Reading room seat intelligent distribution and management system and method based on AI

Abstract

The invention discloses an AI-based reading room seat intelligent distribution and management system and method. The system comprises a seat condition detection module, a seat distribution analysis module and a seat real-time feedback module. According to the invention, after the data updating frequency of the seat distribution optimization model is confirmed, the seat state change condition of each seat area of the reading room in the seat management period is monitored in real time to determine whether the recalculation of seat distribution is executed by the model, if yes, the recalculation time consumption condition of the seat distribution optimization model is analyzed, and the recalculation time consumption condition of the seat distribution optimization model is analyzed. According to the method, a real-time influence strategy is executed, otherwise, seat distribution local optimization analysis and adjustment are carried out, and finally, a reading room seat intelligent distribution scheme updated in real time is automatically fed back based on dynamic adjustment of a seat distribution analysis module on seat state changes, so that the dynamic adjustment real-time performance during seat intelligent distribution is improved, and the user experience is improved. The problem of insufficient real-time performance of dynamic adjustment during intelligent seat allocation in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computing model, and particularly relates to an AI-based reading room seat intelligent allocation and management system and method. BACKGROUND

[0002] In order to realize a seat intelligent allocation system covering seat allocation, management, monitoring, user experience and the like, in combination with the reading room scene, the implementation of the system is refined into multiple function modules, specifically including intelligent seat allocation function, user service function, management and operation function, and AI intelligent analysis and optimization; wherein the intelligent seat allocation function is used to realize reservation allocation, dynamic adjustment, personalized recommendation and fairness optimization, the reservation allocation means that the user reserves a seat through the system, the AI recommends the optimal seat according to the preferences (window / power / quiet area), historical behavior and use time prediction, the dynamic adjustment means that when temporary cancellation, no-show, seat damage and the like occur, the AI can adjust and reallocate the empty seats in real time, the personalized recommendation means recommending appropriate seats (for example, recommending quiet areas for users who learn for a long time, and recommending seats near the exit for users who use for a short time) based on user portraits and learning habits, and the fairness optimization means avoiding that the positions with higher functions are always allocated to the same part of people through scheduling algorithms; the user service function is used to realize multi-channel reservation, real-time state query, check-in and check-out management, temporary leaving function and learning data feedback, the multi-channel reservation is used to support mobile APP (Application, application program), WeChat mini-program and Web (World Wide Web, World Wide Web) reservation, the real-time state query means that the user can check the seat usage (idle / occupied / about to release), the check-in and check-out management means checking in through scanning, facial recognition or NFC (Near Field Communication, near field communication), and automatically releasing if not checked in for a certain period of time, the temporary leaving function is used to support the user to leave for a short time (such as going to the toilet / taking books), the system automatically times, releases when overtime, the learning data feedback is used to count the user learning time, seat usage habits, and form a personal learning file; the management and operation function is used to realize global seat monitoring, abnormality detection, statistics and report, and flexible rule configuration, the global seat monitoring is used for the management end to view the seat state, usage rate, reservation rate, no-show rate and the like in real time, the abnormality detection is used to detect long-time seat occupation without learning, illegal proxy check-in, malicious occupation and the like, the system automatically alarms or releases the seat, the statistics and report are used to generate seat utilization rate, user satisfaction, reservation peak period and the like data report, assist management optimization, and the flexible rule configuration means configuring the upper limit of reservation time, advance reservation time, no-show punishment mechanism; the AI intelligent analysis and optimization is used to realize demand prediction, user behavior modeling, abnormal behavior identification and dynamic optimization model, the demand prediction means predicting the future seat demand peak by using historical data, and making allocation optimization in advance, the user behavior modeling is based on the user's arrival time, stay time, preferred area, and continuously optimizes the recommendation strategy, the abnormal behavior identification identifies whether there are illegal situations such as "occupying seats without using" or "proxy occupying seats" through image recognition or sensor detection, and the dynamic optimization model uses reinforcement learning or heuristic algorithm to realize rapid adjustment, and still can give a better scheme when the resources are limited.

[0003] The above technology at least has the following technical problems: In the intelligent seat allocation function, dynamic adjustment may have deficiencies. In actual application, when temporary cancellation, no-show, etc. occurs, the AI needs to recalculate the allocation scheme. If the algorithm is time-consuming, it may not respond in time, and there is a problem of insufficient real-time performance of dynamic adjustment of seat intelligent allocation. SUMMARY

[0004] The present application provides an AI-based reading room seat intelligent allocation and management system and method, which solves the problem of insufficient real-time performance of dynamic adjustment of seat intelligent allocation in the prior art, and improves the real-time performance of dynamic adjustment of seat intelligent allocation.

[0005] To solve the above-mentioned application purposes, the technical solutions provided by the present application are as follows: an AI-based reading room seat intelligent allocation and management system comprises a seat condition detection module, a seat allocation analysis module and a seat real-time feedback module; wherein the seat condition detection module is used to: after confirming the data update frequency of the seat allocation optimization model used to generate the reading room seat allocation scheme, real-time monitor the seat state change of each seat area in the reading room in the seat management cycle to determine whether to perform recalculation of seat allocation by the model; the seat allocation analysis module is used to: if recalculation of seat allocation by the model is performed, analyze the recalculation time consumption of the seat allocation optimization model, and execute the adaptive real-time performance influence strategy for ensuring seat intelligent allocation, otherwise, perform seat allocation local optimization analysis and adjustment to reduce the real-time performance influence degree of local optimization on seat intelligent allocation; the seat real-time feedback module is used to: based on the dynamic adjustment of seat state change by the seat allocation analysis module, automatically feedback the real-time updated reading room seat intelligent allocation scheme to improve the response ability of the seat allocation optimization model to dynamic changes of seat state.

[0006] The present application also provides an AI-based reading room seat intelligent allocation and management method, and the specific steps are as follows: S1, after confirming the data update frequency of the seat allocation optimization model used to generate the reading room seat allocation scheme, real-time monitor the seat state change of each seat area in the reading room in the seat management cycle to determine whether to perform recalculation of seat allocation by the model; S2, if recalculation of seat allocation by the model is performed, analyze the recalculation time consumption of the seat allocation optimization model, and execute the adaptive real-time performance influence strategy for ensuring seat intelligent allocation, otherwise, perform seat allocation local optimization analysis and adjustment to reduce the real-time performance influence degree of local optimization on seat intelligent allocation; S3, based on the dynamic adjustment of seat state change by S2, automatically feedback the real-time updated reading room seat intelligent allocation scheme to improve the response ability of the seat allocation optimization model to dynamic changes of seat state.

[0007] Compared with the prior art, the above technical solution has at least the following beneficial effects: 1. The above scheme ensures the stability of the update frequency of the system itself after confirming the data update frequency of the seat allocation optimization model used to generate the reading room seat allocation plan. Then, the seat status changes of each seat area in the reading room during the seat management cycle are monitored in real time to determine whether the model should recalculate the seat allocation, thereby improving the real-time response to the seat allocation plan changes caused by seat status changes. If the model recalculates the seat allocation, the recalculation time of the seat allocation optimization model is analyzed, and the time consumption of recalculating the seat allocation plan is quantified more accurately, thereby more accurately determining the impact of recalculating the seat allocation plan on the dynamic state. The real-time impact of dynamic adjustment is calculated, and the adaptive real-time impact strategy is executed to ensure the timeliness of intelligent seat allocation. Otherwise, local optimization analysis and adjustment of seat allocation are performed to reduce the real-time impact of local optimization on intelligent seat allocation, thereby improving the real-time performance of local optimization of seat allocation. Finally, based on the dynamic adjustment of seat status changes by the seat allocation analysis module, the real-time updated intelligent seat allocation plan for the reading room is automatically fed back, thereby improving the responsiveness of the seat allocation optimization model to dynamic changes in seat status, thereby achieving improved real-time performance of dynamic adjustment during intelligent seat allocation, and solving the problem of insufficient real-time performance of dynamic adjustment during intelligent seat allocation in the existing technology.

[0008] 2. The above scheme obtains the recalculation data scale parameter of the seat allocation optimization model to obtain the model calculation time impact coefficient, thereby more accurately quantifying the time consumption of the seat allocation optimization model for recalculation, and then providing a data basis for whether to recalculate the seat allocation plan. Then, the model calculation time impact coefficient is matched with the time consumption mapping table to output the corresponding calculation time estimation value, so as to more accurately evaluate the impact of the recalculation time on the real-time performance of dynamic adjustment. Then, the feedback time limit value and the feedback time error value are extracted to make corresponding judgments. If the calculation time estimation value is less than the feedback time limit value, it means that the recalculation does not affect the real-time performance of dynamic adjustment, and the model continues to perform seat allocation. If the estimated calculation time is not less than the feedback time limit, it means that the recalculation may affect the real-time performance of the dynamic adjustment and further judgment is required. The difference between the estimated calculation time and the feedback time limit is recorded as the corresponding feedback time difference value, which helps to improve the real-time performance of subsequent judgments. Finally, the feedback time difference value is compared with the feedback time error value. If the feedback time difference value is less than the feedback time error value, it means that the real-time performance of the dynamic adjustment is not affected for the time being. The model will continue to recalculate the seat allocation. Otherwise, the real-time impact strategy will be executed to reduce the calculation time of the model to recalculate the seat allocation plan, thereby improving the real-time performance of the seat intelligent allocation plan during dynamic adjustment.

[0009] 3. The above scheme obtains a seat state change number sequence by counting the number of seat state changes in each seat area where the seat state changes occur, so as to facilitate the local optimization of each seat area. Then, based on the number of seat state changes in each seat area in the seat state change number sequence and the seat state change number processing value, the corresponding seat local optimization scheme is determined, thereby ensuring the accuracy of the seat allocation optimization. If the number of seat state changes in a seat area is greater than the seat state change number processing value, the seat area is locally optimized based on the seat state change number to ensure the integrity of the seat allocation scheme, and parallel optimization is obtained based on the seat state change difference query. To save time, the duration of the current calculation's impact on real-time performance is determined, and then the seat areas whose seat status change number is not greater than the seat status change number processing value are optimized in parallel, and the seat areas whose sum of seat status change number is not greater than the seat status change number processing value are combined for parallel optimization, thereby improving the corresponding local optimization efficiency and saving local optimization time. At the same time, the corresponding parallel optimization time is obtained in real time, which is helpful for subsequent analysis. Finally, the real-time performance of local optimization is determined based on the seat management time, and the effectiveness of the local optimization scheme on real-time adjustment is further analyzed, so that the corresponding optimization scheme is selected in time to ensure the scientific nature of seat reallocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A schematic diagram of the structure of an AI-based reading room seat intelligent allocation and management system provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of the long short-term memory network model provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of the decision tree model provided in the embodiment of the present application; Figure 4 A schematic diagram of a flow chart for analyzing the time consumption of recalculation provided in an embodiment of the present application; Figure 5 A schematic diagram of the process of local optimization analysis and adjustment of seat allocation provided in an embodiment of the present application; Figure 6 A flowchart of an AI-based intelligent allocation and management method for reading room seats provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application aims at the problem of insufficient real-time dynamic adjustment of seat intelligent allocation in the prior art, and provides an AI-based reading room seat intelligent allocation and management system and method. After confirming the data update frequency of a seat allocation optimization model, the seat state change of each seat area in the reading room in a seat management period is monitored in real time to determine whether to perform re-computation of seat allocation by the model. If yes, the time consumption of re-computation of the seat allocation optimization model is analyzed, and a real-time impact strategy is executed. Otherwise, local optimization analysis and adjustment of seat allocation are performed. Finally, based on the dynamic adjustment of seat state change by the seat allocation analysis module, a real-time updated seat intelligent allocation scheme of the reading room is automatically fed back, thereby improving the real-time dynamic adjustment of seat intelligent allocation.

[0013] As shown in FIG. 1, a structure schematic diagram of an AI-based reading room seat intelligent allocation and management system provided by an embodiment of the present application is shown. The AI-based reading room seat intelligent allocation and management system comprises a seat condition detection module, a seat allocation analysis module and a seat real-time feedback module. Figure 1

[0014] The seat condition detection module is configured to, after confirming the data update frequency of a seat allocation optimization model used to generate a reading room seat allocation scheme, monitor the seat state change of each seat area in the reading room in a seat management period in real time to determine whether to perform re-computation of seat allocation by the model.

[0015] It should be noted that the seat allocation optimization model involves the following AI technologies according to different targets: 1. Optimization and planning model: integer linear programming (ILP) / constraint programming (CP), which is used for optimal allocation problem that strictly satisfies constraints. When a heuristic algorithm (greedy algorithm, local search) is adopted, a fast approximate solution is achieved, which is suitable for real-time allocation. When a meta-heuristic algorithm (genetic algorithm, simulated annealing, ant colony algorithm) is adopted, it is suitable for large-scale complex optimization problems.

[0016] 2. Prediction and recommendation model: demand prediction model includes time series model, LSTM (Long Short-Term Memory) and Transformer, which is used to predict the seat demand in a certain period, help to reserve resources, and if seat preference modeling (collaborative filtering / recommendation system) is needed, the seat preference is predicted by learning the user's historical behavior, thereby improving the satisfaction.

[0017] ​As Figure 2 shown, a structure diagram of a long short-term memory network model provided by an embodiment of the present application, in the figure, a recurrent neural network model based on a long short-term memory network (LSTM) is shown, which is used to process a serialized seat allocation and optimization problem; wherein, the top part of the picture is a general recurrent neural network unit, x t (input) represents the current input at time step t, A (recurrent unit) represents a general recurrent neural network unit, which receives the current input x t and the hidden state (or internal state) from the previous time step, and performs calculation, h t (output) represents the output (usually the hidden state) at time step t; the self-loop arrow from A to A and the arrow from A to h t indicate that the unit has a "memory" function, that is, the calculation at the current time step depends not only on the current input x t but also on the information of the previous time step; the bottom picture is a long short-term memory network (LSTM) sequence model, the bottom picture specifically shows the general A unit at the top as an LSTM unit, and shows its expansion in time sequence.

[0018] Specifically, time sequence expansion: the figure shows three consecutive time steps: t-1 (previous period), t (current period) and t+1 (next period), each time step has an LSTM unit A; its input (x): x t-1 is the user reservation information and seat state of the previous period, x t is the current user reservation information and seat state, x t+1 is the user reservation information and seat state of the next time, the above input data represents the user demand (such as new reservation, cancellation, user preference, etc.) and the real-time state (such as idle, occupied, failure, cleaning, etc.) of the seat itself in a specific time point (or time period) in a local seat area; its output (h): h t-1 is the seat allocation result, optimization index and user feedback of the previous time, h t is the seat allocation result, optimization index and user feedback of the current time, h t+1 is the seat allocation result, optimization index and user feedback of the next time, the above output is the decision and evaluation given by the LSTM unit after processing the input at each time step, which specifically includes: seat allocation result: specific seat allocation scheme (for example, which user is allocated to which seat), optimization index: index for measuring allocation quality (for example, seat utilization rate, user satisfaction, conflict rate, etc.); user feedback: which can be explicit (user's evaluation of allocation) or implicit (whether the user accepts the allocation, whether the user signs in on time, etc.), which is used for further optimization of the model.

[0019] Note that the internal structure of LSTM unit A (take the middle A unit as an example): LSTM unit solves the gradient vanishing / explosion problem of traditional RNN (Recurrent Neural Network) by introducing "gate" structure, so that it can better learn and remember long-term dependencies. Forget gate: controlled by a sigma (sigmoid) activation function, decides which information to discard from the cell state (not explicitly marked in the figure, but passed through the loop connection) of the previous time step. For example, if a reservation has expired, the forget gate will decide to "forget" the old state of the reservation. Input gate: jointly controlled by a sigma activation function and a tanh activation function. Sigma decides which new information needs to be updated to the cell state, and tanh is responsible for generating new candidate values. For example, when new user reservation information comes in, the input gate decides how to add these new information to the memory. Cell state update: through the calculation of forget gate and input gate, as well as addition and multiplication operations, the cell state is updated. This is the core of LSTM, which carries the network's "memory". Output gate: jointly controlled by a sigma activation function and a tanh activation function. Sigma decides which part of the cell state will be output to the hidden state h t of the current time step, and tanh processes the cell state. For example, according to the current memory, the output gate decides the final seat allocation result and optimization index. In summary, this figure represents a learning and adaptive time series decision system that can dynamically allocate and optimize local seat areas based on historical and current seat and user data.

[0020] 3. Reinforcement Learning (RL): model seat allocation as a sequential decision problem, e.g., if the state is the current seat occupancy and user request, then perform action as user allocation to a certain seat, and get corresponding reward to improve utilization, user satisfaction, and reduce conflicts; RL is suitable for dynamic scenarios.

[0021] For example, Figure 3As shown, the structural diagram of the decision tree model provided by the embodiment of the application is shown, which shows a multi-decision tree-based intelligent seat allocation system architecture. DataSet is the input layer, which is the original input data layer of the whole system. DataSet includes seat-related data, user demand data, and constraint conditions. Specifically, the seat-related data includes the physical attributes of each seat (such as location, type, whether it is near the window, whether it has power supply), current state (idle, occupied, in maintenance), historical use, etc.; the user demand data includes the user's reservation information (required time, duration), preferences (such as quiet area, proximity to specific facilities), user type (ordinary user, VIP user), historical behavior, etc.; the constraint conditions include system-level rules (such as maximum reservation duration, same user can only reserve one seat at the same time, social distancing restrictions), regional capacity restrictions, and operation strategies. Then, the parallel decision tree analysis, i.e., the processing layer, is performed. The system distributes the input data to three parallel decision trees (all marked as "Tree-1, Tree-2, and Tree-3" in the figure). Each decision tree focuses on a specific analysis dimension: Tree-1 is used to analyze seat attributes. This decision tree focuses on evaluating and understanding the inherent characteristics of each seat and its impact on allocation. Its input mainly uses "seat-related data" in DataSet, and its output may include scores or classifications of seat comfort, functionality, and popularity, for example, a seat near the window with power supply may be assigned a higher attribute score. Tree-2 is used to analyze user demand and preferences. This decision tree focuses on understanding the user's individual needs and potential preferences to achieve more accurate matching. Its input mainly uses "user demand data" in DataSet, and its output may include user's inclination score for specific seat attributes, or prediction of the most likely acceptable seat type based on user's historical behavior, for example, a user who often reserves a quiet area will have a higher preference score for quiet seats. Tree-3 is used to analyze time constraints. This decision tree focuses on handling various time-related restrictions and optimizations to ensure the rationality and feasibility of the allocation time. Its input mainly uses "user demand data" (reservation time, duration) and "constraint conditions" (seat available time period, reservation rules) in DataSet, and its output may include judgments or scores such as whether a seat is available in a specific time period, whether it meets the user's duration requirement, whether there is a time conflict, etc. For example, if a seat has been reserved for the user's required period, the decision tree will give an unavailable judgment.

[0022] Secondly, the aggregation mechanism (decision layer) of the decision tree output, which integrates the analysis results (possibly scores, classifications, feasibility judgments, etc.) of the three parallel decision trees, can be implemented through weighted summation, voting mechanism, rule engine, or another meta-learner; this mechanism is the core decision-making part of the system, responsible for weighing all factors, resolving potential conflicts, and finding the best allocation solution. For example, weighted summation assigns weights to the output of each decision tree according to the importance of different dimensions (e.g., user preferences may be more important than seat attributes, or vice versa), and then sums them up to get a comprehensive score; the rule engine defines a series of rules, such as "if the seat attribute score is high and the time is available, then prioritize user preferences"; the optimization algorithm takes the outputs of the decision trees as objective functions or constraints, and finds the best allocation through optimization algorithms.

[0023] Figure 3 The Final Result is the output layer, used to output the final result of seat allocation, which is the final output of the entire system, i.e., the specific seat allocation plan, which specifies which user is allocated to which specific seat at which time period. This result is the optimized product after considering seat attributes, user demand preferences, and time constraints; as described above, this diagram depicts a multi-dimensional, parallel processing intelligent seat allocation system that breaks down complex allocation problems into multiple sub-problems (seat attributes, user demand, time constraints), analyzes them using specialized decision trees, and then integrates these analysis results through an aggregation mechanism to achieve efficient, intelligent, and multi-demand seat allocation, thereby improving the accuracy of seat allocation and user satisfaction.

[0024] The seat allocation analysis module is used to analyze the time-consuming situation of the re-computation of the seat allocation optimization model if the re-computation of the seat allocation is performed by the model, and to perform an adaptive real-time impact strategy for ensuring the real-time performance of the intelligent seat allocation, otherwise to perform local optimization analysis and adjustment to reduce the real-time impact of local optimization on the intelligent seat allocation.

[0025] The seat real-time feedback module is used to automatically feedback the real-time updated intelligent seat allocation scheme of the reading room based on the dynamic adjustment of the seat allocation analysis module to the seat state changes to improve the response capability of the seat allocation optimization model to the dynamic changes of the seat state. Among them, automatic feedback means real-time feedback on multiple terminals, for mobile terminals, push notifications or automatically refresh the seat reservation page to display the new scheme, for WeChat mini programs, through service messages or page automatic update, remind users of seat changes, while the web page automatically refreshes the seat distribution interface and prompts the user to update the scheme on the user reservation management page, and the NFC check-in device synchronously updates the local cache seat allocation state to ensure that the information is consistent when the user checks in.

[0026] In the embodiment, the seat state is monitored in real time by the seat condition detection module, and combined with the data update frequency confirmation of the model, it can be determined in time whether the model needs to be recalculated or locally optimized, so as to avoid unreasonable seat allocation caused by delay; and when the recalculation is needed, the seat allocation analysis module analyzes the calculation time of the model and selects the appropriate influence strategy, and when the recalculation is not needed, the local optimization is adopted, which reduces the overall calculation complexity and resource consumption, realizes the balance between global calculation and local optimization; the seat real-time feedback module automatically outputs the latest seat allocation scheme based on the result of the analysis module, so that the system can quickly respond to the dynamic changes of the seat state, improve the adaptability and accuracy of the scheme, and when it is detected that the overall recalculation is not needed, the system is adjusted through local optimization analysis and adjustment, which is helpful to control the influence degree of the local optimization process on the overall real-time allocation, and further ensures the real-time performance of the dynamic adjustment of the intelligent seat allocation.

[0027] Preferably, the specific process of data update frequency confirmation is as follows: Firstly, the update frequencies of the timing batch data and the dynamic update data of the seat allocation optimization model are obtained respectively, to obtain the corresponding timing batch data update frequency and dynamic update data update frequency. The timing batch data represents data with a set fixed update period, and the dynamic update data represents data with a variable update frequency. The timing batch data is usually updated at a fixed time point every day, and is used to affect the seat allocation of the next batch or the next day, such as user reservation information, blacklist, rule adjustment, etc. The dynamic update data is data that may change in real time, such as seat usage status, user real-time arrival and departure, cancellation and reservation request, etc. The update frequencies of the timing batch data and the dynamic update data are read through the log of the reading room seat intelligent allocation and management system.

[0028] In the second step, the timing batch data update frequency and the dynamic update data update frequency are compared with the corresponding timing batch data update frequency determination value and the dynamic update data update frequency determination value, including two cases. Specifically, the timing batch data update frequency determination value and the dynamic update data update frequency determination value are obtained from the preset database, which are usually set by the preset staff based on the requirements of seat intelligent allocation and historical experience, and are stored in the preset database in advance. In the first case, if the timing batch data update frequency and the dynamic update data update frequency are not less than the corresponding timing batch data update frequency determination value and the dynamic update data update frequency determination value, the seat state of each seat area is continuously detected. In the second case, if the timing batch data update frequency and the dynamic update data update frequency are less than the corresponding timing batch data update frequency determination value and the dynamic update data update frequency determination value, the timing batch data update frequency is adjusted to the timing batch data update frequency determination value, and the dynamic update data update frequency is adjusted to the dynamic update data update frequency determination value, otherwise, if the timing batch data update frequency is less than the corresponding timing batch data update frequency determination value, the timing batch data update frequency is adjusted to the timing batch data update frequency determination value, and if the dynamic update data update frequency is less than the corresponding dynamic update data update frequency determination value, the dynamic update data update frequency is adjusted to the dynamic update data update frequency determination value.

[0029] In the embodiment, by obtaining the update frequency of the timing batch data and the dynamic update data respectively and comparing them with the corresponding determination values, the timeliness and accuracy of the seat allocation optimization model in data driving can be ensured, and the distortion of the allocation result caused by data update lag can be avoided; at the same time, when the update frequency of the timing batch data or the dynamic update data is detected to be lower than the set determination value, the system will automatically adjust it to the determination value level, so as to realize the dynamic regulation of the data update frequency, and thus improve the stability and reliability of the model operation; and only when it is confirmed that the data update frequency meets the determination condition, the seat state detection link is continued, so as to ensure that the detection and subsequent optimization are based on the latest effective data, and the credibility of the detection and analysis result is improved; the automatic judgment and adjustment mechanism of the data update frequency enables the system to dynamically switch according to the actual data update situation, so as to better adapt to the data change demand in different scenarios in the reading room, and improve the real-time response ability of the seat allocation optimization model.

[0030] Preferably, the seat state changes of each seat area in the reading room during the seat management period are monitored in real time to determine whether to perform re-computation of seat allocation by the model, and the specific process is as follows: Step one, detecting the seat state of each seat area in the seat management period and counting the number of changed seats to obtain the seat state change quantity, and accumulating and summing the seat state change quantity of each seat area to obtain the total seat state change quantity, the seat state includes reserved, cancelled reservation, modified reservation, no-show, no check-in, early departure, and new reservation, etc. Step two, comparing the total seat state change quantity with the pre-set seat state change limit value that can trigger the re-computation of the seat allocation optimization model: if the total seat state change quantity is greater than the seat state change limit value, the seat allocation optimization model is executed to re-compute the seat allocation scheme, otherwise, the seat allocation local optimization analysis and adjustment are performed; wherein the seat state change limit value is obtained from the pre-set database, and is specifically determined by the pre-set staff based on the specific calculation performance of the model, and is set and stored in the pre-set database in advance.

[0031] In the embodiment, by detecting the state changes of each seat area in the seat management period and counting the seat state change quantity, it is ensured that the dynamic changes of each seat can be monitored in real time, and then accurate basis is provided for subsequent optimization decision; the system compares the total seat state change quantity with the pre-set trigger limit value, automatically judges whether the re-computation of the seat allocation optimization model needs to be triggered, avoids manual intervention, and thus improves the processing efficiency and system response speed; if the seat state change does not exceed the limit value, local optimization analysis and adjustment are selected, avoiding global re-computation for each small state change, thereby effectively saving computing resources and improving computing efficiency; the process can adjust the model calculation strategy according to the actual situation of seat state change, improve the flexibility and real-time response ability of the seat allocation optimization scheme, thereby enhancing the adaptability of the system to actual scene changes, and thus improving the real-time performance of the dynamic adjustment of the seat allocation optimization model.

[0032] As Figure 4As shown, the flowchart of the re-computation time consumption analysis provided by the embodiment of the application is shown, and the specific logic is as follows: obtaining the re-computation data scale parameter of the seat allocation optimization model to obtain the model computation time consumption influence coefficient; matching the model computation time consumption influence coefficient with the time consumption mapping table to output the corresponding computation time consumption estimation value; extracting the feedback time limit value and the feedback time error value; if the computation time consumption estimation value is less than the feedback time limit value, the re-computation of seat allocation by the model is continued; if the computation time consumption estimation value is not less than the feedback time limit value, the difference between the computation time consumption estimation value and the feedback time limit value is recorded as the corresponding feedback time difference value; comparing the feedback time difference value with the pre-set feedback time error value, if the feedback time difference value is less than the feedback time error value, the re-computation of seat allocation by the model is continued; if the feedback time difference value is not less than the feedback time error value, the real-time influence strategy for reducing the computation time consumption of the model re-computing the seat allocation scheme is executed; through the above process, intelligent prediction of the seat computation re-allocation time consumption length is realized, so that corresponding optimization measures are taken accordingly, and the real-time performance of seat intelligent allocation is improved.

[0033] Preferably, the re-computation time consumption of the seat allocation optimization model is analyzed, and the specific process is as follows: First, the re-computation data scale parameter of the seat allocation optimization model is obtained to obtain the model computation time consumption influence coefficient for quantifying the re-computation time consumption degree of the seat allocation optimization model, the re-computation data scale parameter includes the number of reservation users, the number of seats, the number of idle seats, and the user modification frequency for quantifying the frequency of modification of the corresponding seat state. It should be noted that the number of reservation users, the number of seats, and the number of idle seats can be directly obtained from the system, the number of seats represents the total number of reading room seats, and the user modification frequency for quantifying the frequency of modification of the corresponding seat state is obtained by ratio operation of the number of seats whose state is modified in the seat management period and the time length of the seat management period.

[0034] Then, the model calculates the time-consuming influence coefficient and matches it with the time-consuming mapping table set in advance to reflect the mapping relationship between the time-consuming influence coefficient and the corresponding time-consuming estimation value, and outputs the corresponding time-consuming estimation value. Specifically, the time-consuming mapping table is pre-constructed in the preset database. The time-consuming influence coefficient is input into the trained time-consuming mapping table to output the corresponding time-consuming estimation value, that is, the estimation value of the time-consuming influence coefficient on the time-consuming. The training data used in the mapping table comes from the time-consuming influence coefficient obtained in the historical period and the time-consuming estimation value set by professional technicians according to the experience rule, which is used to fit the mapping relationship between the time-consuming influence coefficient and the time-consuming estimation value, so as to more accurately evaluate the time-consuming. At the same time, the feedback time limit value and the feedback time error value used to ensure the real-time feedback effect of the dynamic adjustment of the seat allocation optimization model are extracted. It needs to be explained that the feedback time limit value and the feedback time error value are read from the preset database and are set in advance by the preset staff.

[0035] Then, if the time-consuming estimation value is less than the feedback time limit value, the model continues to perform the recalculation of seat allocation, otherwise the difference between the time-consuming estimation value and the feedback time limit value is recorded as the corresponding feedback time difference value, that is, the difference between the time-consuming estimation value and the feedback time limit value.

[0036] Finally, the feedback time difference value is compared with the feedback time error value. If the feedback time difference value is less than the feedback time error value, the model continues to perform the recalculation of seat allocation, otherwise the real-time influence strategy for reducing the time-consuming of the model to recalculate the seat allocation scheme is executed.

[0037] In the embodiment, by introducing the re-computation data scale parameters including the number of pre-booking users, the number of seats, the number of idle seats, and the user modification frequency, the model computation time consumption influence coefficient is calculated, the time consumption of the seat allocation optimization model re-computation is quantitatively analyzed more accurately, the model computation time consumption influence coefficient is matched with the time consumption mapping table, the corresponding computation time consumption estimation value is obtained, the system can predict the time consumption level in advance, and a scientific basis is provided for subsequent dynamic feedback regulation; at the same time, by introducing the feedback time limit value and the feedback time error value, it is ensured that the model re-computation time consumption does not exceed the expected feedback time, so that the seat allocation optimization system has the ability of timely dynamic response; and in the case of long computation time, the system will automatically trigger the real-time influence strategy according to the comparison result of the feedback time difference value and the error value, thereby reducing the time consumption pressure of re-computation, and further improving the utilization efficiency of computing resources; by dynamically adjusting the model computation process under the premise of ensuring the feedback timeliness, both full re-computation within the controllable time consumption range and optimization strategy selection when the time consumption is too long are realized, and the stability of the system is further improved.

[0038] Preferably, the specific content of the real-time influence strategy is as follows: X1, the difference between the feedback time difference value and the feedback time error value is the feedback time deviation, that is, the feedback time difference value and the feedback time error value are subjected to difference operation to obtain the feedback time deviation, and according to the feedback time deviation, a projection mapping sequence for reflecting the mapping relationship between the feedback time deviation and the model constraint number reduction ratio is queried to obtain the corresponding model constraint number reduction ratio. It should be noted that the projection mapping sequence is pre-constructed in the preset database, and the feedback time deviation is input into the trained projection mapping sequence to output the corresponding model constraint number reduction ratio, that is, the reduction degree of the model constraint number by the feedback time deviation. The training data of the projection mapping sequence comes from the feedback time deviation obtained in the historical period and the model constraint number reduction ratio set by the professional and technical personnel according to the experience rule, which is used to fit the mapping relationship between the feedback time deviation and the model constraint number reduction ratio, so as to more accurately adjust the projection mapping sequence.

[0039] X2, the model constraint number reduction ratio is compared with the constraint number reduction ratio limit value of the set limit constraint number reduction degree, wherein the constraint number reduction ratio limit value is usually set in advance by the preset professional and technical personnel based on the specification requirements and experience of the constraint number, and stored in the preset database.

[0040] X3, if the model constraint quantity reduction ratio is lower than the constraint quantity reduction ratio limit value, the number of constraints with low-to-high constraint priority in the seat allocation optimization model is reduced by the model constraint quantity reduction ratio. It should be noted that the constraint is a rule and limit that must be followed when performing seat allocation, and reducing the number of constraints with low-to-high constraint priority means removing some low-priority constraints to reduce the complexity of constraints, for example, if the seat preferences of users such as window seats, proximity to aisles, etc. do not need to be considered, it can be simplified to seat area priority, or some complex priority sorting is removed and replaced with a simple "first come first served" or allocation according to a preset area.

[0041] X4, if the model constraint quantity reduction ratio is not lower than the constraint quantity reduction ratio limit value, the number of constraints with low-to-high constraint priority in the seat allocation optimization model is reduced by the constraint quantity reduction ratio limit value.

[0042] In this embodiment, by calculating the feedback time deviation and combining the projection mapping sequence, the feedback time deviation can be mapped with the model constraint quantity reduction ratio, so as to realize fine regulation and control of the model calculation complexity. The system selectively reduces the constraints in the seat allocation optimization model according to different model constraint quantity reduction ratios, and the reduction process gradually reduces according to the constraint priority from low to high, ensuring efficient implementation of the optimization goal within a certain range. By setting the constraint quantity reduction ratio limit value, the distortion or unavailability of the allocation result caused by the excessive reduction of the model constraints is avoided, and the executability of the optimization model is ensured while reducing the time consumption. And dynamically adjusting the number of constraints according to different feedback time deviations can make the seat allocation optimization system better adapt to the time consumption requirements in different scales and different scenarios, and realize more intelligent seat allocation regulation and control.

[0043] As a further embodiment, the method for obtaining the model calculation time consumption influence coefficient includes two steps: The first step operation recalculates the data size parameter for data normalization processing and matches the corresponding recalculated data size parameter influence factor for reflecting the influence degree of the model calculation time consumption influence coefficient. The recalculated size parameter influence factor includes the reservation user number influence factor, the seat number influence factor, the idle seat number influence factor, and the user modification frequency influence factor. It needs to be supplemented that the recalculated size parameter influence factor is obtained based on the influence factor matching data table in the preset database. The influence factor matching data table is a data table reflecting the mapping relationship between the recalculated data size parameter and the corresponding recalculated size parameter influence factor. The real-time recalculated data size parameter is input into the influence factor matching data table, and the corresponding reservation user number influence factor, seat number influence factor, idle seat number influence factor, and user modification frequency influence factor are output. They are used for the influence degree of the reservation user number, seat number, idle seat number, and user modification frequency on the model calculation time consumption influence coefficient, so as to obtain a more accurate model calculation time consumption influence coefficient.

[0044] The second step operation is to obtain the model calculation time consumption influence coefficient by coupling the weighted operation of the recalculated data size parameter based on the recalculated size parameter influence factor. The specific limit expression is as follows: ; In the formula, x1 represents the reservation user number, x2 represents the seat number, x3 represents the idle seat number, x4 represents the user modification frequency, a represents the reservation user number influence factor, b represents the seat number influence factor, c represents the idle seat number influence factor, d represents the user modification frequency influence factor, and y represents the model calculation time consumption influence coefficient.

[0045] In the embodiment, the algorithm combines the re-computation data size parameter and the corresponding re-computation size parameter influence factor for comprehensive analysis to obtain a model computation time influence coefficient, wherein, as the re-computation data size parameter increases, the corresponding model computation time influence coefficient also increases; specifically, when the number of reservation users, the number of seats, and the number of idle seats gradually increase, the higher the requirement for the computing capacity of the seat allocation optimization model, the higher the computing complexity, and the longer the computation time, the larger the corresponding model computation time influence coefficient; similarly, when the user modification frequency is higher, the seat state that the model needs to process may be more complex, the higher the requirement for the computing capacity of the model, the longer the computation time, and the larger the corresponding model computation time influence coefficient; moreover, the re-computation data size parameters are interrelated, for example, when the number of reservation users is not greater than the number of seats, it indicates that there is a seat surplus, otherwise, seat optimization allocation is needed, which determines whether there is seat competition in the reading room; the number of idle seats reflects the resource utilization rate, and a small number of idle seats indicates a resource shortage and a high seat utilization rate, while a large number of idle seats indicates that the seats may not be allocated evenly; in addition, if the number of reservation users is close to the total number of seats, the number of idle seats tends to be 0, and if the user modification frequency is higher, the effective reservation number of the system may be unstable, that is, the number of user reservations is unstable; through the above analysis, it is helpful to more clearly understand the time consumption of model re-computation, thereby more accurately evaluating the influence of re-computation time consumption on the real-time performance of dynamic adjustment, and then taking corresponding optimization measures to improve the real-time performance of dynamic adjustment of seat intelligent allocation.

[0046] As shown in Figure 5 FIG. 1 is a flowchart of seat allocation local optimization analysis and adjustment provided by an embodiment of the present application, and the specific logic is as follows: the number of seat state changes of each seat area where the seat state changes is counted to obtain a seat state change number sequence; the number of seat state changes of each seat area in the seat state change number sequence is determined based on the seat state change number processing value; if the number of seat state changes of the seat area is greater than the seat state change number processing value, the seat area is locally optimized based on the seat state change number, and a seat state change difference value is obtained based on the deviation operation result between the seat state change number and the seat state change number processing value, and the parallel optimization time saving is obtained by querying the time saving mapping table; the seat area whose number of seat state changes is not greater than the seat state change number processing value is parallelly optimized, and each seat area whose sum of seat state changes is not greater than the seat state change number processing value is combined for parallel optimization, while the corresponding parallel optimization time is obtained in real time; the local optimization real-time performance is determined based on the seat management time; through the above analysis, it is helpful to reduce the time consumption of seat allocation local parallel optimization and improve the real-time performance of seat allocation optimization dynamic adjustment.

[0047] Preferably, the specific steps of performing the seat allocation local optimization analysis and adjustment are as follows: P1, the number of seat state changes of each seat area where the seat state change occurs is counted to obtain a seat state change number sequence. P2, based on the number of seat state changes of each seat area in the seat state change number sequence, a seat state change number processing value for limiting the size of parallel optimization is determined, which is usually preset by a pre-set staff. P3, if the number of seat state changes of the seat area is greater than the seat state change number processing value, the seat local optimization is performed based on the number of seat state changes, and the seat state change difference value is obtained based on the deviation operation result between the number of seat state changes and the seat state change number processing value, that is, the result of ratio operation between the difference value between the number of seat state changes and the seat state change number processing value and the seat state change number processing value, and the parallel optimization time saving is obtained by querying in the time saving mapping table which has a mapping relationship between the seat state change difference value and the parallel optimization time saving. It needs to be explained that the seat local optimization means optimization by local area allocation algorithm, which includes dynamic adjustment, priority allocation and conflict detection. The dynamic adjustment is based on the current seat state in the seat area, combined with the seat idle condition of the surrounding seat area, user demand, etc., to calculate the best seat allocation scheme of the seat area. The priority allocation adjusts the seat allocation according to the current reservation condition or user behavior (such as giving priority to high-frequency reservation users or signed-in users for seat allocation). The conflict detection checks whether there is seat conflict or repeated allocation in the current seat area. If there is conflict, the seat arrangement in the seat area is adjusted again. It needs to be supplemented that the local area allocation algorithm includes but is not limited to priority-based allocation algorithm, proximity optimization algorithm, time slice optimization algorithm, conflict resolution algorithm and regional load balancing algorithm, etc., which are set according to the corresponding reading room seat allocation rules and requirements. Among them, the priority-based allocation algorithm is used to sort the idle seats and users to be allocated in the local area according to priority, and match them in order. The proximity optimization algorithm quantifies the physical distance between the current position or target position of the user and the idle seat, and preferentially selects the seat with the lowest cost for allocation. The time slice optimization algorithm divides the use time of the seat into segments, and matches the users with the highest "time demand fit degree" according to the use time length required by the user and the remaining time length of the seat. The conflict resolution algorithm solves the conflict by setting rules (such as reservation sequence, user credit score, use time length, etc.), or by "secondary allocation", which reallocates the conflicting users to other available seats in the same seat area. The regional load balancing algorithm is used to randomly or uniformly distribute the users in the seat area to ensure balanced overall seat utilization.

[0048] It should be noted that the time saving mapping table is pre-constructed in the preset database, and the seat state change difference is input into the trained time saving mapping table to output the corresponding parallel optimization time saving, that is, the time saving to be optimized reflected by the seat state change difference. The training data used by the mapping table comes from the seat state change difference obtained in the historical period, and the parallel optimization time saving set by the professional technician according to the experience rule, which is used to fit the mapping relationship between the seat state change difference and the parallel optimization time saving.

[0049] P4, the seat area with a seat state change number not greater than the seat state change number processing value is parallel optimized, and each seat area with a seat state change number sum not greater than the seat state change number processing value is combined for parallel optimization, while the corresponding parallel optimization time is obtained in real time; wherein, the parallel optimization means that each seat area that can be combined is synchronized to perform seat local optimization. For example, assuming that the seat state change number processing value is 100, and the seat state change number of each seat area in the seat state change number sequence is 15, 8, 26, 9, 2, 10, 25, 14, 5, 12,..., the seat state change numbers of each seat area are added from front to back, 15+8+26+...+25=95, then the corresponding seat area can be combined for parallel optimization, if 14 is added, the value is 109, which exceeds the seat state change number processing value 100, then the seat area corresponding to the seat state change number 14 is included in the next combination, and the seat state change number 14 is added from the back, that is, "14+5+12+...".

[0050] P5, based on the seat management time, the local optimization real-time determination is performed, the seat management time includes the total parallel optimization time, the total optimization time saving and the total local optimization time, the total parallel optimization time is the sum of each parallel optimization time, the total optimization time saving is the sum of each parallel optimization time saving, and the total local optimization time is the time length of each seat area performing the reallocation local optimization.

[0051] In the embodiment, by counting and judging the number of state changes of the seat area, the area with a large number of changes can be focused on for optimization, the optimization range is localized, and the pertinence and refinement of the optimization are improved. For the area with a number of seat state changes less than the processing value, parallel optimization is performed in a combined manner, which can fully utilize parallel computing resources while ensuring reasonable distribution, shorten the overall optimization time, and dynamically estimate the time saving effect of optimization to flexibly select the appropriate optimization strategy under different change scales. By comprehensively considering the total parallel optimization time, the total optimization time saving, and the total local optimization time, a real-time judgment mechanism for seat management time is established, which helps to avoid the feedback effect of affecting system dynamic adjustment due to long optimization time, ensures that large-scale change areas can respond quickly, and ensures that small-scale change areas can be processed in a combined parallel optimization manner, thereby balancing the overall optimization effect and system stability.

[0052] Preferably, the specific conditions of the local optimization real-time judgment based on the seat management time are as follows: in the first case, if the total local optimization time is less than the feedback time limit value, it indicates that the current optimization is effective, and the seat state of each seat area in the next seat management period is detected. In the second case, if the total local optimization time is not less than the feedback time limit value, the difference between the total local optimization time and the total parallel optimization time is recorded as the to-be-verified optimization time saving, which is the difference between the total local optimization time and the total parallel optimization time, and the to-be-verified optimization time saving is compared with the total optimization time saving. In the third case, if the to-be-verified optimization time saving is higher than the total optimization time saving, it indicates that the local optimization is effective, an alarm prompt of seat allocation optimization model computing resource is issued, and a request for a pre-set staff to release a seat reallocation result delay publication is sent. In the fourth case, if the to-be-verified optimization time saving is not higher than the total optimization time saving, it indicates that the local optimization is ineffective, and the local optimization is adjusted.

[0053] In the embodiment, by judging whether the total local optimization time is less than the feedback time limit value, the dynamic monitoring of the optimization real-time is realized, the optimization result can be output within the limited time, thereby avoiding the allocation delay, and the real-time is improved. Through the double judgment mechanism (time comparison and saving time comparison), the optimization can be more accurately distinguished, and the deviation caused by single index judgment is avoided. When it is found that the local optimization is within the time limit but the resource pressure is large, the system can automatically issue an alarm of computing resource, and request the staff to delay the publication, thereby reducing the user expectation gap caused by long optimization time. By taking the delay publication or optimization adjustment measures in the time-consuming over-limit scenario, the balance between user experience and computing resources can be achieved.

[0054] Preferably, the specific content of adjusting the local optimization is as follows: The difference between the to-be-verified saving optimization time length and the total optimization saving time length is quantified to obtain a saving time length difference, which represents that the saving time length difference is obtained by subtracting the to-be-verified saving optimization time length from the total optimization saving time length, and the seat state change quantity processing value amplification corresponding to the saving time length difference is matched in the mapping set constructed for fitting the mapping relationship between the saving time length difference and the seat state change quantity processing value amplification. It should be noted that the seat state change quantity processing value amplification is obtained based on the mapping set in the preset database, and the mapping set is a data set reflecting the mapping relationship between the saving time length difference and the corresponding seat state change quantity processing value amplification. The real-time saving time length difference is input into the mapping set, and the corresponding seat state change quantity processing value amplification is output, which is used to reflect the amplification degree of the seat state change quantity processing value, so as to obtain a more accurate seat state change quantity processing value.

[0055] The seat state change quantity processing value is amplified based on the seat state change quantity processing value amplification, which represents that the seat state change quantity processing value is multiplied by the seat state change quantity processing value amplification. If the to-be-verified saving optimization time length of the next seat management cycle is still not higher than the total optimization saving time length, the amplification of the seat state change quantity processing value is continued until it is higher than the set amplification limit value, and the loop optimization of the amplification of the seat state change quantity processing value is stopped. The amplification limit value is pre-stored in the preset database and is usually set by the preset staff. If the to-be-verified saving optimization time length is higher than the total optimization saving time length within the corresponding time period of the loop optimization, a seat allocation optimization model computing resource alarm prompt is sent, otherwise a seat allocation optimization model is triggered to recalculate the seat allocation scheme.

[0056] In the embodiment, by quantifying the difference between the to-be-verified saving optimization duration and the total optimization saving duration, and constructing a mapping relationship, the adjustment of the seat state change quantity processing value can be more accurately controlled, thereby ensuring the efficiency and accuracy of the optimization process; and based on the mapping result of the seat state change quantity processing value amplification, the system can dynamically amplify the processing value, so that the optimization can continue and cope with different optimization demands, thereby improving the adaptability of the system in a variable environment; and setting the amplification range limit value and stopping the cyclic optimization is beneficial to prevent system resource waste or optimization efficiency reduction caused by excessive amplification, thereby ensuring stable operation of the system. In the cyclic optimization process, if the optimization does not meet the expectation, the system can automatically trigger the recalculation of the optimization scheme to ensure that the final optimization result meets the actual demand, thereby improving the self-adaptation and self-adjustment ability of the system; by mapping and adjusting the saving duration difference in the optimization process, the system can more accurately allocate appropriate computing power resources for each period, thereby improving resource use efficiency and reducing resource waste; at the same time, in the cyclic optimization process, if the to-be-verified saving optimization duration is higher than the total optimization saving duration, the system can timely issue a computing power resource warning prompt to prevent potential bottlenecks and resource overload in the optimization scheme execution process.

[0057] As shown in Figure 6 FIG. 1 is a flowchart of an AI-based reading room seat intelligent allocation and management method provided by an embodiment of the present application. The specific steps are as follows: S1, after confirming the data update frequency of the seat allocation optimization model used to generate the reading room seat allocation scheme, the seat state change of each seat area in the reading room is monitored in real time within the seat management period to determine whether to perform recalculation of seat allocation by the model. S2, if the recalculation of seat allocation by the model is performed, the recalculation time consumption of the seat allocation optimization model is analyzed, and an adaptive real-time influence strategy for ensuring seat intelligent allocation is executed, otherwise, local optimization analysis and adjustment of seat allocation are performed to reduce the real-time influence degree of local optimization on seat intelligent allocation. S3, based on the dynamic adjustment of the seat state change in S2, the real-time updated reading room seat intelligent allocation scheme is automatically fed back to improve the response ability of the seat allocation optimization model to dynamic changes in seat state.

Claims

1. An AI-based reading room seat intelligent allocation and management system, characterized in that, The seat condition detection module, the seat allocation analysis module, and the seat real-time feedback module are included. The seat condition detection module is configured to monitor the seat state changes of each seat area in the reading room within a seat management period in real time to determine whether to perform re-computation of seat allocation by the model after confirming the data update frequency of the seat allocation optimization model used to generate the seat allocation scheme of the reading room. The seat allocation analysis module is configured to analyze the time consumption of re-computation of the seat allocation optimization model and perform an adaptive real-time influence strategy for ensuring the seat intelligent allocation if the re-computation of seat allocation by the model is performed, otherwise, to perform local optimization analysis and adjustment of the seat allocation to reduce the real-time influence degree of local optimization on the seat intelligent allocation. The seat real-time feedback module is configured to automatically feed back the real-time updated seat intelligent allocation scheme of the reading room based on the dynamic adjustment of the seat condition changes by the seat allocation analysis module to improve the response capability of the seat allocation optimization model to the dynamic changes of the seat condition.

2. The AI-based reading room seat intelligent allocation and management system according to claim 1, wherein, The specific process of the data update frequency confirmation is as follows: The update frequencies of the timing batch data and the dynamic update data of the seat allocation optimization model are obtained respectively to obtain the corresponding timing batch data update frequency and the dynamic update data update frequency, the timing batch data represents data with a set fixed update period, and the dynamic update data represents data with a dynamic change in update frequency. If the timing batch data update frequency and the dynamic update data update frequency are both not less than the corresponding timing batch data update frequency determination value and the dynamic update data update frequency determination value, the seat state of each seat area is continuously detected. If the timing batch data update frequency and the dynamic update data update frequency are both less than the corresponding timing batch data update frequency determination value and the dynamic update data update frequency determination value, the timing batch data update frequency is adjusted to the timing batch data update frequency determination value, and the dynamic update data update frequency is adjusted to the dynamic update data update frequency determination value, otherwise, if the timing batch data update frequency is less than the corresponding timing batch data update frequency determination value, the timing batch data update frequency is adjusted to the timing batch data update frequency determination value, and if the dynamic update data update frequency is less than the corresponding dynamic update data update frequency determination value, the dynamic update data update frequency is adjusted to the dynamic update data update frequency determination value.

3. The AI-based reading room seat intelligent allocation and management system according to claim 1, wherein, The specific process of the real-time monitoring of the seat state changes of each seat area in the reading room within the seat management period to determine whether to perform re-computation of seat allocation by the model is as follows: Step one, detecting the seat state of each seat area within the seat management period and counting the number of changed seats to obtain the seat state change number, and accumulating and summing the seat state change number of each seat area to obtain the total seat state change number. Step two, comparing the total seat state change number with the pre-set seat state change limit value triggering the re-computation of the seat allocation optimization model to re-compute the seat allocation scheme: If the total seat state change quantity is greater than the seat state change limit value, the seat allocation optimization model is recalculated to obtain a seat allocation scheme, otherwise, a local optimization analysis and adjustment of the seat allocation is performed.

4. The AI-based reading room seat intelligent allocation and management system according to claim 1, wherein, The recalculation of the seat allocation optimization model is analyzed, and the specific process is as follows: A recalculation data size parameter of the seat allocation optimization model is obtained to obtain a model calculation time consumption influence coefficient for quantifying the recalculation time consumption of the seat allocation optimization model, the recalculation data size parameter includes the number of reservation users, the number of seats, the number of idle seats, and a user modification frequency for quantifying the frequency of corresponding seat state modification; The model calculation time consumption influence coefficient is matched with a time consumption mapping table set in advance to reflect the mapping relationship between the calculation time consumption influence coefficient and the corresponding calculation time consumption estimation value, and the corresponding calculation time consumption estimation value is outputted; A feedback time limit value and a feedback time error value for ensuring the real-time feedback effect of the dynamic adjustment of the seat allocation optimization model are extracted; If the calculation time consumption estimation value is less than the feedback time limit value, the recalculation of the seat allocation by the model is continued; If the calculation time consumption estimation value is not less than the feedback time limit value, the difference between the calculation time consumption estimation value and the feedback time limit value is recorded as a corresponding feedback time difference value; The feedback time difference value is compared with a pre-set feedback time error value: If the feedback time difference value is less than the feedback time error value, the recalculation of the seat allocation by the model is continued; If the feedback time difference value is not less than the feedback time error value, a real-time influence strategy for reducing the calculation time consumption of the model recalculation of the seat allocation scheme is executed.

5. The AI-based reading room seat intelligent allocation and management system according to claim 4, wherein, The specific content of the real-time influence strategy is as follows: The difference between the feedback time difference value and the feedback time error value is recorded as a feedback time deviation, and a model constraint quantity reduction ratio corresponding to the feedback time deviation is obtained by querying a projection mapping sequence stored in advance to reflect the mapping relationship between the feedback time deviation and the model constraint quantity reduction ratio; The model constraint quantity reduction ratio is compared with a constraint quantity reduction ratio limit value set to limit the constraint quantity reduction degree, and the number of constraint conditions with a priority from low to high in the seat allocation optimization model is reduced by the model constraint quantity reduction ratio; If the model constraint quantity reduction ratio is not lower than the constraint quantity reduction ratio limit value, the number of constraint conditions with a priority from low to high in the seat allocation optimization model is reduced by the constraint quantity reduction ratio limit value. The specific acquisition method of the model calculation time consumption influence coefficient is as follows:

6. The AI-based reading room seat intelligent allocation and management system according to claim 4, wherein, The recalculation data size parameter is subjected to data normalization processing, and a corresponding recalculation size parameter influence factor for reflecting the influence degree of the recalculation data size parameter on the model calculation time consumption influence coefficient is matched, the recalculation size parameter influence factor includes a reservation user quantity influence factor, a seat quantity influence factor, an idle seat quantity influence factor, and a user modification frequency influence factor; ​ The model calculation time consumption influence coefficient is obtained by coupling the weighting operation on the re-calculation data scale parameter based on the re-calculation scale parameter influence factor.

7. The AI-based reading room seat intelligent allocation and management system according to claim 1, wherein, The specific steps of performing the seat allocation local optimization analysis and adjustment are as follows: The seat state change quantity of each seat area in which the seat state change occurs is counted to obtain a seat state change quantity sequence; It is determined whether the seat state change quantity of each seat area in the seat state change quantity sequence is greater than a seat state change quantity processing value preset for limiting the parallel optimization scale; If the seat state change quantity of a seat area is greater than the seat state change quantity processing value, the seat of the seat area is locally optimized based on the seat state change quantity, and a seat state change difference value is obtained based on the deviation operation result between the seat state change quantity and the seat state change quantity processing value. The parallel optimization time saving is obtained by querying a time saving mapping table in which a mapping relationship between the seat state change difference value and the parallel optimization time saving is established. The seat areas whose seat state change quantity is not greater than the seat state change quantity processing value are optimized in parallel, and each seat area whose seat state change quantity sum is not greater than the seat state change quantity processing value is combined for parallel optimization, while the corresponding parallel optimization time is obtained in real time. The local optimization real-time determination is performed based on the seat management time, which includes a total parallel optimization time length, a total optimization time saving length, and a total local optimization time length. The total parallel optimization time length is the sum of each parallel optimization time, the total optimization time saving length is the sum of each parallel optimization time saving, and the total local optimization time length is the time length for each seat area to execute the re-allocation local optimization.

8. The AI-based reading room seat intelligent allocation and management system according to claim 7, wherein, The specific process of the local optimization real-time determination based on the seat management time is as follows: If the total local optimization time length is less than a feedback time limit value, it indicates that the current optimization is effective, and the seat state of each seat area in the next seat management period is detected; If the total local optimization time length is not less than the feedback time limit value, the difference between the total local optimization time length and the total parallel optimization time length is recorded as a to-be-verified optimization time saving length, and the to-be-verified optimization time saving length is compared with the total optimization time saving length: If the to-be-verified optimization time saving length is higher than the total optimization time saving length, it indicates that the local optimization is effective, a seat allocation optimization model computing resource warning prompt is issued, and a pre-set staff is requested to release a seat re-allocation result delay public announcement; If the to-be-verified optimization time saving length is not higher than the total optimization time saving length, it indicates that the local optimization is ineffective, and the local optimization is adjusted.

9. The AI-based reading room seat intelligent allocation and management system according to claim 8, wherein, The specific content of the adjustment of the local optimization is as follows: The to-be-verified optimization time saving length and the total optimization time saving length are quantified to obtain a time saving length difference, and a mapping set for fitting the time saving length difference and a seat state change quantity processing value release amount is constructed to match the seat state change quantity processing value release amount corresponding to the time saving length difference. The seat state change quantity processing value is amplified, and if the to-be-verified saving optimization time length of the next seat management cycle is still not higher than the total optimization saving time length, the amplification of the seat state change quantity processing value is continued until the amplification is higher than the set amplification range limit value, and the loop optimization of the amplification of the seat state change quantity processing value is stopped; If the to-be-verified saving optimization time length is higher than the total optimization saving time length in the time period corresponding to the loop optimization, a seat allocation optimization model computing resource alarm prompt is issued, otherwise the seat allocation optimization model is triggered to recalculate the seat allocation scheme.

10. An AI-based reading room seat intelligent allocation and management method, characterized in that, The specific steps are as follows: S1, after confirming the data update frequency of the seat allocation optimization model used to generate the reading room seat allocation scheme, the seat state change of each seat area in the reading room in the seat management cycle is monitored in real time to determine whether the seat allocation is recalculated by the model; S2, if the seat allocation is recalculated by the model, the recalculation time consumption of the seat allocation optimization model is analyzed, and the real-time influence strategy for ensuring seat intelligent allocation is executed, otherwise the seat allocation local optimization analysis and adjustment are performed to reduce the real-time influence degree of local optimization on seat intelligent allocation; S3, based on the dynamic adjustment of the seat state change in S2, the real-time updated reading room seat intelligent allocation scheme is automatically fed back to improve the response ability of the seat allocation optimization model to the dynamic change of the seat state.

Images