CN107274215A - Flight prices Forecasting Methodology, device, equipment and storage medium - Google Patents

Abstract

The embodiment of the invention discloses a kind of flight prices Forecasting Methodology, device, equipment and storage medium.This method can include method：N price data of same flight scheduled time slot before departure from port of T different predetermined date of departure is obtained, wherein, price data includes：Ticket price and the air ticket quantity corresponding with ticket price.P different sample of the sample length for k is obtained according to n price data.Neural network model is trained using p different samples as training sample.The price of predetermined instant of the same flight based on the predetermined date of departure of Neural Network model predictive after training before departure from port.Flight prices Forecasting Methodology, device, equipment and the storage medium, can accurately and efficiently predict the future price of the same flight of identical predetermined date of departure.

Description

Airline price prediction method, device, equipment and storage medium

technical field

The invention belongs to the field of computer technology, and in particular relates to a flight price prediction method, device, equipment and storage medium.

Background technique

Airline prices change dynamically with the market sales situation. Good-selling routes or flights will be sold at an increase in price, but will be sold at a lower price. In addition, in addition to obtaining customers through price competition, airlines also have agreements to protect prices, for example, some price protection agreements that provide no less than 20% off the price three days before departure. Therefore, the flight price change is a very complicated problem, and it is quite difficult to predict the price.

Contents of the invention

Embodiments of the present invention provide a flight price prediction method, device, equipment and storage medium, which can accurately and efficiently predict the future price of the same flight with the same scheduled departure date.

In the first aspect, a flight price prediction method is provided, which may include:

Obtain n pieces of price data of the same flight with T different scheduled departure dates in the scheduled period before departure, wherein the price data includes: ticket price and the number of tickets corresponding to the ticket price.

According to n price data, p different samples with a sample length of k are obtained, wherein the price data included in each sample is the price data of the same flight on the same scheduled departure date in a predetermined period, and each sample includes one or two More than one price trend information of the same flight, k is less than or equal to n.

Use p different samples as training samples to train the neural network model.

Predict the price of the same flight on the scheduled departure date at the scheduled time before departure based on the trained neural network model, where T, n, k, and p are all positive integers.

In a second aspect, a flight price prediction device is provided, which may include: a data collection unit, a sample selection unit, a model training unit, and a price prediction unit.

The data collection unit can be used to obtain n price data of the same flight with T different scheduled departure dates in a predetermined period before departure, wherein the price data includes: ticket price and ticket quantity corresponding to the ticket price.

The sample selection unit can be used to obtain p different samples with a sample length of k according to n price data, wherein the price data that can be included in each sample is the price data of the same flight on the same scheduled departure date in a predetermined period of time, and , each sample includes one or more than two price trend information of the same flight, and k is less than or equal to n.

The model training unit can be used to train the neural network model with p different samples as training samples.

The price prediction unit can be used to predict the price of the same flight on the scheduled departure date at the scheduled moment before departure based on the trained neural network model, where T, n, k, and p are all positive integers.

In a third aspect, a flight price prediction device is provided, which may include a memory and a processor. The memory can be used to store executable program codes; the processor can be used to read the executable program codes stored in the memory to perform the above-mentioned flight price prediction method.

In a fourth aspect, a computer-readable storage medium is provided, which may include instructions, and when the instructions are run on a computer, the computer is made to execute the flight price prediction method as described above.

In a fifth aspect, a computer program product containing instructions is provided, and when the instructions are run on a computer, the computer is made to execute the above-mentioned flight price prediction method.

In a sixth aspect, a computer program is provided. When the computer program is run on a computer, it causes the computer to execute the above-mentioned flight price prediction method.

According to the flight price prediction method, device, equipment and storage medium provided by the embodiments of the present invention. By obtaining n price data of the same flight with T different scheduled departure dates in the scheduled period before departure, according to n price data, p different samples with a sample length of k are obtained, and the price data that each sample can include is The price data of the same flight with the same scheduled departure date in the scheduled period, and each sample includes one or more than two pieces of price trend information of the same flight, k is less than or equal to n, the above sample acquisition method can be obtained to predict the same scheduled departure More effective samples of the same flight on the Hong Kong date, and avoid the defects of difficult sample selection, complex calculation, high difficulty in price prediction and low accuracy that are prone to occur when predicting route prices through route price data. By using the above p different samples as training samples to train the neural network model, the trained neural network model can accurately predict the price of the same flight on the scheduled departure date in the future at the scheduled time before departure.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required in the embodiments of the present invention. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

Fig. 1 is to determine the 3U8996 flight of departure date as an example, the flight price change curve of the 3U8996 flight that should determine the departure date before the departure date;

Fig. 2 is an example with 3U8996 flight, the flight price change curve of this 3U8996 flight in the scheduled time period before the multiple departure days of this flight;

Fig. 3 is a schematic flow chart of the flight price prediction method of an embodiment of the present invention;

Fig. 4 is the schematic flowchart of the sample acquisition process of the flight price prediction method of an embodiment of the present invention;

Fig. 5 is the schematic flow chart of the flight price prediction method of another embodiment of the present invention;

Fig. 6 is the schematic flowchart of the acquisition process of the price data of the flight price prediction method of an embodiment of the present invention;

Fig. 7 is a schematic structural block diagram of a flight price prediction device according to an embodiment of the present invention;

Fig. 8 is a schematic block diagram of a flight price prediction device according to another embodiment of the present invention;

Fig. 9 is a schematic structural block diagram of a flight price prediction device according to an embodiment of the present invention.

detailed description

Based on people's inertial thinking, under normal circumstances, the flight price will increase as the departure time gets closer, so if you book a flight ticket a long time in advance, you will get a lower price. However, big data based on actual flight prices show that the closer the flight departure time is, the higher the flight price is not.

Fig. 1 is the 3U8996 flight whose departure date is determined as an example, the flight price change curve of the 3U8996 flight whose departure date should be determined before the departure date. As shown in Figure 1, the closer the departure time of flight 3U8996, the lower the price. There are many possible reasons for this situation, such as low passenger load factor of the flight, price competition of related flights, and other factors. And, these factors are not accidental.

Fig. 2 is to take 3U8996 flight as an example, the flight price change curve of this 3U8996 flight in the scheduled time period before multiple departure days of this flight. As shown in Figure 2, tracking the price of flight 3U8896 with a departure time period from 2016-10-15 to 2016-10-24, it is easy to see from Figure 2 that the flight price fluctuates significantly 3 to 15 days before departure. Therefore, it is of great significance to accurately predict flight price fluctuations during this period.

Features and exemplary embodiments of various aspects of the invention will be described in detail below. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without some of these specific details. The following description of the embodiments is only to provide a better understanding of the present invention by showing examples of the present invention.

It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. Embodiments will be described in detail below with reference to the accompanying drawings.

Fig. 3 is a schematic flowchart of a flight price prediction method according to an embodiment of the present invention. As shown in Figure 3, a flight price prediction method may include: S310-S340.

S310. Obtain n pieces of price data of the same flight with T different scheduled departure dates in the scheduled period before departure, wherein the price data includes: air ticket price and the number of air tickets corresponding to the air ticket price.

S320. Obtain p different samples with a sample length of k according to n price data, wherein the price data included in each sample is the price data of the same flight on the same scheduled departure date at a predetermined time period, and each sample includes a Or more than two same flight price trend information, k is less than or equal to n.

S330, using p different samples as training samples to train the neural network model.

S340. Predict the price of the same flight on the scheduled departure date at the scheduled time before departure based on the trained neural network model, wherein T, n, k, and p are all positive integers.

In some examples, the same flight with T different scheduled departure dates in S310 may be represented as a specific flight with different departure dates.

For example, the T different scheduled departure dates can be 10 different scheduled departure dates from 2016-10-15 to 2016-10-24, or the departure dates from 2016-10-15 to 2016 - Scheduled 5 departure dates out of 10 different scheduled departure dates for 10-24. The same flight can be flight 3U8896.

The predetermined time period in S310 may be the predetermined time period before the departure of a certain scheduled flight on a certain departure date. For example, 10 to 20 days before the departure of flight 3U8896 on 2016-10-15.

Then, S310 is, for example: select flight 3U8896 with 10 different scheduled departure dates from 2016-10-15 to 2016-10-24, and obtain the 3 to 15 days before the departure of the above flight. Flight price data for the above flights.

In some examples, the price data in S310 may come from China TravelSky Internet Booking Engine (IBE), which is an Internet-based open platform technology that provides access to various user application systems. The way of China TravelSky's traditional reservation business system is to use the application programming interface (Application Programming interface, API) interface. The IBE interface is the interface provided by TravelSky. Reservation, ticket issuance, change, value-added services, basic data and other services.

For example, S310 can connect and communicate with the IBE interface to obtain the status of the flight class.

In other examples, it can also be connected and communicated with other TravelSky Internet reservation interfaces to obtain flight price data.

In some examples, the cabin status obtained above may include cabin type and remaining quantity. For example: FA, I2, J2, A2, YA, BA, TS, HA, GA, SQ, LA, EA, VQ, RA, KQ, NQ, XQ, US, WQ, QQ, MQ, ZQ.

In some examples, the price data of S310 can be obtained by matching and querying the existing basic freight rate database according to the above-mentioned cabin status. This method of obtaining the flight price is the published freight price of the flight. Compared with the flight price obtained by other methods, this method of obtaining the flight price is more original, and avoids the adjustment of the original price by the agent. Using published freight rates to predict flight price changes has the advantages of high prediction accuracy and low computational complexity compared to obtaining flight prices in other ways.

In some examples, S310 may include:

Divide the scheduled period evenly into n time periods, and record the lowest price of the same flight in each time period as the price data of each time period. It should be understood that the obtained price data can be the lowest price of the same flight in each time period, because users usually only pay more attention to the lowest price of the flight, so the obtained price data is the lowest price of the same flight in each time period. The following flight price predictions are more meaningful.

In some examples, the price data included in each sample in S320 is the price data of the same flight on the same scheduled departure date in a predetermined time period.

For example, a certain sample may include the price data of flight 3U8896 whose departure date is 2016-10-17, which was queried 4 to 15 days before departure.

In some examples, each sample in S320 can also include one or more than two same flight price trend information.

For example, it may be the change trend information among the price data of the flight 3U8896 with a departure date of 2016-10-17 queried 4 to 15 days before departure.

In some examples, for example, the change trend information may be the price change trend of flight 3U8896 on the 3rd day before departure relative to the 4th day before departure.

In some examples, the neural network model of S330 may be a BP neural network model, for example: a BP neural network model with a single hidden layer.

There are many parameters of the neural network, and the setting of the neural network parameters has a great influence on the network performance, such as the selection of the number of nodes in the hidden layer, the selection of the activation function, the selection of the learning rate, etc.

in:

The selection of the number of neurons in the hidden layer is very important. Too few neurons will make the neural network "underfitting", and too many neurons will make the neural network "overfitting", and it will increase the amount of calculation, making the training more difficult. slow, reducing the accuracy of model predictions. The choice of the number of nodes in the hidden layer is related to the number of input nodes.

The choice of activation function has a significant impact on both the recognition rate and the convergence speed.

The choice of learning rate affects the speed of network convergence and whether the network can converge. The selection of the learning rate should not be too large or too small. A small setting can ensure network convergence, but the convergence speed is slow; on the contrary, if the setting is too large, the network may not converge and affect the prediction effect.

In some examples, the number of hidden layer nodes of the neural network model in S330 can be obtained according to formula (1):

In formula (1), S is the number of nodes in the hidden layer, n is the number of nodes in the input layer, m is the number of nodes in the output layer, and a is a constant from 1 to 10.

For example: the number of nodes in the input layer is 40, and the number of nodes in the output layer is 3, then according to the formula (1), the number of points in the hidden layer can be 10.

In some examples, because the accuracy of the sigmoid function is much higher than that of the linear function when approaching the high-order curve, the activation function of the neural network model in S330 is selected as the sigmoid function.

In some examples, the learning rate of the neural network model in S330 can be selected to be around 0.1.

In some examples, S340 may predict the future price change trend of flight 3U8896 whose scheduled departure date is 2016-10-28 based on the neural network model trained in the above example.

In some examples, the above-mentioned price change trend can be represented by numbers, for example, 0 means that the price has decreased, 1 means that the price remains unchanged, and 2 means that the price has increased. For example, if the above neural network model predicts that the result is 1, then it can be determined that the future price of flight 3U8896 on October 28, 2016 will be equal to the current price.

In some examples, the above prediction results can be used in various data service demand scenarios, not limited to a single flight price prediction, but can be applied to complex scenarios involving other aviation services that require flight price prediction services.

Therefore, the flight price prediction method provided by the embodiment of the present invention. By obtaining n price data of the same flight with T different scheduled departure dates in the scheduled period before departure, according to n price data, p different samples with a sample length of k can be obtained, and each sample can include price data It is the price data of the same flight with the same scheduled departure date in the scheduled period, and each sample includes one or more than two price trend information of the same flight, k is less than or equal to n, the above sample acquisition method can be obtained to predict the same scheduled More effective samples of the same flight on the departure date, and avoid the defects of difficult sample selection, complex calculation, high difficulty in price prediction and low accuracy that are prone to occur when predicting route prices through route price data. By using the above p different samples as training samples to train the neural network model, the trained neural network model can accurately predict the price of the same flight on the scheduled departure date in the future at the scheduled time before departure.

According to some embodiments, before S330, it also includes:

Compare p to the sample size threshold, and determine that p is greater than or equal to the sample size threshold.

According to some embodiments, S320 may include:

Obtain m×t-k different samples through the price data of the same flight with the same scheduled departure date in the scheduled period before departure, where m is the number of days in the scheduled period, and t is the number of times the price data is obtained per day, where m ×t is greater than or equal to k;

Based on m×t-k different samples obtained from the price data of the same flight with the same scheduled departure date in the scheduled period before departure and the number T of different scheduled departure dates, p different samples are obtained.

Fig. 4 is a schematic flowchart of the sample acquisition process of the flight price prediction method of an embodiment of the present invention. As shown in Fig. 4, the sample acquisition process of the flight price prediction method includes: S401-S410.

S401. Obtain the current date D, and initialize c=0, i=0. Among them, c is the initial sample size, and i is the number of initial data acquisition cycles.

S402, i=i+1, can be understood as adding 1 to the number of initial data acquisition cycles.

S403. Obtain price change and cabin class change data of a scheduled flight whose departure date is D-i day.

S404, filter the data of m days before the scheduled departure from the above data as the basic data.

S405. Select k dimensions in the basic data as time series.

S406, judging the time series result, and constructing a machine learning sample in combination with the basic data of the time series of k-1 dimensions. For example, the change trend of the last data relative to the previous data in the time series can be taken as the time series result.

S407. Accumulate the m×t-k samples into the sample library.

S408, c=c+m×t-k.

S409, judging whether c is greater than the preset quantity threshold x. If c is greater than the preset quantity threshold x, enter S410, and if c is not greater than the preset quantity threshold x, return to S402.

S410, storing data samples. So far, X machine learning training samples have been successfully obtained.

Fig. 5 is a schematic flowchart of a flight price prediction method according to another embodiment of the present invention. As shown in Figure 5, the above flight price prediction method may also include: S510, when missing price data appears in the n price data, select the time node of the next price data collection corresponding to the time node of price data collection corresponding to the missing price data For the corresponding price data, the selected price data is added to the n price data as missing price data.

Fig. 6 is a schematic flowchart of the price data acquisition process of the flight price prediction method according to an embodiment of the present invention. As shown in Figure 6, the acquisition process of the price data of the flight price prediction method may include: S601-S607.

S601. Obtain unprocessed data. Unprocessed flight price data can be obtained through the IBE interface.

S602, separating the cabin code and the number of remaining seats in the cabin. In the obtained unprocessed flight price data, the class code and the remaining seats of the class are separated.

S603. Obtain the lowest freight price of the flight in combination with the cabin price database. It can be understood that the minimum freight rate of the above-mentioned flight can be obtained by matching query in the existing basic freight rate database according to the above-mentioned cabin status.

S604, judging whether the obtained lowest freight rate of the flight, that is, the required price data, has a missing value. If there is a missing value, go to S605, if there is no missing value, go to S607.

S605. Determine whether the missing value is within the allowable range. If the missing value is within the allowable range, proceed to S606. If the missing value is not within the allowable range, discard the data acquired this time and return to S601.

S606. Use adjacent data to fill missing values. It can be understood that when there is missing price data in the price data, select the price data corresponding to the time node of the price data collection corresponding to the missing price data and the price data corresponding to the time node of the next price data collection, and supplement the selected price data as the missing price data to in the price data.

S607. Classify and store the price data in the database according to the time of data collection and the date of departure of the flight.

The flight price prediction method according to the embodiment of the present invention is described in detail above in conjunction with Fig. 3 to Fig. 6, and the flight price prediction device, equipment and storage medium according to the embodiment of the present invention will be described in detail below in conjunction with Fig. 7 to Fig. 9 .

Fig. 7 is a schematic structural block diagram of a flight price prediction device according to an embodiment of the present invention. As shown in Figure 7, the flight price prediction device 700 may include a data collection unit 710, a sample selection unit 720, a model training unit 730 and a price prediction unit 740.

The data collection unit 710 can be used to obtain n pieces of price data of the same flight with T different scheduled departure dates in a predetermined period before departure.

The sample selection unit 720 can be used to obtain p different samples with a sample length of k according to the n price data, wherein the price data that can be included in each sample is the price data of the same flight on the same scheduled departure date in a predetermined period of time, and , each sample includes one or more than two price trend information of the same flight, and k is less than or equal to n.

The model training unit 730 can be used to train the neural network model with p different samples as training samples.

The price prediction unit 740 can be used to predict the price of the same flight on the scheduled departure date at the scheduled moment before departure based on the trained neural network model, where T, n, k, and p are all positive integers.

The flight price prediction device 700 according to the embodiment of the present invention may correspond to the execution subject in the flight price prediction method according to the embodiment of the present invention, and the functions of each unit in the flight price prediction device 700 are respectively to realize each method in FIG. 3 For the sake of brevity, the corresponding process will not be repeated here.

Therefore, an apparatus for predicting flight prices is provided according to an embodiment of the present invention. By obtaining n price data of the same flight with T different scheduled departure dates in the scheduled period before departure, according to n price data, p different samples with a sample length of k are obtained, and the price data that each sample can include is The price data of the same flight with the same scheduled departure date in the scheduled period, and each sample includes one or more than two pieces of price trend information of the same flight, k is less than or equal to n, the above sample acquisition method can be obtained to predict the same scheduled departure More effective samples of the same flight on the Hong Kong date, and avoid the defects of difficult sample selection, complex calculation, high difficulty in price prediction and low accuracy that are prone to occur when predicting route prices through route price data. By using the above p different samples as training samples to train the neural network model, the trained neural network model can accurately predict the price of the same flight on the scheduled departure date in the future at the scheduled time before departure.

In some examples, the data collection unit 710 can also be used to:

Divide the scheduled period evenly into n time periods, and record the lowest price of the same flight in each time period as the price data of each time period.

In some examples, the sample selection unit 720 can also be used to:

Obtain m×t-k different samples through the price data of the same flight with the same scheduled departure date in the scheduled period before departure, where m is the number of days in the scheduled period, and t is the number of times the price data is obtained per day, where m ×t is greater than or equal to k;

Based on m×t-k different samples obtained from the price data of the same flight with the same scheduled departure date in the scheduled period before departure and the number T of different scheduled departure dates, p different samples are obtained.

In some examples, the flight price prediction device may also include a judging unit, which may be used for:

Compare p to the sample size threshold, and determine that p is greater than or equal to the sample size threshold.

Fig. 8 is a schematic block diagram of a flight price prediction device according to another embodiment of the present invention. As shown in Figure 8, the flight price prediction device 800 may include: a data collection unit 810, a missing data processing unit 820, a sample selection unit 830, a model training unit 840 and a price prediction unit 850.

The data acquisition unit 810 can be used to obtain n price data of the same flight with T different scheduled departure dates in a predetermined period before departure, wherein the price data includes: ticket price and ticket quantity corresponding to the ticket price.

The missing data processing unit can be used to select the price data corresponding to the time node of the price data collection corresponding to the missing price data when the missing price data appears in the n price data, and the price data corresponding to the time node of the next price data collection, and the selected price The data is added to n price data as missing price data.

The sample selection unit 820 can be used to obtain p different samples with a sample length of k according to the n price data, wherein the price data that can be included in each sample is the price data of the same flight on the same scheduled departure date in a predetermined period of time, and , each sample includes one or more than two price trend information of the same flight, and k is less than or equal to n.

The model training unit 830 can be used to train the neural network model with p different samples as training samples.

The price prediction unit 840 can be used to predict the price of the same flight on the scheduled departure date at the scheduled moment before departure based on the trained neural network model, where T, n, k, and p are all positive integers.

Fig. 9 is a schematic structural block diagram of a flight price prediction device according to an embodiment of the present invention. As shown in FIG. 9 , the flight price prediction device 900 may include a memory 904 and a processor 903 .

The memory 904 can be used to store executable program codes; the processor 903 can be used to read the executable program codes stored in the memory to execute the above-mentioned flight price prediction method.

Therefore, a flight price prediction device is provided according to an embodiment of the present invention. By obtaining n price data of the same flight with T different scheduled departure dates in the scheduled period before departure, according to n price data, p different samples with a sample length of k are obtained, and the price data that each sample can include is The price data of the same flight with the same scheduled departure date in the scheduled period, and each sample includes one or more than two pieces of price trend information of the same flight, k is less than or equal to n, the above sample acquisition method can be obtained to predict the same scheduled departure More effective samples of the same flight on the Hong Kong date, and avoid the defects of difficult sample selection, complex calculation, high difficulty in price prediction and low accuracy that are prone to occur when predicting route prices through route price data. By using the above p different samples as training samples to train the neural network model, the trained neural network model can accurately predict the price of the same flight on the scheduled departure date in the future at the scheduled time before departure.

In some examples, the flight price prediction device 900 may also include an input device 901, an input port 902, an output port 905, and an output device 906. Wherein, the input port 902, the processor 903, the memory 904, and the output port 905 are connected to each other through the bus 910, and the input device 901 and the output device 906 are respectively connected to the bus 910 through the input port 902 and the output port 905, and then connected to other components of the device 900. Component connections.

In some examples, the output interface and input interface here may also be represented by I/O interface. Specifically, the input device 901 receives input information from the outside, and transmits the input information to the processor 903 through the input port 902 . For example, enter price data.

In some examples, the processor 903 processes the input information based on computer-executable program codes or instructions stored in the memory 904 to generate output information. For example, the processor 904 performs the following steps: Obtain the same n pieces of price data of the scheduled time period before the departure of the flight, wherein the price data includes: the ticket price and the number of tickets corresponding to the ticket price. According to n price data, p different samples with a sample length of k are obtained, wherein the price data included in each sample is the price data of the same flight on the same scheduled departure date in a predetermined period, and each sample includes one or two More than one price trend information of the same flight, k is less than or equal to n. Use p different samples as training samples to train the neural network model. Predict the price of the same flight on the scheduled departure date at the scheduled time before departure based on the trained neural network model, where T, n, k, and p are all positive integers. The output information is stored temporarily or permanently in the memory 904, and then transmitted to the output device 906 via the output port 905 as needed. The output device 906 outputs output information to the outside of the device 900 . For example, rendering on a display device, or uploading to the cloud.

The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.

In some examples, a computer-readable storage medium is provided, which may include instructions, which, when run on a computer, can cause the computer to execute the above-mentioned flight price prediction method.

In some examples, there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform the flight price prediction method described above.

In some examples, a computer program is provided, which, when run on a computer, causes the computer to execute the above-mentioned flight price prediction method.

In the above examples, it may be fully or partially implemented by software, hardware, firmware or any combination thereof. When implemented using software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the processes or functions according to the embodiments of the present invention will be generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website, computer, server or data center Transmission to another website site, computer, server, or data center by wired (eg, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, a data center, etc. integrated with one or more available media. The available medium may be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a Solid State Disk (SSD)), etc.

According to some embodiments, the flight price prediction method can be implemented on a high-performance computing server with a Linux system through Matlab software.

For example, the high-performance computing server obtains the information that needs to be predicted through Webservice, for example: The predicted information is: the change trend of flight ZH9628 from Harbin to Shenzhen in the next 24 hours on May 1, 2017.

In some examples, in the price data collection stage, if the above-mentioned number of times t to obtain the price data per day is 4, it can also be predicted that the flight ZH9628 from Harbin to Shenzhen on May 1, 2017 will be in the next 6 hours. Trend.

In some examples, the high-performance computing server can read historical data and node data used for prediction from the system historical database according to the above-mentioned computer-readable storage medium according to the content of the obtained prediction information, and form corresponding data files. The data files are divided into There are two parts: prediction data and training data.

In some examples, the high-performance computing server can read the training data file through Matlab according to the computer-readable storage medium, and the program automatically trains the neural network model. After the training is completed, the training result is stored as a neural network structure file of Matlab. Calculate the results of the forecast data through the trained neural network model to obtain the forecasted price change trend. The prediction result is transmitted through the communication port, and then released to the user by the Webservice interface.

Claims (12)

1. a kind of flight prices Forecasting Methodology, it is characterised in that methods described includes：

N price data of same flight scheduled time slot before departure from port of T different predetermined date of departure is obtained, wherein, it is described Price data includes：Ticket price and the air ticket quantity corresponding with the ticket price；

P different sample of the sample length for k is obtained according to the n price data, wherein, the price number that each sample includes According to the same flight for identical predetermined date of departure the scheduled time slot price data, also, each sample include it is described Same flight upward price trend information, k is less than or equal to n；

Neural network model is trained using described p different samples as training sample；

Pre- timing of the same flight based on the predetermined date of departure of Neural Network model predictive after training before departure from port The price at quarter, wherein, T, n, k, p are positive integer.

2. flight prices Forecasting Methodology according to claim 1, it is characterised in that acquisition T is different to make a reservation for departure from port N price data of the same flight on date scheduled time slot before departure from port, including：

The scheduled time slot is averagely divided into n period, by the lowest price of the same flight in each period Lattice are recorded as the price data of each period.

3. flight prices Forecasting Methodology according to claim 1, it is characterised in that described according to the n price data P different sample of the sample length for k is obtained, including：

M × t-k are obtained by the price data of same flight scheduled time slot before departure from port of each identical predetermined date of departure Different samples, wherein, m is the number of days of the scheduled time slot, and t is to obtain the number of times of the price data daily, wherein, m × t More than or equal to k；

M × t-k that the price data of same flight scheduled time slot before departure from port based on each identical predetermined date of departure is obtained The quantity T of individual different samples and different predetermined date of departure obtains described p different sample.

4. flight prices Forecasting Methodology according to claim 1, it is characterised in that described to be made with described p different samples Before training sample training neural network model, in addition to：

When occurring missing price data in the n price data, the corresponding price data of the missing price data is selected The corresponding price data of timing node of the collection of price data next time of the timing node of collection, the price data of selection is made Added to for the missing price data in the n price data.

5. flight prices Forecasting Methodology according to any one of claim 1 to 4, it is characterised in that described with the p Before different samples are trained as training sample to neural network model, in addition to：

Compare p and sample size threshold value, and judge that p is more than or equal to the sample size threshold value.

6. a kind of flight prices prediction meanss, it is characterised in that described device includes：

Data acquisition unit, n valency of same flight scheduled time slot before departure from port for obtaining T different predetermined date of departure Lattice data, wherein, the price data includes：Ticket price and the air ticket quantity corresponding with the ticket price；

Sample chooses unit, for obtaining p different sample of the sample length for k according to the n price data, wherein, often The price data that individual sample includes is price data of the same flight in the scheduled time slot of identical predetermined date of departure, and And, each sample includes the same flight upward price trend information, and k is less than or equal to n；

Model training unit, for training neural network model using described p different samples as training sample；

Price expectation unit, exists for the same flight based on the predetermined date of departure of Neural Network model predictive after training The price of predetermined instant before departure from port, wherein, T, n, k, p are positive integer.

7. flight prices prediction meanss according to claim 6, it is characterised in that the data acquisition unit is additionally operable to：

The scheduled time slot is averagely divided into n period, by the lowest price of the same flight in each period Lattice are recorded as the price data of each period.

8. flight prices prediction meanss according to claim 6, it is characterised in that the sample is chosen unit and is additionally operable to：

M × t-k are obtained by the price data of same flight scheduled time slot before departure from port of each identical predetermined date of departure Different samples, wherein, m is the number of days of the scheduled time slot, and t is to obtain the number of times of the price data daily, wherein, m × t More than or equal to k；

M × t-k that the price data of same flight scheduled time slot before departure from port based on each identical predetermined date of departure is obtained The quantity T of individual different samples and different predetermined date of departure obtains described p different sample.

9. flight prices prediction meanss according to claim 6, it is characterised in that returning apparatus includes missing data and handles single Member, is used for：

When occurring missing price data in the n price data, the corresponding price data of the missing price data is selected The corresponding price data of timing node of the collection of price data next time of the timing node of collection, the price data of selection is made Added to for the missing price data in the n price data.

10. the flight prices prediction meanss according to any one of claim 6 to 9, it is characterised in that described device is also wrapped Judging unit is included, is used for：

Compare p and sample size threshold value, and judge that p is more than or equal to the sample size threshold value.

11. a kind of pre- measurement equipment of flight prices, it is characterised in that including memory and processor；

The memory is used to store executable program code；

The executable program code that the processor is stored for reading in the memory is any with perform claim requirement 1 to 5 Flight prices Forecasting Methodology described in.

12. a kind of computer-readable recording medium, including instruction, when the instruction is run on computers so that the meter Calculation machine performs the flight prices Forecasting Methodology as described in claim any one of 1-5.