Conference notification and seat intelligent scheduling system based on adaptive learning

Abstract

The invention discloses a conference notification and seat intelligent scheduling system based on adaptive learning, and the system comprises a behavior input data set construction module which is used for collecting data and generating a behavior input data set; the behavior feature learning module is used for learning to generate participant behavior feature vectors; the relation density calculation module is used for calculating relation density and generating a relation density set; the conference room space constraint generation module is used for constructing a conference room space structure expression and generating a conference room space constraint set; the particle swarm initialization module is used for initializing a particle swarm; the candidate seat layout generation module is used for iteratively generating a candidate seat layout scheme; the seat layout selection module is used for selecting a target seat layout scheme according to the fitness function; and the self-adaptive notification and update module is used for generating and pushing notification data and updating a new target seat layout scheme. According to the invention, incremental learning and particle swarm optimization are adopted, intelligent arrangement of seats is realized, and the method has the advantage of real-time, efficient and adaptive optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent meeting management technology, and in particular to a meeting notification and seat scheduling intelligent scheduling system based on adaptive learning. Background Technology

[0002] As conferences become increasingly larger and more complex, the roles, past participation behaviors, collaborative relationships, and preferences of attendees are playing a growing role in conference organization. Existing conference management systems often rely on static rules or manual settings for seating arrangements, typically configuring them based solely on conference type, attendee level, or simple fixed templates. This fails to dynamically model and intelligently respond to changes in attendee relationships, behavioral shifts, and evolving conference conditions. Furthermore, traditional seating recommendation methods lack comprehensive consideration of spatial constraints, resource distribution, and the location of entrances and exits within the conference room, resulting in significant deficiencies in interpretability, rationality, and flexibility in seating layouts.

[0003] While existing technologies have gradually developed seating allocation schemes based on machine learning or optimization algorithms, these schemes largely rely on statically trained models. They cannot incrementally learn from the behavioral characteristics of attendees, nor can they perform real-time optimization in response to changes in attendees' additions, departures, or meeting conditions. Furthermore, existing particle swarm optimization methods typically do not incorporate attendee relationship characteristics and meeting room space constraints, and lack exclusion mechanisms to enhance solution space exploration capabilities. This leads to a tendency to get stuck in local optima in complex seating layout problems, failing to generate high-quality layout schemes that balance relationship matching and spatial constraints.

[0004] Therefore, how to provide a meeting notification and intelligent seating scheduling system based on adaptive learning is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a meeting notification and seat scheduling intelligent scheduling system based on adaptive learning. This invention employs incremental learning and particle swarm optimization to achieve intelligent seat arrangement, possessing the advantages of real-time, efficient, and adaptive optimization.

[0006] An adaptive learning-based intelligent meeting notification and seating scheduling system according to an embodiment of the present invention includes: The behavior input dataset building module is used to collect data and generate behavior input datasets; The behavioral feature learning module uses an incremental self-organizing neural network to learn from the behavioral input dataset online and generate behavioral feature vectors of the participants. The relationship density calculation module calculates the relationship density based on the behavioral feature vectors of the participants and generates a set of relationship densities. The meeting room space constraint generation module is used to construct a representation of the meeting room space structure and generate a set of meeting room space constraints. The particle swarm initialization module is used to initialize the particle swarm and apply the relation density set, the conference room space constraint set, and the repulsion term to the particle swarm. The candidate seating layout generation module iterates through the particle swarm based on the set of relation density, the set of meeting room space constraints, and the exclusion term to generate candidate seating layout schemes. The seat layout selection module selects the target seat layout scheme from the candidate seat layout schemes based on the fitness function; The adaptive notification and update module is used to generate meeting notification data and push seating arrangement information. When participants are added, leave, or meeting conditions change, the module updates the behavioral input dataset and re-executes learning and search to generate a new target seating layout scheme.

[0007] Optionally, modules can be integrated using the following methods: Collect participants' historical attendance records, role attributes, preference information, interaction data, and meeting type data to construct a behavioral input dataset; The incremental self-organizing neural network is used to learn the behavioral input dataset online, generate behavioral feature vectors of the participants, and update the topology of the incremental self-organizing neural network in real time. Calculate the relationship density between participants based on their behavioral feature vectors, and generate a relationship density set. Construct a spatial structure representation of the conference room, including seating positions, conference room structural parameters, entrance and exit locations, and resource locations, forming a set of spatial constraints for the conference room; Initialize the particle swarm, with each particle corresponding to a seat layout code, and use the relationship density set and the meeting room space constraint set as the basis for particle search, while introducing an exclusion term; Based on the set of relation density, the set of meeting room space constraints, and the exclusion term, the particle swarm optimization is iteratively searched to generate candidate seating layout schemes. The target seating layout is selected from the candidate seating layouts based on the fitness function; Generate meeting notification data and push it to participants. When participants are added, leave, or meeting conditions change, update the behavioral input dataset and generate a new target seating layout.

[0008] Optionally, the construction of the behavioral input dataset includes: Collect historical attendance records of participants, and organize the participants' attendance time, attendance duration, attendance frequency, check-in behavior and departure behavior in each meeting in chronological order to form historical attendance record data; Collect data on the role attributes, preferences, and interaction relationships of participants. Record the job level, department category, and meeting responsibilities of participants in a structured form to form role attribute data, preference information data, and personnel interaction relationship data. Collect meeting type data, organize the type classification information of each meeting according to the meeting records, and form meeting type data; Historical meeting attendance records, role attribute data, preference information data, personnel interaction data, and meeting type data are processed and integrated in a unified format using attendees as the index, to construct a behavioral input dataset.

[0009] Optionally, the generation of the participant behavior feature vector includes: The behavioral input dataset is transformed into an input vector sequence indexed by the participants. The feature values ​​contained in the behavioral input dataset are then used to form the input vectors in sequence. The input vector sequence is then fed into the incremental self-organizing neural network in chronological order. Based on the competition mechanism of the incremental self-organizing neural network, the distance between each input vector in the input vector sequence and the weight vector of each node in the network is calculated, and the node with the smallest distance is determined as the winning node. The weights of the winning node and its neighboring nodes are then updated. According to the topology growth rule of the incremental self-organizing neural network, when the distance between the input vector and the weight of the winning node exceeds the growth threshold, a new node is generated and added to the topology of the incremental self-organizing neural network. According to the topology adjustment rules of the incremental self-organizing neural network, the connection relationship between nodes is added or deleted when the update conditions are met. After all input vectors have been processed online, the node weight vectors corresponding to the participants in the incremental self-organizing neural network are output as the participant behavior feature vectors.

[0010] Optionally, the generation of the relation density set includes: The behavioral feature vectors of the participants are arranged according to the participants as the index to form a sequence of behavioral feature vectors of the participants. All feature values ​​in each participant's behavioral feature vector are then processed in a structured manner according to a preset order. Perform difference calculation on the behavior feature vectors of any two participants to obtain the behavior difference between the two participants, and calculate the corresponding relationship density based on the behavior difference; Perform pairwise combination processing on the behavioral feature vectors of all participants, and perform behavioral difference calculation and relation density calculation on all combinations in sequence, so that all participants in the whole set of participants can obtain the corresponding relation density; Arrange the relationship density between each participant and other participants according to the participant as the index to form a corresponding relationship density arrangement sequence. Merge all relationship density arrangement sequences to generate a relationship density set.

[0011] Optionally, the generation of the meeting room space constraint set includes: Record the positions of all seats in the meeting room, represent the position of each seat with horizontal and vertical coordinates, and obtain the distance between two seats; Record the structural parameters of the conference room, including its length, width, and the location and dimensions of all fixed structures within the room, and obtain the area occupied by each fixed structure. Record the locations of entrances and exits, represent the horizontal and vertical coordinates of each entrance and exit, and obtain the distance between the seats and the entrances and exits; Record the location of resources in the meeting room, represent the location of each resource with horizontal and vertical coordinates, and obtain the distance between the seat and the resource; The seating positions, meeting room structural parameters, entrance and exit positions, and resource positions are organized in a unified format to form a meeting room spatial structure representation. All the contents of the meeting room spatial structure representation are integrated to generate a meeting room spatial constraint set.

[0012] Optionally, the particle swarm initialization includes: A particle swarm is generated based on the number of attendees. Each particle in the swarm is assigned a corresponding seating layout code. The seating layout codes are then organized according to the order in which the attendees are seated. Each particle represents a complete seating layout. The relation density set is used as the reference information for particle initialization, and the meeting room space constraint set is used as the reference information for particle initialization. A repulsion term is introduced into the particle swarm to calculate the repulsive force between particles; The set of relation density, the set of meeting room space constraints, and the repulsion term are applied to each particle in the particle swarm. After initialization, each particle is simultaneously associated with the seat layout code, the set of relation density, the set of meeting room space constraints, and the repulsion force, and the particle swarm is output.

[0013] Optionally, the generation of the candidate seating layout scheme includes: Apply relation density set to particle swarm optimization, and process the particle velocity corresponding to each particle in the particle swarm. Apply the spatial constraints of the conference room to the particle swarm, and process the particle velocity corresponding to each particle in the particle swarm. By applying the repulsion term to the particle swarm, the positions of any two particles in the swarm are compared to form the way the repulsion term acts in the particle swarm. In each iteration, the particle velocity is updated based on the relation density set, the meeting room space constraint set, and the repulsion term. After the particle velocity is updated, the particle position is updated. After a predetermined number of iterations, the seating layout code formed by the particle position of each particle in the particle swarm at the end of the iteration is used as a candidate seating layout scheme.

[0014] Optionally, the selection of the target seating layout scheme includes: Input each candidate seating layout scheme into the fitness function; The positional relationships of attendees in the candidate seating layout schemes are processed based on the relationship matching degree to form the relationship matching degree; The seating arrangement in the candidate seating layout scheme is processed based on the spatial constraint satisfaction to form the spatial constraint satisfaction. The order of attendees in the candidate seating arrangement is processed according to the role priority satisfaction, thus forming the role priority satisfaction. The distribution of seats in the candidate seating layout schemes is processed based on the layout balance index to form the layout balance index. The target seating layout scheme is obtained by combining the relationship matching degree, spatial constraint satisfaction degree, role priority satisfaction degree, and layout balance index according to the requirements of the fitness function.

[0015] Optionally, the generation of the new target seating layout scheme includes: Based on the target seating layout plan, the correspondence between attendees and seats in the target seating layout plan is processed, and the seat corresponding to each attendee is written into the meeting notification data in text form. After the meeting notification data is generated, the meeting notification data will be pushed to the participants included in the target seating layout plan; When participants are added or leave the meeting, the behavior input dataset is processed to add or remove the content related to the addition or departure of participants. When meeting conditions change, the behavior input dataset is processed, and the content related to the change in meeting conditions is written into the behavior input dataset; When the behavioral input dataset changes, the incremental self-organizing neural network learning is re-executed; After the incremental self-organizing neural network learning is completed, the particle swarm iteration search is re-executed to generate a new target seating layout scheme based on the content provided by the incremental self-organizing neural network learning, thereby achieving real-time adaptive optimization.

[0016] The beneficial effects of this invention are: This invention employs an incremental self-organizing neural network to learn from the behavioral input dataset online, enabling the system to automatically adjust its internal structure and generate new participant behavioral feature vectors in response to changes in participant behavior, additions or withdrawals of participants, and updates to meeting conditions. Compared to traditional meeting systems that rely on static models, this invention allows for real-time adaptation to behavioral feature changes without retraining the model, ensuring that seating layout optimization is based on the latest behavioral data, thereby significantly improving the system's adaptability and real-time response efficiency. Furthermore, by quantifying participants' historical participation behavior, role attributes, preference characteristics, and interaction relationships into behavioral feature vectors, and further calculating a relationship density set, this invention imbues seating layout optimization with quantifiable and interpretable behavioral drivers, overcoming the shortcomings of existing technologies that struggle to accurately reflect participant relationships and dynamically adjust the system.

[0017] This invention constructs a set of spatial constraints for meeting rooms and simultaneously introduces a set of relationship densities, a set of spatial constraints for meeting rooms, and an exclusion term during particle swarm optimization (PSO) initialization and iteration. This enables PSO to perform efficient searches under complex multi-dimensional constraints. The inclusion of the exclusion term enhances the search diversity of the PSO, avoiding the problem of traditional PSO optimization easily getting trapped in local optima, and making the generated candidate layout schemes more reasonable and stable. The fitness function integrates relationship matching degree, spatial constraint satisfaction degree, role priority satisfaction degree, and layout balance, enabling the system to select the target seating layout scheme with globally optimal performance. Furthermore, this invention directly uses the target seating layout scheme for meeting notification generation and automatically triggers relearning and re-optimization when meeting conditions change, achieving a linked adaptive update of seating layout and meeting notifications, making meeting organization more intelligent, automated, and efficient. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0019] Figure 1 This is a flowchart of the overall process of a meeting notification and intelligent seat scheduling system based on adaptive learning proposed in this invention. Figure 2 This is a schematic diagram of the overall seat layout optimization mechanism in a meeting notification and intelligent seat scheduling system based on adaptive learning proposed in this invention. Figure 3 This is a schematic diagram of the overall structure of the meeting notification and adaptive update mechanism in an intelligent meeting notification and seat scheduling system based on adaptive learning proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figures 1-3 A meeting notification and intelligent seating scheduling system based on adaptive learning, comprising: The behavior input dataset building module is used to collect data and generate behavior input datasets; The behavioral feature learning module uses an incremental self-organizing neural network to learn from the behavioral input dataset online and generate behavioral feature vectors of the participants. The relationship density calculation module calculates the relationship density based on the behavioral feature vectors of the participants and generates a set of relationship densities. The meeting room space constraint generation module is used to construct a representation of the meeting room space structure and generate a set of meeting room space constraints. The particle swarm initialization module is used to initialize the particle swarm and apply the relation density set, the conference room space constraint set, and the repulsion term to the particle swarm. The candidate seating layout generation module iterates through the particle swarm based on the set of relation density, the set of meeting room space constraints, and the exclusion term to generate candidate seating layout schemes. The seat layout selection module is used to select the target seat layout scheme from the candidate seat layout schemes based on the fitness function; The adaptive notification and update module is used to generate meeting notification data and push seating arrangement information. When participants are added, leave, or meeting conditions change, the module updates the behavioral input dataset and re-executes learning and search to generate a new target seating layout scheme.

[0022] In this embodiment, the modules are interconnected using the following method: Collect participants' historical attendance records, role attributes, preference information, interaction data, and meeting type data to construct a behavioral input dataset; Incremental self-organizing neural networks are used to learn from the behavioral input dataset online, generating behavioral feature vectors of participants, and the topology of the incremental self-organizing neural network is updated in real time according to changes in the behavioral input dataset. Calculate the relationship density between participants based on their behavioral feature vectors, and generate a relationship density set. Construct a spatial structure representation of the conference room, including seating positions, conference room structural parameters, entrance and exit locations, and resource locations, forming a set of spatial constraints for the conference room; Initialize the particle swarm, with each particle corresponding to a seat layout code, and use the relationship density set and the meeting room space constraint set as the basis for particle search. At the same time, introduce a repulsion term into the particle swarm to calculate the repulsive force between particles. Based on the set of relation density, the set of meeting room space constraints, and the exclusion term, the particle swarm is iteratively searched, and candidate seating layout schemes are generated by updating particle velocity and particle position. The target seating layout scheme is selected from the candidate seating layout schemes based on the fitness function, which includes the relationship matching degree, spatial constraint satisfaction degree, role priority satisfaction degree and layout balance index. Based on the target seating layout plan, meeting notification data is generated and the corresponding seating arrangement information is pushed to the participants. When participants are added, leave, or meeting conditions change, the behavioral input dataset is updated and incremental self-organizing neural network learning and particle swarm iterative search are re-executed to generate a new target seating layout plan, thereby achieving real-time adaptive optimization of the seating layout.

[0023] In this embodiment, the construction of the behavioral input dataset includes: Collect historical attendance records of participants, and organize the participants' attendance time, attendance duration, attendance frequency, check-in behavior and departure behavior in each meeting in chronological order to form historical attendance record data; Collect the role attributes of the participants, and record their job level, department category, and meeting responsibilities in a structured form to form role attribute data; Collect the preference information of the participants, and record the participants' preferences for seating areas, their preferences for neighboring persons, and their preferences for different meeting types as preference information data; Collect data on personnel interaction relationships, and organize the communication behavior, collaboration behavior, and historical seat-to-seat information among participants to form personnel interaction relationship data; Collect meeting type data, organize the type classification information of each meeting according to the meeting records, and form meeting type data; Historical meeting attendance records, role attribute data, preference information data, personnel interaction data, and meeting type data are processed and integrated in a unified format using attendees as the index, to construct a behavioral input dataset.

[0024] In this embodiment, the generation of the participant behavior feature vector includes: The behavioral input dataset is transformed into an input vector sequence indexed by the participants. The feature values ​​contained in the behavioral input dataset are then used to form the input vectors in sequence. The input vector sequence is then fed into the incremental self-organizing neural network in chronological order. Based on the competition mechanism of the incremental self-organizing neural network, the distance between each input vector in the input vector sequence and the weight vector of each node in the network is calculated, and the node with the smallest distance is determined as the winning node. The winning node is determined by calculating the distance between the input vector and the weight vectors of all nodes in the network. The distance is obtained by squaring the difference between the input vector and the node weight vector at each corresponding position, adding all the squared values ​​and taking the square root. After the network completes all distance calculations, the node with the smallest distance is determined as the winning node, and node update, neighborhood update and topology adjustment are performed based on the winning node. The winning node and its neighboring nodes are updated with weights. The update is performed by adjusting the new weights to be equal to the old weights plus the learning rate multiplied by the input vector minus the old weights, so that the node weights are updated gradually under the continuous input of the behavioral input dataset. According to the topology growth rule of the incremental self-organizing neural network, a new node is generated when the distance between the input vector and the weight of the winning node exceeds the growth threshold. The weight of the new node is equal to the average of the input vector and the weight of the winning node, and the new node is added to the topology of the incremental self-organizing neural network. The generation of the new node specifically includes: when the input vector enters the incremental self-organizing neural network, the distance between the input vector and the weight vector of the winning node is calculated by squaring the difference between the input vector and the weight vector of the winning node at each corresponding feature position, adding all the squared results and taking the square root. When the distance is greater than the preset growth threshold, it is determined that the current input vector cannot be represented by the existing node, triggering the generation of a new node. The weight vector of the new node is obtained by adding the input vector and the weight vector of the winning node at each corresponding feature position and taking the average. According to the topology adjustment rules of the incremental self-organizing neural network, the topology adjustment rules are used to dynamically modify the connection relationship between nodes based on the continuous arrival of input vectors, so that the network can maintain the correct expression of the data structure. When the update conditions between nodes are met, the connection relationship between nodes is added or deleted, so that the topology remains dynamically adjusted as the behavioral input dataset is continuously updated. After all input vectors have been processed online, the node weight vectors corresponding to the participants in the incremental self-organizing neural network are output as the participant behavior feature vectors.

[0025] In this embodiment, the generation of the relation density set includes: The behavioral feature vectors of the participants are arranged according to the participants as the index to form a sequence of behavioral feature vectors of the participants for relation density calculation. All feature values ​​in each participant's behavioral feature vector are then structured in a preset order. For any two participants, perform a difference calculation on their behavior feature vectors. The difference calculation involves sequentially calculating the difference between the two participants' behavior feature vectors at each corresponding feature position. The behavior difference is equal to the square root of the sum of the squares of all corresponding feature differences, thus obtaining the behavior difference between the two participants. The corresponding relationship density is calculated based on the behavior difference. The relationship density is equal to a constant divided by one plus the behavior difference, so that a corresponding relationship density is generated for each pair of participants. Perform pairwise combination processing on the behavioral feature vectors of all participants, and perform behavioral difference calculation and relation density calculation on all combinations in sequence, so that all participants in the whole set of participants can obtain the corresponding relation density; Arrange the relationship density between each participant and other participants according to the participant as the index to form a corresponding relationship density arrangement sequence; All relationship density sequences are merged to generate a relationship density set, which contains the relationship densities of all participants.

[0026] In this embodiment, the generation of the meeting room space constraint set includes: Record the positions of all seats in the meeting room, represent the position of each seat with horizontal and vertical coordinates, and obtain the distance between two seats; Record the structural parameters of the conference room, including its length, width, and the location and dimensions of all fixed structures within the room, and obtain the area occupied by each fixed structure. Record the locations of entrances and exits, represent the horizontal and vertical coordinates of each entrance and exit, and obtain the distance between the seats and the entrances and exits; Record the location of resources in the meeting room, represent the location of each resource with horizontal and vertical coordinates, and obtain the distance between the seat and the resource; The seating positions, meeting room structural parameters, entrance and exit locations, and resource locations are organized in a unified format to form a representation of the meeting room spatial structure; Integrate all the contents of the conference room space structure representation to generate a set of conference room space constraints.

[0027] In this embodiment, the particle swarm initialization includes: A particle swarm is generated based on the number of attendees. Each particle in the swarm is assigned a corresponding seating layout code. The seating layout codes are then organized according to the order in which the attendees are seated. Each particle represents a complete seating layout. The relation density set is used as reference information for particle initialization. The seat layout code corresponding to the particle is compared with the relation density of each participant in the relation density set, so that each particle can reference the relation density set during the initialization process. The meeting room space constraint set is used as reference information for particle initialization. The seat layout code corresponding to the particle is compared with the seat position, meeting room structure parameters, entrance and exit positions and resource positions contained in the meeting room space constraint set, so that each particle can reference the meeting room space constraint set during the initialization process. A repulsion term is introduced into the particle swarm to calculate the repulsive force between particles. The calculation method is to process the difference between the seat layout codes of the two particles at each seat position, square all the differences one by one, and then add all the square values ​​and take the square root to obtain the degree of difference between the two particles. The repulsive force is generated according to the degree of difference, and the repulsive force increases as the degree of difference between the two particles decreases. The relation density set, the meeting room space constraint set, and the repulsion term are applied to each particle in the particle swarm. After initialization, each particle is simultaneously associated with the seating layout code, the relation density set, the meeting room space constraint set, and the repulsion force, and the particle swarm is output. The particle swarm is in an initial state in which subsequent processing can be performed using the relation density set, the meeting room space constraint set, and the repulsion term.

[0028] In this embodiment, the generation of the candidate seating layout scheme includes: Applying relation density sets to particle swarm optimization allows for processing of the particle velocity for each particle within the swarm, enabling the particle velocity to be adjusted during iteration based on the information recorded in the relation density sets. Apply the meeting room space constraint set to the particle swarm, process the particle velocity corresponding to each particle in the particle swarm, so that the particle velocity can be further adjusted according to the content recorded in the meeting room space constraint set during the iteration process; The repulsion term is applied to the particle swarm. The positions of any two particles in the particle swarm are compared. The comparison method is to process the difference between the corresponding positions of the two particles in the seating layout code one by one, square each difference, add all the squares together, and then take the square root of the sum. This forms the way the repulsion term acts in the particle swarm, so that the repulsion term can affect the particle velocity. In each iteration, the particle velocity is updated based on the relation density set, the meeting room space constraint set, and the repulsion term. The update method is to add the adjustment generated by the relation density set, the adjustment generated by the meeting room space constraint set, and the effect of the repulsion term in sequence on the particle velocity of the previous moment, so that the particle velocity changes continuously in the iteration. After the particle velocity is updated, the particle position is updated. The update method is to combine the particle position of the previous moment with the updated particle velocity, and process the particle position according to the direction and magnitude of the particle velocity, so that the particle position changes continuously with iteration. After completing a predetermined number of iterations, the seating layout formed by the particle positions of each particle in the particle swarm at the end of the iteration is encoded as a candidate seating layout scheme.

[0029] In this embodiment, the selection of the target seating layout scheme includes: Each candidate seating layout scheme is input into the fitness function, so that each candidate seating layout scheme participates in the subsequent processing of each component in the fitness function. The positional relationships of attendees in the candidate seating layout scheme are processed based on the relationship matching degree. The processing method is to match the positional relationships between each attendee in the candidate seating layout scheme with the corresponding relationship content required for the relationship matching degree one by one, and form the relationship matching degree by accumulating each item. The relationship matching degree is used to reflect the matching situation between the positional relationships of attendees and the relationship content in the candidate seating layout scheme. The seating arrangement in the candidate seating layout scheme is processed based on the spatial constraint satisfaction. The processing method is to match the seat positions in the candidate seating layout scheme with the relevant contents of the conference room space required for the spatial constraint satisfaction one by one, and to form the spatial constraint satisfaction by accumulating them item by item. The spatial constraint satisfaction is used to reflect the degree to which the candidate seating layout scheme meets the conference room space constraints. The order of attendees in the candidate seating arrangement is processed according to the role priority satisfaction. The processing method is to match the position of the attendees in the seating sequence with the role order content required for role priority satisfaction one by one, and to form the role priority satisfaction by accumulating the items one by one. The role priority satisfaction is used to reflect the degree to which the candidate seating arrangement satisfies the role priority of the attendees. The distribution of seats in the candidate seat layout scheme is processed according to the layout balance index. The processing method is to compare the relative distribution of each seat in the candidate seat layout scheme one by one, and form the layout balance index by accumulating the items one by one. The layout balance index is used to reflect the overall seat distribution balance of the candidate seat layout scheme. The relationship matching degree, spatial constraint satisfaction degree, role priority satisfaction degree, and layout balance index are combined according to the requirements of the fitness function. The above four items are accumulated one by one. Then, the accumulated contents of all candidate seat layout schemes are compared. The candidate seat layout scheme with the highest accumulated contents is selected as the target seat layout scheme, thus obtaining the target seat layout scheme.

[0030] In this embodiment, the generation of the new target seating layout scheme includes: Based on the target seating layout plan, the correspondence between attendees and seats in the target seating layout plan is processed, and the seat corresponding to each attendee is written into the meeting notification data in text form, so that the meeting notification data contains the seat arrangement information corresponding to the attendees. After the meeting notification data is generated, the meeting notification data will be pushed to the participants included in the target seating layout plan so that the participants can obtain their corresponding seating arrangement information. When new attendees are added, the behavior input dataset is processed to include the content related to the new attendees, so that the behavior input dataset can reflect the situation after the new attendees are added. When a participant leaves the meeting, the behavior input dataset is processed to remove the content related to the participant's departure from the behavior input dataset, so that the behavior input dataset can reflect the situation after the participant leaves the meeting. When meeting conditions change, the behavior input dataset is processed, and the content related to the change in meeting conditions is written into the behavior input dataset so that the behavior input dataset can reflect the situation after the meeting conditions change. When the behavioral input dataset changes, the incremental self-organizing neural network learning is re-executed, so that the incremental self-organizing neural network learning steps can learn according to the changes in the behavioral input dataset; After the incremental self-organizing neural network learning is completed, the particle swarm iteration search is re-executed. Based on the content provided by the incremental self-organizing neural network learning, a new target seating layout scheme is generated, enabling the target seating layout scheme to achieve real-time adaptive optimization when attendees are added, attendees leave, or meeting conditions change.

[0031] Example 1: To verify the feasibility of this invention in practice, it was applied to the conference room of a company headquarters. This company frequently organizes various types of meetings in its daily operations, including departmental collaboration meetings, business seminars, and cross-team coordination meetings. As the company expands, the roles and identities of attendees become more diverse, and their collaborative relationships, seating preferences, past attendance behavior, and participation methods exhibit dynamic changes. In traditional meeting organization methods, seating arrangements often rely on manual experience or fixed templates, which not only fail to accurately reflect the true behavioral characteristics of attendees but also cannot respond quickly to situations such as new attendees joining or leaving, or changes in meeting conditions, significantly impacting meeting efficiency.

[0032] In this scenario, the system proposed in this invention is used to automatically generate meeting seating arrangements and push meeting notifications. The system first retrieves participants' historical meeting records from the enterprise's internal information management platform, including meeting duration, attendance patterns, check-in methods, and departure information. It also obtains participants' role attributes, such as job title, department category, and whether they are responsible for chairing or recording the meeting. The system further retrieves interaction data between participants from the enterprise's collaboration platform, such as jointly undertaken tasks, frequency of historical document collaboration, and instant messaging records. The system also collects participants' seating preferences, such as proximity to equipment areas or exits. All of this data is compiled into a complete behavioral input dataset, providing a foundation for subsequent intelligent analysis.

[0033] Before each meeting, the system feeds the behavioral input dataset into an incremental self-organizing neural network for online learning. Because the company's internal collaborative relationships frequently change due to business adjustments, and the behavioral characteristics of meeting participants are not stable over time, an incremental self-organizing neural network with topological dynamic adjustment capabilities is used. This allows the system to update the behavioral feature vectors of participants based on the latest behavioral data before each meeting. This process can be completed without retraining the entire model, which is key to the system's ability to quickly adapt to business changes.

[0034] Based on the generated behavioral feature vectors, the system calculates the relationship density between participants and organizes it into a relationship density set. Two people with higher relationship density have a history of more frequent or closer collaboration. This information is crucial for seating arrangement, especially in departmental collaborative meetings where communication efficiency is emphasized. Simultaneously, the system performs spatial modeling of the meeting room, incorporating seat locations, meeting room dimensions, entrance and exit locations, and the positions of resources such as whiteboards and projectors into the spatial structure representation, forming a set of meeting room spatial constraints. These constraints ensure that the generated seating layout simultaneously meets the requirements of spatial rationality, safety, and resource accessibility.

[0035] Subsequently, the system enters the seating layout optimization phase. It first incorporates the set of relation density, the set of meeting room space constraints, and the repulsion term into the particle swarm initialization process. Each particle in the swarm represents a seating layout combination; the addition of the repulsion term ensures sufficient difference between particles, thereby expanding the search range. Next, the system iteratively optimizes the seating layout through particle swarm optimization, incorporating the influence of relation density on the layout, the constraints of the meeting room space structure, and the solution space dispersion effect caused by the repulsion force into the particle velocity update and position adjustment process. After multiple iterations, the system generates multiple candidate seating layout schemes.

[0036] These candidate solutions are fed into a fitness function for comprehensive evaluation. The fitness function considers dimensions such as relationship matching, spatial constraint satisfaction, role priority satisfaction, and the overall balance of seating distribution. Based on the comprehensive results of the fitness function, the system selects the most suitable target seating layout for the current meeting. For example, in a cross-departmental collaborative meeting, the system can arrange closely collaborating personnel in adjacent positions, place the facilitator in an area that facilitates control of the meeting's pace, and ensure unobstructed passageways and a reasonable resource allocation.

[0037] Once the target layout is finalized, the system generates meeting notifications and pushes them to each participant's communication device. The notifications include seat numbers and relevant directions, eliminating the need for participants to search for seats at the meeting venue and improving meeting punctuality. In the company's actual usage, the system automatically generates meeting notifications, avoiding the repetitive work of manually editing notification content.

[0038] More importantly, this invention features adaptive updating capabilities. For example, if a new participant is added before the meeting begins, the system immediately updates the behavioral input dataset, re-executes incremental self-organizing neural network learning, and initiates particle swarm optimization to generate a new target seating layout. When participants leave or certain meeting conditions change, such as a change in meeting type from presentation to discussion, the system also recalculates the layout to better suit the new meeting requirements. In multiple consecutive applications, the system can complete updates within a short time, maintaining the effectiveness and rationality of the layout.

[0039] This implementation scenario demonstrates that the present invention significantly improves upon the shortcomings of traditional meeting seating arrangements. The system not only generates intelligent seating layout schemes based on multi-dimensional behavioral data but also maintains rapid responsiveness when meeting conditions change, making meeting organization more efficient, flexible, and intelligent. This embodiment shows that the present invention has significant application value in enterprise office environments, effectively improving meeting preparation efficiency and communication effectiveness.

[0040] Table 1. Performance Comparison between Adaptive Learning Intelligent Scheduling System and Traditional Methods

[0041] As can be seen from the table data, this system significantly outperforms traditional rule-based seating arrangements and manual seating arrangements in several key performance indicators. Firstly, regarding overall scheduling time, this system completes the entire seat optimization process in just 18 seconds, while traditional rule-based seating arrangements take 95 seconds, and manual seating arrangements take a whopping 360 seconds. This time advantage stems from the online learning mechanism of the incremental self-organizing neural network, which eliminates the need for remodeling behavioral features, thus significantly shortening the overall processing time.

[0042] In terms of participant satisfaction, this system achieved 92%, significantly higher than the 71% of traditional methods and 63% of manual seating arrangements. This is because after modeling the behavioral characteristics of participants, this system optimizes seating by combining relationship density with the repulsion term particle swarm optimization algorithm, making the seating arrangement more consistent with real interpersonal interactions and personal preferences, thereby effectively improving the participant experience.

[0043] Regarding the seat conflict rate, this system has a rate of only 1.2%, far lower than the 8.5% of traditional rule-based seating and the 12.7% of manual seating. The repulsion-term particle swarm optimization algorithm actively avoids conflicts and unreasonable arrangements during the optimization process, resulting in a final layout with higher consistency and executability.

[0044] In terms of role priority satisfaction, this system achieved 91%, a significant improvement over the 74% of traditional rule-based seating and the 68% of manual seating. This improvement is mainly due to the inclusion of role priority constraints in the fitness function, which allows the seating distribution to better align with the meeting's organizational structure and role responsibilities.

[0045] In terms of layout balance, this system scored 88, higher than the traditional rule-based seating arrangement of 62 and the manual seating arrangement of 55. Because the repulsive force mechanism prevents the layout from being concentrated or dispersed, this system performs better in terms of space utilization, interactive convenience, and visual balance.

[0046] Furthermore, this system possesses real-time adaptive capabilities. Adaptive updates take only 6 seconds, whereas traditional rule-based seating arrangements cannot achieve real-time updates, and manual seating arrangements require complete replanning. This real-time advantage stems from the incremental self-organizing neural network's ability to quickly update behavioral characteristics when attendees change, giving the entire scheduling system a high degree of dynamic responsiveness.

[0047] In summary, this system significantly improves seat scheduling efficiency, layout quality, user satisfaction, and overall intelligence level, demonstrating comprehensive performance advantages that traditional methods cannot achieve, proving that this invention has significant technological advancement and practical application value.

[0048] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A meeting notification and intelligent seating scheduling system based on adaptive learning, characterized in that, include: The behavior input dataset building module is used to collect data and generate behavior input datasets; The behavioral feature learning module uses an incremental self-organizing neural network to learn from the behavioral input dataset online and generate behavioral feature vectors of the participants. The relationship density calculation module calculates the relationship density based on the behavioral feature vectors of the participants and generates a set of relationship densities. The meeting room space constraint generation module is used to construct a representation of the meeting room space structure and generate a set of meeting room space constraints. The particle swarm initialization module is used to initialize the particle swarm and apply the relation density set, the conference room space constraint set, and the repulsion term to the particle swarm. The candidate seating layout generation module iterates through the particle swarm based on the set of relation density, the set of meeting room space constraints, and the exclusion term to generate candidate seating layout schemes. The seat layout selection module selects the target seat layout scheme from the candidate seat layout schemes based on the fitness function; The adaptive notification and update module is used to generate meeting notification data and push seating arrangement information. When participants are added, leave, or meeting conditions change, the module updates the behavioral input dataset and re-executes learning and search to generate a new target seating layout scheme.

2. The meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 1, characterized in that, The modules are connected in the following way: Collect participants' historical attendance records, role attributes, preference information, interaction data, and meeting type data to construct a behavioral input dataset; The incremental self-organizing neural network is used to learn the behavioral input dataset online, generate behavioral feature vectors of the participants, and update the topology of the incremental self-organizing neural network in real time. Calculate the relationship density between participants based on their behavioral feature vectors, and generate a relationship density set. Construct a spatial structure representation of the conference room, including seating positions, conference room structural parameters, entrance and exit locations, and resource locations, forming a set of spatial constraints for the conference room; Initialize the particle swarm, with each particle corresponding to a seat layout code, and use the relationship density set and the meeting room space constraint set as the basis for particle search, while introducing an exclusion term; Based on the set of relation density, the set of meeting room space constraints, and the exclusion term, the particle swarm optimization is iteratively searched to generate candidate seating layout schemes. The target seating layout is selected from the candidate seating layouts based on the fitness function; Generate meeting notification data and push it to participants. When participants are added, leave, or meeting conditions change, update the behavioral input dataset and generate a new target seating layout.

3. The meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The construction of the behavioral input dataset includes: Collect historical attendance records of participants, and organize the participants' attendance time, attendance duration, attendance frequency, check-in behavior and departure behavior in each meeting in chronological order to form historical attendance record data; Collect data on the role attributes, preferences, and interaction relationships of participants. Record the job level, department category, and meeting responsibilities of participants in a structured form to form role attribute data, preference information data, and personnel interaction relationship data. Collect meeting type data, organize the type classification information of each meeting according to the meeting records, and form meeting type data; Historical meeting attendance records, role attribute data, preference information data, personnel interaction data, and meeting type data are processed and integrated in a unified format using attendees as the index, to construct a behavioral input dataset.

4. The meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The generation of the participant behavior feature vector includes: The behavioral input dataset is transformed into an input vector sequence indexed by the participants. The feature values ​​contained in the behavioral input dataset are then used to form the input vectors in sequence. The input vector sequence is then fed into the incremental self-organizing neural network in chronological order. Based on the competition mechanism of the incremental self-organizing neural network, the distance between each input vector in the input vector sequence and the weight vector of each node in the network is calculated, and the node with the smallest distance is determined as the winning node. The weights of the winning node and its neighboring nodes are then updated. According to the topology growth rule of the incremental self-organizing neural network, when the distance between the input vector and the weight of the winning node exceeds the growth threshold, a new node is generated and added to the topology of the incremental self-organizing neural network. According to the topology adjustment rules of the incremental self-organizing neural network, the connection relationship between nodes is added or deleted when the update conditions are met. After all input vectors have been processed online, the node weight vectors corresponding to the participants in the incremental self-organizing neural network are output as the participant behavior feature vectors.

5. A meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The generation of the relation density set includes: The behavioral feature vectors of the participants are arranged according to the participants as the index to form a sequence of behavioral feature vectors of the participants. All feature values ​​in each participant's behavioral feature vector are then processed in a structured manner according to a preset order. Perform difference calculation on the behavior feature vectors of any two participants to obtain the behavior difference between the two participants, and calculate the corresponding relationship density based on the behavior difference; Perform pairwise combination processing on the behavioral feature vectors of all participants, and perform behavioral difference calculation and relation density calculation on all combinations in sequence, so that all participants in the whole set of participants can obtain the corresponding relation density; Arrange the relationship density between each participant and other participants according to the participant as the index to form a corresponding relationship density arrangement sequence. Merge all relationship density arrangement sequences to generate a relationship density set.

6. The meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The generation of the meeting room space constraint set includes: Record the positions of all seats in the meeting room, represent the position of each seat with horizontal and vertical coordinates, and obtain the distance between two seats; Record the structural parameters of the conference room, including its length, width, and the location and dimensions of all fixed structures within the room, and obtain the area occupied by each fixed structure. Record the locations of entrances and exits, represent the horizontal and vertical coordinates of each entrance and exit, and obtain the distance between the seats and the entrances and exits; Record the location of resources in the meeting room, represent the location of each resource with horizontal and vertical coordinates, and obtain the distance between the seat and the resource; The seating positions, meeting room structural parameters, entrance and exit positions, and resource positions are organized in a unified format to form a meeting room spatial structure representation. All the contents of the meeting room spatial structure representation are integrated to generate a meeting room spatial constraint set.

7. A meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The particle swarm initialization includes: A particle swarm is generated based on the number of attendees. Each particle in the swarm is assigned a corresponding seating layout code. The seating layout codes are then organized according to the order in which the attendees are seated. Each particle represents a complete seating layout. The relation density set is used as the reference information for particle initialization, and the meeting room space constraint set is used as the reference information for particle initialization. A repulsion term is introduced into the particle swarm to calculate the repulsive force between particles; The set of relation density, the set of meeting room space constraints, and the repulsion term are applied to each particle in the particle swarm. After initialization, each particle is simultaneously associated with the seat layout code, the set of relation density, the set of meeting room space constraints, and the repulsion force, and the particle swarm is output.

8. A meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The generation of the candidate seating layout scheme includes: Apply relation density set to particle swarm optimization, and process the particle velocity corresponding to each particle in the particle swarm. Apply the spatial constraints of the conference room to the particle swarm, and process the particle velocity corresponding to each particle in the particle swarm. By applying the repulsion term to the particle swarm, the positions of any two particles in the swarm are compared to form the way the repulsion term acts in the particle swarm. In each iteration, the particle velocity is updated based on the relation density set, the meeting room space constraint set, and the repulsion term. After the particle velocity is updated, the particle position is updated. After a predetermined number of iterations, the seating layout code formed by the particle position of each particle in the particle swarm at the end of the iteration is used as a candidate seating layout scheme.

9. A meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The selection of the target seating layout scheme includes: Input each candidate seating layout scheme into the fitness function; The positional relationships of attendees in the candidate seating layout schemes are processed based on the relationship matching degree to form the relationship matching degree; The seating arrangement in the candidate seating layout scheme is processed based on the spatial constraint satisfaction to form the spatial constraint satisfaction. The order of attendees in the candidate seating arrangement is processed according to the role priority satisfaction, thus forming the role priority satisfaction. The distribution of seats in the candidate seating layout schemes is processed based on the layout balance index to form the layout balance index. The target seating layout scheme is obtained by combining the relationship matching degree, spatial constraint satisfaction degree, role priority satisfaction degree, and layout balance index according to the requirements of the fitness function.

10. A meeting notification and seat scheduling intelligent scheduling system based on adaptive learning according to claim 2, characterized in that, The generation of the new target seating layout scheme includes: Based on the target seating layout plan, the correspondence between attendees and seats in the target seating layout plan is processed, and the seat corresponding to each attendee is written into the meeting notification data in text form. After the meeting notification data is generated, the meeting notification data will be pushed to the participants included in the target seating layout plan; When participants are added or leave the meeting, the behavior input dataset is processed to add or remove the content related to the addition or departure of participants. When meeting conditions change, the behavior input dataset is processed, and the content related to the change in meeting conditions is written into the behavior input dataset; When the behavioral input dataset changes, the incremental self-organizing neural network learning is re-executed; After the incremental self-organizing neural network learning is completed, the particle swarm iteration search is re-executed to generate a new target seating layout scheme based on the content provided by the incremental self-organizing neural network learning, thereby achieving real-time adaptive optimization.

Images