Passenger classification method and device, electronic equipment and storage medium

Abstract

The invention discloses a passenger classification method and device, electronic equipment and a storage medium, and relates to the technical field of data processing or other related fields, and the method comprises the steps: obtaining a passenger sample data set, and generating M sample subsets based on the passenger sample data set, M being a positive integer; generating a classification sub-model based on each sample subset, and evaluating each classification sub-model to obtain an evaluation coefficient; aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; and inputting to-be-classified target passenger data into the target classification model to obtain a classification result. Through the method and the device, the technical problem of inaccurate passenger value classification caused by the fact that a single index analysis method cannot comprehensively reflect the real values of passengers in related technologies is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology or other related fields, and in particular to a passenger classification method and a device, electronic equipment, and storage medium thereof. Background Art

[0002] With the deepening of global economic integration and the significant increase in people's purchasing power, the civil aviation sector has ushered in its golden age of development. Air travel has been widely integrated into daily life and has become one of the preferred modes of business and leisure travel. This trend has not only catalyzed the vigorous development of the civil aviation industry, but also brought unprecedented market challenges. The increasingly fierce competition among airlines requires companies to adopt more refined and innovative operating strategies in order to gain a dominant position in the market. Among them, the accurate evaluation and classification of passenger value has become the core demand of airlines to optimize resource allocation, improve customer experience and achieve business growth.

[0003] In the operational practice of the civil aviation industry, the definition of passenger value goes beyond the simple fare measurement standard, and comprehensively considers multiple dimensions including passenger loyalty, travel frequency, and additional service consumption. These factors together constitute a comprehensive portrait of the passenger's contribution to the airline. Through high-precision passenger value identification, airlines can implement differentiated services, which can not only effectively improve the satisfaction and loyalty of high-value passengers and ensure their continued contribution, but also improve overall operational efficiency through refined management of low-value passengers, optimization of service processes and cost control, thereby consolidating their market position in the fierce competition.

[0004] However, in related technologies, passenger value analysis methods are generally limited by their single-indicator evaluation system. Whether it is fare analysis or flight frequency statistics, they cannot fully capture the complexity and dynamic changes of passenger value, and often ignore the diversity of passenger behavior and potential consumption potential, resulting in a significant reduction in the accuracy of value identification. In addition, the single-indicator method is prone to misjudgment when faced with data noise and outliers, reducing the reliability of strategy formulation.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The main purpose of this application is to provide a passenger classification method and its device, electronic device, and storage medium, so as to at least solve the technical problem that a single indicator analysis method in the related technology cannot fully reflect the true value of passengers, resulting in inaccurate passenger value classification.

[0007] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a passenger classification method is provided, which includes: obtaining a passenger sample data set, and generating M sample subsets based on the passenger sample data set, wherein M is a positive integer; generating a classification sub-model based on each of the sample subsets, and evaluating each of the classification sub-models to obtain an evaluation coefficient; aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; and inputting the target passenger data to be classified into the target classification model to obtain a classification result.

[0008] Furthermore, before generating M sample subsets based on the passenger sample data set, it also includes: obtaining preset feature processing rules, wherein the preset feature processing rules at least include: feature selection rules and feature encoding rules; for each sample data in the passenger sample data set, selecting R passenger features corresponding to each passenger sample in the sample data based on the feature selection rules, wherein R is a positive integer; for each passenger sample, encoding the R passenger features based on the feature encoding rules to obtain the feature encoding corresponding to the passenger sample; using the feature encoding corresponding to each passenger sample as updated sample data, replacing each sample data in the passenger sample data set, to obtain the updated passenger sample data set.

[0009] Furthermore, before generating M sample subsets based on the passenger sample data set, it also includes: testing each sample data in the passenger sample data set to obtain a test result, wherein the test result is used to record missing values ​​and abnormal values ​​in each sample data; for the missing values ​​indicated by the test result, performing missing value processing based on a preprocessing rule, wherein the missing value processing includes one of the following: filling processing, deletion processing; for the abnormal values ​​indicated by the test result, performing abnormal value processing based on the preprocessing rule, wherein the abnormal value processing includes one of the following: correction processing, deletion processing.

[0010] Furthermore, the step of generating M sample subsets based on the passenger sample data set includes: S1, extracting T sample data from the passenger sample data set based on a repeated sampling strategy, wherein the repeated sampling strategy at least specifies a sampling ratio, wherein T is a positive integer; S2, constructing the sample subset based on the T sample data; and looping through steps S1 to S2 to obtain M sample subsets by looping through sampling M times.

[0011] Furthermore, the step of generating a classification submodel based on each of the sample subsets includes: for each of the sample subsets, pre-classifying the T sample data in the sample subset to obtain a pre-classification result, wherein the pre-classification result includes P classification clusters, and P is an integer greater than or equal to 2; mapping each of the classification clusters to the target space to obtain T target points in the target space; determining the core point, neighborhood radius and number of neighborhood points of the classification cluster in the target space, and determining model parameters based on the core point, neighborhood radius and number of neighborhood points of all the classification clusters; obtaining an initial classification submodel, and configuring the initial classification submodel based on the model parameters to obtain the classification submodel.

[0012] Furthermore, the step of evaluating each of the classification sub-models to obtain an evaluation coefficient includes: for the P classification clusters corresponding to each of the classification sub-models, calculating the silhouette coefficient of each of the classification clusters, wherein the silhouette coefficient is used to characterize the closeness of all target points in the classification cluster to the classification cluster; calculating the mean of the P silhouette coefficients, and using the mean as the evaluation coefficient of the classification sub-model; integrating the evaluation coefficients of the M classification sub-models to obtain M evaluation coefficients.

[0013] Furthermore, the step of aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model includes: sorting the M classification sub-models based on the M evaluation coefficients to obtain a sorting result; selecting K classification sub-models based on the sorting result, and marking the K classification sub-models as candidate models; aggregating all the classification clusters in the pre-classification results of the K candidate models to obtain Q aggregated classification clusters, wherein Q is an integer greater than or equal to P; determining the core points, neighborhood radius and number of neighborhood points of the Q aggregated classification clusters in the target space, and determining the target model parameters of the target classification model based on the core points, neighborhood radius and number of neighborhood points; and configuring the initial classification model based on the target model parameters to obtain the target classification model.

[0014] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a passenger classification device is also provided, which includes: an acquisition unit, used to acquire a passenger sample data set, and generate M sample subsets based on the passenger sample data set, wherein M is a positive integer; an evaluation unit, used to generate a classification sub-model based on each of the sample subsets, and evaluate each of the classification sub-models to obtain an evaluation coefficient; an aggregation unit, used to aggregate the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; and an input unit, used to input the target passenger data to be classified into the target classification model to obtain a classification result.

[0015] Furthermore, the passenger classification device also includes: an acquisition module, which is used to acquire preset feature processing rules before generating M sample subsets based on the passenger sample data set, wherein the preset feature processing rules at least include: feature selection rules and feature encoding rules; a selection module, which is used to select R passenger features corresponding to each passenger sample in the sample data for each sample data in the passenger sample data set based on the feature selection rules, wherein R is a positive integer; an encoding module, which is used to encode the R passenger features for each passenger sample based on the feature encoding rules to obtain the feature coding corresponding to the passenger sample; and a replacement module, which is used to replace each sample data in the passenger sample data set with the feature coding corresponding to each passenger sample as the updated sample data to obtain the updated passenger sample data set.

[0016] Furthermore, the passenger classification device also includes: a detection module, which is used to detect each sample data in the passenger sample data set before generating M sample subsets based on the passenger sample data set to obtain a detection result, wherein the detection result is used to record missing values ​​and abnormal values ​​in each sample data; a first processing module, which is used to perform missing value processing on the missing values ​​indicated by the detection result based on preprocessing rules, wherein the missing value processing includes one of the following: filling processing and deletion processing; a second processing module, which is used to perform abnormal value processing on the abnormal values ​​indicated by the detection result based on the preprocessing rules, wherein the abnormal value processing includes one of the following: correction processing and deletion processing.

[0017] Furthermore, the acquisition unit includes: an extraction module, used to execute S1, extract T sample data from the passenger sample data set based on a repeated sampling strategy, wherein the repeated sampling strategy at least specifies a sampling ratio, wherein T is a positive integer; a construction module, used to execute S2, construct the sample subset based on the T sample data; a loop module, used to repeatedly execute steps S1 to S2, and obtain M sample subsets in the case of M sampling cycles.

[0018] Furthermore, the evaluation unit includes: a pre-classification module, used for pre-classifying T sample data in each sample subset to obtain a pre-classification result, wherein the pre-classification result includes P classification clusters, and P is an integer greater than or equal to 2; a mapping module, used for mapping each classification cluster to the target space to obtain T target points in the target space; a first determination module, used for determining the core point, neighborhood radius and number of neighborhood points of the classification cluster in the target space, and determining the model parameters based on the core point, neighborhood radius and number of neighborhood points of all the classification clusters; a first configuration module, used for obtaining an initial classification sub-model, and configuring the initial classification sub-model based on the model parameters to obtain the classification sub-model.

[0019] Furthermore, the evaluation unit includes: a first calculation module, used to calculate the silhouette coefficient of each of the P classification clusters corresponding to each of the classification sub-models, wherein the silhouette coefficient is used to characterize the closeness between all target points in the classification cluster and the classification cluster; a second calculation module, used to calculate the mean of the P silhouette coefficients, and use the mean as the evaluation coefficient of the classification sub-model; an integration module, used to integrate the evaluation coefficients of the M classification sub-models to obtain M evaluation coefficients.

[0020] Furthermore, the aggregation unit includes: a sorting module, which is used to sort the M classification sub-models based on the M evaluation coefficients to obtain a sorting result; a selection module, which is used to select K classification sub-models based on the sorting result, and mark the K classification sub-models as candidate models; an aggregation module, which is used to aggregate all the classification clusters in the pre-classification results of the K candidate models to obtain Q aggregated classification clusters, wherein Q is an integer greater than or equal to P; a second determination module, which is used to determine the core points, neighborhood radius and number of neighborhood points of the Q aggregated classification clusters in the target space, and determine the target model parameters of the target classification model based on the core points, neighborhood radius and number of neighborhood points; a second configuration module, which is used to configure the initial classification model based on the target model parameters to obtain the target classification model.

[0021] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer-readable storage medium is further provided, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the passenger classification methods described above.

[0022] In order to achieve the above-mentioned purpose, according to another aspect of the present application, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the passenger classification methods described above.

[0023] The present invention proposes a passenger classification method, which first obtains a passenger sample data set, and generates M sample subsets based on the passenger sample data set, where M is a positive integer, and then generates a classification sub-model based on each sample subset, and evaluates each classification sub-model to obtain an evaluation coefficient, and then aggregates the M classification sub-models based on the M evaluation coefficients to obtain a target classification model, and finally inputs the target passenger data to be classified into the target classification model to obtain a classification result.

[0024] In the present invention, the method of resampling and model fusion is adopted, and the purpose of improving the accuracy and robustness of passenger value classification is achieved by generating multiple classification sub-models based on different sample subsets and performing comprehensive evaluation and aggregation, thereby achieving the technical effect of refined passenger classification. Specifically, M sample subsets are generated by M times of repeated sampling, and each subset is used to construct an independent classification sub-model. Then, the clustering effect of each sub-model is quantified using evaluation indicators such as silhouette coefficient. Finally, the sub-models are weighted and fused according to the evaluation coefficient to obtain a more stable and accurate target classification model, which can effectively capture the complexity and dynamics of passenger behavior, avoid the limitations of single indicator evaluation, provide airlines with a more accurate passenger value classification tool, help implement differentiated service strategies, enhance market competitiveness, and thus solve the technical problem that the single indicator analysis method in the related technology cannot fully reflect the true value of passengers, resulting in inaccurate passenger value classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0026] Figure 1 is a flow chart of an optional passenger classification method according to an embodiment of the present invention;

[0027] Figure 2 is an architecture diagram of an optional passenger value classification model according to an embodiment of the present invention;

[0028] Figure 3 is a schematic diagram of an optional passenger classification device according to an embodiment of the present invention;

[0029] Figure 44 is a structural block diagram of an electronic device (or mobile device) for executing a passenger classification method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] To facilitate those skilled in the art to understand the present invention, some terms or nouns involved in the embodiments of the present invention are explained below:

[0033] Resampling strategy is a data sampling technique that extracts samples with replacement from the original data set when generating sample subsets to create multiple sample subsets with different data combinations, which helps to improve the diversity and robustness of the model.

[0034] DBSCAN, Density-Based Spatial Clustering of Applications with Noise, is a density-based clustering algorithm that finds clusters of arbitrary shapes in spatial databases. The core idea is to divide different clusters by the density distribution of points. A point with a sufficient number of neighboring points within a specific radius is a "core point", a point without enough points but in the neighborhood of another core point is a "boundary point", and all other points are "noise points". This density-based clustering method can discover hidden clustering structures in data sets and effectively identify and remove noise points.

[0035] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, processing, transmission, provision, disclosure, use and processing of the relevant data comply with the laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or organizations. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0036] The information collection (for example, user voice, video, and text collection) and analysis operations involved in this application have provided users with corresponding operation entrances when they are executed, allowing users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0037] The following embodiments of the present invention can be applied to various systems / applications / devices that require passenger value analysis and refined classification, and can achieve in-depth understanding and efficient classification of passenger attributes and behavior patterns. The present invention uses a sample subset generated based on a resampling strategy to train a classification model, and then weights and aggregates these sub-models through evaluation coefficients, which can better improve the accuracy of the classification results and the robustness of the model, and avoid the common overfitting and classification bias problems in single model or single indicator analysis.

[0038] The present invention also solves the limitation that a single indicator analysis method cannot fully reflect the true value of passengers, and provides a more sophisticated and comprehensive passenger classification solution, which helps airlines optimize resource allocation, improve customer experience and market competitiveness.

[0039] The present invention is described in detail below in conjunction with various embodiments.

[0040] Embodiment 1

[0041] According to an embodiment of the present invention, an embodiment of a passenger classification method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] The present invention provides Figure 1The passenger classification method shown in the figure is implemented by the passenger value analysis system of any airline. It combines data science and machine learning technology and is used in the refined passenger classification scenario of the civil aviation industry, especially to solve the problems of inaccurate passenger value assessment and weak generalization ability of classification models. Through the means of resampling and model fusion, accurate identification and classification of passenger value can be achieved, thereby providing airlines with a scientific basis for differentiated services and enhancing market competitiveness and passenger satisfaction.

[0043] The embodiment of the present invention is described in detail below in conjunction with various specific steps.

[0044] Figure 1 is a flow chart of an optional passenger classification method according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0045] Step S101, obtaining a passenger sample data set, and generating M sample subsets based on the passenger sample data set, where M is a positive integer.

[0046] Specifically, the passenger sample data set contains sample data of N passenger samples, where N is a positive integer. The passenger sample data set is the basic data resource for the implementation of the present invention, which is obtained by collecting and collating historical data of airlines. It covers multiple aspects such as passengers' personal information, travel habits, consumption behavior, etc., and aims to comprehensively reflect the characteristics and value potential of passengers. The passenger sample data set includes but is not limited to the following information: 1. Basic passenger information: name, date of birth, gender, nationality, place of residence, etc., which is used to identify and describe the personal attributes of passengers; 2. Travel records: including the number of historical trips, flight frequency, frequent flight routes, travel purposes, travel time, transfer conditions, cabin preferences, etc. of passengers, which are used to reflect passengers' travel habits and loyalty to airlines; 3. Consumption data: covering the average ticket price of passengers on airlines, the amount of consumption of additional services (such as baggage check-in, in-flight catering, seat upgrades, etc.), the accumulation and redemption of frequent flyer points, whether they are VIP members and their levels, etc., which are used to reveal the economic value and potential consumption capacity of passengers; 4. Behavioral preferences: such as whether passengers tend to book online, use mobile applications, what type of promotional activities they prefer, whether they participate in membership activities organized by airlines, etc., which are helpful to understand the way and preferences of passengers interacting with airlines; 5. Other related data: such as passengers' travel feedback (satisfaction scores, complaint records), response behaviors to flight delays or cancellations (such as rebooking, cancellation, waiting for the next flight), etc., which are used to reflect deeper passenger behaviors and reaction patterns.

[0047] Optionally, before generating M sample subsets based on the passenger sample data set, it also includes: obtaining preset feature processing rules, wherein the preset feature processing rules include at least: feature selection rules and feature encoding rules; for each sample data in the passenger sample data set, selecting R passenger features corresponding to each passenger sample in the sample data based on the feature selection rules, wherein R is a positive integer; for each passenger sample, encoding the R passenger features based on the feature encoding rules to obtain the feature encoding corresponding to the passenger sample; using the feature encoding corresponding to each passenger sample as the updated sample data, replacing each sample data in the passenger sample data set, to obtain an updated passenger sample data set.

[0048] It should be noted that the preset feature processing rules are guidelines for guiding the data preprocessing process, aiming to ensure the data quality and applicability in the subsequent model training and evaluation stages. They mainly include feature selection rules and feature encoding rules, which are used to guide the selection of the most relevant features from the original passenger sample data set and convert these features into numerical forms that can be processed by the machine learning model.

[0049] Among them, the feature selection rules are used to identify and screen out the features that have an impact on the passenger value assessment, and to remove irrelevant, redundant or low-contribution features to the classification based on domain knowledge and data exploration analysis to improve the efficiency and accuracy of the model. The feature selection process in the embodiment of the present invention can be performed using methods such as statistical testing, correlation analysis or feature importance scoring.

[0050] As an exemplary method, initial features may be preset in the feature selection rules, for example, (1) whether the passenger is a VIP user of XX airline; (2) the frequent flyer level of the passenger on XX airline; (3) the fare of the passenger's single trip on XX airline; (4) the price of the paid service of the passenger's single trip on XX airline; (5) the flight mileage of the passenger's single trip on XX airline; (6) the cabin of the passenger's single trip on XX airline; (7) the age of the passenger; (8) the scheduled departure time of the passenger's single trip on XX airline; (9) the scheduled arrival time of the passenger's single trip on XX airline; (10) the actual departure time of the passenger's single trip on XX airline; (11) the actual arrival time of the passenger's single trip on XX airline; (12) whether the passenger's single trip on XX airline is a connecting flight.

[0051] Based on the above feature selection rules, the embodiment of the present invention can derive the following customer features: (1) the unique ID of each passenger sample; (2) the total number of trips each passenger sample chooses to travel on XX airline in the target year; (3) whether each passenger sample is a VIP of XX airline; (4) the frequent flyer level of each passenger sample on XX airline; (5) the total ticket price of each passenger sample on XX airline in the target year; (6) the total flight mileage of each passenger sample on XX airline in the target year; (7) the average mileage price of each passenger sample on XX airline in the target year (total ticket price / total mileage, unit: =(yuan / 1000km); (8) the total price of paid services for each passenger sample on XX airline in the target year; (9) the proportion of economy class trips to total trips for each passenger sample on XX airline in the target year; (10) the average airfare discount for each passenger sample on XX airline in the target year (the discount is calculated based on the class); (11) the average punctuality rate of trips on XX airline in the target year; (12) the proportion of transfer trips to total trips for each passenger sample on XX airline in the target year; and (13) the age range of each passenger sample (divided into age ranges based on experience).

[0052] The feature encoding rule involves converting non-numeric features (such as categorical or textual features) into numeric representations to meet the input requirements of machine learning algorithms. The encoding techniques that can be used in the embodiments of the present invention include: one-hot encoding, label encoding, binary encoding, etc.

[0053] For example, to convert the categorical feature "frequent flyer level" into a numerical value, label encoding can be used to convert different levels into integers (e.g., silver level is encoded as 1, gold level is encoded as 2, and platinum level is encoded as 3); for continuous features (such as ticket prices), scaling may be required to ensure that all features are compared at the same scale.

[0054] An optional embodiment collects the travel records and consumption data of all passengers of an airline in the past year to form a passenger sample data set, and then screens initial passenger characteristics that are highly correlated with passenger value based on preset feature processing rules, including but not limited to: flight frequency, average fare, additional service consumption, frequent flyer points, number of days in advance to purchase tickets, etc.

[0055] In order to make the above features adapt to the input requirements of the machine learning model, each feature is further encoded, including but not limited to: encoding the binary attribute of "frequent flyer" as a numerical value of 0 (no) and 1 (yes); converting the continuous variable of "flight frequency" into categorical features through segmentation processing, such as low frequency (1 time / month), medium frequency (2-4 times / month), and high frequency (5 times or more / month); for "destination preference", one-hot encoding is used to convert each passenger's possible multiple destination preferences into a multi-dimensional binary vector.

[0056] Through the above-mentioned feature engineering, especially the feature selection rules, the most relevant passenger features can be accurately selected in each passenger sample data, effectively reducing the data dimension and avoiding the "dimensionality disaster" caused by data redundancy, while improving the interpretability and prediction accuracy of the model; the conversion of key features based on feature encoding rules ensures that non-numerical features are presented in numerical form. The updated passenger sample data set is more suitable as a high-quality input for model training, which significantly enhances the accuracy and robustness of the classification model, provides a solid data foundation for subsequent clustering analysis and refined division of passenger value, and ensures the efficiency and scientificity of the entire analysis process.

[0057] Optionally, before generating M sample subsets based on the passenger sample data set, it also includes: testing each sample data in the passenger sample data set to obtain a test result, wherein the test result is used to record missing values ​​and abnormal values ​​in each sample data; for missing values ​​indicated by the test result, performing missing value processing based on a preprocessing rule, wherein the missing value processing includes one of the following: filling processing, deletion processing; for abnormal values ​​indicated by the test result, performing abnormal value processing based on a preprocessing rule, wherein the abnormal value processing includes one of the following: correction processing, deletion processing.

[0058] It should be noted that when the sample size is sufficient and the missing values ​​are randomly distributed, deletion processing can be used to directly eliminate records containing missing values; or when the proportion of missing values ​​is small or shows a specific pattern, filling processing can be used, such as using the median, mean or predicted filling to fill the missing values, so as to reduce the negative impact of missing data on model training.

[0059] Similarly, correction or deletion processing is adopted for outliers in the test results. Among them, correction processing is applicable to outliers caused by data entry errors, and they are adjusted to a reasonable range through manual correction or data correction algorithm; deletion processing is used for outliers that obviously deviate from the normal value range, represent extreme situations or erroneous data, and directly remove outliers to prevent misleading model training and ensure the accuracy and reliability of the model.

[0060] The above preprocessing steps can ensure data quality and model effectiveness. Specifically, through detection, detailed records are obtained to clearly identify missing values ​​and outliers in the sample. For missing values, filling or deletion strategies are flexibly adopted according to the preprocessing rules to effectively avoid the negative impact of incomplete data on the analysis results. For outliers, correction or deletion measures are also taken based on the rules to prevent extreme data points from misleading model training. The reliability and applicability of the data set are significantly enhanced, providing high-quality data support for building a stable and accurate passenger value segmentation model, thereby improving the efficiency and scientificity of the entire analysis process.

[0061] Optionally, the step of generating M sample subsets based on the passenger sample data set includes: S1, extracting T sample data from the passenger sample data set based on a repeated sampling strategy, wherein the repeated sampling strategy at least specifies a sampling ratio, wherein T is a positive integer less than or equal to N; S2, constructing a sample subset based on the T sample data; and looping through steps S1 to S2 to obtain M sample subsets by looping through sampling M times.

[0062] It should be noted that the repeated sampling strategy specifies the method of extracting sample data from the passenger sample data set. This strategy at least includes the sampling ratio, that is, the ratio of T to N. This ratio is preset based on experience and the preset range is (0,1) to ensure that the sample subset is consistent with the original data set in order of magnitude while reducing the computational complexity.

[0063] The repeated sampling strategy can be used to extract samples from the passenger sample data set with replacement. Each time, T sample data are extracted to form a subset. By repeating this process M times, M independent sample subsets can be generated. Sampling under this strategy helps to improve the diversity and robustness of model training. Through multiple training and model fusion, the passenger value segmentation model finally obtained can better cope with the complexity and uncertainty of data in the real world.

[0064] As an example, assume that from a large data set containing 10,000 passenger samples, a subset of 7,000 samples is generated based on a 70% sampling ratio. By repeating the extraction process 10 times, 10 different but interrelated passenger sample subsets can be obtained, each of which can be used independently to build a classification sub-model. Finally, through model aggregation and evaluation, a passenger value analysis model with high confidence, accuracy and generalization ability is obtained, which effectively guides the airline's refined operation strategy.

[0065] Step S102: Generate a classification sub-model based on each sample subset, and evaluate each classification sub-model to obtain an evaluation coefficient.

[0066] Optionally, the step of generating a classification submodel based on each sample subset includes: for each sample subset, pre-classifying T sample data in the sample subset to obtain a pre-classification result, wherein the pre-classification result includes P classification clusters, and P is an integer greater than or equal to 2; mapping each classification cluster to the target space to obtain T target points in the target space; determining the core point, neighborhood radius and number of neighborhood points of the classification cluster in the target space, and determining model parameters based on the core point, neighborhood radius and number of neighborhood points of all classification clusters; obtaining an initial classification submodel, and configuring the initial classification submodel based on the model parameters to obtain a classification submodel.

[0067] In the embodiment of the present invention, each sample subset is pre-classified, spatially mapped, and parameter optimized to ensure the accuracy and applicability of the generated classification sub-model. Specifically, each subset containing T sample data is pre-classified using a clustering algorithm (such as DBSCAN), and the data is divided into P classification clusters. Each classification cluster represents a group of passengers with highly similar characteristics, providing airlines with a preliminary perspective on the distribution of passenger value; the cluster is then mapped from the complex original feature space to a target space, which is defined by feature selection and dimensionality reduction techniques, so that T sample data are converted into T target points in the new space, which can more intuitively understand and optimize the classification clusters, simplify data processing, and retain key classification information.

[0068] Determine the core point of each classification cluster in the target space (i.e., the center position of the cluster), the neighborhood radius Eps (used to define the range of neighboring data points around the core point), and the number of neighborhood points MinPts (i.e., the minimum number of data points that must be included in the Eps radius). The above three key parameters together constitute the skeleton of the clustering model, which directly affects the accuracy and stability of the clustering effect.

[0069] In an optional embodiment, when the preset neighborhood radius is ε, for a core point p, the ε-neighborhood of point p is defined as: N ε (p) = {q∈d ′ |dist(p,q)≤ε}, where N ε (p) represents the ε-neighborhood of point p, q represents the neighboring points of point p, and d ′ It represents the target point set corresponding to the classification cluster where point p is located, and dist(p,q) represents the Euclidean distance between point p and point q.

[0070] On this basis, the model parameters are adjusted through the core points, neighborhood radius and number of neighborhood points of all classification clusters, and the initial classification sub-model is deeply configured to obtain a highly customized classification sub-model, which can accurately reflect the passenger value distribution in the sample subset.

[0071] The above steps can greatly enrich the dimensions of model construction. Pre-classification and target space mapping can not only intuitively understand the structure of the data, but also flexibly adjust the model parameters to adapt to the characteristics of different sample subsets, thereby generating more personalized and efficient classification sub-models, improving model accuracy while effectively reducing the risk of overfitting, so that the model can still maintain stable performance when processing complex and changeable passenger data.

[0072] In addition, through fine control of core points, neighborhood radius and number of neighborhood points, the value of different passenger groups can be identified more accurately, providing strong data support for the airline's operating strategy, promoting the optimization of service activities, and ultimately achieving more efficient allocation of resources and maximization of profits.

[0073] Optionally, the step of evaluating each classification sub-model to obtain an evaluation coefficient includes: for the P classification clusters corresponding to each classification sub-model, calculating the silhouette coefficient of each classification cluster, wherein the silhouette coefficient is used to characterize the closeness of all target points in the classification cluster to the classification cluster; calculating the mean of the P silhouette coefficients, and using the mean as the evaluation coefficient of the classification sub-model; integrating the evaluation coefficients of the M classification sub-models to obtain M evaluation coefficients.

[0074] It should be noted that the silhouette coefficient is an indicator used to evaluate the quality of clustering results. It can intuitively reflect the closeness of all target points in a classification cluster to the classification cluster, as well as the separation of the target points from the nearest neighboring classification cluster. The calculation method of the silhouette coefficient is based on the average distance of each target point to its classification cluster (i.e., the closeness within the cluster), and the average distance to the nearest neighboring classification cluster (i.e., the separation between clusters). A higher silhouette coefficient indicates that the target points in the clustering results are closely distributed within the classification cluster, while maintaining a good distance from the neighboring classification clusters, thereby ensuring the clarity and accuracy of the classification.

[0075] In an optional embodiment, a calculation formula for the silhouette coefficient of the target point is provided as follows: Among them, s(i) represents the silhouette coefficient of target point i, a(i) represents the average distance between target point i and other target points in the classification cluster where target point i is located, and b(i) represents the average distance between target point i and each target point in the most adjacent classification cluster. The silhouette coefficient of each classification cluster can be obtained by calculating the mean of the silhouette coefficients of all target points in each classification cluster.

[0076] During the specific implementation process, the silhouette coefficients are calculated one by one for the P classification clusters corresponding to each classification sub-model. The calculation result can reflect the clustering effect of the classification sub-model when processing the corresponding sample subset. Then, the evaluation coefficient of the classification sub-model can be obtained by calculating the average value of these P silhouette coefficients, which directly reflects the overall performance quality of the classification sub-model in the target space. Finally, the evaluation coefficients of all M classification sub-models are integrated to form an evaluation coefficient set, which not only quantifies the effectiveness of each classification sub-model, but also provides strong data support for subsequent model aggregation and selection.

[0077] The above evaluation process provides an objective and quantitative evaluation standard that can accurately measure the clustering performance of each classification sub-model. In addition, calculating the mean of the silhouette coefficient can ensure the comprehensiveness and fairness of the model evaluation, and avoid misleading the overall performance of the model due to the abnormal performance of a single classification cluster. It effectively improves the scientificity and rigor of model construction, provides a basis for subsequent passenger value classification, and also provides precise data guidance for the adjustment of airlines' operational strategies, promotes the optimization of service activities, and ultimately helps airlines achieve more efficient allocation of resources and higher profits.

[0078] Step S103, aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model.

[0079] Optionally, step S103 includes: sorting M classification sub-models based on M evaluation coefficients to obtain a sorting result; selecting K classification sub-models based on the sorting result, and marking the K classification sub-models as candidate models; aggregating all classification clusters in the pre-classification results of the K candidate models to obtain Q aggregated classification clusters, where Q is an integer greater than or equal to P; determining the core points, neighborhood radius and number of neighborhood points of the Q aggregated classification clusters in the target space, and determining the target model parameters of the target classification model based on the core points, neighborhood radius and number of neighborhood points; configuring the initial classification model based on the target model parameters to obtain the target classification model.

[0080] It should be noted that the process of sorting all classification sub-models based on M evaluation coefficients ensures that models with higher performance are given priority. The sorting results intuitively reflect the relative performance of each sub-model in the clustering task. According to the sorting results, the top K classification sub-models with the best performance are selected as candidate models. The basis for selecting K is to balance the relationship between model diversity and performance. Usually K is less than M to focus on the model with the best performance.

[0081] Furthermore, the pre-classification results of the K candidate models are deeply aggregated, which involves comprehensive analysis and integration of all classification clusters, and finally Q aggregated classification clusters are obtained, where Q may be greater than or equal to P (P is the number of classification clusters in a single sub-model). By integrating similar classification clusters in multiple sub-models, classification clusters with poor stability or duplication can be eliminated, reflecting the distribution pattern of passenger value more accurately.

[0082] The core points, neighborhood radius and number of neighborhood points of Q aggregated classification clusters are determined in the target space, and then the target model parameters of the target classification model are defined. The initial classification model is configured and adjusted to obtain the final target classification model.

[0083] An optional embodiment assumes that there are M=10 classification sub-models, and K=3 models with the best performance are selected as candidate models after sorting according to the evaluation coefficients. By performing aggregation analysis on the pre-classification results of these three models, it is found that passengers can be more finely divided into Q=7 groups with different value levels. By optimizing parameters such as core points, neighborhood radius and number of neighborhood points in the target space, a more accurate target classification model is constructed to stably process a large amount of passenger data and accurately identify and divide high-value, medium-value and low-value passenger groups.

[0084] The above steps can be used to optimize model performance and improve model stability. Specifically, the best performing model can be selected and aggregated to avoid resource waste and computational complexity caused by too many models. A more comprehensive, accurate and robust target classification model can be constructed through comprehensive parameter optimization. This model can not only effectively deal with noise and outliers in the data set, but also capture subtle changes in passenger value distribution, providing a real-time and dynamic passenger value segmentation tool, greatly improving the scientific nature and accuracy of market operation strategies, promoting the optimization of service activities, and ultimately achieving efficient resource allocation and higher returns.

[0085] Step S104, inputting the target passenger data to be classified into the target classification model to obtain a classification result.

[0086] It should be noted that before implementing step S104, the target passenger data may be preprocessed to ensure that the format and features are consistent with the data used in model training, including data cleaning, processing missing values, feature encoding, and feature scaling. Subsequently, one or more target passenger data are input into the target classification model, which classifies each passenger by value and finally outputs detailed classification results, such as whether the passenger is marked as a high-value, medium-value, or low-value passenger.

[0087] Through the above steps S101 to S104, a passenger sample data set can be first obtained, and M sample subsets can be generated based on the passenger sample data set, where M is a positive integer, and then a classification sub-model is generated based on each sample subset, and each classification sub-model is evaluated to obtain an evaluation coefficient, and then the M classification sub-models are aggregated based on the M evaluation coefficients to obtain a target classification model, and finally the target passenger data to be classified is input into the target classification model to obtain a classification result.

[0088] In the embodiment of the present invention, the method of resampling and model fusion is adopted, and the purpose of improving the accuracy and robustness of passenger value classification is achieved by generating multiple classification sub-models based on different sample subsets and performing comprehensive evaluation and aggregation, thereby achieving the technical effect of refined passenger classification. Specifically, M sample subsets are generated by M times of repeated sampling, each subset is used to construct an independent classification sub-model, and then the clustering effect of each sub-model is quantified using evaluation indicators such as silhouette coefficient. Finally, the sub-models are weighted and fused according to the evaluation coefficient to obtain a more stable and accurate target classification model, which can effectively capture the complexity and dynamics of passenger behavior, avoid the limitations of single indicator evaluation, provide airlines with a more accurate passenger value classification tool, help implement differentiated service strategies, enhance market competitiveness, and thus solve the technical problem that the single indicator analysis method in the related technology cannot fully reflect the true value of passengers, resulting in inaccurate passenger value classification.

[0089] The present invention is described below in conjunction with another specific implementation manner.

[0090] The implementation mode of the present invention mainly studies the clustering of training set data with the help of civil aviation field experience, statistical results and model algorithms, and establishes and continuously adjusts the classification model for civil aviation passenger value.

[0091] Figure 2 is an architecture diagram of an optional passenger value classification model according to an embodiment of the present invention, such as Figure 2 As shown in the figure, the classification model is constructed through the following process:

[0092] 1. Data preparation and preprocessing: Collect and organize airline passenger data, fill or delete missing values ​​in passenger data, monitor and process outliers, and ensure data quality;

[0093] 2. Feature engineering: select valuable features based on experience, construct new features based on existing data, perform operations such as feature scaling and feature encoding;

[0094] 3. Sample resampling: The passenger data set after preprocessing and feature engineering is sampled with replacement to generate multiple training sets. The number of samples in the training set is smaller than the number of passenger data sets. Each training set is used as a separate input for subsequent model training.

[0095] 4. Construction and evaluation of clustering algorithm: Using the multiple independent training sets obtained above, a classification model is established using a resampling model based on the DBSCAN clustering model (a weak learner is constructed using DBSCAN for each independent training set), and the silhouette coefficient is used for evaluation;

[0096] 5. Classification result interpretation and model aggregation: Based on the experience of the civil aviation industry, a set of rules is specified to interpret and label the different clusters, and the final classification model (strong learner) is obtained according to the specified combination strategy;

[0097] 6. Model application: Whenever a new batch of passengers needs to be classified by value, the kNN algorithm is used to divide the passenger groups;

[0098] 7. Model adaptive update: The model regularly detects the distribution of passenger value. If there is a cluster center shift, the model is automatically trained and updated.

[0099] In a specific implementation scenario, for example, the passenger data and flight data of XX Airlines throughout 2023 are selected as the original data set for the experiment of this patent algorithm. The data set includes passenger personal information, passenger itinerary information, flight information and other data. The construction process of the passenger value classification model is explained below based on the above specific implementation scenario.

[0100] In the data preparation and preprocessing stage, 12 initial features can be selected based on actual business experience, as follows: 1. Whether the passenger is a VIP user of XX airline; 2. The frequent flyer level of the passenger on XX airline; 3. The ticket price of the passenger's single trip on XX airline; 4. The price of the paid service for the passenger's single trip on XX airline; 5. The flight mileage of the passenger's single trip on XX airline; 6. The cabin class of the passenger's single trip on XX airline; 7. The age of the passenger; 8. The planned departure time of the passenger's single trip on XX airline; 9. The planned arrival time of the passenger's single trip on XX airline; 10. The actual departure time of the passenger's single trip on XX airline; 11. The actual arrival time of the passenger's single trip on XX airline; 12. Whether the passenger's single trip on XX airline is a connecting flight.

[0101] The implementation mode of the present invention can process missing values ​​and outliers based on the initial features. Considering that the amount of passenger data is sufficient, a deletion processing scheme is adopted for these two types of data, and some features are quantified (whether it is VIP, frequent flyer level, and passenger age range, etc.).

[0102] In the feature engineering stage, feature construction is performed on the basis of the above initial features, and the following 13 derived features are derived to participate in the subsequent model calculation: passenger unique ID; the total number of trips that the passenger chooses to travel with XX airline in the target year; whether the passenger is a VIP of XX airline; the frequent flyer level of the passenger on XX airline; the total ticket price of the passenger on XX airline in the target year; the total flight mileage of the passenger on XX airline in the target year; the average mileage price of the passenger on XX airline in the target year (total ticket price / total mileage, in yuan / 1000km); the total price of paid services of the passenger on XX airline in the target year; the proportion of the number of economy class trips of the passenger on XX airline in the target year to the total number of trips; the average air ticket discount of the passenger on XX airline in the target year (the discount is calculated from the class); the average punctuality rate of the passenger's trips on XX airline in the target year; the proportion of the number of transfer trips of the passenger on XX airline in the target year to the total number of trips; the age range of the passenger (divided into intervals based on experience).

[0103] In the sample resampling stage, assuming that the total number of samples in the sample data set D is n, n samples can be randomly selected from the data set. ′ The data constitute a data set d ′ According to experience, n ′ The number of n is 70% (sampling ratio), and the sampling operation is repeated m times to obtain m new data sets d1 ′ …d m ′ , change d1 ′ …d m ′ Each dataset in is used as a separate input to a model.

[0104] In the construction and evaluation phase of the clustering algorithm, the DBSCAN clustering model can be used as the basic model to generate m data sets d1 ′ …d m ′ Each data set d ′ Conduct individual training. The specific training steps are as follows:

[0105] S1, initialization, the data set d ′ All points in are marked as unvisited points;

[0106] S2, select unvisited points, randomly select an unvisited point p and mark it as visited;

[0107] S3, check the ε-neighborhood. If the ε-neighborhood of point p contains at least MinPts points, then point p is marked as a core point and a new classification cluster is created. For a data point p, its ε-neighborhood is defined as: N ε (p) = {q∈d ′|dist(p,q)≤ε}where d ′ For a data set, dist(p,q) is the Euclidean distance between point p and point q;

[0108] S4, expand the cluster, add all the points in the ε-neighborhood of point p to the cluster, if these points are also core points, then all the points in their ε-neighborhood are also added to the cluster, and recursively execute until the classification cluster can no longer be expanded;

[0109] S5, process all points and repeat steps S2 to S4 until all points have been processed and points not included in any cluster are marked as noise points.

[0110] It should be noted that, through the above steps, the model parameters to be configured include: Eps (ε neighborhood radius), which defines the neighborhood radius of a point. All points within this radius are considered to be neighborhood points of the point; MinPts (minimum number of samples), which defines the minimum number of points that a point needs to contain in the ε-neighborhood (including the point itself). If the number of points in the ε-neighborhood of a point is ≥MinPts, the point is considered to be a core point.

[0111] Furthermore, the embodiment of the present invention can use the silhouette coefficient to evaluate the clustering algorithm. The silhouette coefficient can measure the closeness of the sample point to its own cluster and the separation from the nearest cluster. The value range of the silhouette coefficient is [-1,1]. The larger the value, the better the clustering effect. A calculation formula for the silhouette coefficient of a sample point is provided as follows: Where a(i) is the average distance from sample point i to other samples in its own cluster, and b(i) is the average distance from sample point i to the samples in its nearest adjacent cluster. The mean of the silhouette coefficients of all sample points in a classification cluster is the silhouette coefficient of the classification cluster.

[0112] In the stage of explaining the classification results and aggregating the models, we select the features that are of core value to the explanation of the clustering results and assign corresponding weights. The features and their weights are as follows: the total number of trips that passengers choose to travel with XX airlines in the target year, 20%; the total flight mileage of passengers on XX airlines in the target year, 20%; the frequent flyer level of passengers on XX airlines, 10%; the total ticket amount of passengers on XX airlines in the target year, 20%; the average ticket discount of passengers on XX airlines in the target year, 20%; the proportion of the number of transfer trips of passengers on XX airlines in the target year, 10%.

[0113] The embodiment of the present invention randomly selects a completed DBSCAN clustering model, calculates the center point of each cluster of the model (the average of all points in the cluster), collects the above key features of the center point, normalizes and weights them, and obtains the weighted value as the core feature value of the cluster. If the core feature value of the cluster is large, the cluster is marked as a high-value passenger, otherwise it is marked as a low-value passenger, and other clusters are marked as ordinary value passengers. The remaining m-1 DBSCAN clustering models are also operated as follows and marked.

[0114] If there is a point c, the number of times it is marked as the same category (high value, low value or normal value) in m DBSCAN clustering models ≥ If the point c is considered to be a point of a certain category (if more than one category appears ≥ This operation is performed on all points in the sample data set D, and all points with determined categories are saved as a new data set D″ as sample data for the final classification model.

[0115] In the model application stage, as time goes by, whenever a new passenger takes a flight of XX airline, the required features are obtained by processing the passenger data and added to the data set D″ after model aggregation. The k points closest to the new passenger’s features (Euclidean distance) are selected for “voting”, and the new passenger’s category is classified as the category of the point with the most votes. The new passenger is retained in the data set D″ to enrich the model’s data set.

[0116] In the model adaptive update phase, the classification model is regularly updated with the enriched dataset.

[0117] The classification model in the embodiment of the present invention is different from the traditional K-means clustering. DBSCAN does not need to specify the number of clusters in advance, and can more effectively ignore and identify noise points. In addition, DBSCAN is also suitable for processing large and uneven data sets, and is more suitable for the value distribution of civil aviation passengers.

[0118] The embodiments of the present invention use a model that is resampled, trained multiple times, and aggregated, which can more effectively evaluate the uncertainty of the model. By classifying the results in the form of voting using multiple models, the overfitting of the model can be effectively reduced, and the generalization ability and stability of the algorithm can be improved.

[0119] The present invention can significantly reduce the need for manual intervention by adaptively updating the model, improve the development efficiency, accuracy and reliability of the model, promote faster and more effective decision-making of the model, and meet practical application needs.

[0120] The present invention is described below in conjunction with another optional embodiment.

[0121] Embodiment 2

[0122] An embodiment of the present invention further provides a passenger classification device. It should be noted that the passenger classification device of the embodiment of the present invention includes multiple implementation units, which can be used to execute the passenger classification method provided in the above-mentioned embodiment one, and each implementation unit corresponds to each implementation step in the above-mentioned embodiment one.

[0123] Figure 3 is a schematic diagram of an optional passenger classification device according to an embodiment of the present invention, such as Figure 3 As shown, the device may include: an acquisition unit 31 , an evaluation unit 32 , an aggregation unit 33 , and an input unit 34 .

[0124] The acquisition unit 31 is used to acquire a passenger sample data set, and generate M sample subsets based on the passenger sample data set, where M is a positive integer.

[0125] The evaluation unit 32 is used to generate a classification sub-model based on each sample subset, and evaluate each classification sub-model to obtain an evaluation coefficient.

[0126] The aggregation unit 33 is used to aggregate the M classification sub-models based on the M evaluation coefficients to obtain a target classification model.

[0127] The input unit 34 is used to input the target passenger data to be classified into the target classification model to obtain a classification result.

[0128] The above-mentioned passenger classification device can first obtain a passenger sample data set through the acquisition unit 31, and generate M sample subsets based on the passenger sample data set, where M is a positive integer, and then generate a classification sub-model based on each sample subset through the evaluation unit 32, and evaluate each classification sub-model to obtain an evaluation coefficient, and then aggregate the M classification sub-models based on the M evaluation coefficients through the aggregation unit 33 to obtain a target classification model, and finally input the target passenger data to be classified into the target classification model through the input unit 34 to obtain a classification result.

[0129] In an embodiment of the present invention, a method of resampling and model fusion is adopted to achieve the purpose of improving the accuracy and robustness of passenger value classification by generating multiple classification sub-models based on different sample subsets and performing comprehensive evaluation and aggregation, thereby achieving the technical effect of refined passenger classification. Specifically, M sample subsets are generated by M times of repeated sampling, each subset is used to construct an independent classification sub-model, and then the clustering effect of each sub-model is quantified using evaluation indicators such as silhouette coefficient. Finally, the sub-models are weighted and fused according to the evaluation coefficient to obtain a more stable and accurate target classification model, which can effectively capture the complexity and dynamics of passenger behavior, avoid the limitations of single indicator evaluation, provide airlines with a more accurate passenger value classification tool, help implement differentiated service strategies, enhance market competitiveness, and thus solve the technical problem that a single indicator analysis method in related technologies cannot fully reflect the true value of passengers, resulting in inaccurate passenger value classification.

[0130] Optionally, the passenger classification device also includes: an acquisition module, which is used to acquire preset feature processing rules before generating M sample subsets based on the passenger sample data set, wherein the preset feature processing rules include at least: feature selection rules and feature encoding rules; a selection module, which is used to select R passenger features corresponding to each passenger sample in the sample data for each sample data in the passenger sample data set based on the feature selection rules, wherein R is a positive integer; an encoding module, which is used to encode the R passenger features for each passenger sample based on the feature encoding rules to obtain the feature coding corresponding to the passenger sample; and a replacement module, which is used to replace each sample data in the passenger sample data set with the feature coding corresponding to each passenger sample as the updated sample data to obtain an updated passenger sample data set.

[0131] Optionally, the passenger classification device also includes: a detection module, which is used to detect each sample data in the passenger sample data set before generating M sample subsets based on the passenger sample data set to obtain a detection result, wherein the detection result is used to record missing values ​​and abnormal values ​​in each sample data; a first processing module, which is used to perform missing value processing based on preprocessing rules for the missing values ​​indicated by the detection result, wherein the missing value processing includes one of the following: filling processing, deletion processing; a second processing module, which is used to perform abnormal value processing based on preprocessing rules for the abnormal values ​​indicated by the detection result, wherein the abnormal value processing includes one of the following: correction processing, deletion processing.

[0132] Optionally, the acquisition unit includes: an extraction module, used to execute S1, extract T sample data from the passenger sample data set based on a repeated sampling strategy, wherein the repeated sampling strategy at least specifies a sampling ratio, wherein T is a positive integer; a construction module, used to execute S2, construct a sample subset based on the T sample data; a loop module, used to repeatedly execute steps S1 to S2, and obtain M sample subsets in the case of M sampling cycles.

[0133] Optionally, the evaluation unit includes: a pre-classification module, used to pre-classify T sample data in each sample subset to obtain a pre-classification result, wherein the pre-classification result includes P classification clusters, and P is an integer greater than or equal to 2; a mapping module, used to map each classification cluster to a target space to obtain T target points in the target space; a first determination module, used to determine the core point, neighborhood radius and number of neighborhood points of the classification cluster in the target space, and determine the model parameters based on the core point, neighborhood radius and number of neighborhood points of all classification clusters; a first configuration module, used to obtain an initial classification sub-model, and configure the initial classification sub-model based on the model parameters to obtain a classification sub-model.

[0134] Optionally, the evaluation unit includes: a first calculation module, used to calculate the silhouette coefficient of each classification cluster for the P classification clusters corresponding to each classification sub-model, wherein the silhouette coefficient is used to characterize the closeness between all target points in the classification cluster and the classification cluster; a second calculation module, used to calculate the mean of the P silhouette coefficients, and use the mean as the evaluation coefficient of the classification sub-model; an integration module, used to integrate the evaluation coefficients of M classification sub-models to obtain M evaluation coefficients.

[0135] Optionally, the aggregation unit includes: a sorting module, used to sort M classification sub-models based on M evaluation coefficients to obtain a sorting result; a selection module, used to select K classification sub-models based on the sorting result, and mark the K classification sub-models as candidate models; an aggregation module, used to aggregate all classification clusters in the pre-classification results of the K candidate models to obtain Q aggregated classification clusters, where Q is an integer greater than or equal to P; a second determination module, used to determine the core points, neighborhood radius and number of neighborhood points of the Q aggregated classification clusters in the target space, and determine the target model parameters of the target classification model based on the core points, neighborhood radius and number of neighborhood points; a second configuration module, used to configure the initial classification model based on the target model parameters to obtain the target classification model.

[0136] The above-mentioned passenger classification device may also include a processor and a memory. The above-mentioned acquisition unit 31, evaluation unit 32, aggregation unit 33, input unit 34, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0137] The processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and the M classification sub-models are aggregated based on M evaluation coefficients by adjusting kernel parameters to obtain a target classification model, and the target passenger data to be classified is input into the target classification model to obtain a classification result.

[0138] The above-mentioned memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0139] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: obtaining a passenger sample data set, and generating M sample subsets based on the passenger sample data set, where M is a positive integer; generating a classification sub-model based on each sample subset, and evaluating each classification sub-model to obtain an evaluation coefficient; aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; and inputting the target passenger data to be classified into the target classification model to obtain a classification result.

[0140] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the passenger classification methods in the above-mentioned embodiment 1.

[0141] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement the passenger classification method of any one of the above-mentioned embodiments.

[0142] Figure 4 is a hardware structure block diagram of an electronic device (or mobile device) for executing a passenger classification method according to an embodiment of the present invention. Figure 4 As shown, the electronic device may include one or more ( Figure 4 402a, 402b, ..., 402n are used to illustrate) a processor (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), and a memory 404 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 4 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 4 More or fewer components as shown, or with Figure 4 Different configurations are shown.

[0143] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0144] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0145] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0146] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0148] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0149] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A passenger classification method, characterized in that: include: Acquire a passenger sample data set, and generate M sample subsets based on the passenger sample data set, where M is a positive integer; Generating a classification sub-model based on each of the sample subsets, and evaluating each of the classification sub-models to obtain an evaluation coefficient; Aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; The target passenger data to be classified is input into the target classification model to obtain a classification result.

2. The passenger classification method according to claim 1, characterized in that: Before generating M sample subsets based on the passenger sample data set, the method further includes: Acquire preset feature processing rules, wherein the preset feature processing rules at least include: feature selection rules and feature encoding rules; For each sample data in the passenger sample data set, selecting R passenger features corresponding to each passenger sample in the sample data based on the feature selection rule, where R is a positive integer; For each of the passenger samples, encoding the R passenger features based on the feature encoding rule to obtain a feature code corresponding to the passenger sample; The feature code corresponding to each of the passenger samples is used as updated sample data, and each of the sample data in the passenger sample data set is replaced to obtain the updated passenger sample data set.

3. The passenger classification method according to claim 1, characterized in that: Before generating M sample subsets based on the passenger sample data set, the method further includes: Performing a test on each sample data in the passenger sample data set to obtain a test result, wherein the test result is used to record missing values ​​and abnormal values ​​in each sample data; For the missing value indicated by the detection result, performing missing value processing based on a preprocessing rule, wherein the missing value processing includes one of the following: filling processing and deletion processing; For the abnormal value indicated by the detection result, abnormal value processing is performed based on the preprocessing rule, wherein the abnormal value processing includes one of the following: correction processing and deletion processing.

4. The passenger classification method according to claim 1, characterized in that: The step of generating M sample subsets based on the passenger sample data set includes: S1, extracting T sample data from the passenger sample data set based on a repeated sampling strategy, wherein the repeated sampling strategy at least specifies a sampling ratio, and T is a positive integer; S2, constructing the sample subset based on T sample data; Steps S1 to S2 are executed in a loop, and M sample subsets are obtained in the case of M sampling cycles.

5. The passenger classification method according to claim 4, characterized in that: The step of generating a classification sub-model based on each of the sample subsets comprises: For each of the sample subsets, pre-classify the T sample data in the sample subset to obtain a pre-classification result, wherein the pre-classification result includes P classification clusters, where P is an integer greater than or equal to 2; Mapping each of the classification clusters to the target space to obtain T target points in the target space; Determine the core point, neighborhood radius and number of neighborhood points of the classification cluster in the target space, and determine model parameters based on the core point, neighborhood radius and number of neighborhood points of all the classification clusters; An initial classification sub-model is obtained, and the initial classification sub-model is configured based on the model parameters to obtain the classification sub-model.

6. The passenger classification method according to claim 5, characterized in that: The step of evaluating each of the classification sub-models to obtain an evaluation coefficient comprises: For the P classification clusters corresponding to each of the classification sub-models, a silhouette coefficient of each classification cluster is calculated, wherein the silhouette coefficient is used to characterize the closeness between all target points in the classification cluster and the classification cluster; Calculate the mean of P silhouette coefficients, and use the mean as the evaluation coefficient of the classification sub-model; The evaluation coefficients of the M classification sub-models are integrated to obtain M evaluation coefficients.

7. The passenger classification method according to claim 6, characterized in that: The step of aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model includes: Sorting the M classification sub-models based on the M evaluation coefficients to obtain a sorting result; Select K classification sub-models based on the ranking result, and mark the K classification sub-models as candidate models; Aggregating all the classification clusters in the pre-classification results of the K candidate models to obtain Q aggregated classification clusters, where Q is an integer greater than or equal to P; Determine the core points, neighborhood radius and number of neighborhood points of Q aggregate classification clusters in the target space, and determine the target model parameters of the target classification model based on the core points, neighborhood radius and number of neighborhood points; The initial classification model is configured based on the target model parameters to obtain the target classification model.

8. A passenger classification device, characterized in that: include: an acquisition unit, configured to acquire a passenger sample data set, and generate M sample subsets based on the passenger sample data set, where M is a positive integer; An evaluation unit, used to generate a classification sub-model based on each of the sample subsets, and evaluate each of the classification sub-models to obtain an evaluation coefficient; an aggregation unit, used for aggregating the M classification sub-models based on the M evaluation coefficients to obtain a target classification model; The input unit is used to input the target passenger data to be classified into the target classification model to obtain a classification result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the passenger classification method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The invention comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the passenger classification method described in any one of claims 1 to 7.

Images