General aviation demand prediction and pricing optimization method based on income management

Abstract

The invention discloses a general aviation demand prediction and pricing optimization method based on income management, and belongs to the technical field of aviation, and the method comprises the steps: constructing and fusing a gray prediction model and a logit model, obtaining summarized demand prediction data, obtaining an optimal pricing strategy through a Markov decision model, and carrying out the calculation of the summarized demand prediction data. And finally, the demand prediction data and the optimal pricing strategy are coupled to obtain a demand-pricing coupling result, so that the technical problems of establishing a coupling relationship between demands and pricing in a general aviation market and realizing self-adaptive updating of the pricing strategy and maximization of long-term income through a dynamic pricing model and demand prediction are solved; according to the method, the price can be adaptively adjusted, the dynamic adaptability is high, the interaction analysis of the demand and the price can be realized, the enterprise pricing strategy is optimized, and the prediction and pricing results are more accurate.

Description

Technical Field

[0001] The present invention belongs to the field of aviation technology, and in particular relates to a general aviation demand forecasting and pricing optimization method based on revenue management. Background Art

[0002] In the existing general aviation pricing and demand forecasting technologies, static demand forecasting and pricing analysis are usually performed based on market surveys and historical data. These methods are difficult to adjust prices in real time under a changing market environment and lack the flexibility to respond to changes in market demand. At the same time, most existing technologies ignore the interaction between demand and price and cannot achieve effective coupling between demand and pricing, making it difficult for companies to maximize long-term profits when responding to market fluctuations. In addition, existing demand forecasting models and pricing strategies are also insufficient in dealing with complex and changing passenger selection behaviors, especially when using diverse passenger data collected through Internet channels, which makes it difficult to accurately predict demand. Summary of the invention

[0003] The purpose of the present invention is to provide a general aviation demand forecasting and pricing optimization method based on revenue management, which solves the technical problem of establishing a coupling relationship between demand and pricing in the general aviation market, and realizing adaptive updating of pricing strategies and maximization of long-term revenue through dynamic pricing models and demand forecasting.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A general aviation demand forecasting and pricing optimization method based on revenue management includes the following steps:

[0006] Step 1: Establish a distributed server cluster, and establish a demand data collection module, a gray demand forecasting module, an unconstrained demand forecasting module, a demand forecasting fusion module and a pricing forecasting module in the distributed server cluster;

[0007] The demand data collection module obtains the basic data of the general aviation market, pre-processes the basic data, generates a structured basic data set, and sends the basic data set to the grey demand forecasting module. The basic data includes the passenger's choice behavior data;

[0008] Step 2: The grey demand forecasting module receives the preprocessed basic data set, establishes a grey model for macro demand forecasting, and outputs the macro trend data of future demand;

[0009] Step 3: The unconstrained demand forecasting module retrieves data related to passenger behavior, analyzes and constructs a nested logit model to predict passenger selection behavior and generate micro-demand data;

[0010] Step 4: The demand forecast fusion module retrieves the macro trend data and micro demand data, and fuses the two using a weighted fusion method to generate comprehensive demand forecast data;

[0011] Step 5: The pricing prediction module defines the relevant factors that affect pricing, establishes a Markov decision model for dynamic pricing analysis, and generates the optimal pricing strategy under each time state;

[0012] Step 6: The coupling relationship analysis module retrieves the comprehensive demand forecast data and the optimal pricing strategy, analyzes the mutual influence between demand and pricing, and outputs the dynamic demand-pricing coupling result, i.e., the pricing strategy;

[0013] Step 7: The pricing strategy application module retrieves the pricing strategy output by the demand-pricing coupling model, applies it to the market, adjusts the price of general aviation products in real time, and feeds back pricing effect data to the demand data collection module for dynamic optimization and strategy update.

[0014] Preferably, when executing step 1, the specific steps performed by the demand data collection module are as follows:

[0015] Step 1-1: Obtain basic data from the general aviation market, including annual flight hours, passenger volume, frequency and content of policy releases that affect market demand, GDP growth rate, residents’ disposable income, historical general aviation airfares or product price data, search volume data, booking conversion rate data, and competitor prices;

[0016] Step 1-2: Preprocess the basic data, including cleaning and structuring the collected data, unifying the units and data formats, and generating a structured basic data set;

[0017] Step 1-3: Send the basic data set to the grey demand forecasting module.

[0018] Preferably, the data type of the basic data includes: structured table data or text data.

[0019] Preferably, when executing step 2, the grey demand forecasting module specifically performs the following steps:

[0020] Step 2-1: Obtain the basic data set and parse out the flight hours, passenger volume and economic indicators, the economic indicator is the GDP growth rate;

[0021] Step 2-2: Construct a grey forecasting model, namely the GM (1,1) model, set the time span and forecast range, and fit past trends to forecast future demand;

[0022] The GM(1,1) model is as follows:

[0023] x(0) (k+1)=ax (0) (k)+b;

[0024] Among them, a and b are model parameters; x (0) (k) represents the original data sequence, which is the original observation of the system, including flight hours, passenger volume or GDP growth rate; k is the sequence number or time step of the data; (0) is used to refer to the "original" data;

[0025] Grey model for the original data x (0) (k) Perform the cumulative generation operation to obtain the following cumulative generation formula:

[0026]

[0027] That is, the flight hours data x (0) (k) After accumulation, a new sequence x is obtained (1) (k) makes the series smooth and is used to establish the trend equation of the grey model;

[0028] Step 2-3: Generate the sequence x by accumulation according to the calculation results of the grey prediction model (1) (k) Output the trend forecast data of macro future demand, i.e. macro trend data;

[0029] Step 2-4: Send the macro trend data to the demand forecast fusion module.

[0030] Preferably, when executing step 3, the specific steps performed by the unconstrained demand prediction module are as follows:

[0031] Step 3-1: retrieve the passenger selection behavior data collected by the data collection module, including passenger preference scores, price sensitivity coefficients and travel frequency;

[0032] The traveler preference score is obtained by weighted calculation based on search volume data;

[0033] The price sensitivity coefficient is obtained from the following regression model formula:

[0034] R=A+β price X+∈;

[0035] Among them, R is the booking conversion rate data; X is the price; A is the intercept, which is the constant term of the preset regression equation; β price is price sensitivity, is the error term;

[0036] Travel frequency is obtained by analyzing the frequency of passengers’ repeat bookings in the booking conversion rate data;

[0037] Step 3-2: Construct a nested logit model to simulate the passenger behavior responses under different circumstances, especially the impact of price changes on the choice behavior data, and obtain the behavior choice probability, which includes the following steps:

[0038] Step 3-2-1: Define the passenger's choice behavior, which includes two levels of choices:

[0039] First level selection: passengers select a specific category j (e.g., route, flight);

[0040] Second level selection: select specific product i in the selected category j (such as a specific price range, cabin space, etc.);

[0041] Step 3-2-2: Construct the utility function V of the traveler choosing product i in category j ij To measure its preference, the specific formula is as follows:

[0042] V ij =β0+β price Price ij +β pref Preference ij +β freq Frequency ij +∈ ij ;

[0043] Among them, Price ij is the price of product i in category j; Preference ij Rate travelers’ preferences for specific products; Frequency ij is the travel frequency; price is the price sensitivity coefficient, reflecting the impact of price on utility; β pref and β freq are the weights of preference and frequency respectively; ij is the random error term, indicating the unobserved influencing factors;

[0044] Step 3-2-3: Construct a nested logit model and calculate the selection probability, including conditional probability calculation and probability calculation of selecting category j. The specific formula is as follows:

[0045] Conditional probability: The probability of selecting product i in category j. The specific formula is as follows:

[0046]

[0047] Among them, C j represents the set of products in category j;

[0048] The probability of selecting category j is:

[0049]

[0050] Among them, λ j is a nested parameter that represents the correlation between products in category j;

[0051] Step 3-2-4: Calculate and output the joint probability of a passenger choosing category j and product i, i.e., the behavior selection probability:

[0052] P(i, j) = P(j) × P(i|j);

[0053] Step 3-3: Use the nested logit model to simulate and predict the demand conditions under different pricing scenarios, and output micro-demand data that reflects passenger selection behavior. The specific steps include the following:

[0054] Step 3-3-1: Set multiple price sensitivity scenarios (such as high sensitivity, medium sensitivity and low sensitivity), and use the methods from step 3-2-1 to step 3-2-4 to obtain the probability of passengers' joint selection of each product under different price sensitivity scenarios;

[0055] Step 3-2-5: Under different price sensitivities, observe the changing trend of the probability of selecting each product, and finally obtain the probability distribution of passengers' selection under various price sensitivity scenarios, thereby obtaining micro-demand data;

[0056] Step 3-4: Send micro demand data to the demand forecasting fusion module.

[0057] Preferably, when executing step 4, the specific steps performed by the demand forecasting fusion module are as follows:

[0058] Step 4-1: Retrieve macro trend data and micro demand data;

[0059] Step 4-2: Through the weight setting method, the macro trend data and micro demand data are weighted and integrated to generate a balanced comprehensive demand forecast data. The specific formula is as follows:

[0060] D 融合 =w1D 灰色 +w2D 无约束

[0061] Among them, D 灰色 It represents the macro trend data obtained through the grey demand forecasting model, indicating the demand at a certain moment in the future; D 无约束 represents the micro-demand data obtained through the nested logit model, reflecting the user's tendency to choose a product or service under given conditions; w1 and w2 are weights, D 融合 is the fused demand forecast value, i.e., the comprehensive demand forecast data;

[0062] Step 4-3: Output the comprehensive demand forecast data and send it to the coupling relationship analysis module.

[0063] Preferably, when executing step 5, the specific steps performed by the pricing prediction module are as follows:

[0064] Step 5-1: Define factors that influence pricing, including price sensitivity, customer demand, or competitor prices;

[0065] Customer demand is obtained by multiplying the comprehensive demand forecast data with the predetermined price;

[0066] Competitor prices are directly input by users into the demand data collection module;

[0067] Step 5-2: Establish a Markov decision model and formulate the optimal pricing strategy for each state according to different states and factors;

[0068] Step 5-3: Generate dynamic pricing strategies under different time states.

[0069] Preferably, when executing step 6, the specific steps performed by the coupling relationship analysis module are as follows:

[0070] Step 6-1: Retrieve comprehensive demand forecast data and optimal pricing strategy;

[0071] Step 6-2: Construct a demand-pricing coupling relationship model. The demand-pricing coupling relationship model is as follows:

[0072] D t+1 =α·P optimal +β·D t ;

[0073] Among them, α and β are the coupling coefficients of the optimal price and current demand, respectively, reflecting their impact on future demand; D t+1 is the demand forecast value at the next time point; optimal is the optimal price obtained from the Markov decision model; D t is the comprehensive demand forecast value at the current time point;

[0074] Step 6-3: Output the optimization results of the demand-pricing coupling model and generate a dynamic pricing strategy.

[0075] Preferably, when executing step 7, the pricing effect data fed back specifically include passenger response, sales data and market share changes.

[0076] The general aviation demand forecasting and pricing optimization method based on revenue management described in the present invention solves the technical problem of establishing a coupling relationship between demand and pricing in the general aviation market, and realizing adaptive updating of pricing strategies and maximization of long-term benefits through dynamic pricing models and demand forecasting. The dynamic pricing model established by the present invention is based on a Markov decision model, which can adaptively adjust prices to adapt to changes in market demand and has strong dynamic adaptability. The present invention can realize the interaction analysis of demand and price and optimize the enterprise pricing strategy through the coupling of the demand forecasting model and the pricing model. The present invention comprehensively considers the impact of multiple factors on market demand and passenger selection, so that the forecasting and pricing results are more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is the main flow chart of the present invention;

[0078] Figure 2 It is a flow chart of establishing a grey prediction model of the present invention;

[0079] Figure 3 is a flow chart of establishing a nested logit model of the present invention;

[0080] Figure 4 is a flow chart of the selection probability calculated by the nested logit model of the present invention;

[0081] Figure 5 It is a flow chart of the present invention for fusing the result of the grey prediction model with the result of the nested logit model;

[0082] Figure 6 It is a flow chart of constructing a demand-pricing coupling relationship model of the present invention. DETAILED DESCRIPTION

[0083] Depend on Figure 1-Figure 6 A general aviation demand forecasting and pricing optimization method based on revenue management is shown, comprising the following steps:

[0084] Step 1: Establish a distributed server cluster, and establish a demand data collection module, a gray demand forecasting module, an unconstrained demand forecasting module, a demand forecasting fusion module and a pricing forecasting module in the distributed server cluster;

[0085] In this embodiment, the data acquisition module, the gray demand forecasting module, the unconstrained demand forecasting module, the demand forecasting fusion module and the pricing forecasting module are all servers in the distributed server cluster.

[0086] The demand data collection module obtains the basic data of the general aviation market, pre-processes the basic data, generates a structured basic data set, and sends the basic data set to the grey demand forecasting module. The basic data includes the passenger's choice behavior data;

[0087] The data type of the basic data includes: structured table data or text data.

[0088] The specific steps performed by the demand data collection module are as follows:

[0089] Step 1-1: Obtain basic data from the general aviation market, including annual flight hours, passenger volume, frequency and content of policy releases that affect market demand, GDP growth rate, residents’ disposable income, and historical general aviation airfare or product price data;

[0090] In this embodiment, the basic data is obtained through general aviation industry reports, statistical yearbooks, and feedback pricing effect data.

[0091] Among them, according to the general aviation industry report and statistical yearbook, the data obtained are divided into the following types:

[0092] Market demand data: annual flight hours, passenger volume, etc., reflecting the overall trend of general aviation market demand;

[0093] Policy data: the frequency and content of policy releases that affect market demand, such as airspace opening policies;

[0094] Economic data: including GDP growth rate, residents’ disposable income, etc., reflecting the impact of the macroeconomic environment;

[0095] Historical price data: general aviation historical air ticket prices or product price data, reflecting market price change trends.

[0096] The above data will be used as input variables of the grey demand forecasting model to analyze the overall trend affecting macro demand.

[0097] Step 1-2: Preprocess the basic data, including cleaning (such as cleaning, missing value filling, etc.) and structuring (classifying by time, location, etc.) the collected data, unifying the units and data formats, and generating a structured basic data set;

[0098] Data cleaning and structural processing are existing technologies for data preprocessing, and their purpose is to ensure that the data is suitable for the input requirements of the grey model.

[0099] Step 1-3: Send the basic data set to the grey demand forecasting module.

[0100] Step 2: The grey demand forecasting module receives the preprocessed basic data set, establishes a grey model for macro demand forecasting, and outputs the macro trend data of future demand;

[0101] The specific steps of the grey demand forecasting module are as follows:

[0102] Step 2-1: Obtain the basic data set and parse out the flight hours, passenger volume and economic indicators; the economic indicator is the GDP growth rate.

[0103] Step 2-2: Construct a grey forecasting model, namely the GM model, set the time span and forecast range, and fit past trends to forecast future demand;

[0104] In this embodiment, the number of flight hours, passenger volume, GDP growth rate, etc. are used as inputs, and the grey prediction model is used to predict future macro demand to obtain the future forecast value of demand trend. The grey prediction formula is as follows:

[0105] The grey model adopts the GM(1,1) model:

[0106] x (0) (k+1)=ax (0) (k)+b;

[0107] Among them, a and b are model parameters; x (0) (k) represents the original data sequence, which is the original observation value of the system, such as flight hours, passenger volume or GDP growth rate, etc. k is the sequence number or time step of the data. Taking the observation data of flight hours per year as an example, x is used to represent the original observation value of the system, such as flight hours, passenger volume or GDP growth rate, etc. (0) (1), x (0) (2), ..., x (0) (n) to represent data points in different years, that is, x (0) (k) represents the original value at the kth year or kth time step, and (0) is used to refer to the “original” data.

[0108] The original data sequence x (0) The data for (k) come from indicators such as flight hours, passenger volume, GDP growth rate, etc. These observations are organized into series, which can be collected by year, quarter, or other time interval.

[0109] In this embodiment, the original data may have missing values ​​or abnormal values, which need to be cleaned and formatted to make the time series data uniform and complete.

[0110] Generate the original sequence x from the sorted flight hours, passenger volume and other data (0) (k) in order to perform cumulative generation operations to adapt to the needs of the grey prediction model.

[0111] Accumulate and generate sequence x (1) (k): Grey model usually uses the original data x (0) (k) Perform cumulative generation operations to reduce the volatility of the data and improve the smoothness of the sequence. The cumulative generation formula is as follows:

[0112]

[0113] For example, the flight hours data x (0) (k) After accumulation, a new sequence x is obtained (1) (k) makes the series smooth and facilitates the establishment of trend equation in grey model.

[0114] Generate sequence x using accumulation (1) (k), the grey demand forecasting model constructs a forecasting model through differential equations to fit future trends.

[0115] In this embodiment, the correlation method may be used, that is, to test the correlation relationship between the predicted value and the original data, so as to test the accuracy of the grey demand forecasting model. The correlation method is a prior art.

[0116] Step 2-3: According to the calculation results of the grey prediction model, output the trend forecast data of macro future demand, i.e. macro trend data;

[0117] Step 2-4: Send the macro trend data to the demand forecast fusion module.

[0118] Step 3: The unconstrained demand forecasting module retrieves data related to passenger behavior, analyzes and constructs a nested logit model to predict passenger selection behavior and generate micro-demand data;

[0119] The specific steps performed by the unconstrained demand forecasting module are as follows:

[0120] Step 3-1: retrieve the passenger selection behavior data collected by the data collection module, including passenger preference scores, price sensitivity coefficients and travel frequency;

[0121] The network of the modern service industry is divided into two network environments:

[0122] Network Environment 1: The network-based (Origin and Destination, O&D) organizational business environment in which the company operates; demand forecasting and optimization control at the O&D network level is known as one of the most important milestones in the development of RMS.

[0123] Network environment 2: Information market environment with the Internet as an emerging distribution channel; search engines, company websites, online travel agents (OTAs) and mobile Internet have become some of the important distribution channels in recent years, making travelers' choices more convenient, thereby intensifying internal and external competition in various RM application industries.

[0124] In this embodiment, the network environment 2 is taken as the object and the following input data is defined:

[0125] Search volume data: passenger search volume from search engines and online travel platforms;

[0126] Booking conversion rate data: indicates the proportion of browsing converted into actual bookings;

[0127] Obtain traveler behavior indicators such as "traveler preference score, price sensitivity coefficient and travel frequency" from "search volume data and booking conversion rate data".

[0128] Traveler preferences reflect travelers' inclination towards specific products or routes, and are strongly correlated with search volume data:

[0129] Search volume data can indicate the level of interest in specific products, destinations or routes. If the search volume for certain destinations is high, it can be inferred that the destination or product type is highly preferred among the traveler group.

[0130] In this embodiment, after weighted calculation of the search volume data, the passenger preference score is obtained.

[0131] Price sensitivity indicates how users respond to price changes, especially consumption habits during peak periods or specific time periods, and is strongly correlated with clickstream data;

[0132] The relationship between booking conversion rate data and price changes can measure price sensitivity. For example, when the price is raised, if the booking rate drops significantly, it can indicate that users are more sensitive to prices. In this embodiment, the price sensitivity can be obtained by the following regression model formula:

[0133] R=A+β price X+∈;

[0134] Among them, R is the booking conversion rate; X is the price; A is the intercept, which is the constant term of the preset regression equation; β price is the price sensitivity and is the error term.

[0135] β price It can be intuitively shown that for every n-unit increase in price, the booking conversion rate will drop by a corresponding percentage.

[0136] Travel frequency reflects the user's travel habits, especially the frequency of choosing a specific route or destination, and has a strong correlation with reservations in booking conversion rate data;

[0137] By analyzing the frequency of repeat bookings by travelers in the booking conversion rate data, we can directly obtain the travel frequency. For example, high-frequency users will have more historical booking records.

[0138] Step 3-2: Construct a nested logit model to simulate the passenger behavior responses under different circumstances, especially the impact of price changes on the choice behavior data, and obtain the behavior choice probability, which includes the following steps:

[0139] Step 3-2-1: Define the passenger's choice behavior, which includes two levels of choices:

[0140] First level selection: passengers select a specific category j (e.g., route, flight);

[0141] Second level selection: select specific product i in the selected category j (such as a specific price range, cabin space, etc.);

[0142] Step 3-2-2: Construct the utility function V of the traveler choosing product i in category j ij To measure its preference, the specific formula is as follows:

[0143] V ij =β0+β price Price ij +β pref Preference ij +β freq Frequency ij +∈ ij ;

[0144] Among them, Price ij is the price of product i in category j; Preference ij Rate travelers’ preferences for specific products; Frequency ij is the travel frequency; price is the price sensitivity coefficient, reflecting the impact of price on utility; β pref and β freq are the weights of preference and frequency respectively; ij is the random error term, indicating the unobserved influencing factors;

[0145] Step 3-2-3: Construct a nested logit model and calculate the selection probability, including conditional probability calculation and probability calculation of selecting category j. The specific formula is as follows:

[0146] Conditional probability: The probability of selecting product i in category j. The specific formula is as follows:

[0147]

[0148] Among them, C j represents the set of products in category j;

[0149] The probability of selecting category j is:

[0150]

[0151] Among them, λ j is a nested parameter that represents the correlation between products in category j;

[0152] Step 3-2-4: Calculate and output the joint probability of a passenger choosing category j and product i, i.e., the behavior selection probability:

[0153] P(i, j) = P(j) × P(i|j);

[0154] Step 3-3: Use the nested logit model to simulate and predict the demand conditions under different pricing scenarios, and output micro-demand data that reflects passenger selection behavior. The specific steps include the following:

[0155] Step 3-3-1: Set multiple price sensitivity scenarios (such as high sensitivity, medium sensitivity and low sensitivity), and use the methods from step 3-2-1 to step 3-2-4 to obtain the probability of passengers' joint selection of each product under different price sensitivity scenarios;

[0156] Step 3-2-5: Under different price sensitivities, observe the changing trend of the probability of selecting each product, and finally obtain the probability distribution of passengers' selection under various price sensitivity scenarios, thereby obtaining micro-demand data;

[0157] Step 3-4: Send micro demand data to the demand forecasting fusion module.

[0158] Step 4: The demand forecast fusion module retrieves the macro trend data and micro demand data, and fuses the two using a weighted fusion method to generate comprehensive demand forecast data;

[0159] The specific steps performed by the demand forecast fusion module are as follows:

[0160] Step 4-1: Retrieve macro trend data and micro demand data;

[0161] Step 4-2: Through the weight setting method, the macro trend data and micro demand data are weighted and integrated to generate a balanced comprehensive demand forecast data. The specific formula is as follows:

[0162] D 融合 =w1D 灰色 +w2D无约束

[0163] Among them, D 灰色 It represents the macro trend data obtained through the grey demand forecasting model, indicating the demand at a certain moment in the future; D 无约束 represents the micro-demand data obtained through the nested logit model, reflecting the user's tendency to choose a product or service under given conditions; w1 and w2 are weights, D 融合 It is the fused demand forecast value, that is, the comprehensive demand forecast data.

[0164] Step 4-3: Output the comprehensive demand forecast data and send it to the coupling relationship analysis module.

[0165] Step 5: The pricing prediction module defines the relevant factors that affect pricing, establishes a Markov decision model for dynamic pricing analysis, and generates the optimal pricing strategy under each time state;

[0166] The specific steps performed by the pricing prediction module are as follows:

[0167] Step 5-1: Define factors that influence pricing, including price sensitivity, customer demand, or competitor prices;

[0168] Among them, price sensitivity is β price ;

[0169] Customer demand is obtained by multiplying the comprehensive demand forecast data and the price:

[0170] Competitor prices are set by user input.

[0171] Step 5-2: Establish a Markov decision model and formulate the optimal pricing strategy for each state according to different states and factors. The specific steps include the following:

[0172] Construct a Markov decision model as follows:

[0173]

[0174] Where T is the total number of time periods; t is the current time period, from 0 to T; R t is the profit in time period t, which usually depends on the set price and related factors; γ is the discount factor (0<γ<1), which represents the present value of future profits and usually reflects the importance of future profits. A smaller γ means more emphasis on current profits, while a larger γ means that future profits are equally important.

[0175] Where R t Calculated by the following formula:

[0176]

[0177] Among them, P t is the price in time period t; D t is the comprehensive demand forecast data in time period t; β price P is the price sensitivity coefficient; competitor Competitor's price

[0178] When P competitor >P t , it means that the company's products are lower in price than its competitors, which usually increases the company's demand because travelers tend to choose the lower-priced products.

[0179] When P competitor <P t , it indicates that a company's product prices are higher than those of its competitors, which generally reduces demand.

[0180] Step 5-3: Generate a dynamic pricing strategy under each time state, specifically using dynamic programming to maximize the revenue in each time period. The specific formula uses the Bellman equation:

[0181]

[0182] Among them, V(s) is the value of state s (i.e., the maximum expected return in this state), which means the maximum expected return that can be obtained by taking the best strategy in state s;

[0183] R(s,u) is the immediate benefit obtained when taking action u in state s. The benefit function R(s,u) can actually be combined with the previously defined benefit R t equal;

[0184] u is the action that can be taken in state s, which usually corresponds to different price settings P in pricing problems. t ;

[0185] γ is the discount factor, and its value range is between 0<γ<1. It indicates the importance of future benefits. A smaller γ indicates that current benefits are more important, while a larger γ indicates that future benefits are equally important.

[0186] P(s′|s,u) is the probability of transitioning from state s to a new state s′ after taking action u, which indicates the possibility of transitioning to the next state after taking an action in the current state;

[0187] The Bellman equation can calculate that the maximum value V(s) of the current state s is equal to the immediate benefit R(s,u) obtained by taking an action u in this state plus the expected value of the discounted future state.

[0188] By calculating the benefit R(s,u) (i.e., Rt) in each state, the Bellman equation can be used to evaluate the long-term benefits of adopting the best pricing strategy in each state, thereby determining the optimal price P optimal .

[0189] Step 6: The coupling relationship analysis module retrieves the comprehensive demand forecast data and the optimal pricing strategy, analyzes the mutual influence between demand and pricing, and outputs the dynamic demand-pricing coupling result, i.e., the pricing strategy;

[0190] The specific steps performed by the coupling relationship analysis module are as follows:

[0191] Step 6-1: Retrieve comprehensive demand forecast data and optimal pricing strategy;

[0192] Step 6-2: Construct a demand-pricing coupling relationship model. The demand-pricing coupling relationship model is as follows:

[0193] D t+1 =α·P optimal +β·D t ;

[0194] Among them, α and β are the coupling coefficients of the optimal price and current demand, respectively, reflecting their impact on future demand; D t+1 is the demand forecast value at the next time point; optimal is the optimal price obtained from the Markov decision model; D t It is the comprehensive demand forecast value at the current time point.

[0195] Step 6-3: Output the optimization results of the demand-pricing coupling model and generate a dynamic pricing strategy.

[0196] Step 7: The pricing strategy application module retrieves the pricing strategy output by the demand-pricing coupling model, applies it to the market, adjusts the price of general aviation products in real time, and feeds back pricing effect data to the demand data collection module for dynamic optimization and strategy update.

[0197] The pricing effect data fed back specifically include passenger response, sales data and market share changes.

[0198] The general aviation demand forecasting and pricing optimization method based on revenue management described in the present invention solves the technical problem of establishing a coupling relationship between demand and pricing in the general aviation market, and realizing adaptive updating of pricing strategies and maximization of long-term benefits through dynamic pricing models and demand forecasting. The dynamic pricing model established by the present invention is based on a Markov decision model, which can adaptively adjust prices to adapt to changes in market demand and has strong dynamic adaptability. The present invention can realize the interaction analysis of demand and price and optimize the enterprise pricing strategy through the coupling of the demand forecasting model and the pricing model. The present invention comprehensively considers the impact of multiple factors on market demand and passenger selection, so that the forecasting and pricing results are more accurate.

Claims

1. A general aviation demand forecasting and pricing optimization method based on revenue management, characterized by: The steps include: Step 1: Establish a distributed server cluster, and establish a demand data collection module, a gray demand forecasting module, an unconstrained demand forecasting module, a demand forecasting fusion module and a pricing forecasting module in the distributed server cluster; The demand data collection module obtains the basic data of the general aviation market, pre-processes the basic data, generates a structured basic data set, and sends the basic data set to the grey demand forecasting module. The basic data includes the passenger's choice behavior data; Step 2: The grey demand forecasting module receives the preprocessed basic data set, establishes a grey model for macro demand forecasting, and outputs the macro trend data of future demand; Step 3: The unconstrained demand forecasting module retrieves data related to passenger behavior, analyzes and constructs a nested logit model to predict passenger selection behavior and generate micro-demand data; Step 4: The demand forecast fusion module retrieves the macro trend data and micro demand data, and fuses the two using a weighted fusion method to generate comprehensive demand forecast data; Step 5: The pricing prediction module defines the relevant factors that affect pricing, establishes a Markov decision model for dynamic pricing analysis, and generates the optimal pricing strategy under each time state; Step 6: The coupling relationship analysis module retrieves the comprehensive demand forecast data and the optimal pricing strategy, analyzes the mutual influence between demand and pricing, and outputs the dynamic demand-pricing coupling result, i.e., the pricing strategy; Step 7: The pricing strategy application module retrieves the pricing strategy output by the demand-pricing coupling model, applies it to the market, adjusts the price of general aviation products in real time, and feeds back pricing effect data to the demand data collection module for dynamic optimization and strategy update.

2. The general aviation demand forecasting and pricing optimization method based on revenue management according to claim 1, characterized in that: When executing step 1, the specific steps performed by the demand data collection module are as follows: Step 1-1: Obtain basic data from the general aviation market, including annual flight hours, passenger volume, frequency and content of policy releases that affect market demand, GDP growth rate, residents’ disposable income, historical general aviation airfares or product price data, search volume data, booking conversion rate data, and competitor prices; Step 1-2: Preprocess the basic data, including cleaning and structuring the collected data, unifying the units and data formats, and generating a structured basic data set; Step 1-3: Send the basic data set to the grey demand forecasting module.

3. A general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: The data type of the basic data includes: structured table data or text data.

4. The general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 2, the grey demand forecasting module specifically performs the following steps: Step 2-1: Obtain the basic data set and parse out the flight hours, passenger volume and economic indicators, the economic indicator is the GDP growth rate; Step 2-2: Construct a grey forecasting model, namely the GM (1,1) model, set the time span and forecast range, and fit past trends to forecast future demand; The GM(1,1) model is as follows: x (0) (k+1)=ax (0) (k)+b; Among them, a and b are model parameters; x (0) (k) represents the original data sequence, which is the original observation of the system, including flight hours, passenger volume or GDP growth rate; k is the sequence number or time step of the data; (0) is used to refer to the "original" data; Grey model for the original data x (0) (k) Perform the cumulative generation operation to obtain the following cumulative generation formula: That is, the flight hours data x (0) (k) After accumulation, a new sequence x is obtained (1) (k) makes the series smooth and is used to establish the trend equation of the grey model; Step 2-3: Generate the sequence x by accumulation according to the calculation results of the grey prediction model (1) (k) Output the trend forecast data of macro future demand, i.e. macro trend data; Step 2-4: Send the macro trend data to the demand forecast fusion module.

5. The general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 3, the specific steps performed by the unconstrained demand forecasting module are as follows: Step 3-1: retrieve the passenger selection behavior data collected by the data collection module, including passenger preference scores, price sensitivity coefficients and travel frequency; The traveler preference score is obtained by weighted calculation based on search volume data; The price sensitivity coefficient is obtained through booking conversion rate data using the following regression model formula: R=A+β price X+∈; Among them, R is the booking conversion rate data; X is the price; A is the intercept, which is the constant term of the preset regression equation; β price is price sensitivity, is the error term; Travel frequency is obtained by analyzing the frequency of passengers’ repeat bookings in the booking conversion rate data; Step 3-2: Construct a nested logit model to simulate the passenger behavior responses under different circumstances, especially the impact of price changes on the choice behavior data, and obtain the behavior choice probability, which includes the following steps: Step 3-2-1: Define the passenger's choice behavior, which includes two levels of choices: First level selection: passengers select a specific category j (e.g., route, flight); Second level selection: select specific product i in the selected category j (such as a specific price range, cabin space, etc.); Step 3-2-2: Construct the utility function V of the traveler choosing product i in category j ij To measure its preference, the specific formula is as follows: V ij =β0+β price ·Price ij +b pref ·Preference ij +b freq ·Frequency ij +∈ ij ; Among them, Price ij is the price of product i in category j; Preference ij Rate travelers’ preferences for specific products; Frequency ij is the travel frequency; price is the price sensitivity coefficient, reflecting the impact of price on utility; β pref and β freq are the weights of preference and frequency, respectively; ij is a random error term, indicating unobserved influencing factors; Step 3-2-3: Construct a nested logit model and calculate the selection probability, including conditional probability calculation and probability calculation of selecting category j. The specific formula is as follows: Conditional probability: The probability of selecting product i in category j. The specific formula is as follows: Among them, C j represents the set of products in category j; The probability of selecting category j is: Among them, λ j is a nested parameter that represents the correlation between products in category j; Step 3-2-4: Calculate and output the joint probability of a passenger choosing category j and product i, i.e., the behavior selection probability: P(i, j) = P(j) × P(i|j); Step 3-3: Use the nested logit model to simulate and predict the demand conditions under different pricing scenarios, and output micro-demand data that reflects passenger selection behavior. The specific steps include the following: Step 3-3-1: Set multiple price sensitivity scenarios (such as high sensitivity, medium sensitivity and low sensitivity), and use the methods from step 3-2-1 to step 3-2-4 to obtain the probability of passengers' joint selection of each product under different price sensitivity scenarios; Step 3-2-5: Under different price sensitivities, observe the changing trend of the probability of selecting each product, and finally obtain the probability distribution of passengers' selection under various price sensitivity scenarios, thereby obtaining micro-demand data; Step 3-4: Send micro demand data to the demand forecasting fusion module.

6. A general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 4, the specific steps performed by the demand forecast fusion module are as follows: Step 4-1: Retrieve macro trend data and micro demand data; Step 4-2: Through the weight setting method, the macro trend data and micro demand data are weighted and integrated to generate a balanced comprehensive demand forecast data. The specific formula is as follows: D 融合 =w1D 灰色 +w2D 无约束 Among them, D 灰色 It represents the macro trend data obtained through the grey demand forecasting model, indicating the demand at a certain moment in the future; D 无约束 represents the micro-demand data obtained through the nested logit model, reflecting the user's tendency to choose a product or service under given conditions; w1 and w2 are weights, D 融合 is the fused demand forecast value, i.e., the comprehensive demand forecast data; Step 4-3: Output the comprehensive demand forecast data and send it to the coupling relationship analysis module.

7. A general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 5, the specific steps performed by the pricing prediction module are as follows: Step 5-1: Define factors that influence pricing, including price sensitivity, customer demand, or competitor prices; Customer demand is obtained by multiplying the comprehensive demand forecast data with the predetermined price; Competitor prices are directly input by users into the demand data collection module; Step 5-2: Establish a Markov decision model and formulate the optimal pricing strategy for each state according to different states and factors; Step 5-3: Generate dynamic pricing strategies under different time states.

8. The general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 6, the specific steps performed by the coupling relationship analysis module are as follows: Step 6-1: Retrieve comprehensive demand forecast data and optimal pricing strategy; Step 6-2: Construct a demand-pricing coupling relationship model. The demand-pricing coupling relationship model is as follows: D t+1 =α·P optimal +β·D t ; Among them, α and β are the coupling coefficients of the optimal price and current demand, respectively, reflecting their impact on future demand; D t+1 is the demand forecast value at the next time point; optimal is the optimal price obtained from the Markov decision model; D t is the comprehensive demand forecast value at the current time point; Step 6-3: Output the optimization results of the demand-pricing coupling model and generate a dynamic pricing strategy.

9. The general aviation demand forecasting and pricing optimization method based on revenue management as claimed in claim 2, characterized in that: When executing step 7, the pricing effect data fed back specifically include passenger response, sales data and market share changes.

Images