Airport landside traffic safety risk assessment early warning method and system

Abstract

The invention relates to the technical field of intelligent management, in particular to an airport landside traffic safety risk assessment and early warning method and system, and the method comprises the steps: collecting multi-source data including passenger features, public traffic real-time states, satisfaction, supply, flights and weather, and carrying out the preprocessing and standardization; calculating a normal safety guarantee index; performing pressure test and index correction through a preset abnormal scene; comparing the corrected safety index with a static threshold value and a dynamic threshold value to trigger graded early warning; calculating the contribution degree of each index to locate a risk short board; generating a disposal suggestion based on short-board index mapping, and generating an intelligent passenger distribution scheme by using an optimization algorithm in combination with real-time passenger characteristics, vehicle states and road condition data; according to the invention, quantifiable, benchmarking and closed-loop evaluation and management of airport landside traffic safety risks are realized, a refined intelligent scheduling decision can be output, and emergency response safety and efficiency are effectively improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of intelligent management, in particular to an airport land-side traffic safety risk assessment and early warning method and system. BACKGROUND

[0002] Airport land-side traffic is a key interface between the aviation transportation system and the urban integrated transportation system, and its safety and order directly affect the passenger door-to-door travel experience and airport operation resilience. In the night-to-departure center, the latest operation time of public transportation is early, the degree of traffic mode selection is insufficient, or the airport is far from the city center, passengers are likely to face long waiting time, inconvenient transfer, and opaque information; in the case of large-area flight delay, extreme weather or major activities, the above problems will be magnified, and there may be safety risks. In the existing management practice, risk research and judgment often rely on field inspection, complaint statistics or experience judgment; even if a small number of statistical indicators are formed, it is difficult to realize unified quantitative benchmarking across airports / time periods, and it is more difficult to form an evaluation-early warning-disposal-review closed loop. SUMMARY

[0003] In order to make up for the shortcomings of the prior art, the application provides an airport land-side traffic safety risk assessment and early warning method and system, which solves the technical problem that it is difficult to realize unified quantitative benchmarking across airports / time periods, and it is more difficult to form an evaluation-early warning-disposal-review closed loop.

[0004] The airport land-side traffic safety risk assessment and early warning method provided by the application comprises the following steps: S1. Collecting and preprocessing multi-source data corresponding to the evaluation unit to generate an original index matrix; the evaluation unit is an airport or a combination of an airport, a region and a time window; the multi-source data at least includes passenger satisfaction data, public transportation supply data and spatial relationship data; S2. The index is processed in a positive direction, so that all index values represent better safety conditions; S3. The index data after positive direction is processed by dimensionless standardization to obtain a standardization matrix; S4. Based on the CRITIC objective weighting method, the data fluctuation of each index in the standardization matrix and the conflict between indexes are used to calculate the information amount, and then the objective weight of each index is determined; S5. According to the objective weight and the standardization matrix, the normal safety guarantee index of the evaluation unit is calculated by weighted synthesis; S6. A set of abnormal scenarios is constructed, when the evaluation unit meets the triggering condition of a specific scenario, the normal safety guarantee index is pressure tested and corrected according to the preset penalty coefficient and the default level, to obtain the corrected safety index; S7. Comparing the modified safety index with a preset threshold value, or monitoring its trend of change within a preset time window, triggering a corresponding risk warning; S8. Based on the objective weight and the standardized value, calculating the contribution of each index to the total index, identifying the short-board index to locate the risk source; S9. According to the identified short-board index, mapping to the preset measure library to generate treatment suggestions; S10. Outputting visual evaluation, warning, diagnosis and suggestion results, and regularly updating model parameters.

[0005] Preferably, in step S1, the multi-source data further includes one or more of flight arrival data, passenger flow prediction data, road congestion data, weather data, event complaint data, passenger feature data, and public transportation real-time state data; the preprocessing includes data quality verification, spatio-temporal alignment, and missing value filling.

[0006] Preferably, the passenger feature data includes luggage quantity statistics, accompanying children and baby quantity statistics of the passengers to be evacuated; the public transportation real-time state data includes real-time position, passenger capacity, vehicle type and planned driving route information of the operating vehicles.

[0007] Preferably, the dimensionless standardization processing is normalization processing, which maps each index data into a unified numerical interval.

[0008] Preferably, in step S5, the lower the score of the normal safety guarantee index, the higher the corresponding safety risk.

[0009] Preferably, in step S6, the stress test correction is to deduct the normal safety guarantee index according to the penalty coefficient and the default level of all triggering scenarios, and ensure that the modified index is not less than zero.

[0010] Preferably, in step S7, the conditions for triggering a risk warning include that the modified safety index is lower than a preset static risk threshold value, or its decline amplitude exceeds a preset dynamic change threshold value within a preset time window.

[0011] Preferably, in step S8, the specific method for identifying the short-board index is to calculate the contribution of each index to the total index, and sort the contribution from low to high, and the index set with the lowest contribution is the short-board index; step S8 further includes: performing gap diagnosis by calculating the weighted standardized value difference between the evaluation unit and the benchmark unit on each index.

[0012] Preferably, in step S9, the generating the treatment suggestion specifically comprises: when the treatment suggestion relates to transport scheduling, further generating an intelligent passenger allocation scheme based on real-time collected passenger feature data, public transportation real-time state data, weather data and road congestion data through an optimization algorithm model; the optimization algorithm model takes the overall evacuation safety and efficiency optimization as the target, and the constraint conditions include the safety carrying upper limit of a single public transportation unit, the real-time risk assessment value of different travel routes, and the allocation rules of special passenger groups.

[0013] The application also provides an airport land-side traffic safety risk assessment and early warning system, which comprises: a data access module for accessing multi-source data and completing cleaning, alignment and preprocessing; an index module for calculating basic indexes and performing normalization and standardization; a calculation module for calculating objective weights and normal safety guarantee indexes based on a CRITIC objective weighting method; a test and early warning module for managing abnormal scenarios, performing index correction and triggering graded early warning according to rules; a diagnosis module for locating risk sources and generating treatment suggestions; a visualization and interface module for result display and data service provision.

[0014] The application has the following beneficial effects: 1. The application solves the problem that safety risks of different airports are difficult to benchmark by constructing a unified and scalable quantitative index system; the CRITIC objective weighting method is adopted to overcome the deviation of subjective weighting and suppress the repeated weighting caused by the correlation between indexes; the abnormal scenario stress test mechanism is introduced to enable the evaluation model to depict the resilience of the system under sudden events, and the evaluation result is more comprehensive; a complete closed loop from evaluation to early warning to diagnosis and suggestion is realized, and the output result has direct action guidance value; 2. The application introduces micro data of passenger features and traffic unit states, and generates an intelligent passenger allocation scheme by using an optimization algorithm, so that the treatment suggestion is further arranged from the macro level of increasing transport capacity to the executable level of how to optimally arrange each passenger, which significantly improves the fine and intelligent level of the emergency plan and the actual evacuation safety. BRIEF DESCRIPTION OF DRAWINGS

[0015] The application will be further described below in combination with the drawings and embodiments.

[0016] Figure 1 is a flowchart of the method in the application. DETAILED DESCRIPTION

[0017] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application is further described below in combination with specific embodiments.

[0018] As shown in Figure 1 , the present application is an airport land-side traffic safety risk assessment and early warning method, comprising the following steps: S1. Collecting and pre-processing multi-source data corresponding to the evaluation unit to generate an original index matrix; the evaluation unit is an airport or a combination of airport-region-time window; the multi-source data at least includes passenger satisfaction data, public transportation supply data and spatial relationship data; further including one or more of flight arrival data, passenger flow prediction data, road congestion data, weather data and event complaint data; further including passenger feature data and public transportation real-time state data; the passenger feature data includes luggage quantity statistics of passengers to be evacuated, number statistics of accompanying children and infants; the public transportation real-time state data includes real-time position, passenger capacity, vehicle type (carrying capacity, whether equipped with child seat, etc.), planned driving route information of various types of operating vehicles (such as airport buses, taxis, online taxis); Collecting multi-source data corresponding to the evaluation unit, and performing quality check and time / space alignment; marking missing values and filling them up by using mean / interpolation / similar airport replacement, etc.; generating an original index matrix , expressed as: , wherein, represents the th evaluation unit, the evaluation unit is an airport, or a combination of airport-region-time window, represents the th evaluation index, represents the th evaluation unit on the th index, is the number of evaluation units, is the number of indexes.

[0019] S2. Positive processing of the index, so that all index values represent better safety condition; index positive processing: identifying index type, positive conversion of cost type index, so that all indexes meet the condition that the larger the value, the better the safety condition.

[0020] S3. Dimensionless standardization processing is performed on the index data after forward transformation to obtain a standardized matrix; specifically, the index data after completing the forward transformation processing is taken as input, and dimensionless standardization is uniformly performed, so that indexes of different dimensions and different orders of magnitude fall into the same numerical interval, thereby ensuring that there is no dimension dominance phenomenon in subsequent weighting and synthesis. In specific implementation, each index is taken as a processing object, the original value of the index is scanned in the range of all evaluation units, the minimum value and the maximum value of the index in the current sample set are determined respectively, and the value span of the index is obtained accordingly; the value span is used to express the "changeable space of the index in the current sample".

[0021] Subsequently, normalization is completed according to the attribute direction of the index: for a benefit type index, the original value of a certain evaluation unit is first subtracted from the minimum value of the index to obtain a relative increment, and then the value span of the index is used to scale the relative increment in proportion, so that the minimum value in the sample corresponds to 0 and the maximum value corresponds to 1; for a cost type index, the maximum value of the index is subtracted from the original value of a certain evaluation unit to obtain a reverse relative increment, and then the value span of the index is scaled in proportion, so that the highest cost in the sample corresponds to 0 and the lowest cost corresponds to 1. If the maximum value and the minimum value of a certain index in the current sample set are the same, it means that the index does not produce distinguishing information between all evaluation units, so scaling is not performed, but the standardized result of the index on all evaluation units is uniformly set to a preset constant to avoid division by zero and maintain numerical stability; the preset constant is determined by business rules, and 1 or 0.5 is preferably used: taking 1 expresses that "the index is in a full score state and does not produce distinction", and taking 0.5 expresses that "the index is in a neutral state and does not produce distinction"; both satisfy the standardization interval constraint and do not introduce relative differences between evaluation units.

[0022] S4. Based on the CRITIC objective weighting method, the data volatility of each index in the standardized matrix and the conflict between indexes are used to calculate the information amount, and then the objective weight of each index is determined; Specifically, the CRITIC objective weighting method is used to calculate the information amount of each index based on data volatility and conflict between indexes, and the information amount is normalized to form an objective weight. The more obvious the fluctuation of a certain index between evaluation units, the more distinguishing ability it has; the weaker the correlation between a certain index and other indexes, the more complementary information it can provide, and the lower the redundancy, so it should obtain a higher weight.

[0023] The implementation process is as follows: First, calculate the standardized mean of each indicator. The mean describes the overall level of the indicator in the sample set. Then, calculate the standard deviation of the indicator. The standard deviation is obtained by summing the squared deviations of the standardized values ​​of each evaluation unit from the mean, averaging them by the number of evaluation units, and finally taking the square root. This is used to characterize the volatility of the indicator. Second, construct the correlation results between indicators: For any two indicators, center the standardized sequences of the two indicators respectively. That is, subtract the mean of the indicator from the standardized value of each evaluation unit to obtain two deviation sequences. Multiply the two deviation sequences by each evaluation unit and sum them to obtain the co-variance. Then, take the square root of the sum of the squared deviations of the two indicators to obtain the discrete scale of the two indicators. Divide the co-variance by the product of the discrete scales of the two indicators to obtain the correlation coefficient between the two indicators. The correlation coefficient ranges from -1 to 1. The closer the value is to 1, the stronger the redundancy in the same direction. The closer the value is to -1, the stronger the inverse correlation. Next, calculate the conflict summation value for each indicator: calculate "1 minus the correlation coefficient" for each indicator and all other indicators, and sum the results. The larger the sum, the stronger the complementarity between the indicator and other indicators. Then, multiply the standard deviation of the indicator by the conflict summation value to obtain the information content of the indicator. The information content reflects both the discriminative ability and the complementary contribution. Finally, sum the information contents of all indicators as the normalization base, and divide the information content of each indicator by the normalization base to obtain the objective weight of the indicator, ensuring that the sum of all weights is 1 and that each weight is not negative. If the sum of the information contents of all indicators is 0, it indicates that the indicators as a whole do not fluctuate under the sample set or that the correlation structure causes objective discrimination to fail. In this case, set all indicator weights to the same value, so that the sum of the weights is 1, thereby ensuring that the subsequent index calculation has determinism and interpretability.

[0024] S5. Based on the objective weights and the standardization matrix, the normal security assurance index of the evaluation unit is obtained through weighted comprehensive calculation. In specific implementation, for a certain evaluation unit, the standardized value of each indicator is multiplied by the corresponding indicator weight to obtain the weighted contribution, and then the weighted contributions of all indicators are summed. The sum is used as the normal security assurance index of the evaluation unit. Since the standardized value is limited to 0 to 1 and the sum of the weights is 1, the normal security assurance index naturally falls within the range of 0 to 1.

[0025] To meet the requirements of both presentation and management, the normal security assurance index is linearly mapped to a range of 0 to 100 as needed. The mapping process uses a scaling factor of "index value multiplied by 100," and the mapping does not change the relative order between assessment units. The index is uniformly defined as a lower value indicating a higher security risk, thus maintaining consistency with the deduction and correction logic in subsequent stress tests.

[0026] S6. Construct an abnormal scenario set. When the evaluation unit meets the triggering conditions of a specific scenario, the normal security assurance index is stress-tested and corrected according to the preset penalty coefficient and the degree of breach, so as to obtain the corrected security index. Abnormal scenario stress test and index correction: a set containing multiple abnormal scenarios is preset, and triggering conditions and penalty coefficients are configured for each scenario.

[0027] Specifically, the construction of the abnormal scenario set follows the principles of scenario enumerability, condition determinability, and consequence quantification: First, enumerate the abnormal situations affecting safety assurance, defining each situation as a scenario entry; configure trigger conditions, penalty coefficients, and rules for calculating the degree of breach for each scenario entry. Trigger conditions are usually manifested as threshold judgments or combined judgments. Threshold determination follows three sources with set priorities: The first source is regulations, standards, and regulatory requirements, directly adopting explicit limits or trigger conditions; the second source is corporate safety commitments and operational control red lines, forming thresholds based on management objectives, available resources, and the upper limit of emergency response capabilities; the third source is historical distribution and similar benchmarks, using quantiles of historical samples or benchmark levels to form thresholds and conducting robustness checks, releasing them as operational thresholds after passing the checks. The penalty coefficient is used to express the severity of the scenario and the intensity of its impact on the index. Its determination follows a risk grading system: first, scenarios are graded based on potential for personal injury, impact of operational disruption, public attention, controllability, and recovery time; then, different penalty intensities are assigned to different levels, and the penalty intensities are calibrated using historical event replays to ensure consistency within the same level and a monotonically increasing trend across levels. The default severity is used to quantify the degree of deviation after triggering, with a value limited to 0 to 1. Its determination uses a segmented mapping: the default severity is 0 when the real-time indicator just triggers the threshold; it is 1 when the real-time indicator reaches the severity upper limit defined by the scenario; values ​​between these two are calculated linearly; if the real-time indicator exceeds the severity upper limit, the default severity remains at 1, thus avoiding unlimited amplification of deductions.

[0028] Trigger determination and correction calculations operate on a time-series basis: At each calculation moment, real-time data from the evaluation unit is collected, and trigger determination is performed scenario by scenario. Scenarios that meet the trigger conditions are added to the trigger set for that moment. For each scenario within the trigger set, the degree of default is calculated according to the default severity rules for that scenario, and then the penalty coefficient is multiplied by the default severity to obtain the deduction amount for that scenario. The deduction amounts of all triggered scenarios are summed to obtain the total deduction amount. The corrected security index is obtained by subtracting the total deduction amount from the normal security assurance index. To ensure reasonable index boundaries, if the value after deduction is lower than 0, the corrected security index is set to 0 to avoid negative values ​​and maintain consistent index interpretation. Through this mechanism, the normal index undertakes the characterization of long-term structural levels, while the abnormal correction undertakes the reflection of short-term risk shocks. The two are connected on the same dimension, maintaining comparability while improving the responsiveness to changes in the security situation.

[0029] S7. Compare the corrected safety index with a preset threshold, or monitor its changing trend within a preset time window, and trigger a corresponding risk warning; Risk grading and warning triggering: Set a static risk threshold and a dynamic change threshold; Compare the corrected safety index with the static risk threshold, and if it is lower than the threshold, trigger a risk warning of the corresponding level; or within a preset time window, calculate the decrease of the corrected safety index, and if it exceeds the dynamic change threshold, trigger a trend deterioration warning; Output warning level, triggering reason and impact range information; Execute grading and warning: The static risk threshold is used to characterize the grading boundary of the safety index, and is usually configured in multiple levels to form a threshold sequence from high to low. The threshold sequence corresponds one-to-one with the warning level. The higher the threshold, the more sensitive the situation, and the lower the level, the more serious the situation. First, the regulatory red lines, organizational safety commitments, and operational control objectives are verified, and boundary values ​​with clear binding force are directly included in the threshold candidate range. Then, historical sample periods are selected under a unified standard, covering peak and off-peak seasons, typical weather conditions, and typical support intensities. After eliminating known data defects, the distribution of the corrected safety index in the historical samples is statistically analyzed, and several quantiles are selected as threshold candidate points. The selection of quantiles is determined based on the warning capacity and the upper limit of response resources; the more abundant the resources, the higher the value is selected to increase sensitivity, and the more strained the resources, the lower the value is selected to reduce false alarms. Finally, similar airports or assessment units of the same operational scale are introduced as references to compare the landing points of the threshold candidate points in the same set. The thresholds are required to have consistent risk meanings within the same set, avoiding the same value corresponding to completely different risk semantics on different assessment objects. A corresponding update mechanism is implemented after the thresholds are released. Update trigger conditions include changes in the indicator system, changes in weight versions, significant changes in sample structure, and adjustments to management objectives. During updates, backtesting is performed first, using warning hit rate, false alarm rate, and response closure timeliness as constraints. Threshold replacement is completed after the backtest passes.

[0030] The dynamic threshold is used to characterize the upper limit of the deterioration rate within a preset time window. Its core purpose is to identify situations where the decline is rapid but has not yet broken the static boundary. The determination of the dynamic threshold follows these principles: First, calculate the index decline sequence in historical samples with a fixed window length. Only the positive decline portion of the decline sequence is retained to avoid misjudging rises or flatness as deterioration. Next, the high quantile position of the decline sequence is used as a candidate threshold. The later the high quantile position, the more likely it is to trigger an alert only for extreme downtrends, thereby reducing frequent alarms. Then, the candidate threshold is aligned with past risk events, operational interruptions, and major default records. The candidate threshold is required to have a significant trigger difference before and after the risk event, avoiding excessively high thresholds leading to a large number of missed alarms or excessively low thresholds leading to a large number of false alarms. Finally, the candidate threshold is tested for stability. The test dimensions include at least the trigger frequency distribution under different seasons, different intraday times, and different operational intensities. If the stability is insufficient, the threshold is adjusted for convergence until the trigger frequency meets management expectations and is interpretable.

[0031] The time window length is used to define the observation span of trend monitoring, and its determination is constrained by the data sampling frequency, business response time, and disturbance propagation speed. There is a matching relationship between the time window length and the sampling frequency; the time window must contain a sufficient number of effective sampling points to support reliable calculation of the decline amplitude, while simultaneously being shorter than the typical closed-loop processing time, thus ensuring that triggering the early warning has practical intervention value. Specifically, the sampling frequency is first determined, and the minimum number of effective samples is estimated to ensure that at least a continuous effective observation sequence is formed within the time window. Then, the shortest time from early warning triggering to the effective implementation of the action is evaluated, and the time window length is set to be no shorter than a multiple of this shortest time to avoid early warning lag. Finally, scenario type calibration is used: a shorter window is used for sudden scenarios to capture rapid downtrends, while a longer window is used for systemic degradation scenarios to suppress short-term noise.

[0032] The early warning calculation process is executed in the following order: data acquisition, cleaning, comparison, classification, interpretation, and output. In the data acquisition phase, the corrected safety index for the current moment is read from the time series, along with the corrected safety index at the start of the time window. If no precise sampling point exists at the start of the time window, the process backtracks to the nearest valid sampling point as the window start value, and records the backtracking span for result interpretation. In the cleaning phase, boundary pruning is performed on the read index values ​​to ensure they are neither lower than 0 nor higher than 1. A missing value strategy is applied, prioritizing not triggering trend judgment in this case and outputting the reason for data missing to avoid false alarms when there is a lack of evidence. In the static threshold triggering phase, the current corrected safety index is compared level by level with the static risk threshold sequence, from severe to mild. First, it is checked whether the index is lower than the most severe threshold. If it is, the index is directly determined to be at that severity level, and the static comparison ends. If it is not, the next higher threshold is checked until a match is achieved, ultimately obtaining the static trigger result and the corresponding level. During the dynamic trend triggering phase, the decline rate within the time window is calculated. This decline rate is obtained by subtracting the current index from the starting index of the window. When this difference is negative, the decline rate is set to 0, ensuring that trend judgment focuses solely on the decline. The decline rate is compared with a dynamic change threshold. If the decline rate exceeds the threshold, it is determined as a trend deterioration trigger, and a trend trigger level is assigned. The trend trigger level is mapped from the proportion of the decline rate exceeding the threshold; the larger the proportion, the more severe the level. To suppress alarm jitter caused by short-term fluctuations, a continuity constraint is introduced, requiring that static triggering or trend triggering occur within two consecutive sampling periods before entering a formal warning state. The number of periods for the continuity constraint is determined based on the sampling frequency; higher sampling frequencies result in a larger number of periods to suppress noise, while lower sampling frequencies result in a smaller number of periods to reduce trigger latency.

[0033] The alert level is synthesized using the highest possible level: when both static and trend triggers are triggered, the higher level is output, and both static breakdown and trend deterioration are noted in the trigger cause; when only static triggers are triggered, the static trigger level is output, and the current index is noted as being below the corresponding boundary; when only trend triggers are triggered, the trend trigger level is output, and the decline within the window exceeds the dynamic boundary; when neither is triggered, no alert is output, and the current index, the starting index of the window, and the decline magnitude are recorded for record keeping. The generation of trigger causes follows the traceability principle. The cause content includes at least the trigger type, corresponding threshold level, current index value, comparison conclusion, window length, window starting value, and whether a backtracking point occurred. The cause is associated with the trigger scenario in the previous stress test. If there is an abnormal scenario deduction at the current moment, the scenario item with the largest deduction contribution and its default degree range are listed in the cause, thus establishing a clear link between index deterioration and its source.

[0034] The impact scope information is used to answer the business boundaries of the warning's impact. Its generation is based on the structural decomposition of the assessment unit: First, the spatial and organizational scope corresponding to the warning subject is determined. The spatial scope is divided according to the terminal, mode of transportation, entrance / exit area, or operational area, while the organizational scope is divided according to the responsible unit and the handling chain. Then, the indicators that have the largest decrease in contribution to the index at the current moment are selected as impact dimensions. The identification of indicators is based on the degree of deviation of the current standardized value from the historical mean multiplied by the weight ranking result. Indicators with greater deviation and higher weight are prioritized into the impact dimension list. Subsequently, the impact dimensions are mapped to the data source to clarify the affected data links and collection locations, forming a locatable handling entry point. The final output includes three parts: warning level, triggering reason, and impact scope. The output format maintains structured fields to facilitate linkage display and consumption by the handling system. At the same time, the threshold version, dynamic threshold version, time window length version, and continuity constraint parameter version used in this calculation are retained to ensure that the warning results at the same time are recalculated and auditable.

[0035] S8. Based on the objective weights and standardized values, calculate the contribution of each indicator to the overall index, identify the weakest indicators to pinpoint the sources of risk; Risk source identification and diagnosis: Calculate the contribution of each indicator to the overall safety index of the assessment unit according to its objective weights and standardized values, sort them from low to high contribution, and identify the set of weakest indicators; Optionally, perform gap diagnosis by comparing the weighted standardized values ​​of the assessment unit and the benchmark unit on each indicator; Perform risk source identification: Calculate the contribution of each indicator, define the first... Each indicator is relevant to the evaluation unit. The index contribution is: Then there is ,Will The set of indicators with the lowest contribution, ranked from lowest to highest, constitutes the bottleneck indicator set. Optionally, let the benchmark unit be b, then the evaluation unit... In terms of indicators The weighted difference is defined as: ,when The larger the value, the greater the potential for improvement relative to the benchmark, and the higher the priority.

[0036] S9. Based on the identified bottleneck indicators, map them to a preset measures library to generate disposal suggestions: Map and match the identified bottleneck indicators with the preset countermeasures knowledge base to automatically generate a list of disposal suggestions including categories such as capacity scheduling, facility optimization, information guidance, and law enforcement control; The capacity scheduling suggestions are further based on passenger characteristic data, real-time public transportation status data, current weather data, and road congestion data collected in step S1, and generate an intelligent passenger allocation scheme through an optimization algorithm model; The optimization algorithm model aims to optimize the overall evacuation safety and efficiency, and the constraints include: the upper limit of the safe carrying capacity of a single public transportation unit (comprehensively considering the number of seats, standing space, luggage storage space, and the special needs of accompanying children / infants), the risk assessment value of different routes under the current weather and road conditions, and the priority or combination allocation rules for special passenger groups (such as those carrying large luggage or infants), thereby outputting specific passenger grouping and vehicle matching schemes, rather than simple suggestions for increasing or decreasing capacity; and the suggestions can be prioritized according to budget constraints or response time requirements.

[0037] S10. Output visualized assessment, early warning, diagnosis and suggestion results, and update model parameters regularly (including intelligent passenger allocation schemes) to the visualization platform or business system interface; the system recalculates indicator weights and optimizes early warning thresholds using accumulated new data according to preset cycles to achieve adaptive updates of the assessment model.

[0038] As one embodiment of the present invention, in step S1, the multi-source data further includes one or more of flight arrival data, passenger flow prediction data, road congestion data, meteorological data, incident complaint data, passenger characteristic data, and real-time public transportation status data; the preprocessing includes data quality verification, spatiotemporal alignment, and missing value imputation.

[0039] As one embodiment of the present invention, the passenger characteristic data includes statistics on the number of luggage belonging to passengers to be evacuated and statistics on the number of accompanying children and infants; the real-time status data of public transportation includes the real-time location, passenger capacity, vehicle type, and planned route information of operating vehicles.

[0040] As one embodiment of the present invention, the dimensionless standardization processing is a normalization processing, which maps the data of each indicator to a unified numerical range.

[0041] In one embodiment of the present invention, in step S5, the lower the score of the normal security assurance index, the higher the corresponding security risk.

[0042] In one embodiment of the present invention, in step S6, the stress test correction is to deduct the normal security assurance index based on the penalty coefficient and the degree of default of all triggered scenarios, and to ensure that the corrected index is not lower than zero.

[0043] As one embodiment of the present invention, in step S7, the conditions for triggering a risk warning include: the corrected safety index is lower than a preset static risk threshold, or its decrease exceeds a preset dynamic change threshold within a preset time window.

[0044] As one embodiment of the present invention, in step S8, the specific method for identifying the bottleneck indicators is as follows: calculate the contribution of each indicator to the total index, and sort the indicators from low to high. The set of indicators with the lowest contribution is the bottleneck indicator. Step S8 also includes: performing gap diagnosis by calculating the difference between the weighted standardized values ​​of the evaluation unit and the benchmark unit on each indicator.

[0045] As one embodiment of the present invention, in step S9, generating disposal suggestions specifically includes: when the disposal suggestions involve capacity scheduling, further generating an intelligent passenger allocation scheme based on real-time collected passenger characteristic data, real-time public transportation status data, meteorological data, and road congestion data through an optimization algorithm model; the optimization algorithm model aims to optimize the overall evacuation safety and efficiency, and the constraints include the safe carrying capacity limit of a single public transportation unit, the real-time risk assessment value of different travel routes, and the allocation rules for special passenger groups.

[0046] An airport landside traffic safety risk assessment and early warning system, the system comprising: The data access module is used to access multi-source data and perform cleaning, noise reduction, missing value verification, and time alignment. The metrics module is used to calculate basic metrics and perform positive conversion and standardization. The calculation module is used to calculate objective weights and normal security assurance index based on the CRITIC objective weighting method; The test early warning module is used to manage the abnormal scenario library, perform stress tests and index corrections based on trigger conditions, and trigger graded early warnings according to preset threshold rules; The diagnostic module is used to locate the source of risk and generate treatment suggestions; it is used to perform contribution analysis and gap breakdown to locate the source of risk, and integrates an optimization algorithm engine to generate targeted treatment suggestions and intelligent passenger allocation schemes based on real-time multi-source data and calling the measures knowledge base. The visualization and interface module is used to display the results of the entire evaluation and early warning chain, especially to display intelligent allocation schemes in an actionable view (such as passenger queue and vehicle matching dashboard), and to provide data service interfaces.

[0047] The specific implementation process of this invention is as follows: Step 1: Identify the assessment target and prepare the data; The evaluation unit is a large domestic hub airport (referred to as Airport A) during a weekday evening from 22:00 to 23:00. Multi-source data is collected for this unit. Satisfaction data: Obtained from the airport service evaluation system, the overall airport transportation score is 4.2 out of 5, and the departure transportation score is 3.8 out of 5. Public transportation supply data: During this period, three airport bus routes are in operation, with the latest departure time being 00:30 the following day; urban rail transit is suspended. Spatial relationship data: The airport is 35 kilometers away from the city center; Extended data: Moderate rain is forecast for this period, with 25% of flights delayed and an estimated 3,000 arriving passengers. Real-time micro-level data: Through the airport departure system and the passenger questionnaire fast track, statistics show that among the 800 passengers currently waiting for evacuation, approximately 150 are carrying large luggage (over 20 kg), 20 are families with infants, and 50 are families with children (2-12 years old). Real-time data is also available on the models (3 standard 45-seat buses and 2 30-seat buses with large luggage compartments), locations, and planned routes of 5 standby airport buses. Real-time traffic conditions for each bus route are obtained from the transportation department's API (one route reported flooding). The above data is cleaned, formatted, and missing values ​​are filled in (e.g., using historical averages of similar airports) to form the original data vector; Step Two: Indicator Processing and Standardization; The indicator system is constructed, including: (1) overall airport traffic score (efficiency type); (2) departure airport traffic score (efficiency type); (3) latest operating time of public transportation (efficiency type, converted to minutes); (4) availability of public transportation modes (efficiency type, based on the number of operating lines); (5) distance of airport from city center (cost type). The cost-related indicator of distance from the city center is positively oriented (e.g., by subtracting the value from the maximum distance); then, data from multiple benchmark airports during the same period are collected to form a sample set, and each indicator is normalized and standardized so that all indicator values ​​fall within a uniform numerical range. Step 3: Objective weighting and normal index calculation; Based on the standardized data matrix, the CRITIC method is used to calculate the weights. This method comprehensively considers the volatility of each indicator data and the conflict between indicators to calculate the amount of information contained in each indicator. The greater the amount of information, the higher the weight. Assume the weight vector obtained by the calculation. Then, the standardized values ​​of each indicator of airport A and their corresponding weights are weighted and combined to obtain the normal safety assurance index for this period. Step 4: Correction for abnormal stress testing; Preset severe weather scenarios, triggered by issuing yellow or higher warnings for rain / snow / fog, and set corresponding penalty coefficients. The degree of breach can be assigned according to the rainfall intensity, visibility, etc.; preset large-scale delay scenarios, triggered by the proportion of delayed flights being greater than 20%, and set corresponding penalty coefficients. The degree of breach is assigned according to the proportion of delays. The system determines that Airport A has triggered the two abnormal scenarios mentioned above at the same time period. According to the preset penalty rules, the index deduction caused by the two scenarios is calculated separately and deducted from the normal safety guarantee index to obtain the corrected safety index. The corrected index reflects the actual risk level under the combined effects of severe weather and flight delays. Step 5: Early Warning Triggering and Risk Identification; The system has a pre-set high warning threshold. The calculated, corrected safety index is compared to this threshold. If the index is below the threshold, the system triggers a high-level warning, clearly stating in the warning message that the triggering cause is the combined effect of severe weather and flight delays. The system further identifies the sources of risk and calculates the contribution of each indicator to the current overall safety index. Ranking reveals specific indicators with the lowest contribution, which are identified as weak points. Comparative analysis with benchmark airports further clarifies the specific areas where gaps exist. Step Six: Generate handling recommendations; Based on identified shortcomings (such as insufficient nighttime public transportation capacity), the system automatically matches corresponding improvement suggestions from a pre-set knowledge base. Furthermore, the system activates the intelligent passenger allocation submodule; this module, aiming for safe and rapid evacuation and improved passenger experience, takes into account the real-time micro-data collected in step one, and considers the risks of moderate rain and flooded roads. After optimizing the algorithm, the following suggestions are output: 1) Prioritize dispatching two buses with large luggage compartments to carry 150 passengers with large luggage and some other passengers traveling the same route, avoiding flooded routes; 2) Arrange specific vehicles equipped with infant seats (or reserve front-row space) to centrally serve 20 families with infants; 3) ... The remaining passengers and families with children are reasonably allocated to other vehicles to ensure that each vehicle is not overloaded and has a safe space. The plan also provides the estimated evacuation time and safety risk score for each vehicle. 4) Flexible scheduling: Based on the specific conditions of the road the bus travels on, the best bus load allocation plan is analyzed to ensure that the bus load is optimally allocated in real time according to the weather, route and road conditions. If it rains heavily and the road is slippery, passengers with a lot of luggage can be allocated to ensure that the bus is allocated with the optimal number of passengers and luggage weight. This ensures that the vehicle is within the safe weight range and can reduce the number of passengers, ensuring the best plan is achieved in the case of heavy rain and slippery roads. Step 7: Output and Update; All assessment results, early warning signals, diagnostic conclusions, intelligent passenger allocation schemes, and handling suggestions are pushed to the airport operations management center's visual decision-making screen; dispatchers can intuitively see passenger grouping and vehicle matching suggestions and issue dispatch instructions with one click; the system automatically uses recently accumulated data every quarter to re-execute the weight calculation and threshold optimization process, and complete the iterative update of model parameters to maintain the timeliness and accuracy of the assessment.

[0048] Furthermore, the system to which this invention pertains consists of six core modules connected via an internal network or bus; The data access module is equipped with multiple data interface adapters for periodically or in real-time data retrieval from sources such as the airport operation database, transportation department API, meteorological service API, and public opinion system. It also adds interfaces with the departure system, passenger information terminal, and fleet management system to obtain passenger characteristics and real-time vehicle status data. Furthermore, it includes a built-in rule-based and statistical anomaly detection and cleaning program. The indicator module has a built-in indicator calculation formula library and standardization processor, capable of automatically calculating indicator values ​​and standardizing them according to the user-configured evaluation granularity (e.g., by hour, by region). The calculation module integrates the CRITIC algorithm engine, which can run automatically periodically and output the latest weights and normal indexes. The test and early warning module includes an editable scenario library management interface, allowing users to define new abnormal scenarios and their parameters. This module monitors the data stream in real-time, determines scenario triggers, calculates correction indices, and... The system matches the data with a pre-defined early warning rule base to generate early warning instructions. The diagnostic module incorporates contribution analysis and gap analysis algorithms, and integrates a multi-objective optimization algorithm engine (such as a genetic algorithm and a linear programming solver). This module is associated with a structured measures knowledge base, which stores suggested measures and their metadata for different bottleneck indicators and risk levels. When generating capacity scheduling suggestions, this module calls the optimization algorithm engine and combines real-time micro-data to generate an executable intelligent passenger allocation scheme. The visualization and interface module provides a web-based visualization dashboard displaying multi-dimensional risk maps, early warning lists, bottleneck analysis diagrams, etc. It also features a smart scheduling sub-view that displays passenger queues, vehicle resources, and matching relationships in chart form. Furthermore, it provides a standardized API interface for other airport management systems (such as traffic scheduling systems and broadcasting systems) to access risk assessment results and allocation schemes.

[0049] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for assessing and warning of traffic safety risks at airport landside, characterized in that, Includes the following steps: S1. Collect and preprocess multi-source data corresponding to the evaluation unit to generate an original indicator matrix; the evaluation unit is an airport or a combination of airport-region-time window; the multi-source data includes at least passenger satisfaction data, public transportation supply data and spatial relationship data. S2. Positively process the indicators so that the larger the value of all indicators, the better the safety status; S3. Perform dimensionless standardization on the positiveized index data to obtain the standardized matrix; S4. Based on the CRITIC objective weighting method, the information content is calculated by utilizing the data volatility and inter-indicator conflict of each indicator in the standardized matrix, and then the objective weight of each indicator is determined. S5. Based on the objective weights and standardized matrix, the normal security assurance index of the evaluation unit is obtained through weighted comprehensive calculation; S6. Construct a set of abnormal scenarios. When the evaluation unit meets the triggering conditions of a specific scenario, stress test the normal security assurance index according to the preset penalty coefficient and the degree of breach, and obtain the corrected security index. S7. Compare the corrected safety index with a preset threshold, or monitor its changing trend within a preset time window, and trigger a corresponding risk warning; S8. Based on the objective weights and standardized values, calculate the contribution of each indicator to the total index, and identify the weakest indicators to pinpoint the sources of risk; S9. Based on the identified shortcomings, map them to the pre-set measures library to generate disposal suggestions; S10. Output visualized assessment, early warning, diagnosis and suggestion results, and update model parameters regularly.

2. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S1, the multi-source data also includes one or more of the following: flight arrival data, passenger flow forecast data, road congestion data, meteorological data, incident complaint data, passenger characteristic data, and real-time public transportation status data; the preprocessing includes data quality verification, spatiotemporal alignment, and missing value imputation.

3. The airport landside traffic safety risk assessment and early warning method according to claim 2, characterized in that, The passenger characteristic data includes statistics on the number of luggage belonging to passengers awaiting evacuation, and statistics on the number of accompanying children and infants; the real-time status data of public transportation includes the real-time location, passenger capacity, vehicle type, and planned route information of operating vehicles.

4. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, The dimensionless standardization process is a normalization process that maps the data of each indicator to a unified numerical range.

5. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S5, the lower the score of the normal security assurance index, the higher the corresponding security risk.

6. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S6, the stress test correction is to deduct the normal security assurance index based on the penalty coefficient and default degree of all triggered scenarios, and ensure that the corrected index is not lower than zero.

7. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S7, the conditions for triggering a risk warning include: the corrected safety index is lower than a preset static risk threshold, or its decrease exceeds a preset dynamic change threshold within a preset time window.

8. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S8, the specific method for identifying the weak link indicators is as follows: calculate the contribution of each indicator to the total index, sort the indicators from low to high, and the set of indicators with the lowest contribution is the weak link indicator; step S8 also includes: perform gap diagnosis by calculating the difference between the weighted standardized values ​​of the evaluation unit and the benchmark unit on each indicator.

9. The airport landside traffic safety risk assessment and early warning method according to claim 1, characterized in that, In step S9, a disposal suggestion is generated, which specifically includes: when the disposal suggestion involves capacity scheduling, a smart passenger allocation scheme is generated based on real-time collected passenger characteristic data, real-time public transportation status data, meteorological data and road congestion data through an optimization algorithm model; the optimization algorithm model aims to optimize the overall evacuation safety and efficiency, and the constraints include the safe carrying capacity limit of a single public transportation unit, the real-time risk assessment value of different routes, and the allocation rules for special passenger groups.

10. An airport landside traffic safety risk assessment and early warning system, characterized in that, An airport landside traffic safety risk assessment and early warning method applicable to any one of claims 1-9, the system comprising: The data access module is used to access multi-source data and complete cleaning, alignment and preprocessing; The metrics module is used to calculate basic metrics and perform positive conversion and standardization. The calculation module is used to calculate objective weights and normal security assurance index based on the CRITIC objective weighting method; The test warning module is used to manage abnormal scenarios, perform index corrections, and trigger tiered warnings according to rules; The diagnostic module is used to locate the source of risk and generate response recommendations; The visualization and interface module is used for result display and data service provision.

Images