JP6850310B2 - Methods and servers for providing quoted prices, such as sets of airfare price quotes - Google Patents

Description

The field of technology is the estimated price, such as airfare, railroad fare, hotel price, goods (actually any type of goods or services, the prices are not fixed, but are variable instead. Thus, with respect to methods, servers, and computer program products for providing a set of price quotes for goods or services for which it is useful to be able to provide quotes for those prices.

In the airfare context, the price of an airfare ticket for an itinerary is complicated to calculate. It is the result of calculating possible routing, certified fares, and the availability of these fares. These calculations are typically performed in the Global Distribution System (GDS), and executing queries against such systems can be slow, expensive, and energy-intensive to perform the calculations. You need to send the results.

Remembering ticket prices and flight availability for all possible airports between all possible routes for a sufficiently long range (eg, one year) requires extremely significant data storage capacity. Would be.

FR284118 (B1) is a method for loading flight schedule changes, in which steps are taken to receive at least one set of flight changes and individual changes within that set are retrieved and stored in registers for future scheduling. The steps to remember in and the steps to simulate the reassignment of reservations related to flight schedule changes by accessing both the record and flight schedule database via the reservation delivery server, and the reservation margin. A method is disclosed that includes, and a step, to finally update the minute flight schedule database.

As an example, in the context of airfare flight pricing, flight pricing has traditionally been derived from GDS (how the GDS system works, and more generally how flight pricing works). See Section C for a more detailed description of.) Flight comparison services (eg Skyscanner) and some airlines pay a fee to get a live bookable price from the GDS. These prices are the actual reservable prices that prospective passengers can book. However, many prospective passengers often only browse and do not require a clearly accurate bookable price and will instead be satisfied with the quote. As a result, it would be very useful if we could devise a method and system for estimating accurate prices, especially for third party resources (eg GDS) (not only expensive to access, but also). This is the case if reliance on access to access (which may not be available to obtain the required information) is removed.

Therefore, obtaining a price and flight availability quote between any two airports in such a way that it is not too late to access one or more remote servers multiple times to get a quote. Therefore, it is desirable to provide a method that does not require the enormous amount of data storage required to pre-store the results for all possible queries.

According to the first aspect of the present invention, there is a method of providing a price estimate, in which (i) a computer server receives a request for a price for the goods or services, eg, an airfare as well as a parameter defining those goods or services. As in step 1 and (ii) determining an estimated price from an incomplete historical price data set at any time to step (i) above by analyzing patterns in that data set, 1 Or the step of configuring multiple processors and the step of configuring one or more processors to calculate an estimate for the requested price for an article or service that meets the (iii) parameters, and (iv) the end user of the price estimate. • Methods are provided that include sending to a computing device, such as a personal computer, smartphone or tablet.

One of the advantages is that this method requires the time or energy required to access one or more remote servers multiple times to actually set the fare availability and calculate the fare price. Do not do. A further advantage is that the method does not require the enormous amount of data storage required to pre-store the results for all possible queries. The computer that performs this method operates in a new way and solves the recognized problem by this method. The steps (ii) and (iii) in this method are not necessarily separate steps, but may be part of the same step instead.

The method includes parameters that define those goods or services, such as airfare, hotel reservation, rail fare; date range; destination; origin; desired weather conditions; star rating; keywords; any other It may include one or more of the user-defined preferences.

The method may determine the estimated price by estimating, deriving, or predicting the estimated price.

The method involves, in step (ii), (a) obtaining past price estimates from a computer data store, (b) grouping past price estimates by category, and (c) for each group. Derivation of statistical data, (d) storing a classifier containing the derived statistical data for each group on a computer, and (e) using the stored classifier, the required price corresponds. It may include identifying a group.

In step (iii), the method calculates a set of quotes for a requested price over a specified date range, using statistical data obtained from a stored classifier corresponding to the identified group. May be included.

The method may include step (ii) analyzing patterns in the dataset using rules.

In the method, step (ii) may include a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model may be used to predict unobserved prices.

The method may train the classifier with one observed price and the corresponding set of features.

Is this method characteristically related to the request, the departure date of the week, the length of stay, and does the trip involve spending a Saturday night at the destination? It may include one or more of the airline, time to travel, route, month.

The method then involves the classifier predicting the price of an unobserved price by being given a set of features and by giving the price most likely to have those features. You may.

In this method, features may be derived by training a plurality of models using different features and comparing the prediction accuracy of different models.

The method includes, in step (ii), building a statistical model from past prices, identifying missing quote candidates, and pricing quote candidates based on the statistical model. It may be the one that is used.

In this method, the price is estimated for each candidate estimate, the step of extracting the category feature value from the estimate, the step of extracting the trained classifier for the extracted category from the database, and the estimation candidate. Steps to extract all feature values from, classify candidate estimates, calculate Bayes posterior probabilities for each price range stored in the classifier, and price ranges with the highest Bayesian posterior probabilities. It may be done in steps performed by selecting a class and attaching a price class to a candidate quote.

The method may include a list of routes, a classifier categorization scheme, past estimates, and a set of supported features with weights in the input to the statistical model.

The method may filter past estimates by age.

In this method, the input to the statistical model may include a reverse path equivalent.

The method may include a step that includes the cashed fare price in the price quote set.

The method may be such that the price quote set is configured without inquiring the distribution system after step (i).

In this method, the price may be for a trip that is a one-way trip.

In this method, the price may be for a trip that is a round trip.

In this method, the price may include the airfare price.

In this method, the price may include the railroad fare price.

In this method, the price may include the rental vehicle price.

In this method, the price may include the hotel price.

In this method, the request may include a flexible search request.

The method may be such that the final result of the method is at most an input date or one estimate per input date.

The method is such that (A) at any point in time to step (i), determining the confidence range of the estimated price from an incomplete historical price dataset is performed by analyzing the patterns in that dataset. Includes steps to configure one or more processors and (B) to configure one or more processors to calculate the confidence range of the estimated price for the requested price for goods or services that meet the parameters. There may be.

The method may further include (C) providing a fare price quote along with a confidence range to an end-user computing device, such as a personal computer, smartphone, or tablet.

The method may further include the step of using a confidence range to determine whether to display the price to the user or to provide the user with an likely price range.

The method may display the likely price range as an error bar.

In this method, the date range may include only one departure date.

In this method, the date range may include only one return date.

The method may include a method of providing the best set of round-trip fare price estimates for a specified departure date range and a specified return date range. Step (i) includes receiving a request for a round-trip fare price for a specified departure date range and a specified return date range for a trip from origin to destination, and step (iv) includes a specified departure. Includes providing the best set of round-trip fare price quotes for the day range and the specified return date range.

The method may provide the best round-trip fare price quote set in bar chart format.

In this method, the estimation process adds the minimum Bayesian posterior probability required to tolerate the classification result, the maximum number of path operators involved in candidate generation, or the random variation added to the Bayesian posterior probability to avoid ties. It may be parameterized by one or more of them.

The method may be performed on a server on which a service for estimating fare availability and price is provided.

According to a second aspect of the present invention, a server configured to provide a price quote, (i) receives a request for a price for the goods or services, eg, an airfare as well as a parameter defining those goods or services. That is, the parameters include the type of action, such as airfare, hotel reservation, railfare; date range; destination; origin; desired weather conditions; star rating; keywords; of any other user-defined preference. Analyzing patterns in a data set that includes one or more of the above, (ii) determining an estimated price from an incomplete past price data set at any time for (i) above. A server is provided that is arranged to do so, to calculate an estimate for the requested price for an article or service that meets the (iii) parameters, and to provide an (iv) price estimate. ..

The server tells (ii) that the server (a) obtains past prices from computer data stores, (b) groups past prices by category, and (c) each group. (D) For each group, store the classifier containing the derived statistical data, and (e) identify the group using the stored classifier corresponding to the required price. It may be arranged to do and to do.

The server, for (iii), uses statistical data obtained from a stored classifier corresponding to a group identified by the server as calculating a set of estimates for the requested price over a specified date range. It may be arranged so as to be performed.

The server may be further arranged to perform the method of any of the methods according to the first aspect of the invention.

According to a third aspect of the present invention, a computer program product embodied on a non-temporary storage medium, the computer program product, when executed on the computer, provides a price quote. When executed on a computer, a computer program product is arranged to (i) receive a request for a parameter that defines the goods or services along with a price for the goods or services, such as an airfare, in the parameters. Includes the type of action, eg, airfare, hotel reservation, railroad fare; date range; destination; origin; desired weather conditions; star rating; keywords; any other user-defined preference. To receive, and (ii) determine an estimated price from an incomplete historical price data set at any time to (i) above, by analyzing the patterns in that data set. A computer program product is provided that is arranged to calculate an estimate for a required price for an article or service that meets (iii) parameters and (iv) provide a price estimate.

The computer program product may be arranged to perform a method that belongs to any of the methods according to the first aspect of the invention.

According to a fourth aspect of the present invention, a computer terminal connected to a server of any of the servers according to the second aspect of the present invention, wherein the computer terminal is designated for a trip from a departure point to a destination. A computer terminal is provided that is arranged to send a request for a price quote for a date range and is further arranged to receive a price quote.

According to a fifth aspect of the present invention, an article or service, for example a ticket, for which a fare estimate for the article or service can be made using the method according to the first aspect of the present invention, or a second aspect of the present invention. The goods or services provided using the server according to the above embodiment are provided.

According to a sixth aspect of the present invention, a web-based price comparison website in which an end user interacts with the web-based price comparison website to request a price for goods or services. And its web-based price comparison website is a server that estimates its request using any of the methods according to the first aspect of the invention. A web-based price comparison website will be provided.

Next, the above-mentioned and other aspects of the present invention will be described with reference to the following figures, merely as an example.

FIG. 5 is a diagram of an embodiment of a system for estimating a price from another price, with the purpose of using that price to increase the cash of the observed fare. It is a figure which shows the example of how to derive the availability of a fare class from a live update using a "difference" engine. It is a figure which shows the example of the formula suitable for use in the naive Bayes classification. It is a figure which shows the Example which the feature value is represented by the numbered area, and the frequency information is represented by the price range (bucket). It is a figure which shows the Example of the whole training process. It is a figure which shows the Example which can price each candidate estimate placeholder using a naive Bayes classifier. It is a figure which shows the Example of the whole process which infers a missing estimate. FIG. 5 illustrates an example of a connection between a server that provides a fare availability and price quote service, an input data source, and an output for partners and users. It is a figure which shows the Example of the user interface. It is a figure which shows the Example of the user interface. It is a figure which shows the Example of the system for performing a search process. It is a figure which shows the Example of the user interface. It is a figure which shows the Example of the user interface output. It is a figure which shows the Example of the user interface output. It is a figure which shows the example of the architecture for browsing the search. FIG. 5 illustrates an example of an overall system architecture for a system that includes the provision of fare availability and price quote services. It is a figure which shows the GDS configuration example. It is a figure which shows the example of the fare about the fare class record, the fare category, and the category data table. It is a figure which shows the embodiment of the process which provides the fare pricing option. It is a figure which shows the Example of the relationship between a pricing engine, AVS / NAVS, seamless (polling), a cache DB, and an airline host. It is a figure which shows the system example including the availability server. It is a figure which shows the graphic display example of the confidence range with respect to the estimated price using a bar, and the bar ranges from the lowest price to the highest price in the range where the price should be found. Bars are provided for outbound and inbound day ranges.

One implementation of the present invention provides "model-based pricing". This is described in Section A below.

A. Model-based pricing Provides flight pricing quotes using model-based data interpolation.

Background / Motivation Ticket prices for itineraries are complex to calculate. It is the result of calculating possible routing, certified fares, and the availability of these fares. These calculations are typically performed in a global distribution system (GDS), and making queries to such a system can be slow and costly.

Airlines that file fares in the GDS typically use a pricing model called "scheduled". The sum of such important things should be noted about the model, if the round-trip travel the day O ₁ to return to the starting day R _1, the price of round-trip ticket P (O _{1, R 1),} the two one-way ticket It is not equal to P (O ₁ ) + P (R _1). An alternative pricing model (referred to as "low budget" pricing) is a model in which _{P (O 1} , R ₁ ) = P (O ₁ ) + P (R _1).

Two important consequences of the scheduled price model are the limits imposed on the reuse of observed prices and the space required to store them.

In the low budget model, if two "round trip" queries are made, one for date O ₁ R ₁ and the other for date O ₂ R ₂ , the resulting price P (O). _{Using 1} ), P (R ₁ ), P (O ₂ ), and P (R ₂ ), up to four round-trip fares (ie P (O ₁ , R ₁ ), P (O ₁ , R ₂₎ ), P (O ₂ , R ₁ ), P (O ₂ , R ₂ )), and the four fares P (O ₁ ), P (R ₁ ), P ( It has O ₂ ) and P (R ₂ )). If the query were run against a "scheduled" model, there would be only two fares: P (O ₁ , R ₁ ) and P (O ₂ , R ₂ ). Thus, in the low-budget model, two queries can answer eight queries, while in the schedule model, two queries can answer two queries.

Assuming that flight prices are available only on that day and the next 364 days, finding all possible one-way and round-trip fares on the route is possible by making 365 round-trip inquiries. Become. For the scheduled model, 365 ^* (365 + 1) / 2 = 66,795 round-trip inquiries and 365 one-way inquiries (in each direction) would be required. Thus, building a complete list of prices for routes requires about 185 times more queries for the scheduled model than for the low-budget model.

Having the results of all these inquiries would inevitably increase the storage space required to store fares by about 185 times.

The above is merely an example for a route between two defined airports. In reality, there are about 5,000 private airports in the world. Therefore, the number of possible trips between airports is about (5,000) ² . Considering the number 365 ^* (365 + 1) / 2 = 66,795 possible round-trip inquiries mentioned above, this results in the total number of possible inquiries between all airports per year (5,000) ^{2 *} 365 ^*. (365 + 1) / 2 is obtained, which is about 10 ¹² . The implied amount of data storage is actually very large.

The significant increase in the number of queries required to form the "big picture" of combining scheduled model paths with the cost of GDS makes it impractical (ie, inconvenient) to build such a cache. It means). And the result is incomplete cash, and there are "holes" where fares have not been observed.

However, the price (returned by GDS) is the result of applying fare rules and availability. As a result, there are patterns and consistency in the observed fares. In this embodiment, it is possible to interpolate the fare missing from the cash based on using these patterns, using a model that reflects the fare rule that generated the price in the first place.

A system for estimating a price from another price A system for estimating a price from another price, which is used to be observed (for example, requesting a price for a live bookable fare by a user). (Observed by) is provided with the purpose of increasing fare cash. A model embodiment is shown in FIG.

The estimator predicts other estimates using the statistical data derived from the estimate. With a very simple estimator, it is possible to calculate that all missing fares are equal to the average of all observed fares. Prediction levels can be improved by making the model more sophisticated to take into account the type of rule used to generate the fare in the first place.

For example, a common "rule" that governs whether a fare is valid for an itinerary is whether the itinerary involves staying on a Saturday night. Typically, the implication of such a rule is that dates that involve not spending Saturday nights at the destination are much more expensive. The rules are usually designed to penalize business travelers who wish to return during the working week. In this way, a better model containing the two mean values can be calculated. That is, the average of all observed fares with a Saturday night stay and the average of all observed fares without a Saturday night stay. The algorithm then predicts the price for a previously unobserved date pair by first calculating whether there was a Saturday night between the date pairs and then returning the corresponding average. Do it. As a result, the accuracy of the estimation is increased.

Similar improvements can be made to the model. For example, travel prices vary depending on the day of the week (Monday is typically more expensive than Wednesday). Therefore, calculating and using an average that also takes into account the days of the week would potentially improve model accuracy.

Therefore, providing a statistical model for predicting fares may be done, for example, using statistical data and rules that may be derived by humans. Alternative methods of forming statistical models are also possible. This approach uses a machine learning approach called the "naive Bayes classifier" to form a probabilistic model of fares and uses that model to predict unobserved fares. Classifier training is performed using the observed price and the corresponding set of "features". Features could include weekly departure dates, length of stay, Saturday stays, airlines, etc. for inquiries that generated fares. The classifier then predicts the price of the unobserved fare by being given a set of features and giving the price most likely to have those features.

With the following reference to a statistical model for predicting prices, you can predict fares that have not been observed before, or you can predict fares that have been observed (if you want to measure the accuracy of the model).

Key Features and Their Computational Features that may be used to train the classifier include:
・ Saturday stay ・ Number of days of stay ・ Weekday ・ Time to travel ・ Airline ・ Website ・ Route ・ Date-day and year-month level ・ Month

These features analyze all fares listed in FROP (Fare Rule Output Product, from ATPCO (Airline Customs Publishing)) and determine which rules apply to the maximum number of fares. It is derived from what you do. For example, 54% of fares have a minimum stay rule, and 31% of all fares have a minimum stay rule that requires Saturday nights to be spent at the destination. Therefore, a feature may be a useful feature to classify based on price. A more accurate estimate of the usefulness of a feature may be derived by training multiple models with different features and comparing the prediction accuracy of the different models.

Profit-Estimate missing flight price data-Form a good forecast model with data generation features

Potential market and / or application area models may be used in the flight industry to estimate previously unobserved prices, display prices to users, or be worth asking to return prices. It may be determined whether or not there is (for example, it may be a good value).

This model may be used in other areas where price generation (eg, pricing rules) is performed by structured means (eg, pricing rules), although there are disadvantages of obtaining prices (eg, inquiry costs). This includes travel products such as hotels, trains, and car rentals, as well as other products with structured pricing, such as insurance.

Other Applications Consumer Price Comparison Website companies attempt to price competitively by comparing prices to those of their competitors.

Model-based pricing-guessing missing quotes Guessing missing quotes may be done in two stages (training and quotes).
・ Training ○ Build a statistical model from past prices ・ Estimate ○ Identify missing estimate candidates ○ Price estimate candidates based on the statistical model

In an example, the statistical model is a set of naive Bayes classifiers trained on a regular basis (eg, daily) using historical prices. The missing estimate is estimated based on the route operator's data and priced using a classifier.

Statistical Model Examples Naive Bayes classifier is a dependency between the probability that an object described by a set of features belongs to a class and the probability that an object from this class will feature a set of features. Based on above. See equations (1) and (2) in FIG.

In FIG. 3, the equations (1) and (2) can be simplified into the equation (3). In equations (1) to (3),
-C is a class in which the input object can be classified (for example, in this case, the price range, for example, 75 to 125 EUR).
_-Fi is a feature value (for example, in this case Friday, Air France website, etc.).

The classification can be expressed using the equation (4) of FIG.

For a set of feature values that wants to find a class (price range), the feature values occur most often, thus maximizing the probability that the class will characterize the set that you enter.

Training Classifier The intent of training is to calculate how often feature values occur for the observed price range. For example, in the process shown in FIG. 4, feature values are represented by numbered regions and price ranges (buckets) represent frequency information (for example, feature value quantities range from 75 to 125 EUR, 126 to 175 EUR, and so on). ..

Enter the following elements for the training process example.
-List of routes-Classifier category classification method-Past estimates-Supported feature sets with weights

The list of routes sets boundaries for training. The category defines the range of a classifier, and the category is represented by a set of feature values. The experimentally selected category consists of city-level routes and airlines (eg London-Paris, Air France), but other options are possible (eg, city-level routes without airlines). There is always one classifier per category. Train the classifier using only past estimates that match that category (for example, only estimates for the London-Paris route (operated by Air France)).

The first step in the training process is to retrieve a certain number of up-to-date historical estimates with a certain age from the database. Evaluate the category for any quote. Group quotes by category. Each group is a training set for one classifier.

The training set can be extended with estimates for reverse path equivalents. Example: For past estimates on Route London Stansted-Rome Ciampino (Air France), the reverse route equivalent is Rome Ciampino-London Stansted (Airline and other details eg Price Will be the same). Include estimates for reverse pathways in training scaled by weights that adjust their importance.

Next, for each group, a classifier with the following statistical information extracted from the estimate is constructed.
-Estimated weights containing different feature values-Estimated weights containing different feature values per granularity price range-Many observed price ranges, estimated price ranges, and all estimates

The weights applied to many estimates can be varied between features to emphasize that some of them have a greater impact (for example, the inclusion of weekdays in the trip may be more important than the length of stay). is there).

The process ends with one classifier for each training set containing statistics (eg, one classifier for the London-Paris route (operated by Air France)). Memorize the model for future use. Retraining the classifier can be done, for example, on a regular basis or when many past estimates change.

The formation of a classifier may only be performed if a certain number of estimates matching that category are available.

An example of the entire training process is shown in FIG.

In addition to the parameters listed above, the training process can be parameterized by:
-Price range size per route (eg 1EUR, 5EUR, 10EUR)
-Weighting applied to the reverse route-Weighting function applied based on the estimated age

Estimate the missing quote

In an embodiment, the user searches for one-way or round-trip flights for a route and date range. The system responds to the query with the set of cheapest quotes available from the database. Up to one per date (or for date-to-round-trip flights). The response may not include an estimate for a date (date vs. date). Guessing the missing estimates can be done using a trained naive Bayes classifier.

The following elements may be entered into the guessing process.
· Route · List of dates (or date pairs) with the lowest quote information missing

Since the classifier can only be used to price existing quotes, candidate quotes must be evaluated first. Candidates are all available airport pairs, dates (date pairs), airlines, and agency combinations for the route of inquiry. The list is generated based on the process input and the route operator database.

Next, each candidate quote placeholder is priced using a naive Bayes classifier. See, for example, FIG.

Price estimation may be performed for each candidate estimate in the following steps.
-Extract category feature values from estimates.
・ Extract the classifier trained for the extracted category from the data base ・ Extract all the feature values from the estimation candidates ・ Classify the candidate estimates by the following ○ In each price range stored in the classifier On the other hand, calculate the Bayes posterior probability ○ Select the price range class with the highest Bayesian posterior probability ・ Attach the price class to the candidate estimate

If there is no classifier for a candidate quote, no price guess is made and the date (date vs.) slot remains empty.

As a result of the process, there can be multiple priced quote candidates per date (date vs. date). For this reason, the next step is to select a single quote for each date (date vs.). The choice can be based on price. For example, select the cheapest quote per date (date vs. date).

The final result of the process is at most one estimate per input date (date vs. date). The inferred estimate can be included in the response returned to the user. An example of the entire process of estimating the missing estimate is shown in FIG.

In addition to the parameters already listed, the guessing process can be parameterized by:
-Minimum Bayesian posterior probability required to allow classification results-Maximum number of path operators involved in candidate generation-Random variation added to Bayesian posterior probabilities to avoid ties

Estimating method that gives confidence in the price As an estimation method, a method that gives confidence in the price may be provided. For example, 99% sure it's under $ 5 and under $ 1000. Therefore, the estimation method may provide the following.
-Calculation and return of these beliefs.
• Use these beliefs to determine whether to show the price to the user or to show the user the likely price range (eg, error bars).

Many methods of estimating prices can return a measure of confidence in the price. For example, in the case of a simple Bayesian classifier, the selected class has the highest probability, but that probability can be used as an indicator of confidence in the estimate. The closer the probability is to 1, the more confident that the predicted price is accurate.

The resulting conviction may be used in many ways. If you are very uncertain about your estimate, you may make the following decisions:
○ Do not give a quote as it is not useful to the user ○ Trigger a live fetch of the price so that the user can be shown the exact value. Confidence can be shown to the user.

Confidence can be shown to the user. This can be shown in many ways.
○ A bar showing the range of possible values for the price (see, for example, Figure 22).
○ Reduce the visibility / intensity of the list price (for example, using coloring, shading, or dotted lines).
The range of confidence in the quoted price ranges from the lowest price to the highest price in the range that leads to the confidence that the price will be found as a result of the quote method. The range of confidence in the estimated price may be shown graphically using bars. Bars may be displayed for each date in the date range. The bar may be displayed for each date in the date range for the outbound and inbound journeys. The bars are color coded, for example, the bars correspond to recent prices, for booking features, for price ranges derived from multiple tickets that share a departure date, and similar itineraries. It may be shown that the price may be stable with respect to the price range estimated from. One bar may be provided for a date. A plurality of bars may be provided for a certain date. An example is shown in FIG.

B. Derived availability Engine background / motivation Motivation is very similar to that in the model-based pricing example.

The meaning of the "scheduled" price model is that two single quotes cannot be used to form a round-trip quote, so queries must be made for every possible date pair, and dates. It is forbidden to do this due to the large number of pairs, due to the cost of obtaining this data and the space required to store the data. However, the round-trip price is the result of combining the available fares available for the outbound and inbound legs, along with some rules. Thus, if the availability and rules are known, it is possible to break down round-trip queries into "reusable" legs and use them to construct new "round-trip" prices, a low-budget price model. It is possible in the same (but different) way as used for airlines with.

Derived Availability Engine Examples The global distribution system uses schedules from OAG (Official Airlines Guide: see, for example, www.oag.com), fares from ATPCO, and fare classes from airlines. Gather possibilities and store this information in its own system. This stored information is updated regularly from those sources. When the GDS is contacted for an estimate for a route and date, the GDS takes the following steps:
1. 1. Determine a valid itinerary from the schedule.
2. Calculate valid fares for your itinerary (some GDSs do not have a pricing engine and are provided by a third party engine, such as SITA (see, eg, http://www.sita.aero/). ) To ensure that the exact price applies to each itinerary).
3. 3. Find the availability of these fares.
4. Add accurate taxes and surcharges.

Step 1 may be implemented in the route service. Specifically, step 1 may be realized by using a timetable and rules regarding which ones can be sold. User inquiries can identify valid itineraries from which tickets can be purchased.

The airline sends the fare to ATPCO, which provides an application for these data. However, it would be a great effort to translate these unprocessed fares and rules into a system for determining the exact fares for a query. Fortunately, ATPCO supplies merged fares and rule data in a data supply called FROP (Fare Rule Output Product). The FROP data is transmitted in a fixed length record file containing rule conditions and fare information summarized for the main categories. Therefore, step 2 can be performed using the FROP data. It can also be achieved by combining fare rules and prices from airlines using a pricing engine.

ATPCO also provides a supply of itinerary and service charge taxes. As a tax and surcharge engine, we will take the data from these applications and use them to calculate the overall price that the customer will pay for a fare and, as a result, build an engine that emulates step 4. be able to.

The only remaining item for the low cost calculation of GDS fares is availability-step 3. Derivation of availability is possible by comparing the results of live updates (final results of steps 1 to 4) with the calculations of steps 1, 2 and 4 outlined above, i.e., route service, FROP. It is possible to calculate all possible fares by combining the fares obtained from, as well as the tax and surcharge engine, and using the observed fares obtained from the live update, which of the possible fares It is possible to infer whether it was used and therefore which of the calculated fares was available. Shown in the chart in FIG. 2 is a method by which such a "difference" engine can be used to derive the availability of fare classes from live updates.

If the available fares are known for both the outbound and inbound legs of the round-trip flight, the round-trip price calculation can be performed even if no particular date pair on the round-trip has been observed. This method of modeling estimates is known as a connectable monopod (SLC). SLC is similar to the "low budget" price model. However, in the case of a low budget, the round-trip outbound and inbound legs are also valid as one-way tickets. For low budgets, constructing a return trip is simply a matter of adding return and outbound leg prices (because the SLC structure needs to apply availability after finding a valid fare for a date pair). Slightly more complicated).

If the price of the fare has increased, but the availability remains the same, it is possible to calculate the new price using the derived availability engine, even if it has not been observed.

There are additional benefits to doing so to provide the derived availability as compared to using GDS. Pure calculations in the GDS do not guarantee that fares can be "found" on airlines or OTA websites, but observed fares are, by definition, found. Also, the pure calculation does not take into account surcharges, margins, discounts, and personal fares (which may be incorporated into the observed fares).

It should be noted that fare prices do not necessarily change, there are many different fares for a particular flight, and the availability of each fare may change. Since revenue management software may continuously change which fares are available, the disappearance of cheaper tickets can occur because revenue management cancels the ticket and increases revenue. ..

It is not necessary to contact the GDS or airline reservation system to find the profit / availability, but to derive the availability.
• Calculate new prices based on estimated availability rather than true availability.
• Estimate new prices based on availability and fare rules, without any observed cases of new prices.

Potential Other Applications Providing airlines with pricing for their own products. This is because airlines are still doing this pricing from the GDS at a high cost at this time.

The company is interested in finding flight data prices at a fairly low cost.

Architecture Outline Example A server (which may be real or virtual) is provided to which a service for estimating fare availability and price is provided. Related websites, smartphone apps, inter-company services, connections with traditional airlines (eg British Airways, Qantas, KLM, etc.), low-budget airlines (eg Ryanair, easyJet, German Wings, etc.) Connections with and travel booking websites (eg, lastminute.com, Opodo, Thomson, etc.) may be provided.

Travel booking websites and airlines may provide input data to servers that provide fare availability and price quote services. The output may be provided by the server to a public application programming interface (API), such as a partner company. The server may provide the output to the search website, the search mobile website, and the app on the smartphone for the user. An example of a connection configuration between a server and an input data source, and outputs to partners and users are shown in FIG.

In a search on the server, the user may specify a place of departure (eg, an individual airport, a city with multiple airports, or a country with multiple airports). In the search on the server, the user may specify a destination (for example, an individual airport, a city containing multiple airports, or a country containing multiple airports). In a search on the server, the user specifies a departure date or range of departure dates (for example, a date range or calendar month, or an unspecified date (which can include a date range, such as any date in the next year)). You may. In a search on the server, the user specifies a return date or range of return dates (for example, a date range or calendar month, or an unspecified date (which can include a date range, such as any date in the next year)). You may. For searches on the server, the user may specify a one-way ticket or a round-trip ticket. The user may specify the number of passengers in a particular age range. The user may specify the class of the ticket (eg Economy, Premium Economy, Business Class or First Class). You may perform a parallel search for hotels. You may perform a parallel search for rental cars. You may indicate your preference for direct flights. A user interface embodiment is shown in FIG. A further user interface embodiment is shown in FIG. This is an example of flexible search. This is because the departure indicates that any airport may be used in London (UK) and the destination indicates that any airport in Australia is permitted.

The output search results may include a list of flights that meet the search criteria. A graphical indicator (eg, a slider bar) may be provided to limit the departure time range for foreign flights. Graphical indicators (eg, slider bars) may be provided to limit the departure time range for round-trip flights. Selectable tabs may be provided to arrange the flights in order of increasing cost. Selectable tabs may be provided to arrange flights in order of increasing travel time. Selectable tabs may be provided to arrange flights in alphabetical order by airline name. A user interface embodiment is shown in FIG.

In the user search example, if the flight between the departure point and the destination (which may be a one-way flight or a round-trip flight) is applicable to a foreign flight date, if applicable. Required for round-trip flight dates. In the search process embodiment, in the first step, the partner search website showing the output is selected. The second step is to identify the relevant price that has already been cashed. The third step is to identify and update the associated prices that are not available. The fourth step is to cleanse and store the new data (eg, tax and toll removal may be removed, and who the actual flight operator may be identified). In the fifth step, the price was returned to the user. An embodiment of the system for performing the search process is shown in FIG.

In flexible search, the output data may provide a list of destination airports that meet flexible search criteria (for example, the destination is any airport in Australia and departs from any airport in London, UK). good. The list may be listed in ascending order of price. An example of user interface output is shown in FIG. Once a particular airport is selected, the lowest price for the outbound trip may be shown as a function of the outbound departure date, for example for selectable calendar months, and the lowest price for the return trip of the trip may be shown as the return departure. As a function of days, it may be shown, for example, for a selectable calendar month, for a selected departure and destination city or airport. Example 14 of user interface output is shown.

Browsing search may be provided. Browsing services may be provided to users of smartphone apps or users of search websites. You may access the browsed price data by the browsing service. The price data viewed may include data stripped from the website content. Browsing services may be connected to public application programming interfaces (APIs). The browsing service does not have to be connected to the airline server, or to the server of another travel information service provider. FIG. 15 shows an example of the architecture for browsing search.

An example of an overall system architecture for a system that includes the provision of fare availability and price quote services is shown in FIG.

The following sections provide background information on how to calculate airfares and the role of GDS.

C. Introduction to the Global Distribution System The Global Distribution System (GDS) may be defined as a system that includes availability, pricing, and related services to airlines through which reservations and tickets can be issued. .. The GDS makes these features available to application travel agencies, booking engines, airlines, and other applicants. A GDS configuration embodiment is shown in FIG. The more general term is "delivery system". This may include the ability to make reservations for modes of transport other than air transport (eg, trains, etc.).

Route Search Historically, as the airline network grew, GDS added route search. When the number of potential and possible connections increases exponentially (SOF-NYC has many connection points, for example SOF-IST-NYC, SOF-FRA-NYC, SOF-MOW-CPH- (NYC, etc.), GDS has added a layer of this "connecting" building to reduce the number of different pricing routes. Therefore, it is not necessary to process many fares and rules. They did it in a very dull way-static routes and connection tables. This method tells them essentially a route-finding city between SOF and NYC to fly between zones (specified by ATPCO-US-EUR, EUR-APAC for simplicity of discussion). Includes connections in potentially wrong directions (eg, IST), but rejects a large number of cities as possible connections. For example, flight SOF-PRG-CPH-NYC may be considered too complex and potentially costly and will therefore be dropped during the route search phase.

After setting up the connection construction routing, the connection construction begins to look up the airline, number of flights, and connection limits (included in the Minimum Connection Time (MCT) table). The MCT (delivered by OAG) is a complex document that specifies general rules about how flights / airlines can connect in terms of time. For example, if a long-distance US-EU connects to the EU-EU in LHR, a 120-minute MCT is applied. The general rules are easy to interpret but important. If someone tries to book outside the range of MCT parameters, manual invalidation is required.

Interesting is the MCT exception (about 200,000). This may require 55 minutes, even when DL Flight # 003 connects to AF Flight # 004. However, if DL flight # 003 arrives at gate 19 of the CDG and connecting flight AF009 departs from gate 49, 65 minutes will be applied as the MCT, and so on. About 70,000 of all exceptions are for Air France (AF) and Paris Charles de Gaulle Airport (CDG). That's why many passengers consider it the worst to fly.

Fares An example of a fare is a pricing contract. This specifies the price (basic fare) for the services provided by the airline. Prices are defined by service class, class type (Y, Q, J, M, H, etc.) and type of travel (one way, round trip, multi-city, round the world (RTW)). ..

In the examples, each fare is accompanied by a number of rules and limits that apply to it, making the fare a 20-30 page long contract rather than just an Excel line (spreadsheet or CSV file). Think of it as. Instead of thinking of the fare as a row in a large excel spreadsheet and the columns showing which features the fare has, instead the fare is associated with many parameters and restrictions by itself. It should be considered a complex matter and therefore more like a contract. Fares are who, when and how the fare-passenger type can be used, when the fare is applicable (NVB, NVA-not valid before, not valid after), and many others. It stipulates the restrictions of. Fares can be public fares (accessible to any reseller) or private fares (accessible to designated resellers / sellers).

Fares are given by the airline. They are defined at a general level for the market and are limited by travel and ticketing dates. Naturally, the fare represents the price point.

Fare Upload and Distribution Upload and Distribution ATPCO is an industry group that handles freight and fare distribution in the industry. There are a small number of airlines that only serve SITA (Bulgaria Air is an example).

Scheduling So far, ATPCO has released fares on a specific schedule, but at the request of airlines, the organization has begun to release fares every hour. This allows airlines to ensure a rapid response to market changes that are not only suitable for availability control.

SITA sells the fare four times a day on weekdays and three times on weekends. There are also surviving fares within the GDS (GDS-specific fares or negotiations). This fare is completely under the control of GDS and does not follow a general delivery schedule.

Fare rules and restrictions and categories Provide fare data, fare class data, and fare restriction data.

How does pricing work (very high level)?
The rules are calculated by the pricing engine in a very strict order of rule records, tables, and subtables. If no particular rule record is found, there are general rules that specify all of them.

Limit Data-The fare class tells the engine that there are rules for a particular fare.
-Category control tells the engine what type of restrictions exist.
-The category data table is last when the engine reads to get relevant information.

For example, the fare class tells the engine that there are rules. The rule is category 15 (sales restrictions), which retrieves additional information about it (what type, what are the rules and restrictions) after examining the category data table.

An example of a fare for a fare class record, a fare category, and a category data table is shown in FIG.

Fare Display vs. Fare Estimates and Processes Many industry experts use the term fare display and fare estimate interchangeably, which is unacceptable to the expert. I will explain it.

Fares Display A comprehensive integrated concept of market fares and related rules. This allows the agency to:
-Compare fare rules (NB! Not price)
-Tell passengers about the contract period (fare rules)

Fare estimate The final price after applying the contract period. It captures most of what we want to give our passengers.
-All taxes-All surcharges-All related rules-Final price

An embodiment of the process of providing the fare pricing option is shown in FIG.

Availability Knowledge of availability is very important in the industry. Availability gives airlines control over their margins (seats), and adding another way to manage how profitable a route is by opening and closing the availability bucket. Be done. It is managed by a revenue management system (RM).

While some RMs are more capable of managing complex rules, they are more complex because most of the availability rules are formed by the internal departments of statisticians and mathematicians.

There are three main ways to receive availability if you want to check if a seat / class service is available.

i) NAVS / AVS
NAVS = availability with simple numbers per seat class. Example: Y9H3M0L8. .. ..
AVS = Availability Status-This is Flight Open / Closed, Open for Cancellation Waiting List, etc.

ii) The polling CXR (airline) can use NAVS / AVS as a display if better connectivity is compromised. These CXRs use revenue management rules and need to be polled for actual availability and all availability information.

Some airlines and / or flights require polling, the order for which is MR (Market Restriction). If the tag is MR when the GDS requests AVA, the GDS needs to poll for correct availability. Unfortunately, NAVS / AVS does not show these tags, so when GDS uses NAVS / AVS for tagged flights, it may display incorrect availability.

iii) Cache DB
To reduce polling and NAVS / AVS, pricing engines have developed complex cash DBs. Although they are not accurate, they achieve some of the goal of reducing calls when storing information.

An embodiment of the relationship between the pricing engine, AVS / NAVS, seamless (polling), cache DB, and airline host is shown in FIG.

Availability Server Examples The caching mechanism may be provided by offline traffic, but is used by online channels. The availability server may be able to update the cached data based on rules (two-way polling) and even warn of slight discrepancies in reservations. An embodiment of a system including an availability server is shown in FIG.

NOTE Performing the steps described above can be performed using standard, well-known programming techniques. The novelty of the embodiments described above involves achieving the results described using the steps described, rather than in a particular programming technique. Software programming code that embodies or forms part of the invention is typically stored in permanent non-temporary storage. In a client / server environment, such software programming code may be stored using a storage device attached to the server. The software programming code may be embodied on any of the various known media used with data processing systems (eg, diskettes, or hard drives, or CD-ROMs). The code may be delivered on such a medium, or delivered to a user from the memory or storage of one computer system to another computer system via one type of network. It may be made available to users of other systems such as. Techniques and methods for embodying software program code on physical media and / or distributing software code over a network are well known and are not described further herein.

Not surprisingly, a general purpose and / or dedicated hardware-based system, or general purpose and / or dedicated hardware, that performs a specified function or step to perform each element of the description and a combination of elements in the description. It can be done by a combination of hardware and computer instructions.

These program instructions may be given to the processor to form a device such that the instructions executed on the processor form means for performing the functions specified in the description. Computer program instructions are executed by the processor to cause the processor to perform a series of operation steps to form a process that is performed on the computer, and when the instructions are executed on the processor, they perform the functions specified in the description. You may want to give a step for it. Accordingly, FIGS. 1-17 and 22 support a combination of means for performing the specified function, a combination of steps for performing the specified function, and a program instruction means for performing the specified function. ..

Of course, the above arrangement merely illustrates application to the principles of the present invention. Many modifications and alternative arrangements can be devised without departing from the spirit and scope of the invention. The present invention has been shown in the drawings, with individuality and details, and fully described above, with respect to what is currently considered to be the most realistic and preferred embodiment of the invention, to those skilled in the art. As will be appreciated, many modifications can be made without starting from the principles and concepts of the invention described herein.

Concepts A plurality of concepts (hereinafter referred to as concepts "AH") are included in this disclosure. The following can be useful in defining these concepts. You may combine aspects of the concept.

A. How to Provide a Price Quote A method of providing a price quote, in which (i) a computer server receives a request for a price for an article or service, eg, an airfare and a parameter that defines those article or service, and (ii). ) To determine the estimated price from an incomplete historical price data set at any time to step (i) above by analyzing the patterns in that data set, one or more processors. The steps to configure and the steps to configure one or more processors to calculate an estimate for the requested price for an article or service that meets the (iii) parameters, and (iv) the price estimate to an end-user computing device. A method is provided that includes, for example, a step of sending to a personal computer, smartphone or tablet.

The above may further include any of the following, alone or in combination.
• In this method, the parameters that define those goods or services include the type of action, such as airfare, hotel reservation, railroad fare; date range; destination; origin; desired weather conditions; star rating; keywords; any other. Includes one or more of the user-defined preferences of.
• Determine the estimated price by estimating, deriving, or predicting the estimated price.
• In step (ii), (a) obtain past price estimates from a computer data store, (b) group past price estimates by category, and (c) statistical data for each group. (D) For each group, store the classifier containing the derived statistical data, and (e) use the stored classifier to select the group to which the requested price corresponds. Includes identifying and.
Step (iii) involves calculating a set of quotes for the requested price over a specified date range, using statistical data obtained from a stored classifier corresponding to the identified group. Is done.
• Step (ii) involves analyzing patterns in the dataset using rules.
Step (ii) includes a naive Bayes classifier machine learning approach that produces a probabilistic price model that is used to predict unobserved prices.
-Train the classifier with the observed prices and the corresponding set of features.
• Features relate to requirements and include one or more of the week's departure dates, length of stay, Saturday stays, airlines, travel times, routes, and months.
The classifier then predicts the price of an unobserved price by being given a set of features and giving the price most likely to have those features.
• Features may be derived by training multiple models with different features and comparing the prediction accuracy of the different models.
-Step (ii) includes building a statistical model from past prices, identifying missing quote candidates, and pricing quote candidates based on the statistical model.
-Estimating the price for each candidate estimate, the step of extracting the category feature values from the estimate, the step of extracting the trained classifier from the database for the extracted category, and all of the estimate candidates. Steps to extract feature values, classify candidate quotes, calculate Bayes posterior probabilities for each price range stored in the classifier, and select the price range class with the highest Bayesian posterior probabilities. And the steps taken by attaching the price class to the candidate quote.
Inputs to the statistical model include a list of routes, classifier categorization schemes, historical estimates, and a set of supported features with weights.
• Filter past estimates by age.
• Inputs to the statistical model include reverse path equivalents.
• The method includes a step that includes the cashed fare price in the price quote set.
The price quote set is configured without contacting the delivery system after step (i).
・ Prices are for trips that are one-way trips.
・ Prices are for trips that are round-trip trips.
・ Prices include airfare prices.
・ The price includes the railway fare price.
・ The price includes the rental car price.
・ The price includes the hotel price.
-Requests include flexible search requests.
The final result of this method is a maximum of one estimate per input date or input date.
• The method involves determining the confidence range of an estimated price from an incomplete historical price dataset at any time point to step (i) (A) by analyzing patterns in that dataset. (B) The step of configuring one or more processors to calculate the confidence range of the estimated price for the requested price for the goods or services satisfying the parameter (B) is included. It has been.
The method further includes (C) providing a fare price quote along with a confidence range to an end-user computing device, such as a personal computer, smartphone, or tablet.
• The method further includes the step of using a confidence range to determine whether to display the price to the user or to provide the user with an likely price range.
-Display the likely price range as an error bar.
-The date range includes only one departure date.
-The date range includes only one return date.
• The method includes a method of providing the best set of round-trip fare price estimates for a specified departure date range and a specified return date range, and step (i) is for a trip from origin to destination. Included in receiving a request for a round-trip fare price for a specified departure date range and a specified return date range, step (iv) is to estimate the best round-trip fare price for a specified departure date range and a specified return date range. Includes providing pairs.
-The best round-trip fare price quote set is provided in bar chart format.
• Of the minimum Bayesian posterior probabilities required for the estimation process to tolerate classification results, the maximum number of path operators involved in candidate generation, or the random variation added to the Bayesian posterior probabilities to avoid ties. Parameterized by one or more.
-This method is performed on the server where the fare availability and price estimation service is provided.

A server configured to provide a price quote, (i) receiving a request for a price for an article or service, eg, a request for a parameter defining those article or service along with an airfare, and the parameter is an action type. , For example air fare, hotel reservation, rail fare; date range; destination; origin; desired weather conditions; star rating; keywords; including one or more of any other user-defined preferences. , (Ii) Determining the estimated price from an incomplete past price data set at any time to (i) above by analyzing the patterns in that data set, and (iii) parameters. A server is provided that is arranged to calculate an estimate for the requested price for an article or service that meets the requirements and (iv) provide a price estimate.

The server may further include any of the following, alone or in combination.
(Ii) The server obtains past prices from the computer data store, groups past prices by category, derives statistical data for each group, and derives statistics for each group. It is arranged to store the classifier containing the data and to identify the group using the stored classifier corresponding to the requested price.
The server is arranged to calculate a set of quotes for the requested price over a specified date range, using statistical data obtained from a stored classifier corresponding to a particular group.
The server is further arranged to perform a method that belongs to any of the aspects of Concept A.

A computer program product embodied on a non-temporary storage medium, the computer program product is arranged to provide a price quote when executed on the computer, and the computer program product is When executed on a computer, (i) receive a request for a price for an article or service, eg, an airfare, as well as a request for a parameter that defines those article or service, and the parameter includes an action type, such as an airfare, a hotel reservation. , Rail fare; date range; destination; origin; desired weather conditions; star rating; keywords; including one or more of any other user-defined preferences, and (ii) described above (ii). Determining an estimated price from an incomplete historical price data set at any time for i) by analyzing patterns in that data set and (iii) requirements for goods or services that meet the parameters. A server is provided that is arranged to calculate an estimate for a price and (iv) provide a price estimate.

The computer program product may be arranged to perform the method of any aspect of Concept A.

A computer terminal connected to the server of any of the servers in Concept A, the computer terminal is arranged to send a request for a price quote for a specified date range for a trip from origin to destination, and a computer. The terminal is further provided with a computer terminal that is arranged to receive a price quote.

Goods or services, such as airline tickets, for which a price quote for the goods or services is provided using the method specified in any aspect of Concept A or the server specified in any aspect of Concept A. Goods or services are provided.

A web-based price comparison website that allows end users to interact with the web-based price comparison website by providing price requests for goods or services. A web-based price comparison website is provided that provides the request to a server that estimates the price using a method belonging to any aspect of Concept A.

B. How to guess which fare class is available for a trip How to guess which fare class is available for a trip from origin to destination on a particular date. The method involves (i) defining steps for fare prices for travel from origin to destination on a particular date, and (ii) estimating estimated fare prices from an incomplete historical price dataset. A step of guessing, estimating, or predicting by analyzing patterns in the dataset, and a step of calculating an estimate for the requested fare price using the results of (iii) step (iii), and ( iv) The step of sending the request to the delivery system for the fare price for the trip from the departure point to the destination on a specific date, and (v) the delivery system for the trip from the departure point to the destination on a specific date from the delivery system. The departure point on a specific date is compared with the step of receiving the fare price of (vi) and the calculated estimate for the requested fare price obtained from step (vi) step (v) with the fare price of the delivery system received in step (v). Outputs the step of guessing which fare class is available for the trip from to destination and (vii) the estimated fare class availability for the trip from origin to destination on a particular date. Steps and are included.

The above may further include any of the following, alone or in combination.
• In this method, step (ii) involves obtaining past fare estimates from a computer data store, grouping past fare estimates by category, and deriving statistical data for each group. For each group, it includes storing a classifier containing the derived statistical data on the computer, and using the stored classifier to identify the group to which the requested fare price corresponds. ..
• In this method, step (iii) is to calculate the set of estimates for the requested fare price over a specified date range with statistical data obtained from a stored classifier corresponding to the specified group. Includes doing with.
-The distribution system is a global distribution system.
-Estimated fare class prices are attached to each estimated fare class availability.
-Includes steps to send output to the server.
-The server is an airline server.
• Step (ii) involves analyzing patterns in the dataset using rules.
Step (ii) includes a naive Bayes classifier machine learning approach that produces a probabilistic model of fares, which is used to predict unobserved prices.
-Train the classifier with the observed prices and the corresponding set of features.
• Features relate to requirements and include one or more of the week's departure date, length of stay, Saturday stay, airline, travel time, route, month.
The classifier then predicts the price of unobserved fares by being given a set of features and giving the price most likely to have those features.
• Features may be derived by training multiple models with different features and comparing the prediction accuracy of the different models.
-Step (ii) includes building a statistical model from past prices, identifying missing quote candidates, and pricing quote candidates based on the statistical model.
-Estimating the price for each candidate estimate, the step of extracting the category feature values from the estimate, the step of extracting the trained classifier from the database for the extracted category, and all of the estimate candidates. Steps to extract feature values, classify candidate quotes, calculate Bayes posterior probabilities for each price range stored in the classifier, and select the price range class with the highest Bayesian posterior probabilities. And the steps taken by attaching the price class to the candidate quote.
Inputs to the statistical model include a list of routes, classifier categorization schemes, historical estimates, and a set of supported features with weights.
• Filter past estimates by age.
• Inputs to the statistical model include reverse path equivalents.
-The method includes a step that includes the cashed fare price in the fare price quote set.
・ The trip is a one-way trip.
・ The trip is a round-trip trip.
・ The fare is an air fare.
-Requests include flexible search requests.
• Of the minimum Bayesian posterior probabilities required for the estimation process to tolerate classification results, the maximum number of path operators involved in candidate generation, or the random variation added to the Bayesian posterior probabilities to avoid ties. Parameterized by one or more.
-This method is performed on the server where the fare availability and price estimation service is provided.

In addition, a server configured to infer which fare classes are available for travel from origin to destination on a particular date, the server is (i) departing on a particular date. The data that prescribes the fare price requirements for travel from place to destination and (ii) infer, estimate, or predict the estimated fare price from an incomplete historical price dataset. Doing this by analyzing the patterns in the set, and using the results of the estimated fares in (iii) and (ii) to calculate an estimate for the requested fare price.
(Iv) Sending a request to a delivery system for fare prices for travel from origin to destination on a particular date.
(V) Receiving from the delivery system the fare price of the delivery system for the trip from the departure point to the destination on a specific date.
(Vi) Compare the calculated estimate for the requested fare price obtained from the calculated estimate in (iii) with the fare price of the delivery system received in (v) for the trip from the departure point to the destination on a specific date. Arranged to guess which fare class is available and (vii) output the estimated fare class availability for travel from origin to destination on a particular date. The server that has been provided is provided.

In the server, for (ii), the server (a) obtains past fare estimates from computer data stores, (b) groups past fare estimates by category, and (c) Derivation of statistical data for each group, (d) for each group, storing a classifier containing the derived statistical data, and (e) using the stored classifier, the requested fare price It is arranged to identify the corresponding group and to do so.

In the server, for (iii), the server calculates a set of estimates for the requested fare price over a specified date range, statistical data obtained from a stored classifier corresponding to the identified group. It is arranged to be done using.

The server is further arranged to perform methods that include any of the method limitations described above.

In addition, a computer program product embodied on a non-temporary storage medium, which, when executed on a computer, for a trip from origin to destination on a particular date. Arranged to guess which fare class is available, the computer program product, when run on a computer, (i) the fare price for a trip from origin to destination on a particular date. And (ii) inferring, estimating, or predicting estimated fare prices from incomplete historical price data sets by analyzing patterns in that data set. And (iii) calculate an estimate for the requested fare price using the results of the estimated fare in (iii), and (iv) the fare for the trip from the origin to the destination on a particular date. Obtained from sending to the delivery system for the price, (v) receiving the fare price of the delivery system for the trip from the departure point to the destination on a specific date from the delivery system, and (vi) (iii) calculation estimates. Compare the calculated estimate for the requested fare price with the fare price of the delivery system received in (v) to determine which fare class is available for the trip from departure to destination on a particular date. Computer program products are provided that are arranged to make inferences and (vii) output inferred fare class availability for travel from origin to destination on a particular date. ..

Computer program products are arranged to perform any of the methods limited to the methods described above.

A computer connected to a server in any of the Server Concepts of Concept B described above, the computer being arranged to send a request defining a trip from origin to destination on a particular date. In addition, a computer is provided that is arranged to receive the output estimated fare class availability for travel from origin to destination on a particular date.

C. Method of estimating fare price This method is a method of estimating fare price, in which (i) a step of obtaining a past fare estimate from a computer data store and (ii) a step of grouping past fare estimates by category. And (iii) deriving statistical data for each group, (iv) storing a classifier containing the derived statistical data for each group on a computer, and (v) at a particular date. The step of receiving a request for the fare price for the trip from the origin to the destination, (vi) the step of identifying the group to which the requested fare price corresponds using the memorized classifier, and (vii) the requested fare price. Includes a step of calculating an estimate for, using statistical data obtained from a stored classifier corresponding to the identified group, and (viii) providing a requested fare price estimate.

D. Method of Estimating Fare Price A method of estimating the fare price, which is (i) the step of receiving a request for the fare price for a trip from the origin to the destination on a specific date, and (ii) the estimation. Requesting a fare to be inferred, estimated, or predicted from an incomplete historical price data set using the steps performed by analyzing the patterns in that data set and the results of (iii) step (ii). It includes the steps of calculating an estimate for the fare price obtained and (iv) providing the requested fare price estimate.

E. How to provide a set of round-trip fare price estimates This method is a method of providing a set of round-trip fare price estimates, in which (i) steps to obtain past fare estimates from a computer data store and (ii) past A step of grouping the fare estimates by category, (iii) a step of deriving statistical data for each group, and (iv) storing a classifier containing the derived statistical data for each group on a computer. Requested using steps, (v) receiving a request for a round-trip fare price for a specified departure date and a specified return date for a trip from origin to destination, and (vi) using a memorized classifier. Obtain from the stored classifier corresponding to the identified group to calculate a set of steps to identify the group to which the round-trip fare price corresponds and (vii) a set of estimates for the requested round-trip fare price for the specified departure and return dates. It includes a step performed using the statistical data obtained and a step of providing a set of (viii) fare price estimates.

F. How to provide the best set of round-trip fare price estimates This method provides the best set of round-trip fare price estimates for the specified departure date range and the specified return date range, which is (i) computer data. -Steps to obtain past fare estimates from the store, (ii) steps to group past fare estimates by category, (iii) steps to derive statistical data for each group, and (iv) each group on a computer. Receives a step to store a classifier containing the derived statistical data and (v) a request for a round-trip fare price for a specified departure date range and a specified return date range for travel from origin to destination. A set of steps, (vi) using a memorized classifier to identify the group to which the requested round-trip fare price corresponds, and (vi) a set of estimates for the requested round-trip fare price for the pair of departure and return dates. Steps to calculate using the statistical data obtained from the stored classifiers corresponding to the identified groups, (viii) to select the best fare price estimate for each pair of departure and return dates, and (ix) departure and return. Includes steps to provide the best set of round-trip fare price estimates for each pair on the return date.

G. How to provide a set of round-trip fare price quotes This method is a method of providing a set of round-trip fare price quotes, in which (i) a designated departure date and a designated return for a trip from origin to destination. The step of receiving a request for a round-trip fare price for a day and (ii) inferring, estimating, or predicting an estimated fare from an incomplete historical price dataset, analyzing the patterns in that dataset. Provides a step to calculate a set of estimates for the requested round-trip fare price for a specified departure and return date, and a set of (iv) fare price estimates using the steps performed by step (iii) and the results of step (iii). Including steps to do.

H. How to provide the best set of round-trip fare price quotes This method provides the best set of round-trip fare price quotes for the specified departure date range and the specified return date range. Steps to receive a request for a round-trip fare price for a specified departure date range and a specified return date range for a trip to a destination, and (ii) infer an estimated fare from an incomplete historical price data set and estimate , Or a set of estimates for the requested round-trip fare price for a pair of departure and return dates, using the steps made by analyzing the patterns in the dataset and the results of (iii) step (ii). Includes (iv) a step of providing the best round-trip fare price quote pair for each pair of departure and return dates.

Claims (18)

A computer-based method of estimating fare prices for travel from origin to destination on a particular date.
(I) In the step of obtaining a past fare estimate from a computer data store , the past fare estimate is an incomplete past fare estimate dataset.
(Ii) Steps to group the past fare estimates by category,
(Iii) In the step of deriving statistical data for each group, a naive Bayes classifier machine learning approach is used to generate a probabilistic price model, which model is used to predict unobserved prices. Being done
(Iv) A step of storing a classifier containing the derived statistical data for each group on a computer.
(V) the steps of: receiving a request for the fare price for the travel of said from the starting point at a particular date to the destination,
(Vi) Using the stored classifier, the step of identifying the group to which the requested fare price corresponds, and
(Vii) A step of calculating an estimate for the requested fare price using statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) The step of providing the requested fare price quote and
How to include. By analyzing the pattern before Symbol dataset comprises configuring one or more processors to determine a price quote from the incomplete historical fare estimate data sets, The method of claim 1, wherein. (A) With the step of configuring the one or more processors to determine the confidence range of the price quote from the incomplete past fare quote dataset by analyzing the pattern of the dataset. ,
(B) The steps of configuring the one or more processors to calculate the confidence range of the price quote, and.
(C) Including the step of providing the fare price quote with the confidence range to an end-user computing device such as a personal computer, smartphone, or tablet.
The method according to claim 2. The method of claim 2 or 3 , wherein the classifier is trained with the observed prices and their corresponding set of features. The feature relates to the request and includes one or more of the week's departure date, length of stay, Saturday stay, airline, travel time, route, month, or price at which the classifier is then not observed. Predicting the above-mentioned prices of a plurality of models by being given a set of features and by giving the price most likely to have those features, or by deriving the features. 4. The method of claim 4 , which may be performed by training with and comparing the prediction accuracy of the different models. The method according to any one of claims 1 to 4 , wherein the price quote includes an airfare price or the price quote includes a railroad fare price. The method according to any one of claims 1 to 6 , wherein the price quote is configured without inquiring about the distribution system. A server configured to provide fare quotes for travel from origin to destination on a particular date.
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) When deriving statistical data for each group, it involves the use of a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model is used to predict unobserved prices. When,
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive a request for the fare price for the travel of said from the starting point at a particular date to the destination,
(Vi) Using the stored classifier, identify the group to which the requested fare price corresponds.
(Vii) To calculate an estimate for the requested fare price using the statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) To provide the requested fare price quote and
A server configured to do. A computer program product embodied on a storage medium, said computer program product, which, when executed on a computer, provides a fare price estimate for a trip from origin to destination on a particular date. is configured, the computer program product, when executed on said computer,
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) When deriving statistical data for each group, it involves the use of a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model is used to predict unobserved prices. When,
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive a request for the fare price for the travel of said from the starting point at a particular date to the destination,
(Vi) Using the stored classifier, identify the group to which the requested fare price corresponds.
(Vii) To calculate an estimate for the requested fare price using the statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) To provide the requested fare price quote and
A computer program product that is configured to do. A how to run the computer that provides an estimate of the set saw fare prices on the fare price for the specified date for the trip from the starting point to the destination,
(I) In the step of obtaining a past fare estimate from a computer data store , the past fare estimate is an incomplete past fare estimate dataset.
(Ii) Steps to group the past fare estimates by category,
(Iii) In the step of deriving statistical data for each group, a naive Bayes classifier machine learning approach is used to generate a probabilistic price model, which model is used to predict unobserved prices.
(Iv) A step of storing a classifier containing the derived statistical data for each group on a computer.
(V) a step of receiving the request of the fare price from a starting point with respect to the specified date for the trip to the destination,
(Vi) Using the stored classifier, the step of identifying the group to which the requested fare price corresponds, and
(Vii) A step of calculating a set of estimates for the requested fare price for the specified date using statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) The step of providing the fare price quote set and
How to include. A server configured to provide a set of fare price quotes for fare prices for a specified date for a trip from origin to destination.
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) When deriving statistical data for each group, it involves the use of a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model is used to predict unobserved prices. And that
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the request of the fare price from a starting point with respect to the specified date for the trip to the destination,
(Vi) Using the stored classifier, identify the group to which the requested fare price corresponds.
(Vii) To calculate a set of estimates for the requested fare price for the specified date using statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) To provide the fare price quote set and
A server configured to do. A computer program product embodied on a storage medium, said computer program product, when executed on a computer, a fare for a specified date for a trip from origin to destination. Configured to provide a set of price quotes, said computer program product, when run on said computer,
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) When deriving statistical data for each group, it involves the use of a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model is used to predict unobserved prices. And that
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the request of the fare price from a starting point with respect to the specified date for the trip to the destination,
(Vi) Using the stored classifier, identify the group to which the requested fare price corresponds.
(Vii) To calculate a set of estimates for the requested fare price for the specified date using statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) To provide the fare price quote set and
A computer program product that is configured to do. A method of running a computer that provides a set of round-trip fare price estimates for round-trip fare prices for a designated departure date and a designated return date for a trip from origin to destination.
(I) In the step of obtaining a past fare estimate from a computer data store , the past fare estimate is an incomplete past fare estimate dataset.
(Ii) Steps to group the past fare estimates by category,
(Iii) In the step of deriving statistical data for each group, a naive Bayes classifier machine learning approach is used to generate a probabilistic price model, which model is used to predict unobserved prices. Being done
(Iv) A step of storing a classifier containing the derived statistical data for each group on a computer.
(V) a step of receiving the said request of the round-trip fare price from the departure point for the designated departure date and the specified return date for the trip to the destination,
(Vi) Using the stored classifier, the step of identifying the group to which the requested round-trip fare price corresponds, and
(Vii) A step of calculating a set of estimates for the requested round-trip fare price for the specified departure date and return date using statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) The step of providing the fare price quote set and
How to include. A server configured to provide a set of round-trip fare price estimates for round-trip fare prices for a designated departure date and a designated return date for a trip from origin to destination.
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) When deriving statistical data for each group, it involves the use of a naive Bayes classifier machine learning approach that produces a probabilistic price model, which model is used to predict unobserved prices. And that
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the said request of the round-trip fare price from the departure point for the designated departure date and the specified return date for the trip to the destination,
(Vi) Using the stored classifier, identify the group to which the requested round-trip fare price corresponds.
(Vii) To calculate the set of estimates of the requested round-trip fare price for the specified departure date and return date using the statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) To provide the fare price quote set and
A server configured to do. A computer program product embodied on a storage medium, said computer program product, when executed on a computer, has a designated departure date and designation for a trip from origin to destination. is configured to provide a set of return fare price estimates for return fare price for return date was, said computer program product, when executed on said computer,
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) For each group, when deriving statistical data, a naive Bayes classifier machine learning approach is used to generate a probabilistic model of price, because the model predicts unobserved prices. To be used for
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the said request of the round-trip fare price from the departure point for the designated departure date and the specified return date for the trip to the destination,
(Vi) Using the stored classifier, identify the group to which the requested round-trip fare price corresponds.
(Vii) To calculate the set of estimates of the requested round-trip fare price for the specified departure date and return date using the statistical data obtained from the stored classifier corresponding to the specified group.
(Viii) To provide the fare price quote set and
A computer program product that is configured to do. A method of running a computer that provides the best set of round-trip fare price quotes for a specified departure date range and a specified return date range for a trip from origin to destination.
(I) In the step of obtaining a past fare estimate from a computer data store , the past fare estimate is an incomplete past fare estimate dataset.
(Ii) Steps to group the past fare estimates by category,
(Iii) For each group, the step of deriving statistical data involves the use of a naive Bayes classifier machine learning approach to generate a probabilistic price model, the model for predicting unobserved prices. Used,
(Iv) A step of storing a classifier containing the derived statistical data for each group on a computer.
(V) a step of receiving the request of the round-trip fare price for the departure date range and return date range above specified above specified for the trip from the departure point to the destination,
(Vi) Using the stored classifier, the step of identifying the group to which the requested round-trip fare price corresponds, and
(Vii) A step of calculating a set of the requested round-trip fare price estimates for a pair of departure and return dates using statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) Steps to select the best fare price quote for each pair of departure and return dates, and
(Ix) A step of providing the best round-trip fare price quote pair for each pair of departure and return dates, and
How to include. A server configured to provide the best set of round-trip fare price quotes for a specified departure date range and a specified return date range for a trip from origin to destination.
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) For each group, when deriving statistical data, a naive Bayes classifier machine learning approach is used to generate a probabilistic model of price, because the model predicts unobserved prices. To be used for
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the request of the round-trip fare price for the departure date range and return date range above specified above specified for the trip from the departure point to the destination,
(Vi) Using the stored classifier, identify the group to which the requested round-trip fare price corresponds.
(Vii) To calculate the set of the requested round-trip fare price estimates for the pair of departure and return dates using the statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) Choosing the best fare price quote for each pair of departure and return dates,
(Ix) To provide the best set of round-trip fare price quotes for each pair of departure and return dates.
A server configured to do. A computer program product embodied on a storage medium, said computer program product, when executed on a computer, has a specified departure date range and a specified departure date range for a trip from origin to destination. is configured to provide a return date set of the most good of the round-trip price quote against the range, the computer program product, when executed on the computer,
(I) When obtaining a past fare estimate from a computer data store, the past fare estimate shall be an incomplete past fare estimate dataset.
(Ii) Grouping the past fare estimates by category and
(Iii) For each group, when deriving statistical data, a naive Bayes classifier machine learning approach is used to generate a probabilistic model of price, because the model predicts unobserved prices. To be used for
(Iv) To store on a computer a classifier containing the derived statistical data for each group.
(V) and to receive the request of the round-trip fare price for the departure date range and return date range above specified above specified for the trip from the departure point to the destination,
(Vi) Using the stored classifier, identify the group to which the requested round-trip fare price corresponds.
(Vii) To calculate the set of the requested round-trip fare price estimates for the pair of departure and return dates using the statistical data obtained from the stored classifier corresponding to the identified group.
(Viii) Choosing the best fare price quote for each pair of departure and return dates,
(Ix) To provide the best set of round-trip fare price quotes for each pair of departure and return dates.
A computer program product that is configured to do.

Images