Yield management system

Abstract

To enhance revenue while minimizing lost profits from unsold seats by selling seats in high-demand sections at higher prices with lower discount rates and seats in low-demand sections at lower prices with higher discount rates.SOLUTION: In a yield management system 1, a sales optimization device 30 determines, based on the demand forecast information created by a demand forecasting device 20 and the seat inventory status managed by a seat sales management device 10, the number of seats to release for each vehicle based on the combination of usage sections and discount rates that maximizes revenue. Consequently, seats for high-demand sections are sold at higher prices with lower discount rates, while seats for low-demand sections are sold at lower prices with higher discount rates. This approach enhances revenue while minimizing lost profits from unsold seats.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a yield management system that maximizes profits by varying the price and number of seats available for sale according to the supply and demand situation, which are treated as transportation products that vary for each section of transportation, such as trains, long-distance buses, and passenger ships, where boarding and disembarking locations are predetermined. [Background technology]

[0002] Currently, the airline industry and other sectors use yield management techniques that change fares according to demand for flights. However, in inventory management in the airline industry, one seat on a flight is one seat from departure to arrival, and seats are not managed by section, such as with trains, long-distance buses, and passenger ships, where boarding and disembarking points are predetermined, and therefore yield management techniques that manage seat inventory by combining sections and discount rates have not been adopted.

[0003] Meanwhile, for railways, yield management has been proposed, which manages seat inventory by combining sections and discount rates (see Patent Document 1). Patent Document 1 describes a train seat reservation system in which a train inventory management file and a sales accumulation file are stored in the memory of a train seat reservation system, a weather information management file is stored in the memory of a weather information management system, and the train seat reservation system takes into account weather information for boarding and disembarking stations and adjusts fares just before the train departs, allowing train seats to be used at a cheaper price than usual on days such as bad weather. The train seat reservation system described in Patent Document 1 can increase occupancy rates and improve profits even on days such as bad weather.

[0004] Furthermore, in order to reduce congestion while keeping operating costs down, an information processing system that predicts future demand for transportation has also been proposed (see, for example, Patent Document 2). Patent Document 2 describes a demand prediction system that has a demand prediction unit that predicts demand for transportation, and predicts future user demand for transportation based on past statistical data, information on the latest events being held around railway lines, sensor data such as footage from surveillance cameras installed in stations, and real-time data such as the status of entrances and exits at automatic ticket gates. The demand prediction system described in Patent Document 2 makes it possible to improve profits by changing timetables and fares based on future demand predictions. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-035611 [Patent Document 2] Japanese Patent Publication No. 2020-023234 Summary of the Invention [Problem to be solved by the invention]

[0006] However, the train seat reservation system described in Patent Document 1 only varies fares based on the weather at the boarding or disembarking station, and is not a yield management technology that manages seat inventory by combining railway sections and discount rates.Furthermore, the information processing system described in Patent Document 2 only generates proposed changes to transportation schedules and fares based on demand forecasts by a demand forecasting system and presents them to managers, etc., and is not a yield management technology that manages seat inventory by combining sections and discount rates.

[0007] In other words, general yield management technology used for inventory management in hotels, airplanes, etc., and the railway seat sales technology described in Patent Documents 1 and 2 cannot predict demand for each train section and discount rate, so it is not possible to build a yield management system that optimizes the sales prices of remaining seats and maximizes profits.

[0008] Therefore, the present invention aims to provide a yield management system that increases profits while suppressing lost profits due to unsold seats by selling seats in sections with high demand at high prices with a low discount rate and seats in sections with low demand at low prices with a large discount rate. [Means for solving the problem]

[0009] In order to solve the above problems, the present invention is a yield management system that maximizes profits by changing the price and number of seats available for sale, which are treated as different vehicle use products for each section of a vehicle with predetermined boarding and disembarking locations, in accordance with supply and demand conditions, and is characterized by including: a seat sales management device that comprehensively manages the sales of seats for the vehicle; a demand forecasting device that creates demand forecast information that predicts customer demand according to the operation days and time periods of the vehicle; and a sales optimization device that performs an optimization process to determine the number of seats available for sale for each vehicle based on a combination of usage section and discount rate that will maximize profits, based on the demand forecast information created by the demand forecasting device and the inventory status of the seats managed by the seat sales management device, and registers the results of the optimization process in the seat sales management device.

[0010] In addition, in the above configuration, the demand prediction device may acquire at least past sales records of the vehicle to be predicted and information on the discount rate applicable to the vehicle to be predicted, apply these to a demand prediction model created by machine learning to predict demand for the vehicle to be predicted, and create the number of seats to be sold for each combination of the usage section and the discount rate for the vehicle to be predicted as the demand prediction information based on the demand prediction.

[0011] Furthermore, in the above configuration, the demand forecasting model provided in the demand forecasting device outputs a demand forecast value that serves as an indicator of the level of demand according to the operation day and time period of the vehicle by machine learning various elements including the day of the week arrangement, the usage section, and sales progress of the vehicle's past sales history, and the demand forecasting device is able to refer to demand forecasting data that defines the upper limit of the number of seats available for sale for each combination of the usage section and the discount rate on the operation day and time period of the vehicle, corresponding to the numerical range of the demand forecast value output by the demand forecasting model, and the demand forecasting value of the vehicle to be predicted output by the demand forecasting model may be fitted to the demand forecast data to create the upper limit of the number of seats available for sale for each combination of the usage section and the discount rate for the vehicle to be predicted as the demand forecast information.

[0012] In addition, in the above configuration, the demand forecasting device may output the demand forecast value updated according to the sales progress of the vehicle to be predicted after the seat sales management device starts selling the seats to the demand forecasting model, and fit the updated demand forecast value to the demand forecast data to update the demand forecast information.

[0013] Furthermore, in the above configuration, the sales optimization device may perform the optimization process by acquiring the demand forecast information for the vehicle to be predicted created by the demand forecasting device and the inventory status for each combination of the usage section and the discount rate for the vehicle to be predicted managed by the seat sales management device, creating corrected demand forecast information by correcting the demand forecast information taking into account the current inventory status, and solving, using any mathematical optimization calculation, the number of seats for sale that will maximize the revenue for the combination of the usage section and the discount rate in this corrected demand forecast information.

[0014] In addition, in the above configuration, the sales optimization device may, at least when the number of seats for sale in the demand forecast information created by the demand forecasting device exceeds the current inventory number managed by the seat sales management device, correct the number of seats for sale so that it is equal to or less than the current inventory number, and create the corrected demand forecast information. [Effects of the Invention]

[0015] According to the yield management system of the present invention, the sales optimization device determines the number of seats to be sold for each vehicle based on the combination of the route and discount rate that maximizes profits, based on the demand forecast information created by the demand forecasting device and the seat inventory status managed by the seat sales management device.Therefore, by selling seats in routes with high demand at a high price with a low discount rate and seats in routes with low demand at a low price with a large discount rate, profits are improved while reducing lost profits due to unsold seats. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a schematic diagram showing an embodiment of a yield management system according to the present invention. [Figure 2] FIG. 1 is a schematic explanatory diagram of demand forecasting using a demand forecasting model. [Figure 3] FIG. 10 is an explanatory diagram of the process of calculating demand forecasts for each train from the total demand for a time period. [Figure 4] (A) is an explanatory diagram of demand forecast data applied to a train with a predicted sales efficiency of 57.1%, (B) is an explanatory diagram of demand forecast data applied to a train with a predicted sales efficiency of 96.4%, and (C) is an explanatory diagram of demand forecast data applied to a train with a predicted sales efficiency of 39.4%. [Figure 5] FIG. 10 is an explanatory diagram of types of demand forecast data prepared in advance for each train and application examples thereof. [Figure 6] (A) is an explanatory diagram of dynamic pricing for trains that are expected to be crowded. (B) is an explanatory diagram of dynamic pricing for trains that are selling less well than expected. [Figure 7] (A) is an explanatory diagram that defines the variables for the optimal number of seats per section on the train that is the target of demand forecasting. (B1) is a schematic explanatory diagram of the mathematical optimization calculation that finds the variables that maximize profits. (B2) is a schematic explanatory diagram of the solution obtained by the mathematical optimization calculation. DETAILED DESCRIPTION OF THE INVENTION

[0017] Next, an embodiment of the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 shows a schematic configuration of an embodiment of a yield management system 1. This yield management system 1 is applicable to vehicles with predetermined boarding and disembarking locations, such as trains, long-distance buses, and cruise ships. Since these vehicles allow passengers to board and disembark along their journeys, seats can be sold as different vehicle use products for each section of travel. Note that the term "seat" in this invention does not simply refer to a place to sit and the associated equipment, but rather to the exclusive right of a specific location given to passengers while traveling on the vehicle. The concept of a seat also includes train berths, compartments, and cabins on cruise ships. The yield management system 1 adjusts the price and number of seats available for sale according to supply and demand conditions, thereby maximizing profits.

[0018] The yield management system 1 of this embodiment mainly comprises a seat sales management device 10, a demand forecasting device 20, and a sales optimization device 30. Note that the seat sales management device 10, the demand forecasting device 20, and the sales optimization device 30 do not necessarily need to be configured separately, and the functions of each device may be realized by a group of servers installed in a large data center or the like. In addition, in this embodiment, the vehicle is a train, and the yield management system 1 is described, which changes the number of seats sold and the sales price for each section of the train according to the demand and supply situation.

[0019] The seat sales management device 10 accesses an inventory management database 11 to comprehensively manage the sales of seats on vehicles. This seat sales management device 10 is connected to ticket vending machines at each station and online seat sales sites via a network (not shown) and receives reservation information specifying the customer's desired operating date, train name, section, facilities, etc. It searches for seats matching this reservation information, and if a seat is available, it presents the seat, along with the pre-set sales price for that seat, to the sender of the reservation information. Note that the sales price is collected by ticket vending machines at each station or online seat sales sites, and purchased tickets are issued by ticket vending machines at each station. The seat sales management device 10 only manages the sales of railway seats. The seat sales management device 10 also provides sales information (individual sales information) for individual seats sold by ticket vending machines or websites, as well as remaining seat information (vacant seat information) for each section and discount rate set for each train on each operating day, to the demand forecasting device 20 and the sales optimization device 30.

[0020] The demand prediction device 20 creates demand prediction information that predicts passenger demand according to the operation date and time period of the vehicle. The demand prediction device 20 can access a discount rate database 21 and a past performance database 22, and has a demand prediction model created by machine learning based on the learning data from these. By applying information that identifies the train to be predicted (operation date, train name, applicable discount rate, etc.) to the demand prediction model, the predicted number of seats for each discount rate on each section of the train can be obtained as demand prediction information. For example, the demand for the section of the train named "Limited Express No. 1" from Station A (starting station) to Station C (terminal station) is predicted to be 10 seats with no discount, 20 seats with 10% discount, and 30 seats with 20% discount; the demand for the section of the train named "Limited Express No. 1" from Station A (starting station) to Station B (intermediate station) is predicted to be 10 seats with no discount, 20 seats with 10% discount, and 30 seats with 20% discount; and the demand for the section of the train named "Limited Express No. 1" from Station B (intermediate station) to Station C (terminal station) is predicted to be 5 seats with no discount, 15 seats with 10% discount, and 30 seats with 20% discount, and this is created as demand forecast information. Note that the clustering of learning data and learning calculation methods required to create a demand forecast model are not particularly limited, and known existing methods such as deep learning using a multi-layered neural network may be used as appropriate. Furthermore, the demand forecasting device 20 may re-learn as learning data accumulates, and update the demand forecasting model.

[0021] The sales optimization device 30 performs an optimization process to determine the number of seats for sale for each vehicle based on a combination of route and discount rate that maximizes profits, based on the demand forecast information created by the demand forecasting device 20 and the seat inventory status managed by the seat sales management device 10. In other words, the sales optimization device 30 does not use the demand forecast information created by the demand forecasting device 20 as is for yield management, but updates the demand forecast information through an optimization process. The content of this optimization process is not particularly limited, but optimized demand forecast information can be created by correcting the demand forecast for each route so that it does not exceed the number of vacant seats known from the vacant seat information managed by the seat sales management device 10, and then using any mathematical optimization calculation to find the number of seats for sale that maximizes sales.

[0022] In the demand forecast information received by the sales optimization device 30 from the demand forecasting device 20, first, focusing on the section from Station A (starting station) to Station C (terminal station) of the train named "Limited Express No. 1," it is found that there are only 10 vacant seats, and that sales are progressing faster than the demand forecast. If the number of seats without discount in the Station A → Station C section is Z1, the number of seats with a 10% discount is Z2, and the number of seats with a 20% discount is Z3, then the cases where "Z1 = 10, Z2 = 0, Z3 = 0" satisfy "Z1 ≦ 10," "Z2 ≦ 20," "Z3 ≦ 30," and "Z1 + Z2 + Z3 ≦ 10" and the maximum profit can be obtained, so the demand forecast can be revised to "No discount: 10 seats, 10% discount: 0 seats, 20% discount: 0 seats."

[0023] Next, if we focus on the section of the train named "Limited Express No. 1" from Station A (starting station) to Station B (an intermediate station), we can see that there are only 30 seats available, and that the seats are selling faster than the demand forecast. All of these seats could be put on sale, but if there is a grace period of about one week to ten days until the train's scheduled departure, the sales situation may improve further, and the demand forecast may result in even more expensive seats being sold, so a certain number (for example, 10 seats) can be stockpiled and excluded from sale. Here, of the 30 available seats, 10 seats will be put on stock, and 20 seats will be revised to be available for sale. If the number of undiscounted seats in the Station A → Station B section is X1, the number of seats with a 10% discount is X2, and the number of seats with a 20% discount is X3, then the case in which "X1≦10", "X2≦20", "X3≦30", and "X1+X2+X3≦20" are satisfied and the maximum profit can be obtained when "X1=10, X2=10, X3=0", so the demand forecast can be revised to "undiscounted: 10 seats, 10% discount: 10 seats, 20% discount: 0 seats".

[0024] Finally, if we focus on the section of the train named "Limited Express No. 1" from Station B (an intermediate station) to Station C (the terminal station), we can see that there are 50 vacant seats, and that sales are proceeding close to the demand forecast. Here again, of the 50 vacant seats, 10 will be put into stock and 40 will be revised to be available for sale. If the number of seats with no discount on the Station B → Station C section is Y1, the number of seats with a 10% discount is Y2, and the number of seats with a 20% discount is Y3, then the cases in which "Y1 = 5, Y2 = 15, Y3 = 20" satisfy "Y1 ≦ 5," "Y2 ≦ 15," "Y3 ≦ 30," and "Y1 + Y2 + Y3 ≦ 40" and the maximum profit can be achieved are "Y1 = 5, Y2 = 15, Y3 = 20," so we can revise the demand forecast to "No discount: 5 seats, 10% discount: 15 seats, 20% discount: 20 seats."

[0025] The sales optimization device 30, having optimized the demand forecast information as described above, registers the results of the optimization process as arrangement information with the seat sales management device 10. For example, the arrangement information is created so that seats with the highest discount rate in each section are prioritized for sale, and when seats with that discount rate are sold, seats with the next highest discount rate are sold. In this way, even if the content of the demand forecast information, including multiple discount prices, is used as arrangement information, if the seat sales management device 10 has the function of sequentially switching between multiple prices (prices with different discount rates) to sell seats, the seats in stock can be sold smoothly. However, if the seat sales management device 10 does not have such a function, multiple prices with different discount rates may be offered at once for seats in a certain section of a certain train, which may confuse potential seat buyers. Therefore, for seat sales management devices 10 that do not have the function of automatically switching to the next highest discount rate when seats with a high discount rate are sold out, arrangement information that limits the sale price to one price may be created and registered in the seat sales management device 10. For example, as shown in the arrangement information in Figure 1, the ticket price and number of seats for the section from Station A to Station C for Limited Express No. 1 are arranged as "10 seats with no discount", the ticket price and number of seats for the section from Station A to Station B are arranged as "10% discount: 10 seats" only, excluding "10 seats with no discount", and the ticket price and number of seats for the section from Station B to Station C are arranged as "20% discount: 20 seats" only, excluding "10% discount: 15 seats" and "5 seats with no discount". In other words, if arrangement information is created so that only the products with the highest discount rate for each section are available for sale, it can be used with a seat sales management device 10 that does not have an automatic sales price switching function.

[0026] According to the yield management system 1 of this embodiment, the sales optimization device 30 determines the number of seats to be sold for each vehicle based on the combination of the route and discount rate that maximizes profits, based on the demand forecast information created by the demand forecasting device 20 and the seat inventory status managed by the seat sales management device 10.Therefore, by selling seats in routes with high demand at a high price with a low discount rate and seats in routes with low demand at a low price with a large discount rate, profits are improved while suppressing lost profits due to unsold seats.

[0027] Note that the arrangement information is not limited to being automatically created and automatically registered by the sales optimization device 30, but may also be configured so that a person with management authority, such as a yield manager, can check the optimized demand forecast information and arrangement information and manually make changes. If the yield manager can reflect the knowledge he or she has gained from many years of experience, it will be possible to take appropriate action in cases such as when the demand forecast model of the demand forecasting device 20 creates abnormal demand forecast information.

[0028] Furthermore, it is desirable that the creation of demand forecast information by the demand forecasting device 20, the correction of demand forecast information by the sales optimization device 30, and the creation and registration of arrangement information be performed daily outside of the sales hours of the seat sales management device 10. In this way, the seat sales price can be dynamically changed according to the progress of seat sales, so that seats in sections where demand has changed to high can be sold at a high price with a reduced discount rate, thereby reducing lost profits due to cheap seat sales, and seats in sections where demand has changed to low can be sold at a low price with a large discount rate, thereby reducing lost profits due to unsold seats, thereby achieving further profit improvement.

[0029] Next, the functions of the demand forecasting device 20 will be described in detail with reference to FIGS.

[0030] As shown in Figure 2, the demand forecasting device 20 creates a demand forecasting model by machine learning using the past sales performance of the train being forecast (such as the day of the week, the section used, and the sales progress as determined by individual sales information from the seat sales management device 10) as learning data. The demand forecast value obtained by the demand forecasting model is not particularly limited; it can be a numerical value that indicates the level of demand depending on the train's operation day and time period, obtained by machine learning various elements of the train's past sales performance, including the day of the week, the section used, and the sales progress. The demand forecasting model in Figure 2 outputs the number of seats predicted to be sold as the demand forecast value. This is a simplified illustration of the processing when the number of seats in demand for the section from Station A to Station B during a specific time period (between 8:00 and 10:00) on an operation day (the first day of the March Golden Week holidays) is obtained two days before the operation day. In reality, the demand forecast is performed for the combination of all facilities, all products, and all sections of the train in question.

[0031] Based on the demand forecast value obtained by the demand forecasting model, it is predicted that there will be a demand of 568,284 seats in the 8:00 a.m. time slot on an operating day. Furthermore, as shown in Figure 3, a timetable is scheduled for four trains in the 8:00 a.m. time slot on an operating day, meaning that the total demand for each train is 568,284 seats. If there were no imbalance in demand for each train, it would be sufficient to allocate 25% to each train. However, in reality, there is imbalance in demand for trains due to factors such as the popularity of train cars and transfer connection times. Therefore, the demand allocation ratio for each train in the 8:00 a.m. time slot can be calculated from past performance data, and the total demand (568,284 seats) can be allocated according to the allocation ratio to obtain the demand forecast value for each train. Alternatively, an allocation ratio prediction model that predicts the allocation ratio for each train can be created and the predicted allocation ratio for each train can be used.

[0032] In Figure 3, the allocation ratio for "Limited Express No. 1" is 28%, so the demand forecast value for Limited Express No. 1 is "568.284 seats x 0.28 ≒ 159.120 seats." Similarly, the allocation ratio for "Limited Express No. 2" is 26%, so the demand forecast value for Limited Express No. 2 is "568.284 seats x 0.26 ≒ 147.754 seats." The allocation ratio for "Limited Express No. 3" is 25%, so the demand forecast value for Limited Express No. 3 is "568.284 seats x 0.25 = 142.071 seats." The allocation ratio for "Limited Express No. 4" is 21%, so the demand forecast value for Limited Express No. 4 is "568.284 seats x 0.21 ≒ 119.339 seats."

[0033] In the above-mentioned demand forecasting model, a simple example was shown in which a demand forecast value was obtained as a single price without considering the discount rate, but it is essential to take the discount rate into consideration for yield management. Therefore, in reality, the number of seats available for sale for each combination of the route and discount rate for the transportation to be forecasted must be calculated using the demand forecasting model, and demand forecast information must be created, which increases the calculation load of the demand forecasting model.

[0034] Therefore, the predicted sales efficiency, which quantifies train demand, is used as a demand forecast value that serves as an indicator of the level of demand according to the train's operation date and time period, and demand forecast data is prepared in advance that defines the upper limit of the number of seats available for sale for each combination of the service section and discount rate on the train's operation date and time period, corresponding to the numerical range of the predicted sales efficiency.In other words, by applying the predicted sales efficiency obtained from the demand forecast model to the demand forecast data, the upper limit of the number of seats available for sale for each combination of the service section and discount rate for the train to be predicted is uniquely determined, making it possible to create the demand forecast information.

[0035] As shown in Figure 4(A), if the predicted sales efficiency of the train "Limited Express No. 10" on a certain day is 57.1%, by applying this to the demand forecast data prepared for "Limited Express No. 10" on that day, demand forecast information can be created from the recommended upper limit number of seats corresponding to a demand forecast efficiency of 40 to 59%. That is, the recommended maximum number of seats for the "10% discount" product between Station A and Station C is "0 seats," the recommended maximum number of seats for the "15% discount" product between Station A and Station C is "20 seats," the recommended maximum number of seats for the "30% discount" product between Station A and Station C is "0 seats," the recommended maximum number of seats for the "40% discount" product between Station A and Station C is "10 seats," the recommended maximum number of seats for the "10% discount" product between Station A and Station B is "0 seats," the recommended maximum number of seats for the "15% discount" product between Station A and Station B is "15 seats," and so on are determined, and therefore the upper limit of the number of seats available for sale for each combination of the route and discount rate for the train being predicted can be created as demand forecast information. Note that the setting range for the predicted sales efficiency for setting the upper limit number of seats is shown in 20% increments as an example, but it is not limited to this and can be set to any range. For example, if it could be set in 5% increments (0% to less than 5%, 5% to less than 10%, 10% to less than 15%, ..., 90% to less than 95%, 95% to 100%), it would be possible to fine-tune the maximum number of seats.

[0036] On the other hand, as shown in Figure 4(B), if the predicted sales efficiency of the train "Limited Express No. 10" on another day is 96.4%, by applying this to the demand forecast data prepared for "Limited Express No. 10" on that day, demand forecast information can be created from the recommended upper limit number of seats corresponding to a demand forecast efficiency of 80% or more. That is, the recommended maximum number of seats for the "10% discount" product between Station A and Station C is "5 seats," the recommended maximum number of seats for the "15% discount" product between Station A and Station C is "0 seats," the recommended maximum number of seats for the "30% discount" product between Station A and Station C is "0 seats," the recommended maximum number of seats for the "40% discount" product between Station A and Station C is "2 seats," the recommended maximum number of seats for the "10% discount" product between Station A and Station B is "5 seats," the recommended maximum number of seats for the "15% discount" product between Station A and Station B is "0 seats," and so on. Demand forecast information can be created using the upper limit of the number of seats available for sale for each combination of the route and discount rate for the train being forecast. In this way, when the predicted sales efficiency is high, demand forecast information is created that increases profits by reducing the sales of products with high discount rates and increasing products with low discount rates (or products with no discount).

[0037] Furthermore, as shown in Figure 4(C), if the predicted sales efficiency of the train "Limited Express No. 18" on a certain day is 39.4%, by applying this to the demand forecast data prepared for "Limited Express No. 18" on that day, demand forecast information can be created from the recommended upper limit number of seats corresponding to a demand forecast efficiency of less than 40%. That is, the recommended maximum number of seats for the "10% discount" product in the Station A-C section is "0 seats," the recommended maximum number of seats for the "15% discount" product in the Station A-C section is "0 seats," the recommended maximum number of seats for the "30% discount" product in the Station A-C section is "40 seats," the recommended maximum number of seats for the "40% discount" product in the Station A-C section is "20 seats," the recommended maximum number of seats for the "10% discount" product in the Station A-B section is "0 seats," the recommended maximum number of seats for the "15% discount" product in the Station A-B section is "30 seats," and so on. Demand forecast information can be created using the maximum number of seats available for sale for each combination of the route and discount rate for the train being forecast. In this way, when the predicted sales efficiency is low, demand forecast information is created that reduces sales of products with low discount rates (or products with no discount) and increases products with high discount rates, thereby reducing lost profits due to unsold seats and improving profits.

[0038] As mentioned above, when using the predicted sales efficiency obtained from a demand forecasting model, it is important to set an appropriate recommended upper limit number of seats in the demand forecast data to which the predicted sales efficiency is applied. Therefore, we prepared demand forecast data for each train for each day of the week. For example, as shown in Figure 5, we prepared 12 types of demand forecast data: "Weekday," "Day before a Holiday," "Friday / Day before a Long Weekend," "Holiday," "Saturday," "Sunday," "First Day of a Long Weekend," "Midday of a Long Weekend," "Last Day of a Long Weekend," "Three Busy Seasons Travel Peak," "Three Busy Seasons Travel Peak," and "Three Busy Seasons Return Peak." For the same Limited Express No. 10, comparing the recommended upper limit number of seats in the weekday demand forecast data (used when the train operates on a weekday) with the recommended upper limit number of seats in the demand forecast data for the three busiest seasons travel peak (used when the train operates on the first day of a long weekend) reveals that there are fewer discounted products in the demand forecast data for the three busiest seasons travel peak. The New Year's holiday, Golden Week, and Obon holidays are considered the three busiest periods when demand for trains increases significantly, and it is expected that even low-discounted products (or no discounted products) will sell well. Therefore, even if the predicted sales efficiency is low (for example, less than 40%), profits can be improved by not selling high-discounted products and instead limiting the number of low-discounted products to a small number (for example, only 20 seats with a 10% discount).

[0039] The demand forecast data prepared for each train is not limited to the weekday arrangement shown in FIG. 5 , but may be narrowed down to a smaller number of types, or a more detailed weekday arrangement may be set. Furthermore, when an increase in train demand is expected due to the attendance of a concert held at a specific date and time or an event planned for a specific period, the demand forecasting device 20's standard functions alone cannot predict demand, but this can be addressed by adding corrections to the demand forecast information. A demand correction factor database 23 (shown by the dashed line in FIG. 1 ) is provided to acquire and manage various events that cause demand corrections from a webpage or the like, and a correction value to be added to the demand forecast is defined as an influence correction value for each correction factor. Before and after an event that causes demand corrections, the number of train passengers increases due to travel to and from the venue. Therefore, the increase rate is determined according to the scale of the event, and the original demand is increased according to the increase rate. By performing appropriate demand corrections, the increase in demand due to the event can be handled. For example, if a large-scale live event is expected to result in a 90% increase in demand when people arrive or return home, the increase rate of 90% is used as the influence correction value, and demand correction is performed by multiplying the normal demand forecast for trains operating during that time period by 1.9. In this way, if influence correction values ​​can be set according to the demand correction factors and the demand forecast can be corrected, it will be possible to appropriately respond to spot demand fluctuations that cannot be saved and managed on an annual calendar.

[0040] Furthermore, it is desirable that the demand forecast information be created by the demand forecasting device 20 not only before the start of sales when all seats are in stock, but also periodically or irregularly after the start of seat sales by the seat sales management device 10. In other words, the demand forecast information can be updated by outputting a demand forecast value that takes into account the sales progress of the train being predicted using the individual sales information and vacant seat information supplied from the seat sales management device 10 to the demand forecast model, and applying the updated demand forecast value to the demand forecast data.

[0041] For example, as shown in Figure 6(A), for a train with an initial demand forecast (forecasted sales efficiency) of 93% and expected to be crowded, a recommended upper limit of seats of over 80% was applied based on the demand forecast data. The daily updated forecasted sales efficiency also remained high, exceeding 80%, so the sales period ended without increasing discounted items. On the other hand, as shown in Figure 6(B), for a train with an initial demand forecast (forecasted sales efficiency) of 65%, a recommended upper limit of seats of 60-79% was applied based on the demand forecast data. However, as the daily updated forecasted sales efficiency decreased, for example, when the updated forecasted sales efficiency fell below 60% (reaching 59%), an automatic response was implemented to change the recommended upper limit of seats from 60-79% to 40-59%, increasing discounted items, and reducing unsold seats. In this way, by updating demand forecast information based on demand forecast values ​​that change depending on daily sales, a dynamic pricing system that flexibly adjusts prices according to seat demand can be easily implemented. The timing for reflecting changes in the demand forecast data in the seat sales price is not particularly limited, but it is desirable to do so daily when sales for the day are completed by the seat sales management device 10. For example, if demand forecasting is performed and the demand forecast information is updated when the individual sales information and vacant seat information for the day are confirmed, it is possible to reflect a seat sales price appropriate for the sales situation on the day, and it is possible to sell seats the next day at an appropriate sales price.

[0042] Next, the function of the sales optimization device 30 will be described in detail with reference to FIG.

[0043] The sales optimization device 30 acquires the demand forecast information for the train to be forecast created by the demand forecasting device 20 and the inventory status for each combination of the service section and discount rate for the train to be forecast managed by the seat sales management device 10, and creates corrected demand forecast information by correcting the demand forecast information taking into account the current inventory status. In creating this corrected demand forecast information, at least when the number of seats for sale in the demand forecast information created by the demand forecasting device 20 exceeds the current inventory number managed by the seat sales management device 10, it is essential to correct the number of seats for sale so that it is equal to or less than the current inventory number. In addition to this, there are also cases where a correction is made to secure stock for later sales.

[0044] Furthermore, the sales optimization device 30 performs an optimization process to determine the number of seats available for sale for each train that maximizes revenue from the combination of the route and discount rate in the corrected demand forecast information using any mathematical optimization calculation. Mathematical optimization calculation is a calculation technique that uses mathematical methods and algorithms to solve the problem of maximizing or minimizing a specific objective function under given constraints, and there are no particular limitations on the setting of the mathematical model for determining the number of seats available for sale or the analysis method used.

[0045] An example of the optimization process will now be described with reference to FIG. 7. As shown in FIG. 7(A), assume a route that runs from Station A to Station C via Station B. If the number of seats on the train is 100, before the seats are sold, there are 100 vacant seats from Station A to Station B, 100 vacant seats from Station B to Station C, and 100 vacant seats from Station A to Station C. Let X be the optimal number of seats to allocate in the section from Station A to Station B, Y be the optimal number of seats to allocate in the section from Station B to Station C, and Z be the optimal number of seats to allocate in the section from Station A to Station C. The sales price for the Station A → Station B section is 11,000 yen, and the predicted value of demand based on the demand forecast information is 50 seats. The sales price for the Station B → Station C section is 6,500 yen, and the predicted value of demand based on the demand forecast information is 30 seats. The sales price for the Station A → Station C section is 15,000 yen, and the predicted value of demand based on the demand forecast information is 100 seats (see FIG. 7(B1)).

[0046] Using the optimal seat numbers X, Y, and Z, the objective function that maximizes profits can be defined as "Sales amount 11,000 yen × Optimal seat number X + Sales amount 6,500 yen × Optimal seat number Y + Sales amount 15,000 yen × Optimal seat number Y." There are three constraints on the variables X, Y, and Z. (1) The first constraint must be satisfied: "X + Z ≦ 100 seats" and "Y + Z ≦ 100 seats" so that the number of available seats in the overlapping sections is not exceeded. (2) The second constraint must be satisfied: "X ≦ 100," "Y ≦ 100," and "Z ≦ 100" so that the optimal number of seats in each section does not exceed the number of available seats. (3) The third constraint must be satisfied: "0 ≦ X ≦ 50," "0 ≦ Y ≦ 30," and "0 ≦ Z ≦ 100" so that the optimal number of seats in each section does not exceed the predicted demand.

[0047] When the variables X, Y, and Z that satisfy these first to third constraints and maximize profit are determined through mathematical optimization, the solution of X = 30, Y = 30, and Z = 70 is obtained (see Figure 7(B2)). In other words, the maximum profit from the optimal number of seats presented as the recommended yield value is "sales amount 11,000 yen x 30 seats + sales amount 6,500 yen x 30 seats + sales amount 15,000 yen x 70 seats = 1,575,000 yen." Note that in the example of calculating the optimal number of seats described above, for simplicity's sake, sales amounts for each discount rate were not set. However, in reality, optimization processing is performed to calculate the optimal number of seats for the combination of all facilities, all products for each discount rate, and all sections, and post-optimization corrected demand forecast information is created.

[0048] Furthermore, the sales optimization device 30 can automatically register the results of optimizing the corrected demand forecast information in the seat sales management device 10, but it may also be equipped with a reporting function that presents the optimization results to a yield manager or the like before automatic registration. For example, if a yield manager or the like modifies the optimization results viewed via a yield management terminal or the like, the modified results are reflected, and the sales optimization device 30 creates arrangement information based on the modified optimization results and registers it in the seat sales management device 10, yield management that reflects the knowledge of the yield manager or the like is possible.

[0049] The above describes an embodiment of the yield management system according to the present invention based on the attached drawings, but the present invention is not limited to this embodiment, and may be implemented by adapting publicly known, existing equivalent technical means as long as the configuration described in the claims is not changed. [Explanation of symbols]

[0050] 1. Yield Management System 10 Seat sales management device 11 Inventory Management Database 20 Demand forecasting device 21 Discount Rate Database 22 Past performance database 23 Demand Correction Factor Database 30 Sales optimization device

Claims

1. A yield management system that maximizes profits by varying the price and number of seats available for sale according to the supply and demand situation, which are treated as transportation products that vary for each section of a transportation where boarding and disembarking locations are predetermined, a seat sales management device that comprehensively manages sales of the seats of the vehicle; a demand forecasting device that generates demand forecast information that predicts passenger demand according to the operation date and time period of the vehicle; a sales optimization device that performs an optimization process to determine the number of seats available for sale for each vehicle based on the demand forecast information created by the demand forecast device and the seat inventory status managed by the seat sales management device, combining the route and discount rate that maximizes profits, and registers the results of the optimization process in the seat sales management device; A yield management system comprising:

2. The yield management system described in claim 1, characterized in that the demand prediction device acquires at least past sales records of the vehicle to be predicted and information on the discount rate applicable to the vehicle to be predicted, applies this to a demand prediction model created by machine learning to predict demand for the vehicle to be predicted, and creates the number of seats to be sold for each combination of the usage section and the discount rate for the vehicle to be predicted as the demand prediction information based on the demand prediction.

3. The demand forecasting model provided in the demand forecasting device outputs a demand forecast value that serves as an index of the level of demand according to the operation day and time period of the vehicle by machine learning various elements including the past sales record of the vehicle, the day of the week arrangement, the use section, and sales progress, The demand forecasting device is capable of referencing demand forecast data that defines an upper limit on the number of seats available for sale for each combination of the usage section and the discount rate on the operation day and time period of the vehicle, corresponding to the numerical range of the demand forecast value output by the demand forecasting model, and fits the demand forecast value of the vehicle to be predicted output by the demand forecasting model to the demand forecast data, thereby creating the upper limit on the number of seats available for sale for each combination of the usage section and the discount rate for the vehicle to be predicted as the demand forecast information.

3. The yield management system of claim 2.

4. The yield management system described in claim 3, characterized in that the demand forecasting device outputs the demand forecast value updated according to the sales progress of the vehicle to be predicted after the seat sales management device starts selling the seats to the demand forecasting model, and fits the updated demand forecast value to the demand forecast data to update the demand forecast information.

5. The sales optimization device acquires the demand forecast information for the vehicle that is the target of prediction created by the demand forecasting device and the inventory status for each combination of the usage section and the discount rate for the vehicle that is the target of prediction managed by the seat sales management device, creates corrected demand forecast information by correcting the demand forecast information in consideration of the current inventory status, and performs the optimization process by solving, by any mathematical optimization calculation, the number of seats for sale that maximizes the revenue for the combination of the usage section and the discount rate in this corrected demand forecast information. A yield management system according to any one of claims 1 to 4.

6. The yield management system described in claim 5, characterized in that the sales optimization device creates the corrected demand forecast information by correcting the number of seats for sale so that it is less than or equal to the current inventory number, at least when the number of seats for sale in the demand forecast information created by the demand forecasting device exceeds the current inventory number managed by the seat sales management device.

Images