A pilot driving risk prediction method and device based on big data, and a medium

Abstract

The present application relates to the technical field of data processing, and discloses a pilot driving risk prediction method and device based on big data, and a medium.The present application forms a four-dimensional correlation database through pilot archive data, time coordinate data, health information parameter data and flight task data; an exposure offset coefficient is calculated through the four-dimensional correlation database; different load data of the pilot performing a task within a target time is corrected using the exposure offset coefficient; the actual age data of the pilot is calculated through the corrected flight exposure factor to obtain physiological age data; the physiological age data is used as the physical age of the pilot for calculation, which can make the obtained physiological disease data more consistent with the current situation of the pilot, and thus make the calculated comprehensive driving ability impairment rate of the pilot more accurate, thereby improving the accuracy of the pilot driving risk prediction result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, device, and medium for predicting pilot flight risks based on big data. Background Technology

[0002] As a special professional group, the health status of civil aviation pilots is not only related to their personal interests but also closely linked to air transport and the safety of people's lives and property. A systematic analysis of the health status and common disease spectrum of pilots is of great significance for clarifying the key directions of current civil aviation health protection work and taking targeted preventive measures. According to civil aviation regulations, pilots must pass an annual medical examination to obtain flight qualifications. The existing health monitoring system mainly relies on aviation medical examination institutions to conduct regular physical examinations of pilots, comparing the results with medical standards for aviation personnel to determine whether pilots meet the flight qualification requirements.

[0003] However, manually comparing examination results with aviation personnel medical standards is not only time-consuming and labor-intensive, but also, manually comparing a pilot's medical examination data can only assess the results of that particular examination and cannot predict the pilot's physical condition for future flight operations. If a pilot develops health problems, they will have no choice but to be grounded, resulting in significant losses. Furthermore, existing algorithmic models predicting pilot flight risks are based on actual age, leading to inaccurate predictions.

[0004] Therefore, a new method is needed to predict pilots' flight risks, thereby improving the accuracy of the prediction results. Summary of the Invention

[0005] This invention provides a method, device, and medium for predicting pilot flight risks based on big data, which can improve the accuracy of pilot flight risk prediction results.

[0006] To address the aforementioned technical problems, the first aspect of this invention discloses a pilot risk prediction method based on big data, the method comprising: A four-dimensional relational database is obtained; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data; the time coordinate data includes the time data corresponding to the physical examination date and the time data corresponding to the flight mission date; the health information parameter data includes physical examination data, disease development pattern parameters, and aviation medical grounding information; Based on the aforementioned four-dimensional relational database, an exposure offset coefficient is obtained; the exposure offset coefficient is used to correct the flight exposure factor; the flight exposure factor is used to describe different load data of pilots performing tasks within a target time. Based on the exposure offset coefficient, the actual age data of the assessed pilot is calculated to obtain physiological age data; based on the physiological age data, physiological disease data in the four-dimensional association database is obtained; based on the physiological disease data and the flight mission data, the comprehensive pilot's impairment rate is obtained. The overall driving capability impairment rate is assessed to determine the risk prediction result.

[0007] As an optional implementation, in the first aspect of the present invention, the disease development pattern parameter includes a list of rising diseases, which is obtained by the following means: Acquire pilot medical examination data; divide the pilot medical examination data according to age as the vertical axis and years of flight experience as the horizontal axis to obtain cross-layered structure data; the cross-layered structure data includes several sub-cross-layered structure data; Perform the following operations on each of the sub-cross-layer structure data: For each target disease, the prevalence rate of the target disease is obtained; the relative change rate of the prevalence rate of the target disease among two consecutive years of the target pilots is calculated to obtain the annual change rate of the target disease; the target pilots are pilots born in the same year selected based on the pilot file data. The prevalence rate and annual change rate of the target disease corresponding to each of the sub-cross-layer structure data are filtered according to preset screening conditions to obtain a list of rising diseases; Furthermore, the disease development pattern parameters also include a comorbidity association strength matrix, which is obtained through the following method: Perform the following operations on each of the sub-cross-layer structure data: Based on the number of people who simultaneously suffer from the target disease and other diseases, the conditional probability between the target disease and other diseases is obtained; based on the conditional probability, the lift of the target disease is obtained; the lift of the target disease is used to represent the degree of association between the target disease and other diseases; based on all the lifts of the target disease, the sub-cross layer association strength matrix is ​​obtained. Based on the correlation strength matrices of all the sub-cross layers, the co-morbidity correlation strength matrix is ​​obtained.

[0008] As an optional implementation, in the first aspect of the present invention, obtaining the exposure offset coefficient based on the four-dimensional relational database includes: When it is determined that the flight exposure factor value is less than the preset low exposure factor threshold, a low-exposure pilot group is obtained based on the four-dimensional association database. The prevalence of low-exposure diseases among adjacent age groups in the low-exposure pilot group is calculated to obtain the natural prevalence rate growth rate; a benchmark conversion factor is obtained based on the natural prevalence rate growth rate. Based on the health information parameter data and the flight exposure factor, a multiple linear regression model was constructed and the regression coefficient of the exposure factor was obtained by calculation using the least squares method; Divide the regression coefficient of the exposure factor by the benchmark conversion coefficient to obtain the target disease offset coefficient; The exposure offset coefficient is obtained based on all the target disease offset coefficients.

[0009] As an optional implementation, in the first aspect of the invention, obtaining the overall pilot capability impairment rate based on the physiological disease data and the flight mission data includes: Based on the flight mission data, a mission intensity factor is determined; based on the mission intensity factor, a mission intensity index is obtained. The task intensity index is matched with a preset acceleration factor mapping table to obtain the disease acceleration factor; the actual disease progression rate is obtained based on the physiological disease data and the disease acceleration factor. When it is determined that the disease has not been diagnosed, the first rate of impairment of driving ability is obtained based on the actual rate of disease progression. When a disease is diagnosed, a second pilot capability impairment rate is obtained based on the aforementioned aviation medical grounding information; Based on the aforementioned comorbidity association strength matrix, the comorbidity synergistic risk bonus term is obtained; The comprehensive driving ability impairment rate is obtained based on the first driving ability impairment rate, the second driving ability impairment rate, and the comorbidity risk bonus item.

[0010] As an optional implementation, in the first aspect of the present invention, the method further includes: Based on the four-dimensional relational database, a historical sample dataset is obtained; the overall pilot capability impairment rate of all pilots in the historical sample dataset is calculated to obtain an overall pilot capability impairment rate dataset. Based on the aforementioned aviation medical grounding information, a grounding tag dataset was determined; The driving ability impairment rate dataset is sorted to obtain an ordered probability sequence; candidate judgment thresholds are determined based on the ordered probability sequence. Based on the candidate decision threshold and the grounding label dataset, a first grounding rate and a second grounding rate are obtained; the first grounding rate is used to describe the ratio of correctly predicted pilots among all grounded pilots; the second grounding rate is used to describe the ratio of incorrectly predicted pilots among all non-grounded pilots. Using the first grounding rate as the vertical axis and the second grounding rate as the horizontal axis, a prediction curve is obtained. Based on the prediction curve, a first determination threshold and a second determination threshold are obtained; wherein, the first determination threshold is greater than the second determination threshold.

[0011] As an optional implementation, in the first aspect of the present invention, the step of judging the overall driving capability impairment rate and determining the risk prediction result includes: Based on the current flight mission data of the assessed pilot, a threshold correction coefficient is determined; the threshold correction coefficient is used to correct the first judgment threshold and the second judgment threshold to obtain a first dynamic threshold and a second dynamic threshold. When it is determined that the overall driving ability impairment rate is greater than the first dynamic threshold, the determination result is marked as unsuitable for continued driving. When it is determined that the overall driving ability impairment rate is less than the second dynamic threshold, the determination result is marked as suitable for continued driving. When it is determined that the overall driving ability impairment rate is between the first dynamic threshold and the second dynamic threshold, the determination result is marked as a restricted driving label.

[0012] As an optional implementation, in the first aspect of the present invention, the method further includes: Based on the flight mission data, the circadian rhythm disruption factor is determined; the circadian rhythm disruption factor is used to describe the degree of circadian rhythm disorder caused by the pilot's occupational characteristics. Based on the aforementioned circadian rhythm disruption factors, a circadian rhythm disruption index is obtained; the circadian rhythm disruption index is used as a comprehensive indicator to measure the loss of consistency between the pilot's internal biological clock and the external environmental time due to occupational exposure. The circadian rhythm disruption index is input into a pre-trained disease exacerbation probability prediction model to obtain the predicted exacerbation probability value corresponding to the target disease; the disease exacerbation probability prediction model is trained using a Logistic regression model, wherein the training formula of the Logistic regression model is:

[0013] In the above formula, The probability of annual deterioration. For the intercept parameter, This is the effect coefficient. This is the index of circadian rhythm disruption.

[0014] Based on the flight mission data, the target route type is obtained; the target route type is compared with a preset route disease database to obtain a targeted monitoring disease list; the targeted monitoring disease list, all the predicted deterioration probability values, and the rising disease list are cross-compared to obtain a sorted disease list. When the risk prediction result is determined to be a restricted driving result, the sorted disease list is matched and filtered to obtain a risk preprocessing plan; wherein, the risk preprocessing plan is used to describe the temporary execution of disease risk tasks that are highly correlated with the route types in the sorted disease list.

[0015] A second aspect of this invention discloses a pilot flight risk prediction device based on big data, the device comprising: The data acquisition module is used to acquire a four-dimensional relational database; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data; the time coordinate data includes time data corresponding to the physical examination date and time data corresponding to the flight mission date; the health information parameter data includes physical examination data, disease development pattern parameters, and aviation medical grounding information; An age correction module is used to obtain an exposure offset coefficient based on the four-dimensional relational database; the exposure offset coefficient is used to correct the flight exposure factor; the flight exposure factor is used to describe different load data of the pilot when performing a mission within a target time; based on the exposure offset coefficient, the actual age data of the pilot is calculated to obtain physiological age data. The risk prediction module is used to obtain physiological disease data in the four-dimensional correlation database based on the physiological age data; to obtain the comprehensive piloting ability impairment rate of the pilot based on the physiological disease data and the flight mission data; and to determine the risk prediction result by judging the comprehensive piloting ability impairment rate.

[0016] As an optional implementation, in a second aspect of the present invention, the disease development pattern parameter in the data acquisition module includes a list of rising diseases, which is obtained through the following method: Acquire pilot medical examination data; divide the pilot medical examination data according to age as the vertical axis and years of flight experience as the horizontal axis to obtain cross-layered structure data; the cross-layered structure data includes several sub-cross-layered structure data; Perform the following operations on each of the sub-cross-layer structure data: For each target disease, the prevalence rate of the target disease is obtained; the relative change rate of the prevalence rate of the target disease among two consecutive years of the target pilots is calculated to obtain the annual change rate of the target disease; the target pilots are pilots born in the same year selected based on the pilot file data. The prevalence rate and annual change rate of the target disease corresponding to each of the sub-cross-layer structure data are filtered according to preset screening conditions to obtain a list of rising diseases; Furthermore, the disease development pattern parameters in the data acquisition module also include a comorbidity association strength matrix, which is obtained through the following method: Perform the following operations on each of the sub-cross-layer structure data: Based on the number of people who simultaneously suffer from the target disease and other diseases, the conditional probability between the target disease and other diseases is obtained; based on the conditional probability, the lift of the target disease is obtained; the lift of the target disease is used to represent the degree of association between the target disease and other diseases; based on all the lifts of the target disease, the sub-cross layer association strength matrix is ​​obtained. Based on the correlation strength matrices of all the sub-cross layers, the co-morbidity correlation strength matrix is ​​obtained.

[0017] As an optional implementation, in a second aspect of the present invention, the specific operation method by which the age correction module obtains the exposure offset coefficient based on the four-dimensional relational database includes: When it is determined that the flight exposure factor value is less than the preset low exposure factor threshold, a low-exposure pilot group is obtained based on the four-dimensional association database. The prevalence of low-exposure diseases among adjacent age groups in the low-exposure pilot group is calculated to obtain the natural prevalence rate growth rate; a benchmark conversion factor is obtained based on the natural prevalence rate growth rate. Based on the health information parameter data and the flight exposure factor, a multiple linear regression model was constructed and the regression coefficient of the exposure factor was obtained by calculation using the least squares method; Divide the regression coefficient of the exposure factor by the benchmark conversion coefficient to obtain the target disease offset coefficient; The exposure offset coefficient is obtained based on all the target disease offset coefficients.

[0018] As an optional implementation, in a second aspect of the invention, the risk prediction module obtains the specific operational method for assessing the pilot's overall flying ability impairment rate based on the physiological disease data and the flight mission data, including: Based on the flight mission data, a mission intensity factor is determined; based on the mission intensity factor, a mission intensity index is obtained. The task intensity index is matched with a preset acceleration factor mapping table to obtain the disease acceleration factor; the actual disease progression rate is obtained based on the physiological disease data and the disease acceleration factor. When it is determined that the disease has not been diagnosed, the first rate of impairment of driving ability is obtained based on the actual rate of disease progression. When a disease is diagnosed, a second pilot capability impairment rate is obtained based on the aforementioned aviation medical grounding information; Based on the aforementioned comorbidity association strength matrix, the comorbidity synergistic risk bonus term is obtained; The comprehensive driving ability impairment rate is obtained based on the first driving ability impairment rate, the second driving ability impairment rate, and the comorbidity risk bonus item.

[0019] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The threshold determination module is used to obtain a historical sample dataset based on the four-dimensional association database; and to calculate the comprehensive flight capability impairment rate of all pilots in the historical sample dataset to obtain a comprehensive flight capability impairment rate dataset. Based on the aforementioned aviation medical grounding information, a grounding tag dataset was determined; The driving ability impairment rate dataset is sorted to obtain an ordered probability sequence; candidate judgment thresholds are determined based on the ordered probability sequence. Based on the candidate decision threshold and the grounding label dataset, a first grounding rate and a second grounding rate are obtained; the first grounding rate is used to describe the ratio of correctly predicted pilots among all grounded pilots; the second grounding rate is used to describe the ratio of incorrectly predicted pilots among all non-grounded pilots. Using the first grounding rate as the vertical axis and the second grounding rate as the horizontal axis, a prediction curve is obtained. Based on the prediction curve, a first determination threshold and a second determination threshold are obtained; wherein, the first determination threshold is greater than the second determination threshold.

[0020] As an optional implementation, in a second aspect of the present invention, the specific operation method by which the risk prediction module judges the overall driving capability impairment rate and determines the risk prediction result includes: Based on the current flight mission data of the assessed pilot, a threshold correction coefficient is determined; the threshold correction coefficient is used to correct the first judgment threshold and the second judgment threshold to obtain a first dynamic threshold and a second dynamic threshold. When it is determined that the overall driving ability impairment rate is greater than the first dynamic threshold, the determination result is marked as unsuitable for continued driving. When it is determined that the overall driving ability impairment rate is less than the second dynamic threshold, the determination result is marked as suitable for continued driving. When it is determined that the overall driving ability impairment rate is between the first dynamic threshold and the second dynamic threshold, the determination result is marked as a restricted driving label.

[0021] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The risk processing module is used to determine the circadian rhythm disruption factor based on the flight mission data; the circadian rhythm disruption factor is used to describe the degree of circadian rhythm disorder caused by the pilot's occupational characteristics. Based on the aforementioned circadian rhythm disruption factors, a circadian rhythm disruption index is obtained; the circadian rhythm disruption index is used as a comprehensive indicator to measure the loss of consistency between the pilot's internal biological clock and the external environmental time due to occupational exposure. The circadian rhythm disruption index is input into a pre-trained disease exacerbation probability prediction model to obtain the predicted exacerbation probability value corresponding to the target disease; the disease exacerbation probability prediction model is trained using a Logistic regression model, wherein the training formula of the Logistic regression model is:

[0022] In the above formula, The probability of annual deterioration. For the intercept parameter, This is the effect coefficient. The index of circadian rhythm disruption; Based on the flight mission data, the target route type is obtained; the target route type is compared with a preset route disease database to obtain a targeted monitoring disease list; the targeted monitoring disease list, all the predicted deterioration probability values, and the rising disease list are cross-compared to obtain a sorted disease list. When the risk prediction result is determined to be a restricted driving result, the sorted disease list is matched and filtered to obtain a risk preprocessing plan; wherein, the risk preprocessing plan is used to describe the temporary execution of disease risk tasks that are highly correlated with the route types in the sorted disease list.

[0023] A third aspect of the present invention discloses an apparatus comprising a memory and a processor, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the big data-based pilot flight risk prediction method described in any of the first aspects of the present invention.

[0024] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the big data-based pilot flight risk prediction method described in any of the first aspects of the present invention.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a four-dimensional relational database using pilot profile data, time coordinate data, health information parameters, and flight mission data. An exposure offset coefficient is calculated from this database. This coefficient is then used to correct for different workloads experienced by the pilot during the target time period. The corrected flight exposure factor is then used to calculate the pilot's actual age, yielding physiological age data. Using this physiological age data as the basis for calculating the pilot's physical age ensures that the obtained physiological disease data better reflects the pilot's current condition, leading to a more accurate assessment of the pilot's overall flight capability impairment rate and ultimately improving the accuracy of pilot risk prediction results. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating a pilot flight risk prediction method based on big data disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of a pilot flight risk prediction device based on big data disclosed in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a device including a memory and a processor disclosed in an embodiment of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0030] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0031] This invention discloses a method, device, and medium for predicting pilot flight risks based on big data, which is used to improve the accuracy of pilot flight risk prediction results. These will be described in detail below.

[0032] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a pilot flight risk prediction method based on big data, as disclosed in an embodiment of the present invention. Figure 1 The described big data-based pilot risk prediction method can be integrated into a big data-based pilot risk prediction device, which can be integrated into a cloud server or a local server. For example... Figure 1 As shown, this big data-based pilot risk prediction method may include the following operations: Step 101: Obtain the four-dimensional relational database; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data.

[0033] In this embodiment of the invention, the pilot file data may include the pilot's unique identifier, date of birth, gender, date of employment, pilot type, aircraft type qualification, and operating base; the pilot file data may be obtained from the airline's human resources management system and pilot technical file system; the pilot file data obtained may be the pilot's inherent attribute data.

[0034] In this embodiment of the invention, the time coordinate data can be established using the pilot's annual physical examination date as the time anchor. The time coordinate data can include: the physical examination date and the statistical cutoff date for flight mission data. The statistical cutoff date for flight mission data can be categorized into: flight mission dates throughout the pilot's career: flight mission dates from the pilot's entry into the current physical examination; flight mission dates in the past 12 months: flight mission dates 12 months prior to the pilot's current physical examination; flight mission dates in the past 3 months: flight mission dates 3 months prior to the pilot's current physical examination; and flight mission dates in the next 3 months: flight mission dates 3 months after the pilot's current physical examination.

[0035] In this embodiment of the invention, flight mission data may include flight data throughout one's entire career, flight data for the past 12 months, flight mission data for the past 3 months, and flight mission data for the next 3 months; Career-long flight data can be any flight mission data from the pilot's start date to the current assessment date. This data is used to calculate long-term cumulative data such as cumulative flight hours, cumulative number of time zone crossings, and cumulative number of high-altitude routes. This data can be used to calculate physiological age. Flight data for the past 12 months can be flight mission data from the current assessment date backwards for 12 months. This data is used to calculate recent routine load indicators such as the percentage of night flights in the past 12 months, the average monthly flight hours in the past 12 months, the number of international long-haul routes in the past 12 months, and the percentage of high-frequency short-haul routes in the past 12 months. This data can be used to calculate the circadian rhythm disruption index and threshold correction coefficient. Flight mission data for the past 3 months can be flight mission data from the current assessment date backwards for 3 months. This data can be used to calculate recent average monthly flight hours, recent frequency of time zone crossings, recent percentage of night flights, and recent consecutive rest days, among other short-term high-intensity exposure indicators. This data can be used to calculate the mission intensity index. Flight mission data for the next three months can be flight mission data planned for the period three months from the current assessment date; it can be used to obtain planning information such as the main types of future routes, the average number of flight segments per day in the future, the proportion of high-altitude airports in the future, the total planned flight hours in the future, the average monthly planned flight hours in the future, the number of planned cross-time zone routes in the future, the main types of future routes, the expected proportion of night flights in the future, and the average number of flight segments per day in the future; this data can be used to identify route types and match targeted monitoring diseases.

[0036] In this embodiment of the invention, the health information parameter data may include: direct collection of individual physical examination data, disease development pattern parameters generated based on individual physical examination data statistics, and aviation medical grounding information; wherein, the direct collection of individual physical examination data may be to obtain the individual physical examination data of each pilot from the aviation personnel physical examination management system and the aviation health department database of the airline, which may include: physical examination indicators, laboratory test indicators, cardiovascular indicators, sensory organ indicators, and diagnosed diseases.

[0037] The specific operation of generating disease development pattern parameters based on individual physical examination data can be as follows: All pilots are divided into age groups at 5-year intervals based on their actual age, and into flight year groups at 5-year intervals based on their flight years, forming a cross-stratified structure. This cross-stratification structure refers to grouping all pilots simultaneously according to both actual age and flight years, creating a two-dimensional tabular hierarchical statistical framework. Each cross cell represents a subgroup of pilots who simultaneously meet the criteria of a certain age group and a certain flight year group. For each cross stratum, the following statistical indicators are calculated for each disease: (1) Prevalence: The number of people with the disease in the cross-stratification divided by the total number of people in the cross-stratification.

[0038] (2) Annual growth rate: Focus on pilots born in the same year and calculate the relative change rate of disease incidence between two consecutive years. Specifically, the method can be as follows: use the birth dates in the pilot file data to filter out pilots born in the same year, use the physical examination dates in the time coordinate data to determine the physical examination records for two consecutive years, calculate the disease incidence of the cohort in the two years respectively, and then calculate the relative change rate; then average the growth rate of each disease in this cross-stratification over several years to obtain the annual growth rate of each disease in this cross-stratification.

[0039] (3) Co-morbidity strength: Calculate the conditional probability and lift between each pair of diseases to form a co-morbidity strength matrix; where lift can be used to represent the strength of the association between disease A and disease B. Lift equals the probability of having disease B while having disease A divided by the probability of having disease B among all pilots. A lift greater than 1 indicates a positive correlation, equal to 1 indicates independence, and less than 1 indicates a negative correlation.

[0040] Identify rising diseases based on preset statistical screening criteria and establish a list of rising diseases. Preset statistical screening criteria may include: an annual growth rate of morbidity greater than 5% for three consecutive years and / or Spearman correlation coefficients and association coefficients between morbidity and years of flight experience within preset ranges and / or a morbidity rate greater than a preset probability value among pilots of a target age, where the target age can be 45 years old and the preset probability value can be 15%.

[0041] Those skilled in the art will understand that the Spearman correlation coefficient is a nonparametric statistic calculated using the Spearman rank correlation test. It measures the strength and direction of the monotonic association between two variables, with a value ranging from -1 to 1. A value of 1 indicates a completely positive monotonic relationship, meaning that an increase in one variable leads to an increase in the other; a value of -1 indicates a completely negative monotonic relationship; and a value of 0 indicates no monotonic relationship. The Spearman correlation coefficient does not require variables to follow a normal distribution or to have a linear relationship; it only requires that the variables exhibit monotonicity. Therefore, it can be applied to the association analysis of ordered categorical and continuous variables in this scheme. In the screening criteria of this scheme, the Spearman correlation coefficient is used to measure the degree of monotonic association between years of flight experience and the prevalence of the target disease. For example, a Spearman correlation coefficient greater than 0.6 indicates a strong positive monotonic association between years of flight experience and the prevalence, meaning that the prevalence shows a continuous upward trend with increasing years of flight experience.

[0042] The correlation coefficient can be calculated using the Spearman rank correlation test to determine whether the association is statistically significant. First, we assume the opposite: that there is no monotonic association between years of flight experience and the prevalence of the target disease. Statistically, this is called the null hypothesis. Assuming the null hypothesis is true, we calculate the probability of observing this strong association (i.e., a Spearman correlation coefficient greater than 0.6) purely due to random fluctuations in the data, assuming no relation between the two. This calculated probability is the correlation coefficient. When the correlation coefficient is small, for example, less than 0.05, it means that if years of flight experience and prevalence are truly unrelated, the probability of observing this strong association is less than 5%. This situation is highly unlikely, so we can reject the null hypothesis and consider the association to be real and reliable, not random error. When the correlation coefficient is large, for example, greater than or equal to 0.05, it means that even if the two are truly unrelated, there is a relatively high probability (more than 5%) of randomly observing this association. Therefore, the current data cannot reject the null hypothesis, and the existing data is insufficient to definitively establish an association between the two.

[0043] Furthermore, selecting an annual prevalence rate growth rate greater than 5% for three consecutive years is to identify a continuously worsening trend and exclude random fluctuations. This is because an increase in the prevalence rate in a single year may be caused by changes in physical examination standards, random events, or statistical errors. Only by maintaining an increase for many consecutive years can it be confirmed that this is a real and systematic worsening trend, rather than a fleeting phenomenon. Setting the threshold to 5% is to exclude diseases that grow extremely slowly and have no clinical intervention value in the short term.

[0044] The Spearman correlation coefficient and association coefficient between morbidity and years of flight experience were selected within a preset range to distinguish between "occupation-related diseases" and "age-related diseases," thus identifying occupational risks. This is because the morbidity of many diseases increases with age (e.g., presbyopia, benign prostatic hyperplasia), which is a natural part of aging and not the focus of this study. This study is truly concerned with health damage caused by exposures specific to the flight profession (e.g., crossing time zones, night flights, hypoxia). If the morbidity of a disease is strongly correlated with years of flight experience, it largely indicates that occupational exposure is driving disease progression. The Spearman correlation coefficient and association coefficient being within the preset range indicates that the conclusion is statistically reliable; for example, a Spearman correlation coefficient greater than 0.6 indicates a strong correlation, and a p-value less than 0.05 indicates that the conclusion is statistically reliable. Only when both requirements are met can a disease be confirmed as an "occupational disease."

[0045] Selecting pilots of the target age group with a prevalence rate higher than the preset probability value ensures the disease has sufficient monitoring value and avoids wasting resources, as the purpose of this plan is to conduct population-level risk prediction. If a disease, although showing an upward trend and related to the flying profession, is rare in the entire pilot population (such as a rare tumor), then investing heavily in building predictive models and monitoring systems is not cost-effective. Setting a 15% prevalence rate threshold ensures that the selected diseases have a certain degree of prevalence, and focused monitoring can cover a sufficiently large high-risk population, thereby generating practical health management and safety protection benefits.

[0046] In this embodiment of the invention, the aviation medical grounding information can be derived from the aviation personnel medical standards issued by the Civil Aviation Administration of China and the airline's historical medical grounding event records. This includes aviation medical grounding standard parameters, such as: the medical grounding criteria and quantitative thresholds corresponding to each disease, the average time interval from the first diagnosis of each disease to the actual triggering of the medical grounding criteria, and the weight of each disease's impact on flight safety and piloting ability. The weight of piloting ability's impact can be determined comprehensively based on the frequency of each disease as the primary cause in historical medical grounding events and the consensus of aviation medical experts. For example, if a pilot is first diagnosed with hypertension and their systolic blood pressure remains ≥160 mmHg or diastolic blood pressure remains ≥100 mmHg, or if their blood pressure cannot be stably controlled at a safe level despite taking antihypertensive medication, they need to undergo clinical treatment and be temporarily grounded. Therefore, for hypertension, the medical grounding criteria and quantitative thresholds are determined to be: systolic blood pressure ≥160 mmHg or diastolic blood pressure ≥100 mmHg. The study analyzes the time it takes for a pilot to progress from being first diagnosed with hypertension to triggering the aforementioned grounding criteria, assuming it's 20 months. It then analyzes all past medical grounding incidents of the airline, examining the percentage caused by hypertension. Assuming hypertension ranks first among all grounding incidents, accounting for 20%, it's considered an "extremely high-risk" condition because it can lead to disability, fainting, or serious cardiovascular accidents during flight, directly threatening flight safety. Hypertension is then assigned a high weight; if the weight range is set from 0 to 1 (1 representing the highest risk), the impact weight of hypertension on piloting ability can be determined as 0.90. Finally, a combined assessment of systolic blood pressure consistently ≥160 mmHg or diastolic blood pressure consistently ≥100 mmHg, with a time interval of 20 months and a piloting ability impact weight of 0.90, constitutes the aviation medical grounding information.

[0047] In this embodiment of the invention, the four-dimensional association database can be constructed by using the pilot's unique identifier and physical examination date as a composite primary key to associate the pilot's file data, time coordinate data, health information parameter data, and flight mission data at the same time segment, forming a four-dimensional association data record; multiple four-dimensional association data records of the same pilot along the time axis are arranged to finally form a four-dimensional association database, and all the data in the four-dimensional association database has been encoded and corresponded.

[0048] Step 102: Based on the four-dimensional relational database, obtain the exposure offset coefficient; based on the exposure offset coefficient, calculate the actual age data of the pilots to obtain the physiological age data.

[0049] In this embodiment of the invention, the exposure offset coefficient can be used to correct the flight exposure factor; the flight exposure factor can be used to describe different load data of the pilot performing the mission within the target time.

[0050] In this embodiment of the invention, flight exposure factors can refer to several key indicators selected from flight mission data. Their function is to quantify the pilot's occupational load into calculable data, serving as input variables for calculating physiological age. These key indicators can include cumulative flight hours factor, cross-time zone frequency factor, night flight load factor, high-altitude route factor, and duty density factor.

[0051] The cumulative flight hours factor measures the total cumulative load on the body from flight time. This factor comprehensively reflects the cumulative effects of multiple factors that pilots experience in the cockpit environment, including low pressure, low oxygen, cosmic radiation, prolonged sitting, vibration, and noise. The higher the total flight hours, the more significant this cumulative effect. This is because pilots maintain a certain cabin altitude during flight missions, placing them in a prolonged low-pressure, low-oxygen environment. Cosmic radiation doses are significantly higher at high altitudes than at ground level. Prolonged sitting leads to obstructed venous return in the lower limbs and pressure on intervertebral discs. Engine and airflow noise continuously stimulate the auditory system, and fuselage vibration continuously affects the musculoskeletal system. These factors accumulate with increasing flight time, with no lower threshold for single exposure. In other words, every additional hour of flight means an additional hour of combined effects from these factors. The total flight hours, as an objectively measurable, tamper-proof record covering the entire career, accurately quantifies each pilot's total exposure time and can therefore serve as an indicator of the cumulative load from flight time. Since the original total flight hours can be as high as thousands or even tens of thousands of hours, directly using this value in subsequent multiple regression calculations would cause it to dominate the regression results due to its excessive size, drowning out the contributions of other factors. Therefore, by dividing the total flight hours by 1000 to convert the unit from hours to thousands of hours, the value of the cumulative flight hours factor falls within the range of single digits to tens of digits, placing it on the same order of magnitude as the values ​​of other factors.

[0052] The frequency factor for crossing time zones can be used to measure the cumulative number of times the circadian rhythm is repeatedly disrupted. Each flight across multiple time zones causes the body's internal biological clock to become out of sync with the external environment, leading to jet lag. Frequent misalignments can repeatedly disrupt endocrine, sleep, and metabolic functions, and are a significant contributing factor to metabolic diseases. The body's endogenous circadian rhythm is regulated by the suprachiasmatic nucleus of the hypothalamus, with an endogenous cycle of approximately 24 hours, relying on the light-dark cycle of the external environment for daily calibration. When pilots perform cross-time zone flights, there is a several-hour difference between the local time at the departure and destination points, causing a sudden change in the external light-dark signal, while the body's internal biological clock continues to operate according to the departure time, failing to synchronize with the destination time immediately. This disconnect between internal and external time is the physiological essence of circadian rhythm disorder. Each flight crossing three or more time zones is an independent event sufficient to trigger the aforementioned disorder. Therefore, the cumulative number of flights crossing three or more time zones is the most direct measure of the frequency of such events. The calculation can be done by dividing the cumulative number of times that have crossed three or more time zones by 50. Dividing by 50 is to change the unit of measurement from "times" to "50 times" to make the measurement standard consistent.

[0053] The night flight workload factor measures the workload deviation during natural sleep periods. Working during the nighttime hours when the body should be resting causes sleep deprivation and conflicts with the biological clock; this directly interferes with the secretion of rhythm hormones such as melatonin, having a profound impact on long-term health. This factor measures the difference between an individual pilot's nighttime workload and the industry average; because the human body is naturally in a physiological state of peak melatonin secretion, decreased core body temperature, and reduced alertness during the nighttime hours (usually from 10 PM to 6 AM the next day); performing flight missions during this period directly conflicts with endogenous rhythmic signals, forming a "behavior-rhythm" conflict; working in this conflicting state for a long time leads to cumulative sleep deprivation, continuous suppression of melatonin secretion rhythm, and autonomic nervous system dysfunction, which has a clear physiological link with the occurrence of metabolic syndrome and cardiovascular disease. The percentage of night flights in the past 12 months is an objective measure of this workload intensity. The calculation method can be the individual's night flight percentage over the past 12 months minus the industry benchmark night flight percentage. The industry benchmark night flight percentage is usually taken as 15%. This is to introduce a reference system, convert the absolute proportion into a deviation value relative to the normal level of the industry, so that the positive or negative sign of the factor directly reflects whether the load is higher or lower than the industry average.

[0054] The high-altitude flight factor measures the cumulative additional load from the hypoxic environment at high altitudes. Takeoffs and landings at airports at altitudes of 2400 meters and above mean repeated exposure to hypoxia, which places an additional strain on cardiopulmonary function, blood pressure, and blood oxygen-carrying capacity, producing an independent cumulative effect. Because the atmospheric pressure and oxygen partial pressure at high-altitude airports (2400 meters and above) are significantly lower than at sea level, pilots are in a hypoxic environment both inside and outside the cockpit during takeoffs and landings, resulting in decreased blood oxygen saturation. The cardiopulmonary system must compensate by increasing ventilation and heart rate. Each takeoff and landing on a high-altitude flight is an independent cardiopulmonary hypoxic stress event. Long-term, repeated high-altitude exposure can lead to pulmonary vascular remodeling and increased right ventricular load, producing independent cumulative effects on the cardiovascular and respiratory systems. Therefore, the cumulative number of high-altitude flights is the most direct measure of the frequency of such hypoxic exposure events. The calculation can be done by dividing the cumulative number of high-altitude flights by 30 to ensure dimensional uniformity.

[0055] The duty density factor measures the deviation of recent duty intensity from normal levels. High-intensity continuous duty encroaches on the body's recovery time, leading to the accumulation of fatigue and stress. This factor can capture the health risks associated with excessive short-term work intensity, complementing the factor reflecting long-term cumulative flight hours. Higher average monthly flight hours mean less time available for post-duty recovery, resulting in a more significant cumulative effect of fatigue and stress. Long-term duty intensity exceeding normal levels can lead to decreased autonomic nervous system regulation, abnormal cortisol rhythms, and immunosuppression—all contributing factors to the development and progression of chronic diseases. The 12-month time window smooths out short-term fluctuations while reflecting recent normalized work intensity, providing a timescale that captures recent workload without being overly sensitive. The calculation method involves subtracting a baseline average monthly flight hour from an individual's average monthly flight hour over the past 12 months. The baseline average monthly flight hour is typically 60 hours, introduced to provide a reference point, converting the absolute hours into a deviation value relative to normal levels, allowing the factor's sign to directly reflect whether the duty intensity is higher or lower than normal.

[0056] In this embodiment of the invention, the exposure offset coefficient can be used to eliminate the dimensional differences between different flight exposure factors and convert them into the same unit. Because different types of flight loads have different mechanisms and degrees of wear and tear on the body, they cannot be directly added together. Therefore, an exposure offset coefficient is designed to quantify different flight exposure factors into a unified and comparable unit. For example, the unit of the cumulative flight hours factor is thousands of hours, the unit of the cross-time zone frequency factor is 50 times, the unit of the night flight load factor is percentage points, the unit of the high-altitude route factor is 30 times, and the unit of the duty density factor is hours per month; the original dimensions of these five factors are completely different and cannot be calculated uniformly. Therefore, the exposure offset coefficient is introduced, with different exposure offset coefficients for each of the five factors. Their unit is "years / unit of flight exposure factor," meaning that each increase of one standardized unit in each flight exposure factor corresponds to an increase in physiological age by a certain number of years.

[0057] In this embodiment of the invention, by calculating physiological age data, the actual situation of the pilot's current state can be reflected more accurately; and the calculation process of physiological age data can be the pilot's actual age plus the exposure offset coefficient multiplied by the flight exposure factor.

[0058] Step 103: Based on physiological age data, obtain physiological disease data from the four-dimensional correlation database; based on physiological disease data and flight mission data, obtain the comprehensive piloting ability impairment rate; judge the comprehensive piloting ability impairment rate and determine the risk prediction result.

[0059] In this embodiment of the invention, because a pilot's actual age does not reflect their actual physical condition, physiological age data is calculated and matched in a four-dimensional association database to obtain physiological disease data corresponding to the physiological age data. For example, if a pilot's actual age is 40, the calculated physiological age is 46. However, the four-dimensional association database contains only disease data corresponding to the actual age. If the actual age is used directly for matching, the disease data will not accurately reflect the pilot's physical condition. Therefore, the calculated physiological age of 46 is used instead of the actual age of 40 to match the corresponding disease data in the four-dimensional association database.

[0060] In this embodiment of the invention, the overall pilot capability impairment rate can be the total probability of a pilot being grounded due to illness within a preset assessment period, taking into account all known disease risks. The overall pilot capability impairment rate calculated by combining physiological disease data determined by physiological age with flight mission data is more meaningful for assessment. Finally, the overall pilot capability impairment rate is determined to obtain a risk prediction result.

[0061] For example, the specific calculation process for the overall impaired flight capability rate can be as follows: First, calculate the risk probability of a single disease causing flight suspension in the physiological disease data. For diseases that have not yet been diagnosed, first use the actual rate of disease progression to calculate the expected diagnosis time required from the current deterioration to diagnosis; add this to the average time interval between diagnosis and flight suspension for that disease to obtain the total expected flight suspension time; then calculate the probability of flight suspension caused by that disease within the preset assessment period. For diseases that have been diagnosed, directly use the average time interval between diagnosis and flight suspension, and correct it according to the deviation of the current indicator value from the flight suspension threshold, and then calculate the flight suspension probability. Then, multiply the flight suspension probability of each disease by its respective flight capability impact weight to obtain a weighted risk value; then, combine all weighted risk values ​​using a probability merging formula; then calculate the comorbidity risk bonus, that is, iterate through all disease pairwise combinations, query the improvement degree in the comorbidity association strength matrix, filter out high-association combinations, and calculate additional risk bonus; finally, combine the merged probability with the bonus to obtain the final overall impaired flight capability probability. As can be seen, the embodiments of the present invention construct a four-dimensional relational database using pilot file data, time coordinate data, health information parameter data, and flight mission data; an exposure offset coefficient is calculated using the four-dimensional relational database; the exposure offset coefficient is used to correct different load data of the pilot performing missions within the target time; the corrected flight exposure factor is used to calculate the pilot's actual age data to obtain physiological age data; using the physiological age data as the pilot's physical age for calculation makes the obtained physiological disease data more consistent with the pilot's current condition, thereby making the calculated comprehensive pilot's impairment rate more accurate, and thus improving the accuracy of the pilot's flight risk prediction results.

[0062] In an optional embodiment, the disease progression pattern parameter includes a list of rising diseases, which is obtained through the following means: Obtain pilot medical examination data; divide the pilot medical examination data according to age as the vertical axis and years of flight experience as the horizontal axis to obtain cross-layered structure data; the cross-layered structure data includes several sub-cross-layered structure data; Perform the following operations on each sub-cross-layer structure data: For each target disease, the prevalence rate of the target disease is obtained; the relative change rate of the prevalence rate of the target disease among the target pilots in two consecutive years is calculated to obtain the annual change rate of the target disease; the target pilots are selected from pilots born in the same year based on pilot file data. The target disease prevalence and annual change rate corresponding to each sub-cross-layer structure data are filtered according to preset screening conditions to obtain a list of rising diseases. In addition, the parameters for disease progression also include the comorbidity strength matrix, which is obtained in the following way: Perform the following operations on each sub-cross-layer structure data: Based on the number of people who simultaneously suffer from the target disease and other diseases, the conditional probability between the target disease and other diseases is obtained; based on the conditional probability, the boost of the target disease is obtained. Based on the elevation of all target diseases, the sub-cross-layer correlation strength matrix is ​​obtained; Based on the correlation strength matrix of all sub-cross layers, the comorbidity correlation strength matrix is ​​obtained.

[0063] In this optional embodiment, the pilot's physical examination data can be obtained from the aviation personnel physical examination management system and the airline's aviation health department database, which can include: physical examination indicators, laboratory test indicators, cardiovascular indicators, sensory organ indicators, and diagnosed diseases.

[0064] In this optional embodiment, the cross-layered data structure can be a two-dimensional tabular hierarchical statistical framework formed by grouping all pilots simultaneously according to two dimensions: actual age and years of flight experience. For example, actual age can be grouped in 5-year intervals, such as 20-24, 25-29, 30-34, 35-39, 40-44, etc.; and years of flight experience can be grouped in 5-year intervals, such as 0-4 years, 5-9 years, 10-14 years, 15-19 years, 20-24 years, etc.

[0065] After the two dimensions intersect, a two-dimensional table is formed. Each intersecting cell represents a subgroup of pilots who "simultaneously meet a certain age group and a certain number of years of flying experience," which is the sub-intersecting layer structure data; as shown in Table 1: Table 1 Cross-layered structure data table

[0066] In the table above, "—" indicates that the combination does not exist in reality. For example, a pilot aged 20-24 is unlikely to have 5-9 years of flying experience because the legal entry age is usually over 18. The reason for not stratifying solely by age or solely by years of flying experience is to distinguish the impact of "natural aging" and "occupational exposure" on diseases. If stratified only by age, it would mask the differences in years of flying experience; for example, pilots aged 40-44 who have flown for 20 years and those who have only flown for 5 years may have completely different disease risks.

[0067] If we only stratify by years of flying experience, it will mask the differences in age. For example, pilots who have flown for 15 years may be 35 years old or 55 years old, with significant differences in their basic physical condition. Therefore, this scheme adopts a cross-stratification method, which can avoid these unexpected situations and improve the accuracy of subsequent calculations.

[0068] In this optional embodiment, the target disease can be hyperlipidemia, fatty liver, hyperuricemia, diabetes, obesity, hypertension, cervical spondylosis, sleep disorders, etc. The prevalence of the target disease can be calculated by recording the number of people suffering from the target disease as the number of patients, counting the total number of pilots in the sub-cross-layer structure, and then expressing the target disease prevalence as a percentage equal to the number of patients divided by the total number of pilots.

[0069] In this optional embodiment, the target pilots can be pilots born in the same year, selected based on pilot profile data; the annual change rate of the target disease can be calculated by calculating the prevalence of the target disease in two consecutive years for the target pilots, and then using the prevalence rates of the target disease in those two years to derive the annual change rate of the target disease. For example, if the prevalence rate in the first year is M and the prevalence rate in the second year is N, then the annual change rate of the target disease is (NM) / M. Selecting pilots born in the same year for calculation ensures that the comparison is of the prevalence change of pilots of the same age at two points in time, rather than the difference in prevalence between different populations, thus accurately reflecting the true rate of disease evolution.

[0070] In this optional embodiment, the preset screening conditions may be: the annual change rate of the target disease is greater than 5% for three consecutive years; the Spearman correlation coefficient and association coefficient between the prevalence of the target disease and the years of flight are within a preset range; and the prevalence of the target disease among pilots of the target age is greater than a preset probability value. The target age may be 45 years old, and the preset probability value may be 15%.

[0071] A target disease annual change rate greater than 5% for three consecutive years can be defined as follows: This involves checking whether the annual growth rate of the target disease in each of the three most recent statistical years, within the same cross-stratification, is greater than 5%. This condition is used to exclude random fluctuations and screen for diseases with a sustained upward trend.

[0072] The Spearman correlation coefficient and association coefficient between the prevalence of the target disease and years of flight experience are within a preset range. This preset range can be a Spearman correlation coefficient greater than 0.6 and an association coefficient less than 0.05. Those skilled in the art will understand that the Spearman correlation coefficient and association coefficient are calculated using the Spearman rank correlation test. The Spearman correlation coefficient is calculated with the median years of flight experience for each flight experience group on the horizontal axis and the corresponding prevalence of the target disease on the vertical axis. This measures whether the prevalence shows a continuously monotonically increasing trend with increasing years of flight experience. A Spearman correlation coefficient greater than 0.6 indicates a strong positive monotonically increasing association. The Spearman rank correlation test is performed on the above correlation coefficients to calculate the association value. An association value less than 0.05 indicates that the probability of the association being caused by random sampling error is less than five percent, and the association is statistically significant. This condition is used to distinguish between diseases driven by occupational exposure to flight and diseases driven solely by natural aging, ensuring that the selected diseases have a clear statistical association with the flight occupation.

[0073] If the prevalence of the target disease among pilots of the target age exceeds a preset probability value, it can be determined by calculating the overall prevalence of the target disease among all pilots aged 45 or older and judging whether it exceeds 15%. This condition is used to ensure that the selected diseases are sufficiently prevalent in the older pilot population, avoid including rare diseases in the list of groups requiring key monitoring, and ensure the rational allocation of monitoring resources.

[0074] In this optional embodiment, a disease can only be collected and a rising disease list can be formed if three of the preset screening conditions are met simultaneously.

[0075] In this optional embodiment, the conditional probability can be obtained as follows: For each sub-cross-layer structure data, from all diseases involved in the sub-cross-layer, first select one as the target disease A, and uniformly define each disease other than A as disease B; determine the number of pilots with target disease A, the number of pilots with disease B, and the number of pilots with both target disease A and disease B in the sub-cross-layer structure data; calculate the conditional probability P(B|A), where P(B|A) represents the proportion of pilots with target disease A who also have disease B; then calculate the proportion P(B) of all pilots with disease B in the sub-cross-layer structure data; where P(B|A) and P(B) can be considered as the conditional probabilities in this scheme.

[0076] In this optional embodiment, the target disease lift can represent the degree of association between the target disease and other diseases. In other words, knowing that a pilot has target disease A increases the accuracy of predicting that he also has disease B by how many times. For example, a lift of 1 means that the probability of having disease B in the group with disease A is exactly the same as the probability of having B in all pilots; knowing A does not help in predicting B. The two diseases are independent and have no association. A lift greater than 1 means that the probability of having B in the group with disease A is higher than the overall average. The larger the value, the stronger the association. For example, Lift=2.0 means that the probability of having disease B after having disease A is twice the baseline level; Lift=3.0 means three times. Such highly associated combinations, if occurring simultaneously in the same pilot, will produce a health risk superposition effect greater than the sum of its parts. A lift less than 1 means that the probability of having disease B in the group with disease A is actually lower than the overall average level. The two diseases are negatively correlated; having disease A may mean a lower risk of developing disease B.

[0077] The calculation method is as follows: Lift(A,B)=P(B|A) / P(B); where Lift(A,B) represents the lift of the target disease, which can be the lift at the intersection of the row of "target disease A" and the column of "disease B". Then, using disease B as the target disease B and defining other diseases as disease C, the lift of all diseases is calculated repeatedly to obtain the lift of all diseases, where disease C includes disease A calculated in the previous step.

[0078] In this optional embodiment, the sub-cross-layer association strength matrix can be obtained by collecting the target disease elevation degrees of all diseases in the sub-cross-layer structure data and filling them into the matrix formed between diseases. The calculated sub-cross-layer association strength matrices are then filled into the table formed by the cross-layer structure data to obtain the comorbidity association strength matrix.

[0079] As can be seen, in this optional embodiment, by dividing the pilot's physical examination data into cross-layered structure data according to age and years of flight experience, the influence of "natural aging" and "occupational exposure" on diseases can be avoided. The prevalence and annual variation rate of target diseases for each sub-cross-layered structure data are calculated. The prevalence and annual variation rates of target diseases are then filtered according to preset screening criteria to obtain a list of rising diseases. By determining the list of rising diseases, the disease data used in calculating the overall pilot capability impairment rate can be clearly identified. The target disease elevation degree of each disease in the sub-cross-layered structure data is calculated to obtain the sub-cross-layered correlation strength matrix, and finally, the comorbidity correlation strength matrix is ​​obtained. The accuracy of calculating the overall pilot capability impairment rate can be improved through the comorbidity correlation strength matrix.

[0080] In another alternative embodiment, obtaining the exposure offset coefficient based on a four-dimensional relational database may include: When it is determined that the flight exposure factor value is less than the preset low exposure factor threshold, a low-exposure pilot group is obtained based on the four-dimensional association database. The prevalence of low-exposure diseases among adjacent age groups in the low-exposure pilot population was calculated to obtain the natural prevalence rate increase; based on the natural prevalence rate increase, a baseline conversion factor was obtained. Based on health information parameters and flight exposure factors, a multiple linear regression model was constructed and the regression coefficients of the exposure factors were obtained by calculation using the least squares method. Divide the regression coefficient of the exposure factor by the benchmark conversion factor to obtain the target disease offset coefficient; The exposure offset coefficient is obtained based on the offset coefficients of all target diseases.

[0081] In this optional embodiment, the preset low exposure factor threshold can be calculated in real time. The specific process is as follows: All pilots are sorted according to the value of each flight exposure factor. After sorting each flight exposure factor from smallest to largest, the value at the 25th percentile for each flight exposure factor is taken. A value below the 25th percentile means that the pilot's exposure load for that factor is among the lowest of all pilots. At this point, the value at the 25th percentile is the low exposure factor threshold. When all factors meet this condition, the pilot's total occupational exposure to flight is extremely low, and the increase in morbidity is almost entirely driven by natural aging, with the interference of flight load being negligible.

[0082] In this optional embodiment, pilots whose own flight exposure factors are less than the low exposure factor threshold are selected from the four-dimensional association database to obtain a low-exposure pilot group.

[0083] In this optional embodiment, the prevalence of each disease in the low-exposure pilot group is calculated, which is the low-exposure disease prevalence; then the increase in the low-exposure disease prevalence of a disease among adjacent ages of the same pilot is calculated to obtain the natural prevalence increase; the arithmetic mean of the natural prevalence increase of a disease among different ages of the same pilot is calculated to obtain the baseline conversion factor.

[0084] In this optional embodiment, the construction process of the multiple linear regression model can be as follows: The same disease for which the baseline conversion coefficient is calculated is processed to determine whether all pilots at a given age have the disease. The presence or absence of the disease is used as the dependent variable in the multiple linear regression model. The flight exposure factors of all pilots at the given age are used as independent variables, and multiple linear regression is performed. Because the regression is performed within a group of the same age, the age variable is naturally controlled. Since everyone is the same age, the difference in prevalence cannot be caused by age, but only by differences in flight exposure. The regression coefficient of the exposure factor obtained from the regression represents how many percentage points the prevalence of the disease increases for each unit increase in a certain exposure factor among people of the same age. For example, the formula for the multiple linear regression model could be...

[0085] In the above formula, The value indicates whether the pilot is ill; 1 indicates ill and 0 indicates not ill. For the first The exposure factor regression coefficients corresponding to each flight exposure factor; For the first One flight exposure factor; The intercept term represents the baseline prevalence level when all exposure factors are zero. This is the random error term, representing random fluctuations that the model cannot explain; Let be a positive integer. Here, α is a probability value between 0 and 1. In actual calculations, since the least squares method cannot automatically guarantee that the predicted value falls between 0 and 1, α may exceed the range of 0 to 1 after adding the contributions of each factor. In this case, values ​​greater than 1 are usually truncated to 1, and values ​​less than 0 are truncated to 0.

[0086] In this optional embodiment, the model formula is calculated using the least squares method to obtain the regression coefficients of each flight exposure factor, i.e., the exposure factor regression coefficients. Those skilled in the art will understand that the least squares method can be a mathematical optimization method that finds the most suitable regression coefficient by minimizing the sum of the squares of the errors between the predicted and actual values ​​of all samples. In other words, it finds a straight line that best approximates all data points, minimizing the sum of the squares of the vertical distances from all data points to this line.

[0087] In this optional embodiment, the target disease offset coefficient can be calculated by dividing the exposure factor regression coefficient by the baseline conversion coefficient, which yields the target disease offset coefficient for a given disease. The target disease offset coefficients for different diseases are calculated by selecting multiple diseases. The prevalence of each disease is retrieved from the disease development pattern parameters, and using the prevalence as a weight, the weighted sum of the target disease offset coefficients for each disease is calculated and divided by the total weight, resulting in the final exposure offset coefficient. Because different diseases have different sensitivities to flight exposure; circadian rhythm disorders have a significant impact on hyperuricemia, and noise exposure has a significant impact on hearing loss, if only one disease is used for calibration, the resulting offset coefficient may be particularly sensitive to that disease but not applicable to others. By weighting and averaging multiple diseases, the final offset coefficient is a comprehensive set of conversion parameters suitable for assessing overall physiological age.

[0088] For example, as shown in Table 2, Table 2 shows the calculated prevalence of various diseases and the target disease offset coefficient; Table 2. Prevalence of various diseases and target disease offset coefficients

[0089] Therefore, the exposure offset coefficient for the first flight exposure factor can be calculated as follows: first exposure offset coefficient = (28% × 0.60 + 35% × 0.55 + 25% × 0.65) ÷ (28% + 35% + 25%) = 0.59; and so on, calculate the exposure offset coefficients for other flight exposure factors, and finally obtain the overall exposure offset coefficient.

[0090] As can be seen, in this optional embodiment, the exposure offset coefficient is obtained by calculating the baseline conversion coefficient and the exposure factor regression coefficient; by calculating the exposure offset coefficient, the pilot's actual age can be corrected to obtain physiological age data that is more consistent with the pilot's physical condition; by calculating the pilot's comprehensive flight capability impairment rate through physiological age data, the accuracy of predicting the flight risk of aircraft pilots can be improved.

[0091] In yet another optional embodiment, the overall pilot's impairment rate is assessed based on physiological disease data and flight mission data, including: Based on flight mission data, determine the mission intensity factor; based on the mission intensity factor, obtain the mission intensity index. The task intensity index is matched with a preset acceleration factor mapping table to obtain the disease acceleration factor; based on physiological disease data and the disease acceleration factor, the actual disease development rate is obtained. When it is determined that the disease has not been diagnosed, the first driving ability impairment rate is obtained based on the actual rate of disease progression. When the disease is diagnosed, the second pilot capability impairment rate is obtained based on aviation medical grounding information; Based on the comorbidity association strength matrix, the comorbidity synergistic risk bonus term is obtained; The overall driving ability impairment rate is obtained based on the first driving ability impairment rate, the second driving ability impairment rate, and the comorbidity risk bonus.

[0092] In this optional embodiment, the mission intensity factor can be determined as follows: the mission intensity factor is obtained by filtering the recent monthly average flight hours (total flight hours in the past 3 months divided by 3), the recent cross-time zone frequency (number of flights crossing three or more time zones in the past 3 months), and the recent night flight ratio (night flight hours in the past 3 months divided by total flight hours in the past 3 months) from the flight mission data.

[0093] The mission intensity index can be used to quantify the workload of pilots' recent flight missions. The mission intensity index can be determined as follows: a weighted sum is obtained after normalization using the monthly average flight hours high-risk reference threshold, the cross-time zone frequency high-risk reference threshold, and the night flight percentage high-risk reference threshold as the denominator. The reference threshold for each mission intensity factor can be obtained by arranging all the statistically collected mission intensity factor data in order, using the values ​​at the top 25% as the reference threshold. The mission intensity index uses 1.0 as the baseline value; values ​​greater than 1.0 indicate overload, and values ​​less than 1.0 indicate underload. The weighted summation after normalization can be achieved by first dividing each mission intensity factor by its corresponding high-risk reference threshold to obtain different dimensionless ratios, thus eliminating dimensional differences. Then, each ratio is multiplied by its respective preset weighting coefficient and summed to obtain the mission intensity index value. The weighting coefficients are also determined through historical data regression analysis, using changes in metabolic indicators as the dependent variable and mission intensity factors as the independent variable in a multiple linear regression, with the normalized absolute values ​​of the standardized regression coefficients used as weights.

[0094] In this optional embodiment, the acceleration factor mapping table can be constructed as follows: Pilot sample data with two consecutive years of physical examination records and complete flight mission data are selected from a four-dimensional relational database; the mission intensity index of each pilot sample is calculated; the actual rate of change of the target disease index of each pilot sample from the base year to the next year is calculated, where the target disease index can be the specific values ​​of uric acid or blood pressure; the physiological age of the pilot sample is calculated, thereby obtaining the physiological disease data of the pilot sample; this physiological disease data includes the annual rate of change of the disease calculated in the previous sub-cross-layer structure data. By dividing the actual rate of change by the annual rate of change of the disease, the acceleration factor of the individual disease can be obtained; then, the mission intensity index of the pilot sample is divided into different intervals; for example, a mission intensity index less than 0.8 indicates a low-load region; (0.8, 1.2) is a normal-load region; (1.2, 1.6) is a relatively high-load region; (1.6, 2.0) is a high-intensity-load region; and greater than 2.0 is an ultra-high-intensity-load region. The individual disease acceleration factors of all pilot samples within the mission intensity index range are collected to form a numerical set. Then, the median of this set is taken as the disease acceleration factor falling within this mission intensity index range. Finally, all data are calculated and statistically analyzed to obtain an acceleration factor mapping table. This acceleration factor mapping table is continuously updated based on annual pilot medical examination data. For example, as shown in Table 3, Table 3 shows a partial acceleration factor mapping table.

[0095] Table 3 Partial Acceleration Factor Mapping Table

[0096] In this optional embodiment, the actual disease progression rate can be obtained by multiplying the annual rate of change of the disease in the physiological disease data by the disease acceleration factor.

[0097] In this optional embodiment, the specific calculation method for the overall driving capability impairment rate may be as follows: Firstly, the calculation of the impairment rate of primary driving ability caused by undiagnosed diseases: First, obtain the current indicator values ​​for the disease to be calculated and the clinical diagnostic thresholds from the aviation medical grounding information. Then, calculate the expected diagnosis time based on the actual disease progression rate. The calculation formula is as follows:

[0098] In the above formula, For the expected time of diagnosis, For the current indicator value of the disease to be calculated, The threshold for clinical diagnosis. The actual rate of disease progression. This is to unify the unit of measurement to the month.

[0099] The average time interval from diagnosis to grounding for the disease is obtained from aviation medical grounding information. The calculated expected diagnosis time and the average time interval from diagnosis to grounding are added together to obtain the total expected grounding time. Future assessment times are set, including predictions for the next six months or the next year. The probability of grounding risk for a single disease is calculated using an exponential survival model. Those skilled in the art will understand that the exponential survival model is used to predict the time, probability, and risk from the observation point to the occurrence of the target event. The calculation formula may be:

[0100] In the above formula, The probability of flight suspension due to a single disease. For future evaluation time, This represents the total expected grounding time.

[0101] Secondly, regarding the calculation of the impairment rate of second driving ability caused by a diagnosed disease: The average time interval from diagnosis to flight suspension for this disease is obtained. The most recent quantitative indicator value for this disease is retrieved from medical examination data and compared with the corresponding medical suspension threshold. The deviation ratio between the current indicator value and the suspension threshold is calculated; a larger deviation ratio indicates a longer expected suspension period. The deviation ratio can be calculated by subtracting the suspension threshold from the current indicator value and then dividing by the suspension threshold. The average time interval is divided by this deviation ratio to obtain the corrected expected suspension time. Using an exponential survival model, the probability of flight suspension for a single disease is calculated based on the future assessment time and the corrected expected suspension time. .

[0102] Thirdly, the comorbidity association strength matrix is ​​used to determine the comorbidity risk-additional items: Comorbidity risk enhancement factors can be identified by screening high-association combinations in the comorbidity association strength matrix where the target disease enhancement exceeds a preset threshold, according to the formula. Calculate, where, As a risk factor for comorbidity, For the target disease enhancement score, 0.1 is the preset addition coefficient. Setting the addition coefficient transforms the dimensionless statistical indicator of enhancement score into a modulatory measure of probability. The preset threshold can be set to 1.2, 2, 3, etc. If there are no highly correlated combinations, Set to 0. The comorbidity risk bonus is obtained by calculating the elevation of all target diseases selected through screening.

[0103] Finally, the probability of impaired overall driving ability is calculated: First, analyze all past medical grounding incidents involving airlines to see how many were caused by a single disease. For example, if hypertension ranks first among all grounding incidents, accounting for 20%, then this disease is generally considered "extremely high-risk" because it can lead to disability, fainting, or serious cardiovascular accidents during flight, directly threatening flight safety. Based on this percentage, a weight can be assigned to this disease. This weighting of pilot ability can be based on this percentage or manually assigned based on statistical data. For example, assign a weight of 0.7 to the top 10% of diseases, and a weight of 0.2 to the 10%-20% of diseases, etc. The probability of grounding due to a single disease is then weighted according to the impact of pilot ability. Weighted summation is performed, and the probability of impaired overall driving ability is calculated by using the comorbidity association strength matrix to determine the comorbidity synergistic risk additive term. The final formula for calculating the probability of impaired overall driving ability is as follows:

[0104] In the above formula, To assess the probability of impaired driving ability, The weighting of driving ability corresponding to a single disease. The probability of flight suspension due to a single disease. You can choose 1 or 2. For the types of diseases, This is an added risk factor for comorbidities.

[0105] As can be seen, in this optional embodiment, the mission intensity factor is determined through flight mission data, thereby obtaining the mission intensity index; by discarding the practice of directly using the population average disease rate to assess all individuals, a disease acceleration factor is introduced to accurately quantify the pilot's recent high-intensity, cross-time zone, and other specific flight mission loads as an accelerating effect on disease progression, thus obtaining a personalized individual disease progression rate; the first pilot capability impairment rate is calculated using the disease acceleration factor, and the second pilot capability impairment rate is calculated based on aviation medical grounding information; finally, the comprehensive pilot capability impairment rate is obtained by combining the comorbidity association strength matrix; this can make risk capture more sensitive and improve the accuracy of predicting aircraft pilot flight risks.

[0106] In yet another optional embodiment, the method may further include: Based on the four-dimensional relational database, a historical sample dataset is obtained; the overall pilot capability impairment rate of all pilots in the historical sample dataset is calculated to obtain the overall pilot capability impairment rate dataset. Based on aviation medical grounding information, a grounding tag dataset was determined; The driving ability impairment rate dataset is sorted to obtain an ordered probability sequence; based on the ordered probability sequence, candidate judgment thresholds are determined. Based on the candidate decision threshold and the grounding label dataset, the first grounding rate and the second grounding rate are obtained; Using the first grounding rate as the vertical axis and the second grounding rate as the horizontal axis, the prediction curve is obtained. Based on the prediction curve, the first and second decision thresholds are obtained.

[0107] In this optional embodiment, the historical sample dataset can be data from a pilot four-dimensional correlation database of the previous ten years, or data from the previous 20 years; all subsequent data are known data, including the results of whether pilots in the historical sample dataset were grounded. The overall pilot capability impairment rate of the pilots in the historical sample dataset is calculated to obtain the overall pilot capability impairment rate dataset.

[0108] In this optional embodiment, the grounding tag dataset can be a dataset of all pilots in the historical sample dataset who were subsequently grounded; the grounded tag can be 1, and the non-grounded tag can be 0.

[0109] In this optional embodiment, the candidate decision threshold can be obtained by sorting the data in the pilot capability impairment rate dataset from largest to smallest to obtain an ordered probability sequence. The candidate decision threshold can then be all the data in the ordered probability sequence. Furthermore, to eliminate the extreme case where none of the pilots are grounded, and to ensure the integrity of the threshold sequence and the subsequent calculation of the ROC curve, the candidate decision threshold also needs to include 1. For example, with 5 samples... The values ​​are 0.92, 0.78, 0.45, 0.23, and 0.07 respectively, so the candidate decision threshold set is {1.0, 0.92, 0.78, 0.45, 0.23, 0.07}. Pilots whose scores are greater than or equal to the candidate decision threshold are judged to be grounded, while those whose scores are less than the candidate decision threshold are judged not to be grounded. This is then compared with the actual grounding labels in the grounding label dataset to calculate the first and second grounding rates for that threshold. By trying all possible candidate decision thresholds, the ROC curve, or prediction curve, can be plotted.

[0110] In this optional embodiment, the first grounding rate is used to describe the percentage of pilots who correctly predicted the flight among all grounded pilots; the second grounding rate is used to describe the percentage of pilots who incorrectly predicted the flight among all non-grounded pilots. In this optional embodiment, the prediction curve can be calculated as follows: For each candidate decision threshold, perform the following operations: Based on the candidate determination threshold, the predicted number of grounded passengers and the predicted number of passengers who will not fly are predicted. Compared with the actual grounded passenger labels, four values ​​are calculated: the number of passengers who are actually grounded and are predicted to be grounded is denoted as TP; the number of passengers who are actually grounded but are predicted to not fly is denoted as FN; the number of passengers who are actually not flying but are predicted to be grounded is denoted as FP; and the number of passengers who are actually not flying and are predicted to not fly is denoted as TN.

[0111] The first grounding rate and the second grounding rate are calculated based on four values. The first grounding rate (TP / (TP+FN)) represents the percentage of those who are actually grounded and are correctly identified; the higher this value, the fewer people are missed, and the safer the system. The second grounding rate (FP / (FP+TN)) represents the percentage of those who are actually not grounded but are incorrectly identified as grounded; the lower this value, the fewer people are wrongly accused, and the less disruption to operations.

[0112] Plot the first and second deactivation rates corresponding to each candidate decision threshold on a coordinate system. Plot the second deactivation rate on the horizontal axis (range 0 to 1) and the first deactivation rate on the vertical axis (range 0 to 1). Connect all these points with a line to obtain the ROC curve, also known as the prediction curve. Those skilled in the art will understand that the ROC curve is used to determine the strength of a classification model's ability to distinguish between diseased / healthy or abnormal / normal conditions, demonstrating the model's accuracy and ability to avoid misclassification at different thresholds.

[0113] The process involves determining whether the Area Under the Curve (AUC) of the predicted curve meets preset requirements. If it does, a judgment threshold can be determined using the predicted curve. These preset requirements might be that the AUC is greater than 0.85 or 0.9, indicating a strong discriminative ability of the predicted curve to distinguish between grounding indicators; this also suggests that the calculated overall flight capability impairment rate is valid. If the requirements are not met, the calculated data is flawed, and data needs to be collected and recalculated until the requirements are met.

[0114] In this optional embodiment, the first determination threshold and the second determination threshold can be determined as follows: To ensure no pilots who would actually be grounded are overlooked, the algorithm starts from the point on the ROC curve with the highest first grounding rate and moves along the curve towards a decreasing first grounding rate. It finds the point where the first grounding rate first reaches a preset probability requirement, for example, 95%. If the preset probability requirement is 95%, the first grounding rate could be 96%, 97%, etc. The candidate threshold corresponding to the point with a first grounding rate of 96% is the first judgment threshold. This value serves as the boundary for determining whether a pilot is unfit to continue flying, ensuring that high-risk pilots are not missed. To avoid classifying healthy pilots as unfit to fly, the algorithm starts from the point on the ROC curve with the second lowest grounding rate and moves along the curve towards a increasing second grounding rate. It finds the point where the second grounding rate first reaches a preset low level requirement, for example, no more than 10%. If the second grounding rate is 7%, 8%, or 9%, the candidate threshold corresponding to the point with a second grounding rate of 9% is the second judgment threshold. This value serves as the boundary for determining whether a pilot is fit to continue flying, ensuring that healthy pilots are not incorrectly restricted.

[0115] In this optional embodiment, the first judgment threshold is greater than the second judgment threshold because the purpose of the first judgment threshold is to ensure that no pilots who are truly likely to be grounded are missed, requiring a high first grounding rate; in order to screen out all those who are truly at risk, the threshold must be set relatively high; the higher the threshold, the more accurate the identification of high-risk individuals. The purpose of the second judgment threshold is to minimize the involvement of healthy pilots, requiring a low false positive rate, i.e., a low second grounding rate; in order to ensure that only those with extremely low risk are identified as safe, the threshold must be set relatively low; the lower the threshold, the lower the possibility of misjudgment.

[0116] Furthermore, the first and second decision thresholds can also be determined in the following way: when it is determined that the area under the curve (AUC) of the prediction curve meets the preset requirements, two thresholds are manually set according to the different standards of the airlines to determine whether to continue flying; for example, if the airline has very strict requirements and wants to add a safe range above the standard, then the first decision threshold can be manually set to 0.95 and the second decision threshold can be manually set to 0.20; if the airline's requirements are standard requirements, then the first decision threshold can be manually set to 0.85 and the second decision threshold can be manually set to 0.40.

[0117] As can be seen, in this optional embodiment, a batch of historical pilot sample datasets is obtained through a four-dimensional correlation database, and a comprehensive pilot capability impairment rate dataset is calculated; then, a grounding tag dataset is determined based on aviation medical grounding information; the first grounding rate and the second grounding rate are calculated using these two datasets, and a prediction curve is established; the prediction curve is judged to see if the calculated comprehensive pilot capability impairment rate meets the requirements; finally, a first judgment threshold and a second judgment threshold can be set according to the prediction curve, or the first judgment threshold and the second judgment threshold can be manually set when the prediction curve meets the requirements; by setting the first judgment threshold and the second judgment threshold when the prediction curve meets the requirements, the accuracy of predicting the risk of aircraft pilots flying aircraft can be improved.

[0118] In another optional embodiment, determining the overall driving capability impairment rate and identifying the risk prediction result may include: Based on the assessment of the pilot's current flight mission data, a threshold correction coefficient is determined; the threshold correction coefficient is used to correct the first decision threshold and the second decision threshold to obtain the first dynamic threshold and the second dynamic threshold. When it is determined that the overall driving ability impairment rate is greater than the first dynamic threshold, the determination result will be marked as unsuitable for continued driving. When it is determined that the overall driving ability impairment rate is less than the second dynamic threshold, the determination result is marked as suitable for continued driving. When it is determined that the overall driving ability impairment rate is between the first dynamic threshold and the second dynamic threshold, the determination result will be marked as a restricted driving label.

[0119] In this optional embodiment, the first dynamic threshold may be obtained by correcting the first decision threshold using a threshold correction coefficient, and the second dynamic threshold may be obtained by correcting the second decision threshold using a threshold correction coefficient.

[0120] In this optional embodiment, the threshold correction coefficient can be obtained in the following way: The system obtains the number of consecutive days the pilot has rested up to the assessment date and the average monthly flight hours over the past 12 months from flight mission data. This means the pilot's average monthly flight hours over the past 12 months are used to determine whether the pilot is in a rest period. If the current consecutive rest days reach a preset number of days, for example, a preset number of days greater than or equal to 5 days, it is determined to be a rest period. The threshold correction coefficient can be set to 1.10. Since the pilot has sufficient rest time, the pilot's flying risk will be reduced. Therefore, the threshold can be increased to reduce the probability of the pilot being judged as unsuitable to continue flying or restricted from flying.

[0121] If the rest period conditions are not met, the load level is determined based on the average monthly flight hours over the past 12 months. If the average monthly flight hours are below the first preset value, it is considered a low-load period, with a threshold correction factor of 1.10, slightly relaxing the standard. If the average monthly flight hours are between the first and second preset values, it is considered a normal load period, with a threshold correction factor of 1.00, maintaining a fixed threshold. If the average monthly flight hours are between the second and third preset values, it is considered a high-load period, with a threshold correction factor of 0.85, tightening the standard and strengthening early warnings. If the average monthly flight hours exceed the third preset value, or the longest consecutive duty days in the past 6 months exceed the preset number of days, it is considered an extremely high-load period, with a threshold correction factor of 0.70, ensuring a safety margin during high-intensity mission periods. The first, second, and third preset values ​​are set with reference to civil aviation regulations regarding pilot flight time limits and actual industry operating data; for example, the first preset value can be 50 hours per month, the second preset value can be 80 hours per month, and the third preset value can be 100 hours per month.

[0122] In this optional embodiment, when it is determined that the overall driving ability impairment rate is greater than the first dynamic threshold, the determination result is marked as unsuitable for continuing to drive; when it is determined that the overall driving ability impairment rate is less than the second dynamic threshold, the determination result is marked as suitable for continuing to drive; when it is determined that the overall driving ability impairment rate is between the first dynamic threshold and the second dynamic threshold, the determination result is marked as restricted driving.

[0123] As can be seen, in this optional embodiment, by statistically processing the flight mission data, a correction coefficient for the threshold can be obtained; after correcting the first and second judgment thresholds using the correction coefficient, the overall flight capability impairment rate is then judged to obtain the risk prediction result, which can improve the accuracy of the prediction result.

[0124] In yet another optional embodiment, the method may further include: Based on flight mission data, the circadian rhythm disruption factor is determined; based on the circadian rhythm disruption factor, the circadian rhythm disruption index is obtained. The circadian rhythm disruption index is input into a pre-trained disease exacerbation probability prediction model to obtain the predicted exacerbation probability value for the target disease. The disease exacerbation probability prediction model is trained using a Logistic regression model, wherein the training formula for the Logistic regression model is:

[0125] In the above formula, The probability of annual deterioration. For the intercept parameter, This is the effect coefficient. The index of circadian rhythm disruption; Based on flight mission data, the target route type is obtained; the target route type is compared with a preset route disease database to obtain a list of targeted monitoring diseases; the list of targeted monitoring diseases, all predicted deterioration probabilities, and the list of rising diseases are cross-compared to obtain a sorted list of diseases. When the risk prediction result is determined to be a restricted driving result, the sorted disease list is matched and filtered to obtain a risk pretreatment plan; the risk pretreatment plan is used to describe the temporary execution of disease risk tasks that are highly correlated with the route types in the sorted disease list.

[0126] In this optional embodiment, the circadian rhythm disruption factor is used to describe the degree of circadian rhythm disorder caused by the pilot's occupational characteristics; the circadian rhythm disruption index is used to measure the comprehensive index of the loss of consistency between the pilot's internal biological clock and the external environmental time due to occupational exposure. In this optional embodiment, the circadian rhythm disruption factor can be determined in the following way: it can be selected from the flight mission data by the target pilot's cross-time zone frequency density (the number of times crossing three or more time zones in the past 12 months divided by 12), night flight duration ratio (the number of night flight hours in the past 12 months divided by the total number of flight hours), cross-International Date Line frequency (the number of flights crossing the International Date Line in the past 12 months), duty rhythm irregularity (the standard deviation of take-off and landing times in the past 6 months), etc. These are the circadian rhythm disruption factors.

[0127] In this optional embodiment, the circadian rhythm disruption index can be calculated as follows: Min-Max normalization is performed on the circadian rhythm disruption factor data. Those skilled in the art will know that Min-Max normalization is a method of linearly mapping data to the 0-1 interval. After normalization, the data is weighted and summed according to preset weights, then multiplied by 10 to map to the 0-10 interval, yielding the circadian rhythm disruption index. The preset weights can be obtained by performing a multiple linear regression with changes in metabolic indicators of the disease in the population data as the dependent variable and circadian rhythm disruption factors as independent variables, and calculating the original regression coefficients using the least squares method. The changes in metabolic indicators are quantitative indicators that objectively reflect metabolic health status and are used as the dependent variable, such as changes in blood uric acid, triglyceride, or fasting blood glucose levels. Because different circadian rhythm disruptors have different dimensions, the original regression coefficients cannot be directly compared. To eliminate the influence of dimensions, the original regression coefficients are converted into standardized regression coefficients. The standardized regression coefficient is calculated using the formula: Standardized coefficient = Original regression coefficient × (Standard deviation of the independent variable ÷ Standard deviation of the dependent variable). The standardized regression coefficient describes how many standard deviations the dependent variable changes when the independent variable changes by one standard deviation. Because "standard deviation" is used as the unified measure, the standardized regression coefficients can be directly compared between independent variables with different dimensions. Standardized regression coefficients can be positive or negative. A positive sign indicates that an increase in the circadian rhythm disruptor also increases metabolic indicators, while a negative sign indicates that an increase in the circadian rhythm disruptor actually decreases metabolic indicators. However, only the intensity of the effect is considered for each factor, not its direction. Therefore, the absolute value of the standardized regression coefficient is used to eliminate the influence of the positive or negative sign. Finally, the different standardized regression coefficients are normalized so that their sum equals 1. The calculated value at this point is the preset weight.

[0128] In this optional embodiment, the training method for the disease exacerbation probability prediction model can be as follows: the baseline model used for training is a Logistic regression model, and the training formula for the model is the annual exacerbation probability. ,in, The probability of annual deterioration. For the intercept parameter, This is the effect coefficient. The index of circadian rhythm disruption; and It is obtained by fitting the data to the training sample data using the maximum likelihood estimation method. Indicates when When the value is zero, it represents the logarithmic probability of the disease worsening. The larger the value, the higher the risk of the disease itself worsening in the absence of any rhythm disturbances; This indicates how much the log odds of the risk of disease worsening increases for every unit increase in the circadian rhythm disruption index; The higher the value, the more sensitive the disease is to circadian rhythm disruption, and the stronger the effect of rhythm disruption on the probability of disease deterioration. For each disease in the rising disease list, an independent Logistic regression model is established to prevent influence between different diseases. The input of the model is the pilot's circadian rhythm disruption index value, and the output is the probability of the disease deteriorating in the following year.

[0129] The training sample data was constructed as follows: Pilot samples with at least two complete annual physical examination records, with a one-year interval between the two examinations, were extracted from the population data. For each pilot sample, the circadian rhythm disruption index value at the time of the base year's physical examination was calculated as the input variable, and the output label was determined whether the target disease worsened at the time of the next year's physical examination. Worsening was defined as: for diseases with continuous quantitative indicators, the indicator value changing in an abnormal direction with a change exceeding a preset clinical significance threshold; for diseases with diagnostic conclusions, the change from an undiagnosed state in the base year to a diagnosed state in the next year. The output label value of 1 indicates worsening, and 0 indicates no worsening. For each disease in the rising disease list, the disease worsening probability prediction model was trained using the constructed training sample data, yielding different results. and This allows us to obtain different Logistic regression models for each disease in the rising disease list, and then obtain a trained disease exacerbation probability prediction model.

[0130] In this optional embodiment, the target route type can be determined as follows: obtaining the main route types for the next three months, the average daily number of flight segments for the next three months, the percentage of high-altitude airports for the next three months, the percentage of nighttime departures for the next three months, and the percentage of long-haul routes for the next three months; routes with a long-haul percentage exceeding a preset long-haul threshold are classified as international long-haul routes, routes with a high-altitude airport percentage exceeding a preset high-altitude airport threshold are classified as high-altitude routes, routes with an average daily number of flight segments exceeding a preset high-frequency threshold are classified as high-frequency short-haul routes, and routes with a nighttime departure percentage exceeding a preset nighttime cargo threshold are classified as nighttime cargo routes. The long-haul threshold, high-altitude airport threshold, high-frequency threshold, and nighttime cargo threshold can all be manually set; for example, their thresholds can be determined by sorting all pilots' data on long-haul thresholds, high-altitude airport thresholds, high-frequency thresholds, and nighttime cargo thresholds, and taking the top 20% as their respective thresholds.

[0131] In this optional embodiment, the process of constructing the route disease database can be as follows: In each sub-cross-layer structure data, pilots are grouped according to their target route type, and the prevalence of each disease under each route type group is statistically analyzed; the ratio of the prevalence of each group to the total prevalence of the industry is calculated, and diseases with a ratio exceeding a preset threshold are included in the targeted disease monitoring list for that route type. The higher the ratio, the stronger the association between the disease and the route type; the preset threshold can be a manually set value, such as 1.5, 1.8, or 2. Further stratification of the sub-cross-layer structure data is to control confounding factors; because age and years of flight experience are fundamental factors affecting disease prevalence, if the cross-layer stratification is ignored and statistics are directly performed by route type grouping, the statistical differences in prevalence may partly come from the route type and partly from the different distributions of age and years of flight experience among the groups; by first locking age and years of flight experience within the cross-layer stratification, and then comparing the differences in prevalence of different route types, these differences can only be caused by the route type itself, eliminating the interference of fundamental factors.

[0132] In this optional embodiment, the disease list can be sorted as follows: diseases appearing simultaneously in both the targeted monitoring disease list and the rising disease list are designated as the highest priority; these diseases are highly relevant to upcoming flight missions and show a continuous worsening trend among pilots, making them the most pressing targets for monitoring. These diseases are then sorted from highest to lowest according to their predicted deterioration probability values ​​from the disease deterioration probability prediction model. A higher annual deterioration probability indicates a greater likelihood of the disease worsening within the next year, thus emphasizing the urgency of monitoring.

[0133] Diseases that only appear in the targeted surveillance disease list but are no longer in the rising disease list are classified as secondary priority; these diseases are related to the current flight mission, but do not show a sustained upward trend at the population level, and their surveillance urgency is lower than the highest priority.

[0134] Diseases that appear only on the list of rising diseases but not on the list of targeted surveillance diseases are defined as general priority; these diseases, although showing a continuous worsening trend in the population, are not directly related to the current flight mission and are subject to routine surveillance.

[0135] In this optional embodiment, the risk pretreatment scheme may be that when the risk prediction result is determined to be a restricted driving result, the highest priority disease in the ranked disease list should be dealt with in a timely manner to try to intervene in the budding stage of the disease. At the same time, the pilot should be prevented from engaging in flight missions related to such diseases again in the near future, and such flight missions should be carried out after the body has been fully recovered.

[0136] It is evident that by establishing different disease exacerbation probability prediction models for diseases in the rising disease list, a predicted exacerbation probability value for each disease in the rising disease list can be obtained. By cross-comparing the targeted monitoring disease list, the predicted exacerbation probability values, and the rising disease list, a ranked disease list is obtained. When the risk prediction result is a restricted driving outcome, the ranked disease list is matched and filtered to obtain the corresponding risk pretreatment plan. By formulating a risk pretreatment plan, impending diseases can be dealt with, the prediction results can be refined, and the accuracy of pilots' risk assessment for flying aircraft can be improved.

[0137] Example 2 Please see Figure 2 , Figure 2 This invention discloses a pilot flight risk prediction device based on big data, which may include: The data acquisition module 201 is used to acquire a four-dimensional relational database; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data; the time coordinate data includes time data corresponding to the physical examination date and time data corresponding to the flight mission date; the health information parameter data includes physical examination data information, disease development pattern parameters, and aviation medical grounding information. The age correction module 202 is used to obtain an exposure offset coefficient based on the four-dimensional relational database; the exposure offset coefficient is used to correct the flight exposure factor; the flight exposure factor is used to describe the different load data of the pilot when performing the mission within the target time; based on the exposure offset coefficient, the actual age data of the pilot is calculated to obtain physiological age data. The risk prediction module 203 is used to obtain physiological disease data in the four-dimensional correlation database based on the physiological age data; to obtain the comprehensive pilot's impairment rate based on the physiological disease data and the flight mission data; and to determine the risk prediction result by judging the comprehensive pilot's impairment rate.

[0138] As can be seen, the embodiments of the present invention construct a four-dimensional relational database using pilot file data, time coordinate data, health information parameter data, and flight mission data; an exposure offset coefficient is calculated using the four-dimensional relational database; the exposure offset coefficient is used to correct different load data of the pilot performing missions within the target time; the corrected flight exposure factor is used to calculate the pilot's actual age data to obtain physiological age data; using the physiological age data as the pilot's physical age for calculation makes the obtained physiological disease data more consistent with the pilot's current condition, thereby making the calculated comprehensive pilot's impairment rate more accurate, and thus improving the accuracy of the pilot's flight risk prediction results.

[0139] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a device including a memory and a processor, as disclosed in an embodiment of the present invention. Figure 3 As shown, the device including memory and processor may include: Memory 301 storing executable program code; Processor 302 coupled to memory 301; The processor 302 calls the executable program code stored in the memory 301 to execute some or all of the steps in any of the big data-based pilot flight risk prediction methods in Embodiment 1 of the present invention.

[0140] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0141] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0142] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting pilot flight risks based on big data, characterized in that, The method includes: A four-dimensional relational database is obtained; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data; the time coordinate data includes the time data corresponding to the physical examination date and the time data corresponding to the flight mission date; the health information parameter data includes physical examination data, disease development pattern parameters, and aviation medical grounding information; When it is determined that the flight exposure factor value is less than the preset low exposure factor threshold, a low-exposure pilot group is obtained based on the four-dimensional association database. The prevalence of low-exposure diseases among adjacent age groups in the low-exposure pilot group is calculated to obtain the natural prevalence rate growth rate; a benchmark conversion factor is obtained based on the natural prevalence rate growth rate. Based on the health information parameter data and the flight exposure factor, a multiple linear regression model was constructed and the regression coefficient of the exposure factor was obtained by calculation using the least squares method; Divide the regression coefficient of the exposure factor by the benchmark conversion coefficient to obtain the target disease offset coefficient; Based on all the target disease offset coefficients, an exposure offset coefficient is obtained; the exposure offset coefficient is used to correct the flight exposure factor; the flight exposure factor is used to describe the different load data of the pilot performing the mission within the target time. Based on the exposure offset coefficient, the actual age data of the pilots is calculated to obtain physiological age data; based on the physiological age data, physiological disease data in the four-dimensional association database is obtained; based on the flight mission data, the mission intensity factor is determined; based on the mission intensity factor, the mission intensity index is obtained. The task intensity index is matched with a preset acceleration factor mapping table to obtain the disease acceleration factor; the actual disease progression rate is obtained based on the physiological disease data and the disease acceleration factor. When it is determined that the disease has not been diagnosed, the expected diagnosis time is obtained based on the actual rate of disease progression; the average time interval is obtained, which is the average time interval from disease diagnosis to flight suspension obtained from aviation medical flight suspension information; the expected diagnosis time is added to the average time interval to obtain the total expected flight suspension time; the total expected flight suspension time and the preset future assessment time are calculated using an exponential generation model to obtain the first pilot capability impairment rate; When a disease is diagnosed, the quantitative index value of the most recent test for the disease is obtained based on the physical examination data; the flight suspension threshold for the disease is obtained based on the aviation medical flight suspension information; the deviation ratio is obtained based on the quantitative index value and the flight suspension threshold; the corrected expected flight suspension time is obtained by dividing the average time interval by the deviation ratio; the corrected expected flight suspension time and the preset future assessment time are calculated using an exponential generation model to obtain the second pilot capability impairment rate. Obtain pilot medical examination data; divide the pilot medical examination data according to age as the vertical axis and years of flight experience as the horizontal axis to obtain cross-layered structured data; the cross-layered structured data includes several sub-cross-layered structured data; perform the following operations on each sub-cross-layered structured data: Based on the number of people suffering from both the target disease and other diseases, the conditional probability between the target disease and other diseases is obtained; based on the conditional probability, the lift of the target disease is obtained; the lift of the target disease is used to represent the degree of association between the target disease and other diseases; based on all the lifts of the target disease, the sub-cross layer association strength matrix is ​​obtained. Based on the correlation strength matrices of all the sub-cross layers, the co-disease correlation strength matrix is ​​obtained; In the comorbidity association strength matrix, association elevations greater than a preset elevation threshold are selected. These association elevations are then calculated using an additive term formula to obtain a comorbidity synergistic risk additive term. The specific calculation method for the additive term formula is as follows: In the above formula, As a risk factor for comorbidity, To enhance the correlation, 0.1 is the preset bonus coefficient; The impact weight of flight capability is obtained, which is the proportion of the impact of a disease on all medical grounding events, and a corresponding weight is assigned according to the proportion of the impact. The first driving ability impairment rate and / or the second driving ability impairment rate are weighted and fused based on the driving ability influence weight, and the comprehensive driving ability impairment rate is obtained by constraining the comorbidity risk bonus term. The overall driving capability impairment rate is assessed to determine the risk prediction result.

2. The pilot flight risk prediction method based on big data according to claim 1, characterized in that, The disease development pattern parameters include a list of rising diseases, which is obtained through the following methods: Perform the following operations on each of the sub-cross-layer structure data: For each target disease, the prevalence rate of the target disease is obtained; the relative change rate of the prevalence rate of the target disease among the target pilots in two consecutive years is calculated to obtain the annual change rate of the target disease. The target pilots were selected from those born in the same year based on the pilot profile data. The prevalence rate and annual change rate of the target disease corresponding to each of the sub-cross-layer structure data are filtered according to preset screening conditions to obtain a list of rising diseases.

3. The pilot flight risk prediction method based on big data according to claim 2, characterized in that, The method further includes: Based on the four-dimensional relational database, a historical sample dataset is obtained; the overall pilot capability impairment rate of all pilots in the historical sample dataset is calculated to obtain an overall pilot capability impairment rate dataset. Based on the aforementioned aviation medical grounding information, a grounding tag dataset was determined; The driving ability impairment rate dataset is sorted to obtain an ordered probability sequence; candidate judgment thresholds are determined based on the ordered probability sequence. Based on the candidate decision threshold and the grounding label dataset, a first grounding rate and a second grounding rate are obtained; the first grounding rate is used to describe the ratio of correctly predicted pilots among all grounded pilots; the second grounding rate is used to describe the ratio of incorrectly predicted pilots among all non-grounded pilots. Using the first grounding rate as the vertical axis and the second grounding rate as the horizontal axis, a prediction curve is obtained. Based on the prediction curve, a first determination threshold and a second determination threshold are obtained; wherein, the first determination threshold is greater than the second determination threshold.

4. The pilot flight risk prediction method based on big data according to claim 3, characterized in that, The process of determining the overall driving capability impairment rate and identifying the risk prediction result includes: Based on the current flight mission data of the assessed pilot, a threshold correction coefficient is determined; the threshold correction coefficient is used to correct the first judgment threshold and the second judgment threshold to obtain a first dynamic threshold and a second dynamic threshold. When it is determined that the overall driving ability impairment rate is greater than the first dynamic threshold, the determination result is marked as unsuitable for continued driving. When it is determined that the overall driving ability impairment rate is less than the second dynamic threshold, the determination result is marked as suitable for continued driving. When it is determined that the overall driving ability impairment rate is between the first dynamic threshold and the second dynamic threshold, the determination result is marked as a restricted driving label.

5. The pilot flight risk prediction method based on big data according to claim 4, characterized in that, The method further includes: Based on the flight mission data, the circadian rhythm disruption factor is determined; the circadian rhythm disruption factor is used to describe the degree of circadian rhythm disorder caused by the pilot's occupational characteristics. Based on the aforementioned circadian rhythm disruption factors, a circadian rhythm disruption index is obtained; the circadian rhythm disruption index is used as a comprehensive indicator to measure the loss of consistency between the pilot's internal biological clock and the external environmental time due to occupational exposure. The circadian rhythm disruption index is input into a pre-trained disease exacerbation probability prediction model to obtain the predicted exacerbation probability value corresponding to the target disease; the disease exacerbation probability prediction model is trained using a Logistic regression model, wherein the training formula of the Logistic regression model is: In the above formula, The probability of annual deterioration. For the intercept parameter, This is the effect coefficient. The index of circadian rhythm disruption; Based on the flight mission data, the target route type is obtained; the target route type is compared with a preset route disease database to obtain a targeted monitoring disease list; the targeted monitoring disease list, all the predicted deterioration probability values, and the rising disease list are cross-compared to obtain a sorted disease list. When the risk prediction result is determined to be a restricted driving result, the sorted disease list is matched and filtered to obtain a risk preprocessing plan; wherein, the risk preprocessing plan is used to describe the temporary execution of disease risk tasks that are highly correlated with the route types in the sorted disease list.

6. A pilot flight risk prediction device based on big data, characterized in that, The device includes: The data acquisition module is used to acquire a four-dimensional relational database; the four-dimensional relational database includes pilot file data, time coordinate data, health information parameter data, and flight mission data; the time coordinate data includes time data corresponding to the physical examination date and time data corresponding to the flight mission date; the health information parameter data includes physical examination data, disease development pattern parameters, and aviation medical grounding information; The age correction module is used to obtain a low-exposure pilot group based on the four-dimensional association database when the flight exposure factor value is determined to be less than the preset low exposure factor threshold. The prevalence of low-exposure diseases among adjacent age groups in the low-exposure pilot group is calculated to obtain the natural prevalence rate growth rate; a benchmark conversion factor is obtained based on the natural prevalence rate growth rate. Based on the health information parameter data and the flight exposure factor, a multiple linear regression model was constructed and the regression coefficient of the exposure factor was obtained by calculation using the least squares method; Divide the regression coefficient of the exposure factor by the benchmark conversion coefficient to obtain the target disease offset coefficient; Based on all the target disease offset coefficients, an exposure offset coefficient is obtained; the exposure offset coefficient is used to correct the flight exposure factor; the flight exposure factor is used to describe the different load data of the pilot performing the mission within the target time. Based on the exposure offset coefficient, the physiological age data of the pilots is calculated. The risk prediction module is used to obtain physiological disease data in the four-dimensional correlation database based on the physiological age data; determine the mission intensity factor based on the flight mission data; and obtain the mission intensity index based on the mission intensity factor. The task intensity index is matched with a preset acceleration factor mapping table to obtain the disease acceleration factor; the actual disease progression rate is obtained based on the physiological disease data and the disease acceleration factor. When it is determined that the disease has not been diagnosed, the expected diagnosis time is obtained based on the actual rate of disease progression; the average time interval is obtained, which is the average time interval from disease diagnosis to flight suspension obtained from aviation medical flight suspension information; the expected diagnosis time is added to the average time interval to obtain the total expected flight suspension time; the total expected flight suspension time and the preset future assessment time are calculated using an exponential generation model to obtain the first pilot capability impairment rate; When a disease is diagnosed, the quantitative index value of the most recent test for the disease is obtained based on the physical examination data; the flight suspension threshold for the disease is obtained based on the aviation medical flight suspension information; the deviation ratio is obtained based on the quantitative index value and the flight suspension threshold; the corrected expected flight suspension time is obtained by dividing the average time interval by the deviation ratio; the corrected expected flight suspension time and the preset future assessment time are calculated using an exponential generation model to obtain the second pilot capability impairment rate. Obtain pilot medical examination data; divide the pilot medical examination data according to age as the vertical axis and years of flight experience as the horizontal axis to obtain cross-layered structured data; the cross-layered structured data includes several sub-cross-layered structured data; perform the following operations on each sub-cross-layered structured data: Based on the number of people suffering from both the target disease and other diseases, the conditional probability between the target disease and other diseases is obtained; based on the conditional probability, the lift of the target disease is obtained; the lift of the target disease is used to represent the degree of association between the target disease and other diseases; based on all the lifts of the target disease, the sub-cross layer association strength matrix is ​​obtained. Based on the correlation strength matrices of all the sub-cross layers, the co-disease correlation strength matrix is ​​obtained; In the comorbidity association strength matrix, association elevations greater than a preset elevation threshold are selected. These association elevations are then calculated using an additive term formula to obtain a comorbidity synergistic risk additive term. The specific calculation method for the additive term formula is as follows: In the above formula, As a risk factor for comorbidity, To enhance the correlation, 0.1 is the preset bonus coefficient; The impact weight of flight capability is obtained, which is the proportion of the impact of a disease on all medical grounding events, and a corresponding weight is assigned according to the proportion of the impact. The first driving ability impairment rate and / or the second driving ability impairment rate are weighted and fused based on the driving ability influence weight, and the comprehensive driving ability impairment rate is obtained by constraining the comorbidity risk bonus term. The overall driving capability impairment rate is assessed to determine the risk prediction result.

7. An apparatus comprising a memory and a processor, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the pilot flight risk prediction method based on big data as described in any one of claims 1-5.

8. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by a processor, are used to execute the big data-based pilot flight risk prediction method as described in any one of claims 1-5.

Images