Reinforcement learning-based domestic air ticket dynamic pricing method and system, and electronic equipment
By constructing and optimizing the dynamic pricing model of air tickets based on reinforcement learning, the problems of consumer demand and market environment in domestic air ticket pricing are solved, more accurate and efficient price adjustments are achieved, and airlines' market share and brand influence are improved.
Patent Information
- Application Number
- CN202411943886.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-26
- Publication Date
- 2025-04-22
AI Technical Summary
The existing technology is difficult to effectively solve the complexity of consumer demand, strategic behavior and market environment in dynamic pricing of domestic air tickets, resulting in airlines facing the risk of unsold air tickets in the price decision process and may have a negative impact on service quality.
Using a reinforcement learning-based approach, we build dynamic pricing models by obtaining historical air ticket data, train and optimize models to generate more accurate air ticket pricing decisions. This method defines the environment, state space, action space and reward mechanisms, and uses reinforcement learning algorithms to optimize pricing strategies.
It has achieved rapid response to market changes, precise price adjustments, long-term profit optimization, adapting to different market segments and continuous learning and improvement, helping airlines increase market share and overall revenue, while enhancing brand influence.
Smart Images
Figure CN119863262A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of civil aviation data analysis technology, and in particular relates to a domestic air ticket dynamic pricing method, system and electronic equipment based on reinforcement learning. Background Art
[0002] Air transport plays an indispensable role as a bridge in international political exchanges and economic cooperation. On April 1, 2005, the Civil Aviation Administration of China passed the "Regulations on Domestic Investment in the Civil Aviation Industry" at the Administration Meeting, which was implemented on August 15 of the same year, lowering the entry threshold of the civil aviation market. After the implementation of the policy, many airlines were established one after another, promoting the rapid development of the industry. The civil aviation industry quickly entered a stage of rapid growth and became one of the important pillars of China's economy.
[0003] In the airline's price decision-making process, consumer demand forecasting, consumer behavior patterns, and market competition are key factors. With the continued economic growth and improved living standards, the aviation market is expanding. The opening of information resources makes consumer behavior more rational. They may expect price drops and delay purchases to obtain maximum utility. In this case, airlines may face the risk of unsold tickets, so they often choose price reduction strategies to ensure that all seats are sold before the flight takes off, otherwise they will face the dilemma of zero income. This strategy may have a negative impact on the service quality of the airline. If the complexity and diversity of consumer behavior are ignored, the company's interests may be damaged. Therefore, accurate demand forecasting, analysis of consumer behavior characteristics, and market competition research are important bases for formulating reasonable pricing strategies. These efforts can effectively help airlines maximize their profits.
[0004] Foreign countries started research on demand modeling earlier and developed a variety of models to deal with demand dependencies between flights. In the problem of dynamic pricing of airline tickets, it is necessary not only to consider consumer demand and consumer strategic behavior, but also to pay attention to market competition, demand dependencies in the route network, and demand linkage effects between different flights, especially between high-density routes and hub airports.
[0005] At present, many scholars have recognized the important influence of consumer demand, strategic behavior and market environment on pricing strategy. However, when exploring this issue, many studies usually regard consumers as a single group with strategic behavior, and there is still a lack of research on markets where multiple consumers coexist. Even in the literature that considers the coexistence of multiple consumers, consumers are usually simply divided into strategic and short-sighted types. However, consumer behavior is highly complex, and its demand is not only affected by price, but also by factors such as product quality, service level, and the price and quality of alternative products.
[0006] Therefore, it is necessary to conduct in-depth research on dynamic pricing of domestic air tickets to increase route revenue. Summary of the invention
[0007] In response to the above problems, the present application provides a domestic air ticket dynamic pricing method, system and electronic equipment based on reinforcement learning, aiming to use reinforcement learning technology to provide airlines with a more intelligent and efficient air ticket pricing solution.
[0008] According to a first aspect of the present application, a domestic air ticket dynamic pricing method based on reinforcement learning is provided, the method comprising the following steps:
[0009] Get historical air ticket data;
[0010] Use historical air ticket data to build a dynamic pricing model based on reinforcement learning algorithms;
[0011] Train and optimize the constructed dynamic pricing model;
[0012] Utilize the trained dynamic pricing model to generate domestic air ticket pricing decisions based on the current market environment and flight information.
[0013] Optionally, the use of historical air ticket data to construct a dynamic pricing model based on a reinforcement learning algorithm includes:
[0014] Define the environment, which includes competing airlines’ pricing, as well as passenger arrivals, purchasing behavior, and preferences for different products;
[0015] Define a state space, the state space being used to represent the number of remaining seats for the current dynamically priced flight, the current time, and the flight departure time;
[0016] defining an action space, the action space being used to represent a set of pricing combinations for different service classes in a flight;
[0017] Rewards are defined such that the adjusted price of the ticket purchased by the passenger is the positive reward and the cost of the unsold seats at the time of flight departure is the negative reward.
[0018] Optionally, the training and optimization of the constructed dynamic pricing model includes:
[0019] Training model: Use the preprocessed data to train the constructed dynamic pricing model;
[0020] Optimize the model: Improve model performance and generalization by adjusting dynamic pricing model hyperparameters, increasing training data diversity, and using regularization methods.
[0021] Optionally, the method further includes monitoring and providing feedback on the execution of the pricing decision.
[0022] Optionally, the monitoring and feedback of the execution of the pricing decision includes:
[0023] Collect various data related to air ticket pricing in real time;
[0024] Compare and analyze the actual monitored data and calculated indicators with the expected results of the pricing model to form feedback information;
[0025] Based on feedback information, the pricing model is optimized and adjusted.
[0026] Optionally, when generating pricing decisions, prices are gradually adjusted according to preset price adjustment ranges and adjustment rules to avoid excessive price fluctuations that may cause dissatisfaction among passengers.
[0027] Optionally, the generated pricing decision meets pre-set actual constraints, which include passenger load factor, price positioning to maintain brand image, and price limits agreed with partners.
[0028] According to a second aspect of the present application, a domestic air ticket dynamic pricing system based on reinforcement learning is provided, the system comprising:
[0029] The acquisition module is used to obtain historical air ticket data and perform preprocessing;
[0030] Building blocks that can build dynamic pricing models based on reinforcement learning algorithms;
[0031] The training and optimization module is used to train and optimize the constructed dynamic pricing model;
[0032] The generation module is used to use the trained dynamic pricing model to generate pricing decisions based on the current status information.
[0033] Optionally, the system further includes a monitoring and feedback module for monitoring and providing feedback on the execution of pricing decisions.
[0034] According to a third aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.
[0035] The beneficial effects of this application are:
[0036] 1. Rapid response to market changes. This application can perceive various dynamic factors in the market environment in real time, such as fluctuations in passenger demand, adjustments in competitor fares, seasonal changes, emergencies, etc. By constantly interacting with the environment and learning based on real-time feedback, it can quickly adjust ticket prices to adapt to immediate changes in the market.
[0037] 2. Accurate price adjustment. This application is based on an in-depth understanding and study of the market environment, which can achieve more accurate price adjustments; by analyzing a large amount of historical data and real-time market feedback, the model can identify changes in passengers' price sensitivity under different market conditions.
[0038] 3. Long-term revenue optimization. The reinforcement learning in this application focuses on maximizing long-term cumulative rewards, not just short-term profit. In air ticket pricing, this means that the scheme will take into account the impact of current pricing decisions on future market demand and passenger behavior.
[0039] 4. Adapt to different market segments. The domestic aviation market is diverse, including different routes, seasons, regions and other market segments. This application can carry out targeted pricing according to the characteristics of different market segments. This ability to adapt to different market segments can enable airlines to more accurately meet the needs of each market and increase market share and overall revenue.
[0040] 5. Continuous learning and improvement. The reinforcement learning model has the ability to continuously learn and improve itself. With the continuous influx of new data, the model can continuously update its knowledge and strategies to adapt to changes in the market environment; this continuous learning mechanism enables pricing plans to keep pace with the times and continuously optimize pricing decisions. Compared with traditional pricing methods, it does not require frequent manual adjustments to models and parameters, but continuously improves the accuracy and effectiveness of pricing through automatic learning. At the same time, by monitoring and analyzing the model learning process, new market trends and potential laws can also be discovered, providing more valuable information for airlines' market decisions.
[0041] 6. Enhance market share and brand influence. The dynamic pricing method based on reinforcement learning in this application can help airlines provide a better user experience in terms of price and attract more passengers to choose their flights. At the same time, with the increase in market share and the improvement of passenger satisfaction, the brand influence of airlines will also be enhanced accordingly, forming a virtuous circle and further consolidating their competitive position in the market.
[0042] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or be understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures indicated in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0044] Figure 1 A flowchart of the steps of a domestic air ticket dynamic pricing method based on reinforcement learning according to an embodiment of the present application is shown.
[0045] Figure 2 A route network diagram according to an embodiment of the present application is shown.
[0046] Figure 3 A flowchart of the steps of monitoring and providing feedback on the execution of pricing decisions according to an embodiment of the present application is shown.
[0047] Figure 4 A schematic diagram of the architecture of a domestic air ticket dynamic pricing system based on reinforcement learning according to an embodiment of the present application is shown.
[0048] Figure 5 An exemplary diagram of the architecture of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0050] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.
[0051] Figure 1FIG. 1 shows a flowchart of a method for dynamic pricing of domestic air tickets based on reinforcement learning according to an embodiment of the present application. Figure 1 As shown, the method may include the following steps S1 to S4.
[0052] Step S1: Obtain historical air ticket data.
[0053] Among them, historical air ticket data may include flight model, air ticket price, departure and destination, departure time, passenger type, etc.
[0054] Step S2: Utilize historical airfare data to build a dynamic pricing model based on a reinforcement learning algorithm.
[0055] In step S2, the present application considers that in the route network set in the environment, the probability of passengers arriving at any time, their preference for each product, and the probability of choosing each price ticket are known.
[0056] In some embodiments, using historical airfare data to construct a dynamic pricing model based on a reinforcement learning algorithm includes defining an environment, defining a state space, defining an action space, and defining a reward.
[0057] The environment refers to a virtual market environment, including the pricing of competing airlines, as well as passengers' arrival, purchasing behavior, and preferences for different products. The state space is used to represent the number of remaining seats on the current dynamically priced flight, as well as the current time and the time of flight departure. The action space is used to represent a set of pricing combinations for different service classes in the flight. The rewards include positive rewards and negative rewards. The adjusted ticket price purchased by the passenger is used as a positive reward, and the cost of the unsold seats when the flight takes off is used as a negative reward.
[0058] The construction of a dynamic pricing model is further explained below through a specific embodiment.
[0059] 1. Define the environment
[0060] Figure 2 FIG. 1 shows a route network diagram according to an embodiment of the present application. Figure 2 As shown in the figure, the route network includes four destinations with Beijing as the hub, and the specific flights are Harbin->Beijing, Beijing->Chongqing, and Beijing->Shanghai. The three flights can form five itinerary combinations: Harbin->Beijing, Beijing->Shanghai, Beijing->Chongqing, Harbin->Shanghai, and Harbin->Chongqing.
[0061] In each combination, there are three ticket products with different prices, depending on the refund, change and additional services. The basic prices of these three products (product A, product B, product C) are shown in Table 1.
[0062] Table 1 Basic prices of three products
[0063] Products / Routes Harbin->Beijing Beijing->Shanghai Beijing->Chongqing Harbin->Shanghai Harbin->Chongqing Product A 1000 1300 800 2200 2000 Product B 1200 1400 900 2300 2100 Product C 1300 1500 1000 2400 2200
[0064] It is known that at a specific time before the flight takes off, the arrival rate of each passenger and the probability of choosing each destination and each product are known.
[0065] For example, if the flight takes off on September 15, then on September 10, there are 10 passengers on each route, and a total of 50 passengers. Passengers are divided into business and leisure types, with 6 business passengers and 4 leisure passengers on each route.
[0066] In the Harbin->Beijing route, each product can have 3 adjusted prices, as shown in Table 2 below.
[0067] Table 2 Original prices and adjustable prices of three products
[0068] Harbin->Beijing Original Price Adjustable prices Product A 1000 900，1000，1100 Product B 1100 1000，1100，1200 Product C 1200 1100，1200，1300
[0069] The purchase probability of leisure passengers for the Harbin->Beijing route 5 days before the flight is shown in Table 3. The purchase probability of business passengers for the Harbin->Beijing route 5 days before the flight is shown in Table 4.
[0070] Table 3 Purchase probability of three products for leisure travelers
[0071] Product A Price Product B Price Product C Price Probability of buying A Purchase B probability Purchase C probability 900 1000 1100 35％ 35％ 30％ 1100 1100 1000 20％ 35％ 45％ ...
[0072] Table 4 Purchase probability of three products for business travelers
[0073] Product A Price Product B Price Product C Price Probability of buying A Purchase B probability Purchase C probability 900 1000 1100 35％ 25％ 40％ 1100 1100 1000 15％ 35％ 50％ ...
[0074] 2. Define the state space
[0075] In this embodiment, the state space s can be expressed as [t, f1, f2, f3], where t is the time difference between the departure date and the purchase date, indicating t days before departure. f1 represents the number of remaining seats on the Harbin->Beijing flight, f2 represents the number of remaining seats on the Beijing->Shanghai flight, and f3 represents the number of remaining seats on the Beijing->Chongqing flight.
[0076] For example, state s1[5,100,99,80] means that 5 days before the flight departure, the number of remaining seats for the Harbin->Beijing flight is 100, the number of remaining seats for the Beijing->Shanghai flight is 99, and the number of remaining seats for the Beijing->Chongqing flight is 80.
[0077] 3. Define the action space
[0078] For example, an action a1 can be represented as shown in Table 5:
[0079] Table 5 Action a1
[0080] Products / Routes Harbin->Beijing Beijing->Shanghai Beijing->Chongqing Harbin->Shanghai Harbin->Chongqing Product A 1000 1300 800 2200 2000 Product B 1200 1400 900 2300 2100 Product C 1300 1500 1000 2400 2200
[0081] 4. Define Rewards
[0082] In this application, the adjusted ticket price purchased by the passenger is used as the positive reward, and the cost of the unsold seats when the flight departs is used as the negative reward.
[0083] For example: in state s1[5,100,99,80], action a1 is taken, and the passenger buys 2 Harbin->Beijing product B and 3 Harbin->Shanghai product C.
[0084] At this time, the income should be accumulated as follows: income R = income R of the previous state + 1200*2 + 2400*3;
[0085] The state transition is s2[4,95,96,80] (after each calculation, the time difference t=t-1, which means one day has passed. The ticket from Harbin to Shanghai includes both the flights from Harbin to Beijing and Beijing to Shanghai).
[0086] After the calculation is completed on the last day, the flight takes off. At this time, all remaining seats cannot be sold, and the cost of all remaining seats should be used as negative rewards.
[0087] For example: Assume that the cost of each seat is 200 yuan. When the final state is s0[0,10,9,5], the final profit R0 = the profit of the previous state R-(10+5+9)*200.
[0088] In some embodiments, based on the problem and data characteristics, a deep Q network (DQN) algorithm is selected to try to solve the optimal strategy, i.e., pricing combination, so that the final reward, i.e., the airline's profit, is maximized.
[0089] Step S3: Train and optimize the constructed dynamic pricing model.
[0090] In some embodiments, the training and optimizing of the constructed dynamic pricing model includes training the model and optimizing the model.
[0091] The training model includes training the constructed dynamic pricing model using preprocessed data. The agent continuously interacts with the simulated market environment, tries different pricing strategies, obtains feedback based on the reward function, and adjusts the strategy to maximize the cumulative reward. During the training process, GPUs are used on a large scale for training, depending on the complexity of the flight network and pricing space.
[0092] Optimizing the model involves adjusting the dynamic pricing model hyperparameters (learning rate, discount factor, etc.), increasing the diversity of training data, and using regularization methods to improve model performance and generalization capabilities, ensuring that the model can obtain a near-optimal strategy.
[0093] Step S4: Utilize the trained dynamic pricing model to generate pricing decisions based on the current status information.
[0094] In step S4, the dynamic pricing model will select an appropriate action under a given state based on the strategy it has learned. This action corresponds to the specific direction and magnitude of the price adjustment.
[0095] For example, if the dynamic pricing model determines that the current market demand is high and the number of remaining seats on the flight is limited, it may choose to increase the ticket price; on the contrary, if the market demand is weak and there are many remaining seats on the flight, the dynamic pricing model may recommend lowering the price to attract more passengers. When making decisions, the model will comprehensively consider the long-term and short-term impact of various factors on the airline's revenue, with the goal of maximizing accumulated rewards.
[0096] As a preferred embodiment of the present application, the present application pre-sets price adjustment range limits and adjustment rules, and gradually adjusts prices according to the preset price adjustment range and adjustment rules, rather than changing prices drastically at one time to smooth price adjustments, thereby avoiding excessive price fluctuations that may cause dissatisfaction among passengers.
[0097] As a preferred embodiment of the present application, some actual constraints need to be considered when generating a price strategy. The actual constraints may be the airline's own price policy and objectives, such as ensuring a certain passenger load factor, maintaining the brand image price positioning (avoiding the price being too low or too high to affect the brand image), and the agreed price limits with partners.
[0098] In some embodiments, the reinforcement learning-based domestic air ticket dynamic pricing method of the present application may include the steps of: monitoring and providing feedback on the execution of pricing decisions.
[0099] Figure 3 FIG. 1 is a flowchart showing the steps of monitoring and providing feedback on the execution of pricing decisions according to an embodiment of the present application. Figure 3 As shown, the monitoring and feedback of the execution of the pricing decision includes the following steps S51 to S53.
[0100] Step S51: Collect various data related to air ticket pricing in real time.
[0101] In some embodiments, various data related to air ticket pricing collected in real time include actual air ticket sales data (such as sales price, sales volume, sales time, passengers' personal information and ticket purchasing behavior characteristics, etc.), market competition data (competitors' real-time price changes, route adjustments, promotional activities, etc.), flight operation data (actual flight take-off and landing times, punctuality rate, passenger satisfaction, etc.) and external environment data (changes in macroeconomic data, the impact of emergencies on the tourism market, etc.).
[0102] Step S52: Compare and analyze the actual data monitored and the calculated indicators with the expected results of the pricing model to form feedback information.
[0103] In step S52, if there is a large deviation between the actual sales situation and the model expectations, for example, the actual sales volume is far lower than expected, or the revenue does not reach the expected target, then an in-depth analysis of the reasons is required.
[0104] Step S53: Optimize and adjust the pricing model based on the feedback information.
[0105] In step S53, if it is found that the prediction of the dynamic pricing model is inaccurate in some cases, it may be necessary to retrain the model or adjust the parameters of the model. For example, if the seasonal variation pattern of market demand has changed, resulting in the original model's pricing strategy being ineffective in a specific season, the weight or feature representation of seasonal factors in the model can be readjusted to better adapt to the new market situation.
[0106] Figure 4 The following is an example diagram of the architecture of a domestic air ticket dynamic pricing system based on reinforcement learning according to an embodiment of the present application, which is used to implement the above-mentioned domestic air ticket dynamic pricing method based on reinforcement learning. Figure 4 As shown, the system includes an acquisition module 1, a construction module 2, a training and optimization module 3, and a generation module 4. Among them, the acquisition module 1 is used to obtain historical air ticket price data; the construction module 2 can build a dynamic pricing model based on the reinforcement learning algorithm; the training and optimization module 3 is used to train and optimize the constructed dynamic pricing model; the generation module 4 is used to use the trained dynamic pricing model to generate pricing decisions according to the current state information.
[0107] In some embodiments, the system further includes a monitoring and feedback module 5, which is used to monitor and provide feedback on the execution of pricing decisions, and specifically includes the following steps S51 to S53.
[0108] Step S51: collecting various data related to air ticket pricing in real time;
[0109] Step S52: Compare and analyze the monitored actual data and the calculated indicators with the expected results of the pricing model to form feedback information;
[0110] Step S53: Optimize and adjust the pricing model based on the feedback information.
[0111] The domestic air ticket dynamic pricing method and system based on reinforcement learning in this application has the following features:
[0112] Beneficial effects:
[0113] 1. Rapid response to market changes. This application can perceive various dynamic factors in the market environment in real time, such as fluctuations in passenger demand, adjustments in competitor fares, seasonal changes, emergencies, etc. By constantly interacting with the environment and learning based on real-time feedback, it can quickly adjust ticket prices to adapt to immediate changes in the market.
[0114] 2. Accurate price adjustment. This application is based on an in-depth understanding and study of the market environment, which can achieve more accurate price adjustments; by analyzing a large amount of historical data and real-time market feedback, the model can identify changes in passengers' price sensitivity under different market conditions.
[0115] 3. Long-term revenue optimization. The reinforcement learning in this application focuses on maximizing long-term cumulative rewards, not just short-term profit. In air ticket pricing, this means that the scheme will take into account the impact of current pricing decisions on future market demand and passenger behavior.
[0116] 4. Adapt to different market segments. The domestic aviation market is diverse, including different routes, seasons, regions and other market segments. This application can carry out targeted pricing according to the characteristics of different market segments. This ability to adapt to different market segments can enable airlines to more accurately meet the needs of each market and increase market share and overall revenue.
[0117] 5. Continuous learning and improvement. The reinforcement learning model has the ability to continuously learn and improve itself. With the continuous influx of new data, the model can continuously update its knowledge and strategies to adapt to changes in the market environment; this continuous learning mechanism enables pricing plans to keep pace with the times and continuously optimize pricing decisions. Compared with traditional pricing methods, it does not require frequent manual adjustments to models and parameters, but continuously improves the accuracy and effectiveness of pricing through automatic learning. At the same time, by monitoring and analyzing the model learning process, new market trends and potential laws can also be discovered, providing more valuable information for airlines' market decisions.
[0118] 6. Enhance market share and brand influence. The dynamic pricing method based on reinforcement learning in this application can help airlines provide a better user experience in terms of price and attract more passengers to choose their flights. At the same time, with the increase in market share and the improvement of passenger satisfaction, the brand influence of airlines will also be enhanced accordingly, forming a virtuous circle and further consolidating their competitive position in the market.
[0119] In summary, the domestic air ticket dynamic pricing method and system based on reinforcement learning in this application dynamically adjusts the air ticket price by real-time analysis of multiple factors such as market environment, passenger demand and competitor ticket prices to maximize the airline's revenue. The domestic air ticket dynamic pricing method and system based on reinforcement learning in this application breaks through the limitations of traditional pricing methods, can better adapt to the rapidly changing market, and bring new opportunities and challenges to the development of the domestic aviation industry.
[0120] The present application also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.
[0121] Figure 5 FIG. 1 shows an example diagram of the architecture of an electronic device according to an embodiment of the present application. Figure 5 As shown, the electronic device includes a processor 110, a communication interface 120, a memory 130 for storing processor executable instructions, and a communication bus 140. The processor 110, the communication interface 120, and the memory 130 communicate with each other via the communication bus 140. The processor 110 implements the above method by running an executable computer program.
[0122] Among them, the computer program in the memory 130 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, 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 various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0123] The system embodiments described above are merely illustrative, wherein 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, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected based on actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.
[0124] Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that it is still possible to modify the technical solutions described in the aforementioned embodiments, or to make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A domestic air ticket dynamic pricing method based on reinforcement learning, characterized in that: The following steps are involved: Get historical air ticket data; Use historical air ticket data to build a dynamic pricing model based on reinforcement learning algorithms; Train and optimize the constructed dynamic pricing model; Utilize the trained dynamic pricing model to generate domestic air ticket pricing decisions based on the current market environment and flight information.
2. A domestic air ticket dynamic pricing method based on reinforcement learning according to claim 1, characterized in that: The use of historical air ticket data to build a dynamic pricing model based on a reinforcement learning algorithm includes: Define the environment, which includes competing airlines’ pricing, as well as passenger arrivals, purchasing behavior, and preferences for different products; Define a state space, the state space being used to represent the number of remaining seats for the current dynamically priced flight, the current time, and the flight departure time; defining an action space, the action space being used to represent a set of pricing combinations for different service classes in a flight; Rewards are defined such that the adjusted price of the ticket purchased by the passenger is the positive reward and the cost of the unsold seats at the time of flight departure is the negative reward.
3. The method for dynamic pricing of domestic air tickets based on reinforcement learning according to claim 1, characterized in that: The training and optimization of the constructed dynamic pricing model includes: Training model: Use the preprocessed data to train the constructed dynamic pricing model; Optimize the model: Improve model performance and generalization by adjusting dynamic pricing model hyperparameters, increasing training data diversity, and using regularization methods.
4. The method for dynamic pricing of domestic air tickets based on reinforcement learning according to claim 1, characterized in that: The method also includes monitoring and providing feedback on the execution of the pricing decision.
5. A domestic air ticket dynamic pricing method based on reinforcement learning according to claim 4, characterized in that: The monitoring and feedback of the execution of pricing decisions include: Collect various data related to air ticket pricing in real time; Compare and analyze the actual monitored data and calculated indicators with the expected results of the pricing model to form feedback information; Based on feedback information, the pricing model is optimized and adjusted.
6. A domestic air ticket dynamic pricing method based on reinforcement learning according to claim 1, characterized in that: When making pricing decisions, prices are gradually adjusted based on the preset price adjustment range and adjustment rules to avoid excessive price fluctuations that may cause dissatisfaction among passengers.
7. The method for dynamic pricing of domestic air tickets based on reinforcement learning according to claim 1, characterized in that: The generated pricing decision meets the preset practical constraints, which include passenger load factor, price positioning to maintain brand image, and price limits agreed with partners.
8. A domestic air ticket dynamic pricing system based on reinforcement learning, characterized in that: include: The acquisition module is used to obtain historical air ticket data and perform preprocessing; Building blocks that can build dynamic pricing models based on reinforcement learning algorithms; The training and optimization module is used to train and optimize the constructed dynamic pricing model; The generation module is used to use the trained dynamic pricing model to generate pricing decisions based on the current status information.
9. A domestic air ticket dynamic pricing system based on reinforcement learning according to claim 8, characterized in that: The system also includes a monitoring and feedback module for monitoring and providing feedback on the execution of pricing decisions.
10. An electronic device, characterized in that: The electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method according to any one of claims 1 to 7 when executing the program.