CN112307341A - Flight pushing method and device - Google Patents

Flight pushing method and device Download PDF

Info

Publication number
CN112307341A
CN112307341A CN202011202346.1A CN202011202346A CN112307341A CN 112307341 A CN112307341 A CN 112307341A CN 202011202346 A CN202011202346 A CN 202011202346A CN 112307341 A CN112307341 A CN 112307341A
Authority
CN
China
Prior art keywords
flight
neural network
network model
training
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011202346.1A
Other languages
Chinese (zh)
Inventor
杨洪伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenyang Ne Cares Co ltd
Original Assignee
Shenyang Ne Cares Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenyang Ne Cares Co ltd filed Critical Shenyang Ne Cares Co ltd
Priority to CN202011202346.1A priority Critical patent/CN112307341A/en
Publication of CN112307341A publication Critical patent/CN112307341A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/90335—Query processing
    • G06F16/90344—Query processing by using string matching techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a flight pushing method and a flight pushing device, wherein the flight pushing method comprises the following steps: scanning seat information of each flight, and judging whether inventory flights exist or not based on the seat information; if the flight characteristic data exists, flight information of the inventory flights is obtained, and flight characteristic data corresponding to the inventory flights are generated based on the flight information; acquiring user characteristic data of each user in a database; inputting flight characteristic data and each user characteristic data into a neural network model to obtain a matching value which is output by the neural network model and corresponds to each inventory flight and each user; determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user; and acquiring flight activity information of the inventory flights, and pushing the flight activity information to each target user. By applying the method, the users matched with the stock flights are determined through the neural network model, and the flight activity information of the stock flights is pushed to the users, so that the attendance rate of the stock flights is improved.

Description

Flight pushing method and device
Technical Field
The invention relates to the technical field of computers, in particular to a flight pushing method and device.
Background
With the rapid development of passenger transport aviation, airplanes gradually become the choice of many people when going out. To increase the availability of flights, airlines typically attract customers to purchase tickets for a flight in the form of discounts or the like. Although airline companies have discounted airline tickets for flights, there are still situations in most cases where stock tickets are not sold for the flight.
At present, when a user needs to go out, the ticket of each flight can be purchased on a ticket purchasing website according to a user travel plan, and when the flight has discounted stock tickets, the user also needs to actively search and then know the discounting condition of the ticket. Therefore, in the prior art, a user cannot know which flight has a discounted air ticket in time, and an airline company cannot search for potential customers to push according to discount information of each flight, so that the attendance rate of the flight is affected.
Disclosure of Invention
In view of this, the present invention provides a flight pushing method, by which a neural network model is applied to analyze flight feature data and each user feature data, a potential customer of a flight is actively searched, and activity information of the flight is pushed to a user, so as to improve the attendance rate of the flight.
The invention also provides a flight pushing device used for ensuring the realization and the application of the method in practice.
A flight pushing method, comprising:
scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
if the stock flights exist in each flight, acquiring flight information of the stock flights, and generating flight characteristic data corresponding to the stock flights based on the flight information;
acquiring pre-stored user IDs of all users, and inquiring user characteristic data corresponding to each user ID in a preset database;
inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
and acquiring the flight activity information of the stock flight, and pushing the flight activity information of the stock flight to each target user.
Optionally, in the above method, the generating flight characteristic data corresponding to the inventory flight based on the flight information includes:
obtaining a departure place, a destination and departure time corresponding to the stock flight in the flight information;
acquiring preset route identification numbers corresponding to the departure place and the destination;
determining a travel season, a first date type and a second date type of the inventory flight based on the departure time, wherein the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
acquiring a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type and a second date identification number corresponding to the second date type;
and generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
The method described above, optionally, a process of training the neural network model, includes:
acquiring a preset training data set and a training label corresponding to each training data in the training data set, wherein each training data is combined data composed of user characteristic data and flight characteristic data used for model training;
sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer, a characteristic interaction layer and an output layer;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
Optionally, in the method, the sequentially inputting the training data in the training data set into the neural network model, and training the training data through a plurality of training layers of the neural network model includes:
sequentially inputting each training data in the training data set into an input layer of the neural network model, triggering the input layer to apply a preset one-hot code, and converting each training data into a feature vector;
sequentially inputting the feature vector corresponding to each training data into an embedding layer of the neural network model, and triggering the embedding layer to sequentially convert each feature vector into a dense vector;
sequentially inputting the dense vector corresponding to each piece of training data into a feature interaction layer of the neural network model, triggering the feature interaction layer to apply a preset DNN algorithm, a preset FM algorithm and a preset Multi-Head Attention algorithm, calculating the interaction relation between the user feature data and the flight feature data in the piece of training data corresponding to each dense vector, and obtaining a first parameter value, a second parameter value and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and sequentially inputting the first parameter value, the second parameter value and the third parameter value corresponding to each training data into an output layer of the neural network model, triggering the output layer to finally process the first parameter value, the second parameter value and the third parameter value corresponding to each training data, and outputting a training result corresponding to each training data.
Optionally, the method further includes, after finishing the training of the neural network model:
acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
inputting the test data into the neural network model in sequence, triggering the neural network model to output the test result corresponding to each test data in sequence until the test of the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
A flight pushing device, comprising:
the judging unit is used for scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
the generating unit is used for acquiring flight information of the inventory flights if the inventory flights exist in the flights, and generating flight characteristic data corresponding to the inventory flights based on the flight information;
the query unit is used for acquiring the pre-stored user IDs of all users and querying the user characteristic data corresponding to each user ID in a preset database;
the matching unit is used for inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
the device comprises a determining unit, a judging unit and a judging unit, wherein the determining unit is used for determining each matching value meeting a preset threshold value and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
and the pushing unit is used for acquiring the flight activity information of the stock flight and pushing the flight activity information of the stock flight to each target user.
The above apparatus, optionally, the generating unit includes:
the first acquiring subunit is configured to acquire a departure place, a destination, and departure time corresponding to the stock flight in the flight information;
the second obtaining subunit is used for obtaining preset route identification numbers corresponding to the departure place and the destination;
a first determining subunit, configured to determine, based on the departure time, a travel season of the inventory flight, a first date type and a second date type, where the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
a third obtaining subunit, configured to obtain a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type, and a second date identification number corresponding to the second date type;
and the generating subunit is used for generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
The above apparatus, optionally, further comprises:
the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a preset training data set and a training label corresponding to each training data in the training data set, and each training data is combined data consisting of user characteristic data and flight characteristic data used for model training;
the training unit is used for sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer, a characteristic interaction layer and an output layer;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
The above apparatus, optionally, the training unit, includes:
the first input subunit is configured to sequentially input each training data in the training data set to an input layer of the neural network model, trigger the input layer to apply a preset one-hot code, and convert each training data into a feature vector;
the second input subunit is used for sequentially inputting the feature vectors corresponding to each training data into the embedding layer of the neural network model, and triggering the embedding layer to sequentially convert the feature vectors into dense vectors;
the third input subunit is configured to sequentially input the dense vectors corresponding to each piece of training data into a feature interaction layer of the neural network model, trigger the feature interaction layer to apply a preset DNN algorithm, an preset FM algorithm, and a preset Multi-Head Attention algorithm, calculate an interaction relationship between user feature data and flight feature data in the piece of training data corresponding to each dense vector, and obtain a first parameter value, a second parameter value, and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and the fourth input subunit is configured to sequentially input the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data into an output layer of the neural network model, trigger the output layer to perform final processing on the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data, and output a training result corresponding to each piece of training data.
The above apparatus, optionally, further comprises:
the second acquisition unit is used for acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
the test unit is used for sequentially inputting the test data into the neural network model, triggering the neural network model to sequentially output a test result corresponding to each test data until the test on the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
A storage medium comprising stored instructions, wherein the instructions, when executed, control a device on which the storage medium is located to perform the above-mentioned flight pushing method.
An electronic device includes a memory, and one or more instructions, wherein the one or more instructions are stored in the memory and configured to be executed by one or more processors to perform the flight pushing method.
Compared with the prior art, the invention has the following advantages:
the invention provides a flight pushing method, which comprises the following steps: scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats; if the stock flights exist in each flight, acquiring flight information of the stock flights, and generating flight characteristic data corresponding to the stock flights based on the flight information; acquiring pre-stored user IDs of all users, and inquiring user characteristic data corresponding to each user ID in a preset database; inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively; determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user; and acquiring the flight activity information of the stock flight, and pushing the flight activity information of the stock flight to each target user. By applying the method provided by the invention, the users matched with the stock flights are determined through the neural network model, and the boarding rate of the stock flights is improved by pushing the flight activity information of the stock flights to the users.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a flight pushing method according to an embodiment of the present invention;
fig. 2 is a flowchart of another flight pushing method according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of each training layer of the neural network model according to the embodiment of the present invention;
fig. 4 is a structural diagram of a flight pushing device according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In this application, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions, and the terms "comprises", "comprising", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The invention is operational with numerous general purpose or special purpose computing device environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multi-processor apparatus, distributed computing environments that include any of the above devices or equipment, and the like.
The embodiment of the present invention provides a flight pushing method, which can be applied to a plurality of system platforms, an execution subject of the flight pushing method can be a computer terminal or a processor of various mobile devices, and a flow chart of the method is shown in fig. 1, and specifically includes:
s101: scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
in the embodiment of the invention, the seat information of each flight contains the current selling condition of each seat of the corresponding flight, and whether the flight in stock exists is determined through the seat information of each flight, namely whether the flight with unsold seats exists in each flight is judged.
S102: if the stock flights exist in each flight, acquiring flight information of the stock flights, and generating flight characteristic data corresponding to the stock flights based on the flight information;
in the embodiment of the invention, if the stock flight exists in each flight, the flight information of the stock flight is obtained. The flight information of the inventory flight includes the departure place, the destination, the departure time, the flight number, the seat type and the like of the flight. And generating corresponding flight characteristic data according to the flight information of the inventory flights.
S103: acquiring pre-stored user IDs of all users, and inquiring user characteristic data corresponding to each user ID in a preset database;
in the embodiment of the present invention, the database includes user feature data of a plurality of users, and the user feature data may specifically include: the user's name, gender, age, city of living, travel season preference, weekend travel preference, holiday travel preference, cabin preference, frequent destination and the like. And searching user characteristic data corresponding to each user in the database through the user ID.
When the user buys the air ticket for the first time, a user ID is assigned to the user or the name or other special characteristics of the user are used as the user ID of the user. And generating user characteristic data of the user based on the personal information filled in by the user when the ticket is purchased for the first time, and storing the user characteristic data into a database. And when the user purchases the air ticket of any flight again, updating the user characteristic data of the user in the database according to the air ticket of the flight purchased by the user each time. Wherein, in the user characteristic data, the gender of the user is divided into male, female or missing three conditions; according to the age of the user, the user can be classified into young, middle or old age; the resident city of the user is represented by a symbol corresponding to the city, for example, HK in hong kong, SZX in Shenzhen, and the like. The season that the user likes the travel and whether the user likes weekend travel and holiday travel are determined by counting historical ticket booking information of the user according to the season preference, weekend travel preference and holiday travel preference of the user. And the bin preference and the frequent destination determine the bin which the user prefers to purchase and the destination with the maximum ticket booking frequency according to the historical ticket booking information of the user.
S104: inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
in the embodiment of the invention, the neural network model is trained by a large amount of data, and the trained neural network model can calculate the matching value between each user and the stock flight according to the input user characteristic data and flight characteristic data. And predicting the possibility of purchasing the air ticket of the stock flight by the user corresponding to each user characteristic data in the database through the neural network model, wherein the higher the matching value of the user and the stock flight is, the higher the possibility of purchasing the air ticket of the stock flight is.
S105: determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
in the embodiment of the invention, the high matching value and the low matching value are divided by the preset threshold value, each matching value meeting the preset threshold value is a high matching value, and the probability that the user corresponding to the high matching value purchases the air ticket of the stock flight is higher, so that the user corresponding to each matching value meeting the preset threshold value is set as the target user for pushing the air ticket of the stock flight.
S106: and acquiring the flight activity information of the stock flight, and pushing the flight activity information of the stock flight to each target user.
In the embodiment of the invention, in order to improve the purchase rate of the ticket of the stock flight for the user, the flight activity information of the stock flight is obtained, and the flight activity information is pushed to each target user. The flight activity information comprises flight number, travel time, departure place, destination, cabin price, activity cabin, meal condition and other information.
In the flight pushing method provided by the embodiment of the invention, the seat information of each flight is scanned, so that whether the flight corresponding to each seat information has the remaining seat or not is judged based on each seat information, that is, whether stock flights exist in each flight or not is determined. If the inventory flight exists, acquiring flight information of the inventory flight, and generating flight characteristic data corresponding to the inventory flight based on the flight information. And according to the stored user ID of each user, acquiring user characteristic data corresponding to each user ID in the database, and inputting the flight characteristic data and the user characteristic data into a neural network model to obtain a matching value between each user and the inventory flight output by the model. And selecting the user corresponding to the matching value meeting the preset threshold value as a target user for pushing the stock flight from the users corresponding to the matching values, acquiring flight activity information of the stock flight, and pushing the flight activity information to each target user.
It should be noted that the neural network model can match a plurality of inventory flights and a plurality of users simultaneously according to a plurality of flight feature data and a plurality of user feature data. When multiple flight profiles and multiple user profiles are input to the neural network model, the neural network model will output a match value between each user and the respective inventory flight. And if the matching values between any plurality of stock flights and one user all meet the preset threshold value, pushing each stock flight meeting the matching value of the preset threshold value to the user.
It should be further noted that, in the flight pushing method provided in the embodiment of the present invention, in order to increase the seat availability of each flight, a scanning period corresponding to seat information of each flight may be preset, the seat information of each flight is scanned according to the scanning period, so as to determine the sales condition of each flight seat, and a flight with an unsold seat is pushed to the user in time.
By applying the method provided by the embodiment of the invention, flight characteristic data of the stock flight and user characteristic data of each user in the database are obtained, the flight characteristic data and the user characteristic data are analyzed by the neural network model, the matching value between the stock flight and each user is determined, and flight activity information of the stock flight is pushed to the users meeting the conditions according to the matching value, so that the attendance rate of the stock flight is improved.
In the method provided by the embodiment of the present invention, based on the content of S102, after determining that an inventory flight exists in each flight, flight feature data of the inventory flight needs to be obtained, so as to analyze the flight feature data and each user feature data through a neural network model, and determine a matching value between each user and the inventory flight. Referring to fig. 2, the process of generating flight feature data corresponding to the inventory flights based on the flight information may include:
s201: obtaining a departure place, a destination and departure time corresponding to the stock flight in the flight information;
in the embodiment of the present invention, the flight information of the inventory flights may include flight number, flight type, seat category, and other information besides the departure place, destination, and departure time.
S202: acquiring preset route identification numbers corresponding to the departure place and the destination;
in the embodiment of the invention, the route identification number consists of a city symbol corresponding to a starting place and a city symbol corresponding to a destination; for example: if the starting place of the stock flight is Shanghai, the corresponding city symbol is SHH, and the destination is Shenzhen, the corresponding city symbol is SZX, so that the airline identification number of the stock flight can be SHH-SZX.
S203: determining a travel season, a first date type and a second date type of the inventory flight based on the departure time, wherein the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
in the embodiment of the invention, according to the departure time, determining whether the travel season of the stock flight is in spring, summer, autumn or winter; determining whether the travel time of the inventory flight is a holiday or a non-holiday; it is determined whether the travel time for the inventory flight is weekend or non-weekend.
S204: acquiring a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type and a second date identification number corresponding to the second date type;
in the embodiment of the present invention, after determining the travel season, the first date type, and the second date type corresponding to the inventory flight, identification numbers corresponding to the travel season, the first date type, and the second date type, that is, a season identification number, a first date identification number, and a second date identification number, are respectively obtained.
For example, the travel seasons include spring, summer, fall and winter, and the identification numbers corresponding to each season may be a1, a2, A3 and a 4; if the first date type is a holiday and a non-holiday, the identification numbers corresponding to the holiday and the non-holiday can be B1 and B2; the second date type is weekend and non-weekend, and the identification numbers corresponding to weekend and non-weekend may be C1 and C2. If the travel time of the inventory flight is 10/1 of 2020, the corresponding travel season is autumn, the first date type is holiday, the second date type is non-weekend, and the season identification number at this time is a3, the first date identification number is B1, and the second date identification number is C2.
S205: and generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
In the embodiment of the invention, the flight characteristic data consists of a route identification number, a season identification number, a first date identification number and a second date identification number. For example, if the departure place of the inventory flight is Shanghai, the destination is Shenzhen, and the travel time is 10/1/10/2020, the flight feature data may be SHH-SZX/A3/B1/C2.
It should be noted that, in the embodiment provided by the present invention, the specific representation form of each identification number and the sequence of each identification number in the flight feature data may be adjusted according to the actual application environment, and this is not limited herein.
In the flight pushing method provided by the embodiment of the invention, the departure place, the destination and the departure time corresponding to the stock flight in the flight information are obtained, so that the route identification number corresponding to the stock flight is determined according to the departure place and the destination; and determining the season identification number, the first date identification number and the second season identification number corresponding to the inventory flight according to the departure time. Flight feature data corresponding to the inventory flights are generated based on the airline identification number, the season identification number, the first date identification number and the second season identification number.
Optionally, in the process of generating the flight feature data, in order to improve the matching degree between the user and the flight, the flight feature data may be determined by flight features of airlines, seasons, holidays, and weekends, and may also be determined by flight features such as flight numbers and cabin types.
By applying the method provided by the embodiment of the invention, flight characteristic data are generated based on flight information, so that the flight characteristic data are matched with the user characteristic data through a neural network model.
In the method provided by the embodiment of the invention, after flight characteristic data of the inventory flights and user characteristic data of each user in the database are obtained, a training completion neural network model is required to be applied to analyze the input data so as to obtain a matching value between each user and the inventory flights. The neural network model is trained for multiple times through a large amount of data, and a matching value between a user and a flight can be basically and accurately analyzed. Specifically, the process of training the neural network model may include:
acquiring a preset training data set and a training label corresponding to each training data in the training data set, wherein each training data is combined data composed of user characteristic data and flight characteristic data used for model training;
sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer and a characteristic interaction layer respectively;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
In the flight pushing method provided by the embodiment of the invention, when a neural network model needs to be trained, a training data set for training the neural network model and a training label corresponding to each training data in the training data set are obtained. The training data set comprises a plurality of training data, and each training data is historical data, namely combination data of corresponding user characteristic data and flight characteristic data when a plurality of users purchase tickets historically. The training labels are preset matching values between the user characteristic data and flight characteristic data corresponding to each piece of training data. And inputting the training data into the neural network model in sequence, and training and calculating the training data by each training layer in the neural network model. The neural network model comprises a plurality of training layers, wherein each training layer is an input layer and an embedded layer which are characteristic interaction layers. The input layer is used for preprocessing the currently input training data, the embedded layer converts the preprocessed training data, the feature interaction layer performs feature calculation on the data, and the final output layer performs final processing on the calculation result of the feature interaction layer and outputs the result.
Specifically, the training process of the neural network model is as follows: and sequentially inputting each training data in the training data set into the neural network model, and finally outputting a current training result corresponding to the training data currently input into the neural network model after each training layer in the neural network model processes, converts, interacts with features and the like the training data. The current training result is a predicted matching value between the user characteristic data and flight characteristic data corresponding to the training data predicted by the neural network model. And calculating a loss value corresponding to the training result based on the training label of the training data, namely comparing the real matching value of the training data with a predicted matching value output by the neural network model to obtain a corresponding loss value. Judging whether the loss value meets a preset training condition, wherein the training condition is a preset threshold value; if the loss value does not meet the training condition, namely the loss value does not reach the set threshold value, updating model parameters of the neural network model through the loss value and a gradient descent algorithm, and continuing to train the neural network model; and if the loss value meets the training condition, determining that the neural network model can basically and accurately predict the matching value of the training data, finishing the training of the neural network model, and obtaining the trained neural network model.
Optionally, after determining that the loss value corresponding to the current training result meets the training condition, the neural network model may be trained again according to the training process for multiple times, until the loss values obtained in the multiple training processes all meet the training condition, and then the training of the neural network model is ended.
By applying the method provided by the embodiment of the invention, the neural network model is trained through a large amount of training data in the training data set to obtain the trained neural network model, the matching value between the user and the flight is calculated through the neural network model, the flight is pushed for the user through the matching value, and the attendance rate of the flight is improved.
In the method provided by the embodiment of the present invention, in the process of training the neural network model, each training data input to the neural network model is processed and calculated by each training layer in the neural network model, so that each training data in the training data set is sequentially input to the neural network model, and each training data is trained through a plurality of training layers of the neural network model, including:
sequentially inputting each training data in the training data set into an input layer of the neural network model, triggering the input layer to apply a preset one-hot code, and converting each training data into a feature vector;
sequentially inputting the feature vector corresponding to each training data into an embedding layer of the neural network model, and triggering the embedding layer to sequentially convert each feature vector into a dense vector;
sequentially inputting the dense vector corresponding to each piece of training data into a feature interaction layer of the neural network model, triggering the feature interaction layer to apply a preset DNN algorithm, a preset FM algorithm and a preset Multi-Head Attention algorithm, calculating the interaction relation between the user feature data and the flight feature data in the piece of training data corresponding to each dense vector, and obtaining a first parameter value, a second parameter value and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and sequentially inputting the first parameter value, the second parameter value and the third parameter value corresponding to each training data into an output layer of the neural network model, triggering the output layer to finally process the first parameter value, the second parameter value and the third parameter value corresponding to each training data, and outputting a training result corresponding to each training data.
In the flight pushing method provided by the embodiment of the invention, the neural network model comprises a plurality of training layers, wherein each training layer is an input layer, an embedded layer, a characteristic interaction layer and an output layer. Referring to fig. 3 in particular, fig. 3 is a schematic structural diagram of each training layer in the neural network model, where the input layer 301 is configured to preprocess input training data and convert discrete features in the training data into one-hot-form feature vectors by applying one-hot. After the input layer 301 outputs the feature vector of the training data, the feature vector is input to the embedding layer 302. The embedding layer 302 is used for converting the one-hot form feature vector into a dense vector, and inputting the dense vector output by the embedding layer 302 into the feature interaction layer 303, wherein the feature interaction layer 303 comprises three sub-layers, namely a DNN layer 3031, an FM layer 3032 and an Attention layer 3033; respectively inputting the dense vectors into the three sublayers, calculating the dense vectors by applying a DNN algorithm through the DNN layer 3031, and outputting a first parameter value corresponding to the training data; the FM layer 3032 calculates the dense vector by applying an FM algorithm and outputs a second parameter value corresponding to the training data; the Attention layer 3033 calculates the dense vector by using a Multi-Head Attention algorithm and outputs a third parameter value corresponding to the training data. After the feature interaction layer 303 outputs the first parameter value, the second parameter value and the second parameter value corresponding to the training data, the first parameter value, the second parameter value and the third parameter value are input into the output layer 304, the output layer adds the first parameter value, the second parameter value and the third parameter value, and then the final processing is performed by applying a calculation method of a sigmoid function, so that a training result of the training data is obtained.
Specifically, the formula corresponding to the feature vector for converting the training data into one-hot in the input layer is as follows: x ═ X0,x1,…,xn) Wherein x is0,x1,…,xnFor each discrete feature in the training data;
the formula corresponding to the conversion of the feature vectors into dense vectors in the embedding layer is: v ═ embedding (x);
the DNN algorithm in the DNN layer in the feature interaction layer is as follows: y isDNNDnn (v); the FM algorithm in the FM layer is as follows: y isFMFM (X, V); the Multi-HeadAttention algorithm in the Attention layer is as follows: y isAttention＝Attention(V)；
The training result output by the output layer is as follows: y is sigmoid (y)DNN+yFM+yAttention)。
It should be noted that, after the neural network model is trained, when the trained neural network model is applied to the above S104, the algorithms corresponding to the above training layers are applied to calculate the flight feature data and the user feature data, so as to obtain the matching value corresponding to each user and the inventory flight.
The training data are calculated through each training layer and the corresponding algorithm thereof to predict the matching degree between the user and the flight, the training result corresponding to the current training data is accurately output through the calculation of each training layer, and whether the training of the neural network model reaches the preset condition or not is determined through the training result.
In the method provided by the embodiment of the present invention, after the training of the neural network model is finished, in order to improve the accuracy of the neural network model in matching the user characteristic data and the flight characteristic data, the neural network model may be tested after training, and therefore, the testing process of the neural network model may specifically include:
acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
inputting the test data into the neural network model in sequence, triggering the neural network model to output the test result corresponding to each test data in sequence until the test of the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
In the flight pushing method provided by the embodiment of the invention, the training of the neural network model is finished, and the neural network model is tested by applying each test data in the test data set again. Wherein the process of testing the neural network model is substantially similar to the training process. The test data set contains a plurality of test data, and all the test data are required to be applied to test the neural network model. And sequentially inputting each test data in the test data set into the trained neural network model, and outputting a currently output test result corresponding to the currently input test data by the neural network model every time one test data is input. Judging whether the currently output test result meets preset test conditions or not according to the test label corresponding to the currently input test data, if not, adjusting each model parameter of the neural network model again, after the model parameters are adjusted, judging whether the currently input test data is the last test data or not; continuing to input next test data to the neural network model and continuing to test the neural network model; if the currently input test data is the last test data, the neural network model is stored, and the test of the neural network model is finished.
Optionally, in the process of testing the neural network model, if the currently output test result of the neural network model meets the test condition, it is still determined whether the test data corresponding to the currently output test result is the last test data; if not, continuing the test corresponding to the next test data; if yes, ending the test of the neural network model.
It should be noted that the test condition is whether the currently output test result is within the value range of the test tag of the corresponding test data; if the value is within the value range, the test condition is met, and if the value is not within the value range, the test condition is not met.
It can be understood that, through the test of the neural network model by the plurality of test data, the neural network model can basically meet the test condition when being tested subsequently. The method provided by the embodiment of the invention can further improve the precision of the neural network model.
The specific implementation procedures and derivatives thereof of the above embodiments are within the scope of the present invention.
Corresponding to the method described in fig. 1, an embodiment of the present invention further provides a flight pushing device, which is used for specifically implementing the method in fig. 1, where the flight pushing device provided in the embodiment of the present invention may be applied to a computer terminal or various mobile devices, and a schematic structural diagram of the flight pushing device is shown in fig. 4, and specifically includes:
a determining unit 401, configured to scan seat information of each flight, and determine whether there is an inventory flight in each flight based on the seat information of each flight, where the inventory flight is a flight with unsold seats;
a generating unit 402, configured to, if an inventory flight exists in each flight, acquire flight information of the inventory flight, and generate flight feature data corresponding to the inventory flight based on the flight information;
an inquiring unit 403, configured to obtain user IDs of users stored in advance, and inquire, in a preset database, user feature data corresponding to each user ID;
a matching unit 404, configured to input the flight feature data and the user feature data into a neural network model that is trained in advance, and obtain matching values that correspond to the inventory flight and each user output by the neural network model;
a determining unit 405, configured to determine each matching value meeting a preset threshold, and set a user corresponding to each matching value meeting the preset threshold as a target user;
the pushing unit 406 is configured to obtain flight activity information of the stock flight, and push the flight activity information of the stock flight to each target user.
In the flight pushing device provided by the embodiment of the invention, the seat information of each flight is scanned, so that whether the flight corresponding to each seat information has the remaining seat or not is judged based on each seat information, that is, whether stock flights exist in each flight or not is determined. If the inventory flight exists, acquiring flight information of the inventory flight, and generating flight characteristic data corresponding to the inventory flight based on the flight information. And according to the stored user ID of each user, acquiring user characteristic data corresponding to each user ID in the database, and inputting the flight characteristic data and the user characteristic data into a neural network model to obtain a matching value between each user and the inventory flight output by the model. And selecting the user corresponding to the matching value meeting the preset threshold value as a target user for pushing the stock flight from the users corresponding to the matching values, acquiring flight activity information of the stock flight, and pushing the flight activity information to each target user.
The device provided by the embodiment of the invention is applied to obtain the flight characteristic data of the stock flight and the user characteristic data of each user in the database, the neural network model analyzes the flight characteristic data and the user characteristic data, the matching value between the stock flight and each user is determined, and the flight activity information of the stock flight is pushed to the users meeting the conditions according to the matching value, so that the attendance rate of the stock flight is improved.
In the apparatus provided in the embodiment of the present invention, the generating unit includes:
the first acquiring subunit is configured to acquire a departure place, a destination, and departure time corresponding to the stock flight in the flight information;
the second obtaining subunit is used for obtaining preset route identification numbers corresponding to the departure place and the destination;
a first determining subunit, configured to determine, based on the departure time, a travel season of the inventory flight, a first date type and a second date type, where the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
a third obtaining subunit, configured to obtain a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type, and a second date identification number corresponding to the second date type;
and the generating subunit is used for generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
The device provided by the embodiment of the invention further comprises:
the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a preset training data set and a training label corresponding to each training data in the training data set, and each training data is combined data consisting of user characteristic data and flight characteristic data used for model training;
the training unit is used for sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer, a characteristic interaction layer and an output layer;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
In the apparatus provided in the embodiment of the present invention, the training unit includes:
the first input subunit is configured to sequentially input each training data in the training data set to an input layer of the neural network model, trigger the input layer to apply a preset one-hot code, and convert each training data into a feature vector;
the second input subunit is used for sequentially inputting the feature vectors corresponding to each training data into the embedding layer of the neural network model, and triggering the embedding layer to sequentially convert the feature vectors into dense vectors;
the third input subunit is configured to sequentially input the dense vectors corresponding to each piece of training data into a feature interaction layer of the neural network model, trigger the feature interaction layer to apply a preset DNN algorithm, an preset FM algorithm, and a preset Multi-Head Attention algorithm, calculate an interaction relationship between user feature data and flight feature data in the piece of training data corresponding to each dense vector, and obtain a first parameter value, a second parameter value, and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and the fourth input subunit is configured to sequentially input the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data into an output layer of the neural network model, trigger the output layer to perform final processing on the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data, and output a training result corresponding to each piece of training data.
The device provided by the embodiment of the invention further comprises:
the second acquisition unit is used for acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
the test unit is used for sequentially inputting the test data into the neural network model, triggering the neural network model to sequentially output a test result corresponding to each test data until the test on the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
The specific working processes of each unit and sub-unit in the flight pushing device disclosed in the above embodiment of the present invention may refer to corresponding contents in the flight pushing method disclosed in the above embodiment of the present invention, and are not described again here.
The embodiment of the invention also provides a storage medium, which comprises a stored instruction, wherein when the instruction runs, the device where the storage medium is located is controlled to execute the flight pushing method.
An electronic device is provided in an embodiment of the present invention, and the structural diagram of the electronic device is shown in fig. 5, which specifically includes a memory 501 and one or more instructions 502, where the one or more instructions 502 are stored in the memory 501, and are configured to be executed by one or more processors 503 to perform the following operations according to the one or more instructions 502:
scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
if the stock flights exist in each flight, acquiring flight information of the stock flights, and generating flight characteristic data corresponding to the stock flights based on the flight information;
acquiring pre-stored user IDs of all users, and inquiring user characteristic data corresponding to each user ID in a preset database;
inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
and acquiring the flight activity information of the stock flight, and pushing the flight activity information of the stock flight to each target user.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both.
To clearly illustrate this interchangeability of hardware and software, various illustrative components and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A flight pushing method, comprising:
scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
if the stock flights exist in each flight, acquiring flight information of the stock flights, and generating flight characteristic data corresponding to the stock flights based on the flight information;
acquiring pre-stored user IDs of all users, and inquiring user characteristic data corresponding to each user ID in a preset database;
inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
determining each matching value meeting a preset threshold value, and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
and acquiring the flight activity information of the stock flight, and pushing the flight activity information of the stock flight to each target user.
2. The method of claim 1, wherein generating flight characteristic data corresponding to the inventory flights based on the flight information comprises:
obtaining a departure place, a destination and departure time corresponding to the stock flight in the flight information;
acquiring preset route identification numbers corresponding to the departure place and the destination;
determining a travel season, a first date type and a second date type of the inventory flight based on the departure time, wherein the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
acquiring a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type and a second date identification number corresponding to the second date type;
and generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
3. The method of claim 1, wherein training the neural network model comprises:
acquiring a preset training data set and a training label corresponding to each training data in the training data set, wherein each training data is combined data composed of user characteristic data and flight characteristic data used for model training;
sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer, a characteristic interaction layer and an output layer;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
4. The method of claim 3, wherein the sequentially inputting each of the training data in the set of training data into the neural network model, each of the training data being trained by a plurality of training layers of the neural network model, comprises:
sequentially inputting each training data in the training data set into an input layer of the neural network model, triggering the input layer to apply a preset one-hot code, and converting each training data into a feature vector;
sequentially inputting the feature vector corresponding to each training data into an embedding layer of the neural network model, and triggering the embedding layer to sequentially convert each feature vector into a dense vector;
sequentially inputting the dense vector corresponding to each piece of training data into a feature interaction layer of the neural network model, triggering the feature interaction layer to apply a preset DNN algorithm, a preset FM algorithm and a preset Multi-Head Attention algorithm, calculating the interaction relation between the user feature data and the flight feature data in the piece of training data corresponding to each dense vector, and obtaining a first parameter value, a second parameter value and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and sequentially inputting the first parameter value, the second parameter value and the third parameter value corresponding to each training data into an output layer of the neural network model, triggering the output layer to finally process the first parameter value, the second parameter value and the third parameter value corresponding to each training data, and outputting a training result corresponding to each training data.
5. The method of claim 3, wherein after finishing the training of the neural network model, further comprising:
acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
inputting the test data into the neural network model in sequence, triggering the neural network model to output the test result corresponding to each test data in sequence until the test of the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
6. A flight pushing device, comprising:
the judging unit is used for scanning seat information of each flight, and judging whether stock flights exist in each flight or not based on the seat information of each flight, wherein the stock flights are flights with unsold seats;
the generating unit is used for acquiring flight information of the inventory flights if the inventory flights exist in the flights, and generating flight characteristic data corresponding to the inventory flights based on the flight information;
the query unit is used for acquiring the pre-stored user IDs of all users and querying the user characteristic data corresponding to each user ID in a preset database;
the matching unit is used for inputting the flight feature data and the user feature data into a pre-trained neural network model to obtain matching values which are output by the neural network model and correspond to the inventory flights and the users respectively;
the device comprises a determining unit, a judging unit and a judging unit, wherein the determining unit is used for determining each matching value meeting a preset threshold value and setting a user corresponding to each matching value meeting the preset threshold value as a target user;
and the pushing unit is used for acquiring the flight activity information of the stock flight and pushing the flight activity information of the stock flight to each target user.
7. The apparatus of claim 6, wherein the generating unit comprises:
the first acquiring subunit is configured to acquire a departure place, a destination, and departure time corresponding to the stock flight in the flight information;
the second obtaining subunit is used for obtaining preset route identification numbers corresponding to the departure place and the destination;
a first determining subunit, configured to determine, based on the departure time, a travel season of the inventory flight, a first date type and a second date type, where the first date type is a holiday or a non-holiday, and the second date type is a weekend or a non-weekend;
a third obtaining subunit, configured to obtain a season identification number corresponding to the trip season, a first date identification number corresponding to the first date type, and a second date identification number corresponding to the second date type;
and the generating subunit is used for generating flight characteristic data corresponding to the stock flights based on the airline identification number, the season identification number, the first date identification number and the second date identification number.
8. The apparatus of claim 6, further comprising:
the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a preset training data set and a training label corresponding to each training data in the training data set, and each training data is combined data consisting of user characteristic data and flight characteristic data used for model training;
the training unit is used for sequentially inputting each training data in the training data set into the neural network model, training each training data through a plurality of training layers of the neural network model, and ending the training of the neural network model until a training result output by the last training layer in the neural network model meets a preset training condition; each training layer of the neural network model is an input layer, an embedding layer, a characteristic interaction layer and an output layer;
when each training data in the training data set is sequentially input into the neural network model, acquiring a current training result of the training data currently input into the neural network model; calculating a loss value corresponding to the current training result based on a training label corresponding to the training data; judging whether the loss value meets a preset training condition or not; if not, updating each model parameter in the neural network model based on the loss value and a preset gradient descent algorithm, and continuing to train the neural network model; and if so, finishing the training of the neural network model.
9. The apparatus of claim 8, wherein the training unit comprises:
the first input subunit is configured to sequentially input each training data in the training data set to an input layer of the neural network model, trigger the input layer to apply a preset one-hot code, and convert each training data into a feature vector;
the second input subunit is used for sequentially inputting the feature vectors corresponding to each training data into the embedding layer of the neural network model, and triggering the embedding layer to sequentially convert the feature vectors into dense vectors;
the third input subunit is configured to sequentially input the dense vectors corresponding to each piece of training data into a feature interaction layer of the neural network model, trigger the feature interaction layer to apply a preset DNN algorithm, an preset FM algorithm, and a preset Multi-Head Attention algorithm, calculate an interaction relationship between user feature data and flight feature data in the piece of training data corresponding to each dense vector, and obtain a first parameter value, a second parameter value, and a third parameter value corresponding to each piece of training data output from the feature interaction layer;
and the fourth input subunit is configured to sequentially input the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data into an output layer of the neural network model, trigger the output layer to perform final processing on the first parameter value, the second parameter value, and the third parameter value corresponding to each piece of training data, and output a training result corresponding to each piece of training data.
10. The apparatus of claim 8, further comprising:
the second acquisition unit is used for acquiring a preset test data set and a test label corresponding to each test data in the test data set, wherein each test data is combined data consisting of user characteristic data and flight characteristic data for testing the performance of the neural network model;
the test unit is used for sequentially inputting the test data into the neural network model, triggering the neural network model to sequentially output a test result corresponding to each test data until the test on the neural network model is finished;
judging whether a test result currently output by the neural network model meets a preset test condition or not based on a test label corresponding to test data currently input into the neural network model; if the test result currently output by the neural network model does not meet the test condition, adjusting each model parameter in the neural network model based on the test result, and judging whether the test data currently input into the neural network model is the last test data; if not, continuing to input next test data to test the neural network model; if yes, the neural network model is saved, and the test of the neural network model is finished.
CN202011202346.1A 2020-11-02 2020-11-02 Flight pushing method and device Pending CN112307341A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011202346.1A CN112307341A (en) 2020-11-02 2020-11-02 Flight pushing method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011202346.1A CN112307341A (en) 2020-11-02 2020-11-02 Flight pushing method and device

Publications (1)

Publication Number Publication Date
CN112307341A true CN112307341A (en) 2021-02-02

Family

ID=74333516

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011202346.1A Pending CN112307341A (en) 2020-11-02 2020-11-02 Flight pushing method and device

Country Status (1)

Country Link
CN (1) CN112307341A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113590933A (en) * 2021-06-07 2021-11-02 海南太美航空股份有限公司 Flight pushing method and system and electronic equipment

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015225619A (en) * 2014-05-30 2015-12-14 株式会社野村総合研究所 Vacant seat sales system
CN108038508A (en) * 2017-12-12 2018-05-15 携程旅游网络技术（上海）有限公司 Method for pushing, system, storage medium and the electronic equipment of connecting flight
CN110213325A (en) * 2019-04-02 2019-09-06 腾讯科技（深圳）有限公司 Data processing method and data push method
CN110348968A (en) * 2019-07-15 2019-10-18 辽宁工程技术大学 A kind of recommender system and method analyzed based on user and project coupled relation
CN110400169A (en) * 2019-07-02 2019-11-01 阿里巴巴集团控股有限公司 A kind of information-pushing method, device and equipment
CN110751395A (en) * 2019-10-21 2020-02-04 中国民航信息网络股份有限公司 Passenger travel state determination method, device and server
CN111027714A (en) * 2019-12-11 2020-04-17 腾讯科技（深圳）有限公司 Artificial intelligence-based object recommendation model training method, recommendation method and device
CN111553759A (en) * 2020-03-25 2020-08-18 平安科技（深圳）有限公司 Product information pushing method, device, equipment and storage medium
CN111586162A (en) * 2020-05-06 2020-08-25 支付宝(杭州)信息技术有限公司 Information pushing method and system

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015225619A (en) * 2014-05-30 2015-12-14 株式会社野村総合研究所 Vacant seat sales system
CN108038508A (en) * 2017-12-12 2018-05-15 携程旅游网络技术（上海）有限公司 Method for pushing, system, storage medium and the electronic equipment of connecting flight
CN110213325A (en) * 2019-04-02 2019-09-06 腾讯科技（深圳）有限公司 Data processing method and data push method
CN110400169A (en) * 2019-07-02 2019-11-01 阿里巴巴集团控股有限公司 A kind of information-pushing method, device and equipment
CN110348968A (en) * 2019-07-15 2019-10-18 辽宁工程技术大学 A kind of recommender system and method analyzed based on user and project coupled relation
CN110751395A (en) * 2019-10-21 2020-02-04 中国民航信息网络股份有限公司 Passenger travel state determination method, device and server
CN111027714A (en) * 2019-12-11 2020-04-17 腾讯科技（深圳）有限公司 Artificial intelligence-based object recommendation model training method, recommendation method and device
CN111553759A (en) * 2020-03-25 2020-08-18 平安科技（深圳）有限公司 Product information pushing method, device, equipment and storage medium
CN111586162A (en) * 2020-05-06 2020-08-25 支付宝(杭州)信息技术有限公司 Information pushing method and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113590933A (en) * 2021-06-07 2021-11-02 海南太美航空股份有限公司 Flight pushing method and system and electronic equipment

Similar Documents

Publication Publication Date Title
US20190205939A1 (en) Using Machine Learned Visitor Intent Propensity to Greet and Guide a Visitor at a Physical Venue
US20170364933A1 (en) User maintenance system and method
US20170278173A1 (en) Personalized bundle recommendation system and method
US20190205936A1 (en) Symbiotic Reporting Code and Location Tracking Infrastructure for Physical Venues
CN110097287B (en) Group portrait method for logistic drivers
CN107274257A (en) Method and device is recommended in a kind of hotel service
US11449770B2 (en) Action selection learning device, action selection learning method, and storage medium
US11107109B2 (en) Method and system for personalizing offers
JP2016540319A5 (en)
CN112365283B (en) Coupon issuing method and device, terminal equipment and storage medium
AU2018202369A1 (en) A method and system for providing a loyalty program
US11947632B2 (en) Training a classification model using labeled training data that does not overlap with target classifications for the classification model
JP7078357B2 (en) Distribution device, distribution method and distribution program
CN113139769B (en) Intelligent purchasing scheme recommending method and device, computer equipment and storage medium
JP6101620B2 (en) Purchase forecasting apparatus, method, and program
CN110956303A (en) Information prediction method, device, terminal and readable storage medium
CN111367575A (en) User behavior prediction method and device, electronic equipment and storage medium
CN112307341A (en) Flight pushing method and device
CN111104585B (en) Question recommending method and device
CN114997916A (en) Prediction method, system, electronic device and storage medium of potential user
CN110751575A (en) Method and device for processing civil aviation passenger service data
CN113034168A (en) Content item delivery method and device, computer equipment and storage medium
CN113874890A (en) Subscription-based travel service
Kolbeinsson et al. Galactic air improves ancillary revenues with dynamic personalized pricing
CN111476613A (en) Shopping guide auxiliary method and device based on passenger flow analysis, server and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination