Civil aviation passenger demand forecasting method and system based on artificial intelligence

Abstract

The present invention discloses a civil aviation passenger demand forecasting method and system based on artificial intelligence. The method includes data collection, abnormal demand elimination, civil aviation demand side sampling, establishment of a civil aviation passenger demand forecasting model, and civil aviation passenger demand forecasting. The present invention belongs to the field of civil aviation demand forecasting, and specifically refers to a civil aviation passenger demand forecasting method and system based on artificial intelligence. This solution introduces time-series weighted distance to ensure that samples with off-season and low-demand characteristics are not mistakenly eliminated; based on route similarity weighting, trustworthy neighbors are generated to supplement scarce low-demand samples in a targeted manner; feature importance scores are introduced to calculate feature sensitivity scores of civil aviation samples, and the impact of feature-time points on demand forecasting is evaluated from different angles; the ability to adapt to holiday peaks and extreme weather scenarios is guided by a penalty mechanism; based on the introduction of holiday-related weights to enhance the scheduling strategy, the penalty weight growth rate is dynamically increased, thereby improving the reliability of civil aviation passenger demand forecasting.

Description

Technical Field

[0001] The present invention relates to the technical field of civil aviation demand forecasting, and specifically to a civil aviation passenger demand forecasting method and system based on artificial intelligence. Background Art

[0002] Civil aviation passenger demand forecasting methods analyze historical travel data, passenger behavior characteristics, and real-time flight information, combining big data to predict future passenger flow trends in order to optimize flight schedules and improve service matching. However, typical civil aviation passenger demand forecasting methods are sensitive to abnormal events and easily interfered with by extreme values. They can easily mistake small, normal fluctuations before and after seasonality for noise, mixing up vastly different route data. The resulting pseudo-civil aviation samples lack real meaning, leading to poor demand forecast accuracy. Typical civil aviation passenger demand forecasting methods also have weak forecasting capabilities for extreme scenarios such as holiday peaks and abnormal weather, failing to capture these periodic, concentrated surges, resulting in poor demand forecast reliability. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a civil aviation passenger demand forecasting method and system based on artificial intelligence. In view of the fact that the general civil aviation passenger demand forecasting method is sensitive to abnormal events, easily disturbed by extreme values, and easily mistakes small normal fluctuations before and after seasonality as noise, mixes together route data with huge differences, and generates pseudo civil aviation samples that have no real meaning, which leads to poor demand forecasting accuracy, this solution introduces time-series weighted distance to reduce the probability of eliminating points with similar time and normal fluctuations, retains the real fluctuations before and after holidays, and ensures that civil aviation samples with off-season and low-demand characteristics are not mistakenly eliminated; based on route similarity weighting, trustworthy neighbors are generated to supplement scarce low-demand civil aviation in a targeted manner. Samples; thereby improving the accuracy of civil aviation passenger demand forecasts; in view of the fact that general civil aviation passenger demand forecasting methods have weak forecasting capabilities for extreme scenarios such as holiday peaks and abnormal weather, and are unable to capture such periodic concentrated sudden increase scenarios, which in turn leads to poor reliability of demand forecasts, this scheme introduces feature importance scores, calculates feature sensitivity scores for civil aviation samples, and combines interaction impact values ​​to evaluate the impact of features-time points on demand forecasts from different angles; guides the adaptability to holiday peaks and extreme weather scenarios through a penalty mechanism; based on the introduction of holiday-related weights to enhance the scheduling strategy, dynamically increases the penalty weight growth rate, and adjusts the penalty coefficient according to the holiday distance and the validation set error ratio; thereby improving the reliability of civil aviation passenger demand forecasts.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an artificial intelligence-based civil aviation passenger demand forecasting method, which includes the following steps:

[0005] Step S1: data collection;

[0006] Step S2: Eliminate abnormal demands;

[0007] Step S3: Supplementary sampling on the civil aviation demand side;

[0008] Step S4: Establish a civil aviation passenger demand forecasting model;

[0009] Step S5: Civil aviation passenger demand forecast.

[0010] Furthermore, in step S1, the data collection is to collect historical civil aviation passenger demand data; mark the demand level as a label; and perform standardization on the collected data to obtain an initial civil aviation sample set.

[0011] Furthermore, in step S2, the abnormal demand elimination is to use fuzzy mean clustering to eliminate abnormal points for all low-demand civil aviation samples and high-demand civil aviation samples, and introduce time series weighted distance to calculate the abnormal value of civil aviation samples. , the formula used is: ; ;in, It is a civil aviation sample With cluster center distance; and Civil Aviation Samples With cluster center time point; is the time decay factor; F is the total feature dimension, f is the dimension index, and i is the civil aviation sample index; and They are and The value of the fth dimension; K is the total number of clusters, j is the cluster index; for The civil aviation samples were eliminated; is the noise threshold.

[0012] Furthermore, in step S3, the civil aviation demand side sampling is to introduce the route similarity weight, perform civil aviation sample sampling, and perform fuzzy mean clustering again on the low demand civil aviation samples and high demand civil aviation samples that have been eliminated by abnormal demand; calculate the route feature similarity , expressed as: ; In each cluster, based on the similarity of route characteristics, the remaining low-demand civil aviation samples and high-demand civil aviation samples are selected as neighbors to generate new civil aviation samples , the formula used is: ;in, and It is the original civil aviation sample; is the generation coefficient; the final civil aviation sample set is obtained.

[0013] Furthermore, in step S4, the establishment of the civil aviation passenger demand forecasting model specifically includes the following:

[0014] Step S41: Model architecture design; based on LSTM, the final civil aviation sample set is split into three types of input: static features, historical time series features, and future exogenous indication sequences; the original feature vector is output for subsequent encoding; the static feature encoder receives the static features and outputs a fixed-length static context vector; the encoder receives the historical time series vector with the static context vector concatenated at each step from a multi-layer LSTM, with a length of T, and outputs a hidden state sequence; the Attention mechanism receives the hidden state sequence and the previous hidden state of the decoder and outputs a context vector; the decoder receives the previous prediction value, static context, current context, and future exogenous indications from a multi-layer LSTM and outputs the decoder hidden state; the output layer receives the decoder hidden state and outputs the passenger flow prediction result for the kth step; the basic loss adopts the mean square error loss function;

[0015] Step S42: Loss introduces feature importance score to evaluate the impact of feature-time point on prediction from different perspectives and calculate the feature sensitivity score of civil aviation samples , use dynamic adjustment of the integration step size, calculate the feature sensitivity score according to the change of the feature, and adjust the integration step size As a dynamic variable, it is adjusted according to the rate of change of the characteristics. ; Feature sensitivity score is expressed as: ;in, is the actual eigenvalue of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic change rate of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic mean; is the value of the kth dimension at time point t; is the passenger flow prediction result; x is the civil aviation sample input feature vector, is the mean eigenvector; is the integral variable; calculate the interaction effect value of the civil aviation sample , used to capture the discrete interaction effects of holidays and weather, expressed as: ; Where F is the total set of all feature-time points; S is the subset of the total set F excluding the feature-time point (k, t); is the passenger flow prediction result that receives the subset as input;

[0016] Step S43: Two-stage training strategy; stage one uses pure mean square error training Round, minimize the mean squared error loss, calculated and normalized for each batch, expressed as: ; Where ImSe is the normalized feature importance score; batch is the number of civil aviation samples in each training batch; G is the normalization constant; and is a weight hyperparameter; in the second stage, importance-guided penalty is used to guide and then train Wheel, minimize ; Where L is the total loss function; is the dynamic penalty coefficient; is the mean square error loss function;

[0017] Step S44: Holiday-related weighted enhanced scheduling; introduce holiday-related weighted enhanced scheduling strategy and define the nearest major holiday distance , expressed as: ; Construct holiday enhancement factor , expressed as: ; Among them, h is the average sampling date of the current batch of civil aviation samples The nearest holiday date; H is the set of all holiday dates; is the magnitude of enhancement; is a smoothing scale; calculate the validation set error ratio of the current training round in the key window of 7 days before and after the holiday , expressed as: ; Construct error driver , expressed as: ; The final weight scheduling is expressed as: ;in, is the validation set loss of the current epoch on the 7-day window before and after the holiday, and g is the current number of training times; is the average loss of the validation set within the window; is the error ratio of the last training validation set; is the error amplification factor; and are the minimum penalty coefficient and the maximum penalty coefficient respectively.

[0018] Furthermore, in step S5, the civil aviation passenger demand forecast is to collect civil aviation passenger demand data in real time and input it into a civil aviation passenger demand forecast model, and use the demand level output by the civil aviation passenger demand forecast model as the civil aviation passenger demand forecast result.

[0019] The civil aviation passenger demand forecasting system based on artificial intelligence provided by the present invention includes a data acquisition module, an abnormal demand elimination module, a civil aviation demand side sample replenishment module, a civil aviation passenger demand forecasting model establishment module and a civil aviation passenger demand forecasting module;

[0020] The data collection module collects historical civil aviation passenger demand data to construct an initial civil aviation sample set;

[0021] The abnormal demand elimination module eliminates abnormal demand records in the initial civil aviation sample set;

[0022] The civil aviation demand-side sample supplement module selects neighbors based on the similarity of route characteristics from the civil aviation samples after the abnormal demand is eliminated to generate new civil aviation samples, thereby obtaining the final civil aviation sample set;

[0023] The civil aviation passenger demand forecasting model building module is based on the final civil aviation sample set, LSTM as the basis, introduces feature importance scoring, and combines holiday-related weights to enhance the scheduling strategy to build a civil aviation passenger demand forecasting model;

[0024] The civil aviation passenger demand prediction module performs civil aviation passenger demand prediction on the real-time collected civil aviation passenger demand data based on the civil aviation passenger demand prediction model.

[0025] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0026] (1) In view of the fact that general civil aviation passenger demand forecasting methods are sensitive to abnormal events and easily disturbed by extreme values, they are prone to mistaking small normal fluctuations before and after seasonality as noise, mixing together route data with huge differences, and generating pseudo civil aviation samples that have no real meaning, which in turn leads to poor demand forecasting accuracy. This scheme introduces time series weighted distance to reduce the probability of eliminating points with similar time and normal fluctuations, retain the real fluctuations before and after holidays, and ensure that civil aviation samples with off-season and low demand characteristics are not mistakenly eliminated; based on route similarity weighting, trustworthy neighbors are generated to supplement the scarce low-demand civil aviation samples in a targeted manner; thereby improving the accuracy of civil aviation passenger demand forecasting.

[0027] (2) In view of the fact that general civil aviation passenger demand forecasting methods have weak forecasting capabilities for extreme scenarios such as holiday peaks and abnormal weather, and are unable to capture such periodic concentrated sudden surge scenarios, which in turn leads to poor reliability of demand forecasting, this scheme introduces feature importance scoring, calculates the feature sensitivity score of civil aviation samples, and combines the interaction impact value to evaluate the impact of feature-time points on demand forecasting from different angles; guides the adaptability to holiday peaks and extreme weather scenarios through a penalty mechanism; based on the introduction of holiday-related weights to enhance the scheduling strategy, dynamically improves the penalty weight growth rate, and adjusts the penalty coefficient according to the holiday distance and the validation set error ratio; thereby improving the reliability of civil aviation passenger demand forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of the process of predicting civil aviation passenger demand based on artificial intelligence provided by the present invention;

[0029] Figure 2 A schematic diagram of the civil aviation passenger demand forecasting system based on artificial intelligence provided by the present invention;

[0030] Figure 3 It is a flowchart of step S4.

[0031] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0033] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0034] Example 1, see Figure 1 The present invention provides a method for predicting civil aviation passenger demand based on artificial intelligence, which comprises the following steps:

[0035] Step S1: Data collection: Collect historical civil aviation passenger demand data and construct an initial civil aviation sample set;

[0036] Step S2: Abnormal demand elimination: Abnormal demand records in the initial civil aviation sample set are eliminated;

[0037] Step S3: Resampling on the civil aviation demand side: For the civil aviation samples after removing abnormal demands, select neighbors based on the similarity of route characteristics to generate new civil aviation samples, and obtain the final civil aviation sample set;

[0038] Step S4: Establish a civil aviation passenger demand forecasting model; based on the final civil aviation sample set, using LSTM as the foundation, introducing feature importance scoring, and combining holiday-related weights to enhance the scheduling strategy to build a civil aviation passenger demand forecasting model;

[0039] Step S5: civil aviation passenger demand forecasting: forecasting civil aviation passenger demand based on the real-time collected civil aviation passenger demand data based on the civil aviation passenger demand forecasting model.

[0040] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, data collection is to collect historical civil aviation passenger demand data; the historical civil aviation passenger demand data includes route passenger flow data, external influence data and timestamp; the route passenger flow data includes historical passenger load factor, passenger capacity, ticket issuance, flight segment, take-off and landing airports, flight distance, aircraft model and fare information; the external influence data includes weather, holiday data, promotional events and flight changes; the demand level is marked as a label; the demand level includes low demand, normal demand and high demand; an indicator weight is set for each dimension of the historical civil aviation passenger demand data, and the value range is (0, 1), the indicator weight is 1; set the threshold for low demand and high demand, and the value range is; multiply each dimension of the historical civil aviation passenger demand data by the indicator weight and sum it to obtain the data demand assessment value; if the data demand assessment value is lower than the low demand grade classification threshold, it is marked as a low-demand civil aviation sample; if the data demand assessment value is not lower than the low demand grade classification threshold but lower than the high demand grade classification threshold, it is marked as a normal demand civil aviation sample; if the data demand assessment value is not lower than the high demand grade classification threshold, it is marked as a high-demand civil aviation sample; standardize the collected data to obtain the initial civil aviation sample set.

[0041] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, abnormal demand is eliminated. The initial civil aviation sample set contains abnormal ultra-low or sudden demand records caused by temporary promotions, flight cancellations or system failures, which will distort the demand forecast judgment. Therefore, fuzzy mean clustering is used for all low-demand civil aviation samples and high-demand civil aviation samples to eliminate abnormal points closest to the high-demand area, and time series weighted distance is introduced to reduce the elimination probability of normal fluctuations with similar time, ensure seasonality and smoothness before and after holidays, and calculate the abnormal value of civil aviation samples. , the formula used is: ; ;in, It is a civil aviation sample With cluster center distance; and Civil Aviation Samples With cluster center time point; is the time decay factor, ranging from [0.01, 0.1]; F is the total feature dimension, f is the dimension index, and i is the civil aviation sample index; and They are and The value of the fth dimension; K is the total number of clusters, ranging from [0.01, 0.1], j is the cluster index; for The civil aviation samples were eliminated; is the noise threshold, ranging from 3 to 10. Only civil aviation samples with stable off-season and low demand characteristics are retained, and oversampling is not involved to avoid copying boundary noise to new civil aviation samples.

[0042] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, the civil aviation demand side sampling is to introduce the route similarity weight to perform civil aviation sample supplementation, balance the quantity distribution of low-demand civil aviation samples and mainstream high-demand civil aviation samples, and avoid excessive interpolation to pseudo neighbors with large differences from the target route. Therefore, the low-demand civil aviation samples and high-demand civil aviation samples that have been eliminated by abnormal demand are clustered again with fuzzy mean; calculate the route feature similarity , expressed as: ; In each cluster, based on the similarity of route characteristics, the remaining low-demand civil aviation samples and high-demand civil aviation samples are selected as neighbors to generate new civil aviation samples , the formula used is: ;in, and It is the original civil aviation sample; is the generation coefficient, and its value range is [0.1,1.0]; the final civil aviation sample set is obtained.

[0043] By performing the above operations, we can address the problem that general civil aviation passenger demand forecasting methods are sensitive to abnormal events, easily disturbed by extreme values, and easily mistake small normal fluctuations before and after seasonality as noise, mixing together route data with huge differences, and generating pseudo civil aviation samples that have no real meaning, which in turn leads to poor demand forecasting accuracy. This solution introduces time-series weighted distance to reduce the probability of eliminating points with similar time and normal fluctuations, retain the real fluctuations before and after holidays, and ensure that civil aviation samples with off-season and low-demand characteristics are not mistakenly eliminated; based on route similarity weighting, we generate trusted neighbors and supplement the scarce low-demand civil aviation samples in a targeted manner, thereby improving the accuracy of civil aviation passenger demand forecasting.

[0044] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, establishing a civil aviation passenger demand forecasting model specifically includes the following:

[0045] Step S41: Model architecture design; based on LSTM, the final civil aviation sample set is split into three types of input: static features, historical time series features, and future exogenous indication sequences; the original feature vector is output and prepared for subsequent encoding; the static feature encoder receives the static features and outputs a fixed-length static context vector, which is used to map the category and continuous static information to the same dimension and provide a global route background; the encoder receives the historical time series vector with a length of T and a value range of [7,30] at each step, which is spliced ​​with the static context vector, from a multi-layer LSTM, and outputs a hidden state sequence to capture the historical time series dependency and form a compressed representation of the past T steps; the Attention mechanism receives the hidden state sequence and the previous hidden state of the decoder and outputs a context vector, focusing on the most important historical moment in the encoder for the current prediction for each decoding step; the decoder receives the previous prediction value, static context, current context, and future exogenous indication from a multi-layer LSTM, and outputs the decoder hidden state, which is used to combine historical information and context to gradually generate an internal representation of each future moment; the output layer receives the decoder hidden state and outputs the passenger flow prediction result for the kth step; the basic loss adopts the mean square error loss function;

[0046] Step S42: Loss introduces feature importance score to evaluate the impact of feature-time point on prediction from different perspectives and calculate the feature sensitivity score of civil aviation samples , which is used to evaluate the continuous features of ticket prices and passenger load factors. However, the impact of different features on prediction at different time points may change dynamically. Therefore, the dynamic adjustment of the integral step size can more flexibly calculate the feature sensitivity score according to the change of the feature. As a dynamic variable, it is adjusted according to the rate of change of the characteristics. ; Feature sensitivity score is expressed as: ;in, is the actual eigenvalue of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic change rate of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic mean; is the value of the kth dimension at time point t; is the passenger flow prediction result; x is the civil aviation sample input feature vector, is the mean eigenvector; is the integral variable; calculate the interaction effect value of the civil aviation sample , used to capture the discrete interaction effects of holidays and weather, expressed as: ; Where F is the total set of all feature-time points; S is the subset of the total set F excluding the feature-time point (k, t); is the passenger flow prediction result that receives the subset as input;

[0047] Step S43: Two-stage training strategy; in the first stage, pure mean square error is used to allow the model to robustly learn the association between time series and features to prevent the penalty from being interfered by noise. Round, minimize the mean squared error loss, calculated and normalized for each batch, expressed as: ; Where ImSe is the normalized feature importance score; batch is the number of civil aviation samples in each training batch; G is a normalization constant with a value range of [1,50]; and is a weight hyperparameter with a value range of [0.1, 1.0]. In the second stage, importance-oriented penalty is used to improve the prediction accuracy of holiday peaks and abnormal weather extreme scenes, and then training Wheel, minimize ; Where L is the total loss function; is the dynamic penalty coefficient; is the mean square error loss function;

[0048] Step S44: Holiday-related weighted enhanced scheduling; Airline passenger flow often shows obvious peaks before and after major holidays. Linear or cosine scheduling alone cannot capture this periodic concentrated surge scenario. Therefore, the holiday-related weighted enhanced scheduling strategy is introduced to dynamically increase the penalty weight growth rate and guide the model to focus on the key features of that period earlier and more strongly; define the distance to the nearest major holiday , expressed as: ; Construct holiday enhancement factor , expressed as: ; Among them, h is the average sampling date of the current batch of civil aviation samples The nearest holiday date, the date is normalized; H is the set of all holiday dates; is the enhancement amplitude, ranging from [0.5, 2.0]; is a smoothing scale with a value range of [0.01, 0.2]; when holidays are approaching, Approaching 0, Reach maximum; when the distance is far Approaching 1; if the validation error is still high on the eve of holidays or in bad weather windows, it means that the model has not learned enough about these extreme scenarios and needs to be further amplified; otherwise, it can be appropriately slowed down; therefore, the validation set error ratio of the current training round in the key window of 7 days before and after holidays is calculated. , expressed as: ; Construct error driver , expressed as: ; The final weight scheduling is expressed as: ;in, is the validation set loss of the current epoch on the 7-day window before and after the holiday, and g is the current number of training times; is the average loss of the validation set within the window; is the error ratio of the last training validation set; is the error amplification factor, and its value range is [0.1, 2.0]; and are the minimum penalty coefficient and the maximum penalty coefficient, and their value ranges are [0.1, 1.0] and [1.0, 10.0] respectively.

[0049] By performing the above operations, we can address the problem that general civil aviation passenger demand forecasting methods have weak prediction capabilities for extreme scenarios such as holiday peaks and abnormal weather, and are unable to capture such periodic concentrated surge scenarios, which in turn leads to poor reliability of demand forecasting. This solution introduces feature importance scoring, calculates the feature sensitivity score of civil aviation samples, and combines the interaction influence value to evaluate the impact of feature-time points on demand forecasting from different angles; guides the adaptability to holiday peaks and extreme weather scenarios through a penalty mechanism; based on the introduction of holiday-related weights to enhance the scheduling strategy, dynamically increases the penalty weight growth rate, and adjusts the penalty coefficient according to the holiday distance and the validation set error ratio; thereby improving the reliability of civil aviation passenger demand forecasting.

[0050] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the civil aviation passenger demand forecast is to collect civil aviation passenger demand data in real time and input it into the civil aviation passenger demand forecast model, and use the demand level output by the civil aviation passenger demand forecast model as the civil aviation passenger demand forecast result.

[0051] Example 7, see Figure 2 This embodiment is based on the above embodiment. The civil aviation passenger demand forecasting system based on artificial intelligence provided by the present invention includes a data acquisition module, an abnormal demand elimination module, a civil aviation demand side sample replenishment module, a civil aviation passenger demand forecasting model establishment module and a civil aviation passenger demand forecasting module;

[0052] The data collection module collects historical civil aviation passenger demand data to construct an initial civil aviation sample set;

[0053] The abnormal demand elimination module eliminates abnormal demand records in the initial civil aviation sample set;

[0054] The civil aviation demand-side sample supplement module selects neighbors based on the similarity of route characteristics from the civil aviation samples after the abnormal demand is eliminated to generate new civil aviation samples, thereby obtaining the final civil aviation sample set;

[0055] The civil aviation passenger demand forecasting model building module is based on the final civil aviation sample set, LSTM as the basis, introduces feature importance scoring, and combines holiday-related weights to enhance the scheduling strategy to build a civil aviation passenger demand forecasting model;

[0056] The civil aviation passenger demand prediction module performs civil aviation passenger demand prediction on the real-time collected civil aviation passenger demand data based on the civil aviation passenger demand prediction model.

[0057] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0058] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0059] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based civil aviation passenger demand forecasting method, characterized by: The method comprises the following steps: Step S1: Data collection: Collect historical civil aviation passenger demand data and construct an initial civil aviation sample set; Step S2: Abnormal demand elimination: Abnormal demand records in the initial civil aviation sample set are eliminated; Step S3: Resampling on the civil aviation demand side: For the civil aviation samples after removing abnormal demands, select neighbors based on the similarity of route characteristics to generate new civil aviation samples, and obtain the final civil aviation sample set; Step S4: Establish a civil aviation passenger demand forecasting model; based on the final civil aviation sample set, using LSTM as the foundation, introducing feature importance scoring, and combining holiday-related weights to enhance the scheduling strategy to build a civil aviation passenger demand forecasting model; Step S5: civil aviation passenger demand forecasting: forecasting civil aviation passenger demand based on the real-time collected civil aviation passenger demand data based on the civil aviation passenger demand forecasting model; In step S4, the establishment of a civil aviation passenger demand forecasting model specifically includes the following: Step S41: Model architecture design; based on LSTM, the final civil aviation sample set is split into three types of input: static features, historical time series features, and future exogenous indication sequences; the original feature vector is output for subsequent encoding; the static feature encoder receives the static features and outputs a fixed-length static context vector; the encoder receives the historical time series vector with the static context vector concatenated at each step from a multi-layer LSTM, with a length of T, and outputs a hidden state sequence; the Attention mechanism receives the hidden state sequence and the previous hidden state of the decoder and outputs a context vector; the decoder receives the previous prediction value, static context, current context, and future exogenous indications from a multi-layer LSTM and outputs the decoder hidden state; the output layer receives the decoder hidden state and outputs the passenger flow prediction result for the kth step; the basic loss adopts the mean square error loss function; Step S42: Loss introduces feature importance score to evaluate the impact of feature-time point on prediction from different perspectives and calculate the feature sensitivity score of civil aviation samples , use dynamic adjustment of the integration step size, calculate the feature sensitivity score according to the change of the feature, and adjust the integration step size As a dynamic variable, it is adjusted according to the rate of change of the characteristic. ; Feature sensitivity score is expressed as: ;in, is the actual eigenvalue of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic change rate of the kth dimension of the i-th civil aviation sample at time point t; is the characteristic mean; is the value of the kth dimension at time point t; is the passenger flow prediction result; x is the civil aviation sample input feature vector, is the mean eigenvector; is the integral variable; calculate the interaction effect value of the civil aviation sample , used to capture the discrete interaction effects of holidays and weather, expressed as: ; Where F is the total set of all feature-time points; S is the subset of the total set F excluding the feature-time point (k, t); is the passenger flow prediction result that receives the subset as input; Step S43: two-stage training strategy; Step S44: Enhanced scheduling based on holiday-related weights.

2. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 1, characterized in that: In step S2, the abnormal demand elimination is to use fuzzy mean clustering to eliminate abnormal points for all low-demand civil aviation samples and high-demand civil aviation samples, and introduce time series weighted distance to calculate the abnormal value of civil aviation samples. , the formula used is: ; ;in, It is a civil aviation sample With cluster center distance; and Civil Aviation Samples With cluster center time point; is the time decay factor; F is the total feature dimension, f is the dimension index, and i is the civil aviation sample index; and They are and The value of the fth dimension; K is the total number of clusters, j is the cluster index; for The civil aviation samples were eliminated; is the noise threshold.

3. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 2, characterized in that: In step S3, the civil aviation demand side sampling is to introduce the route similarity weight, perform civil aviation sample sampling, and perform fuzzy mean clustering again on the low demand civil aviation samples and high demand civil aviation samples that have been eliminated by abnormal demand; calculate the route feature similarity , expressed as: ; In each cluster, based on the similarity of route characteristics, the remaining low-demand civil aviation samples and high-demand civil aviation samples are selected as neighbors to generate new civil aviation samples , the formula used is: ;in, and It is the original civil aviation sample; is the generation coefficient; the final civil aviation sample set is obtained.

4. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 3, characterized in that: In step S43, the two-stage training strategy is to use pure mean square error training in stage 1. Round, minimize the mean squared error loss, calculated and normalized for each batch, expressed as: ; Where ImSe is the normalized feature importance score; batch is the number of civil aviation samples in each training batch; G is the normalization constant; and is a weight hyperparameter; in the second stage, importance-guided penalty is used to guide and then train Wheel, minimize ; Where L is the total loss function; is the dynamic penalty coefficient; is the mean square error loss function.

5. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 4, characterized in that: In step S44, the holiday-related weighted enhanced scheduling is to introduce the holiday-related weighted enhanced scheduling strategy, define the nearest major holiday distance , expressed as: ; Construct holiday enhancement factor , expressed as: ; Among them, h is the average sampling date of the current batch of civil aviation samples The nearest holiday date; H is the set of all holiday dates; is the magnitude of enhancement; is a smoothing scale; calculate the validation set error ratio of the current training round in the key window of 7 days before and after the holiday , expressed as: ; Construct error driver , expressed as: ; The final weight scheduling is expressed as: ;in, is the validation set loss of the current epoch on the 7-day window before and after the holiday, and g is the current number of training times; is the average loss of the validation set within the window; is the error ratio of the last training validation set; is the error amplification factor; and are the minimum penalty coefficient and the maximum penalty coefficient respectively.

6. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 5, characterized in that: In step S1, the data collection is to collect historical civil aviation passenger demand data; mark the demand level as a label; and standardize the collected data to obtain an initial civil aviation sample set.

7. The method for predicting civil aviation passenger demand based on artificial intelligence according to claim 6, characterized in that: In step S5, the civil aviation passenger demand forecast is to collect civil aviation passenger demand data in real time and input it into a civil aviation passenger demand forecast model, and use the demand level output by the civil aviation passenger demand forecast model as the civil aviation passenger demand forecast result.

8. An artificial intelligence-based civil aviation passenger demand forecasting system, configured to implement the artificial intelligence-based civil aviation passenger demand forecasting method according to any one of claims 1 to 7, characterized in that: It includes data collection module, abnormal demand elimination module, civil aviation demand side sampling module, civil aviation passenger demand forecasting model establishment module and civil aviation passenger demand forecasting module; The data collection module collects historical civil aviation passenger demand data to construct an initial civil aviation sample set; The abnormal demand elimination module eliminates abnormal demand records in the initial civil aviation sample set; The civil aviation demand-side sample supplement module selects neighbors based on the similarity of route characteristics from the civil aviation samples after the abnormal demand is eliminated to generate new civil aviation samples, thereby obtaining the final civil aviation sample set; The civil aviation passenger demand forecasting model building module is based on the final civil aviation sample set, LSTM as the basis, introduces feature importance scoring, and combines holiday-related weights to enhance the scheduling strategy to build a civil aviation passenger demand forecasting model; The civil aviation passenger demand prediction module performs civil aviation passenger demand prediction on the real-time collected civil aviation passenger demand data based on the civil aviation passenger demand prediction model.

Images