accepted with a certain probability (this probability decreases as the temperature decreases). After thousands of iterations, the temperature drops to the minimum, and the algorithm converges.
[0033] Preferably, the multi-objective optimization function includes: Where S represents a seating arrangement scheme, and T is the total number of tables. Let be the variance of the combined weighted scores of all people at the t-th table, and let Adj be the set of all adjacent seat pairs. To evaluate the social harmony function of individuals a and b sitting next to each other, its value is given by a pre-trained affinity model, where SoftC is the set of all soft constraints. The penalty function measures the degree to which scheme S violates the soft constraint c. α, β, and γ are adjustable weighting coefficients used to balance the importance of different optimization objectives. Specifically, in this embodiment, α is 0.5 and β is 0.3. The target value is 0.2, aiming to balance the importance of each table, promote communication among recommenders, and reduce violations of soft constraints.
[0034] Preferably, after step S3, step S4 is further included: The final seating allocation plan is presented graphically on the user interface, with different weighted personnel marked with different colors or sizes, and constraints indicated by lines or icons. It provides drag-and-drop functionality, allowing users to manually adjust the seating of any person; Record the complete adjustment sequence from the user's automated plan to the final confirmed plan; The adjusted sequence is used as reinforcement learning feedback to update the parameters of the deep neural network model and the graph neural network model. Specifically, in this embodiment, the final solution is displayed in the form of a color seating chart (e.g., Figure 5 The hierarchy of positions is represented by shades of color, and constraints are connected by dashed lines. If the vice president in charge believes that a technical expert should be closer to the main table, he will drag the expert to exchange seats with a manager at the secondary table. This drag will be recorded, generating a feedback data pair: (automatic solution, user-adjusted solution). This data will be added to the training set for subsequent updates to the DNN and GNN models, enabling the models to better understand the implicit preference for such technical experts to be valued in strategic meetings.
[0035] Second Embodiment Based on the same concept, this embodiment also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor performs the steps of a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model as described in the embodiment.
[0036] Based on the same concept, the present invention also provides a storage medium storing computer-readable instructions, characterized in that, when the computer-readable instructions are executed by one or more processors, the one or more processors cause the one or more processors to perform the steps of a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model as described in any one of the embodiments.
[0037] It is understood that, regarding the aforementioned collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model, if all components are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, 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 several instructions to cause a computer device (which may be a personal computer server or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.
[0038] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.
[0039] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.
Claims
1. A collaborative decision-making method for an intelligent seating system based on a machine learning model, characterized in that, Includes the following steps: S1: Obtain the scenario type selected by the user, load the corresponding weight parameter configuration template according to the scenario type, and obtain the set of personnel attribute features, table layout attribute set, and personnel relationship constraint set including the personnel to be seated; S2: Based on the scenario type, the weight parameter configuration template, and the personnel attribute feature set, a machine learning model is used to dynamically calculate and output the comprehensive weight score of each person waiting to be seated. A sequence of ranked personnel is generated based on the comprehensive weight score. S3: Using the sorted personnel sequence as the initial allocation order, the table layout attribute set as the spatial constraint, and the personnel relationship constraint set as the constraint condition that must be satisfied, a hybrid allocation strategy combining rules and metaheuristic optimization algorithms is adopted to allocate seats and iteratively optimize the personnel to be seated. By optimizing the preset multi-objective function, the final seat allocation scheme is output.
2. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 1, characterized in that, Step S2 includes: Extract the personnel attribute feature vector corresponding to the personnel attribute features, and the scene feature vector corresponding to the scene type; The personnel attribute feature vector is concatenated with the scene feature vector to form a combined feature vector; The combined feature vectors are input into a pre-trained deep neural network model, which performs nonlinear mapping and regression prediction to output the comprehensive weight score for each person waiting to be seated.
3. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 1, characterized in that, In step S1, the set of relationship constraints between personnel is obtained, including: Construct a personnel relationship diagram, in which nodes correspond to personnel to be seated and edges correspond to the relationships between personnel. The relationships include at least one of historical co-seat records, organizational affiliation, or communication collaboration frequency. The personnel relationship graph is input into the graph neural network model along with the current scene features selected by the current user to obtain the potential relationship strength score between each pair of personnel and the corresponding constraint type suggestion. Set a relationship strength score threshold, and present relationships with potential relationship strength scores exceeding the threshold to the user. After user confirmation, add them to the set of interpersonal relationship constraints.
4. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 2, characterized in that, The pre-training process of a deep neural network model includes: Collect historical seating arrangement cases, including personnel attribute characteristics, scenario types, and corresponding effective seating allocation schemes; Based on the seating arrangement scheme corresponding to each historical seating case, the relative importance of each person in the scheme is deduced in reverse, and this relative importance is used as the weight score label for training the deep neural network model. Using the personnel attribute characteristics and scene type in historical seating cases as model input, and the corresponding weight score labels as supervision signals, the deep neural network model is trained end-to-end. During training, regularization and cross-validation techniques are used to optimize the deep neural network model.
5. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 1, characterized in that, In step S3, a hybrid allocation strategy is used for seat allocation, including: For each person waiting to be seated, check each of their corresponding candidate seats one by one to determine whether the candidate seat meets all the hard constraints related to that person; If the person belongs to a group that requires them to sit together, then the candidate seat must be at the same table and adjacent to the person who has already been assigned a seat in that group. If the person belongs to a group that cannot be seated together, the candidate seat cannot be at the same table or adjacent to any person already assigned in that group; After filtering out all candidate seats that meet the above hard constraints, the target seat for the person is selected from them according to preset heuristic rules. The preset heuristic rules include prioritizing filling tables that are not full, prioritizing seats near the head or with a good view, and prioritizing making the average comprehensive weight of the people at the current table close to the overall average comprehensive weight of all people. If a person waiting to be seated has no candidate seats that meet the hard constraints, the backtracking adjustment mechanism will be activated to adjust the seats of people who have already been assigned, or the user will be directly prompted that there is a conflict of relationship constraints between people and the user needs to adjust the constraint rules.
6. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 1, characterized in that, The metaheuristic optimization algorithm includes: Define a state space and neighborhood operations, where the state space corresponds to a seating arrangement scheme for each state, and the neighborhood operations are to randomly swap the seats of people at two different tables or randomly rotate the seating order of several consecutive people at the same table. Take the initial feasible allocation scheme as the current state S_curr, and calculate the objective function value E_curr of the current state S_curr, E_curr = F(S_curr); In each iteration, a new seat allocation state S_new is generated by applying a random neighborhood operation to the current state S_curr, ensuring that the new seat allocation state S_new satisfies all hard constraints. A new objective function value E_new is calculated, E_new=F(S_new). The target energy difference ΔE is calculated based on the objective function value and the new objective function value, ΔE = E_new - E_curr. If the target energy difference ΔE<0, the new seat allocation state S_new is accepted as the current state, i.e., S_curr=S_new and E_curr=E_new. If the target energy difference ΔE≥0, the new seat allocation state S_new is accepted with probability P=exp(-ΔE / T'), where T' is the current temperature parameter. During the iteration process, the temperature parameter T' is gradually reduced according to the predefined annealing schedule; When the temperature parameter T' drops to the preset minimum threshold, or when no new state S_new is received for several consecutive iterations, the iteration loop terminates. At this time, the allocation scheme corresponding to the current state S_curr is the optimized final seat allocation scheme.
7. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 6, characterized in that, The multi-objective optimization function includes: Where S represents a seating arrangement scheme, and T is the total number of tables. Let be the variance of the combined weighted scores of all people at the t-th table, and let Adj be the set of all adjacent seat pairs. To evaluate the social harmony function of individuals a and b sitting next to each other, its value is given by a pre-trained affinity model, where SoftC is the set of all soft constraints. The penalty function is used to measure the degree to which scheme S violates the soft constraint c. α, β, and γ are adjustable weighting coefficients used to balance the importance of different optimization objectives.
8. The collaborative decision-making method for intelligent seating arrangement system based on machine learning model according to claim 1, characterized in that, Following step S3, step S4 is also included: The final seating allocation scheme is presented graphically on the user interface, with different weighted personnel marked with different colors or sizes, and constraints indicated by lines or icons. It provides drag-and-drop functionality, allowing users to manually adjust the seating of any person; Record the complete adjustment sequence from the user's automated plan to the final confirmed plan; The adjusted sequence is used as reinforcement learning feedback to update the parameters of the deep neural network model and the graph neural network model.
9. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, cause the processor to perform the steps of a collaborative decision-making method for an intelligent seating arrangement system based on a machine learning model as described in any one of claims 1 to 8.
10. A storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors perform the steps of a collaborative decision-making method for an intelligent seating system based on a machine learning model as described in any one of claims 1 to 8.