JP6865192B2

JP6865192B2 - Programs, devices and methods for estimating user boarding information from mobile terminal communication logs

Info

Publication number: JP6865192B2
Application number: JP2018074436A
Authority: JP
Inventors: 亜令小林; 茂莉黒川
Original assignee: KDDI Corp
Current assignee: KDDI Corp
Priority date: 2018-04-09
Filing date: 2018-04-09
Publication date: 2021-04-28
Anticipated expiration: 2038-04-09
Also published as: JP2019186713A

Description

本発明は、ユーザによって所持される携帯端末の通信ログから、当該ユーザの移動経路を推定する技術に関する。 The present invention relates to a technique for estimating a movement route of a user from a communication log of a mobile terminal possessed by the user.

人の移動量の実態を把握するために、「パーソントリップ調査」がある。例えば不特定多数のユーザが、「いつ」「どの移動経路を」「どの運行会社の」「どの移動機関で」移動したか、をビッグデータとして収集し分析することができる。
特に近年、ユーザは、スマートフォンやタブレットのような携帯端末を常に所持しているために、通信事業者は、通信ログを自動的に収集することによって、ユーザの移動経路を推定することもできる。 There is a "person trip survey" to grasp the actual situation of the amount of movement of people. For example, it is possible to collect and analyze as big data whether an unspecified number of users have moved "when", "which movement route", "which operating company", and "which mobile organization".
Particularly in recent years, since users always have mobile terminals such as smartphones and tablets, telecommunications carriers can also estimate the user's movement route by automatically collecting communication logs.

従来、携帯端末のＧＰＳ(Global Positioning System)等の測位機能を起動させることなく、通信事業設備によって取得可能な、空間的粒度が粗く且つ時間間隔が一定でない基地局位置情報を用いて、ユーザの有意位置を推定する技術がある（例えば特許文献１参照）。
また、通信事業設備によって取得可能な基地局位置情報を用いて、ユーザにとって有意な滞在地を高い精度で推定する技術もある（例えば特許文献２参照）。
更に、移動端末の通信履歴に基づいて、ユーザが利用した路線を推定する技術もある（例えば特許文献３参照）。 Conventionally, without activating a positioning function such as GPS (Global Positioning System) of a mobile terminal, a user can use base station position information that can be acquired by a communication business facility and has a coarse spatial grain size and a non-constant time interval. There is a technique for estimating a significant position (see, for example, Patent Document 1).
There is also a technique for estimating a significant place of stay for the user with high accuracy by using the base station position information that can be acquired by the communication business equipment (see, for example, Patent Document 2).
Further, there is also a technique of estimating the route used by the user based on the communication history of the mobile terminal (see, for example, Patent Document 3).

特開２０１２−０８５０９５号公報Japanese Unexamined Patent Publication No. 2012-085095 特開２０１４−１１６８０８号公報Japanese Unexamined Patent Publication No. 2014-116808 特開２０１６−１３４７３１号公報Japanese Unexamined Patent Publication No. 2016-134731

前述したいずれの従来技術も、移動経路における通信ログを用いたものあって、その移動経路は、基地局との通信可能範囲に存在することを前提としている。即ち、移動経路は、鉄道や道路のような陸上を前提としたものである。 All of the above-mentioned conventional techniques use the communication log in the mobile path, and it is premised that the mobile path exists in the communicable range with the base station. That is, the movement route is premised on land such as railroads and roads.

これに対し、ユーザが航空路線を利用する場合、その空港間では、基地局との間で通信不可能となるために、移動中の通信ログを取得することはできない。
また、同一の出発空港と到着空港との間では、同一の航空路線に、異なる航空会社の複数の航空便が運航しており、いずれの航空便を利用したかを特定することは難しい。
尚、航空路線では、搭乗機の搭乗者数は、各航空会社でのみ管理されており、パーソントリップ調査自体がなされることはない。そのために異なる航空会社の各搭乗機について、その搭乗者数などを分析することも難しい。 On the other hand, when the user uses an air route, it is not possible to acquire the communication log during movement because communication between the airports and the base station becomes impossible.
In addition, it is difficult to identify which flight was used because multiple flights of different airlines operate on the same flight route between the same departure airport and arrival airport.
In addition, on airline routes, the number of passengers on boarding aircraft is managed only by each airline, and the person trip survey itself is not conducted. Therefore, it is difficult to analyze the number of passengers of each boarding aircraft of different airlines.

そこで、本発明は、ユーザによって所持される携帯端末の通信ログを用いて、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンの学習モデルを構築するプログラム、学習装置及び方法を提供することを目的とする。 Therefore, the present invention provides a program, a learning device, and a method for constructing a learning model of a machine learning engine that estimates boarding information of a user's boarding machine by using a communication log of a mobile terminal possessed by the user. With the goal.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、搭乗機の搭乗情報を推定する機械学習エンジンとしてコンピュータを機能させるプログラムであって、
拠点毎に、各搭乗機の出発時刻及び搭乗情報が規定された出発時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積手段と、
通信ログ蓄積手段を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出するトリップログ抽出手段と
して機能させ、
機械学習エンジンは、教師データとして、出発拠点となる「先方通信ログ」と、当該先方通信ログの時刻が、出発時刻表の出発時刻の所定前時間範囲に含まれる「搭乗情報」とを対応付けて学習する
ようにコンピュータを機能させることを特徴とする。 According to the present invention, it is a program that inputs a communication log of a mobile terminal possessed by a user and causes a computer to function as a machine learning engine that estimates boarding information of a boarding machine.
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, the pair of the destination communication log and the backward communication log whose distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length is made to function as a trip log extraction means for extracting as a trip log.
The machine learning engine associates the "destination communication log", which is the departure base, with the "boarding information" in which the time of the destination communication log is included in the predetermined time range before the departure time in the departure timetable as teacher data. It is characterized by making the computer function so as to learn by learning.

本発明のプログラムにおける他の実施形態によれば、
拠点毎に、機械学習エンジンによって通信ログ及び搭乗情報を対応付けて既に学習済みとなった「基準拠点」と、機械学習エンジンに対して学習対象となる「非基準拠点」とに区分した際に、
トリップログ抽出手段は、非基準拠点が出発拠点となり且つ基準拠点が到着拠点となるトリップログを抽出する
ようにコンピュータを機能させることも好ましい。 According to other embodiments in the program of the present invention
When each base is divided into a "reference base" that has already been learned by associating communication logs and boarding information with the machine learning engine and a "non-standard base" that is the target of learning for the machine learning engine. ,
It is also preferable that the trip log extraction means has a computer function so as to extract trip logs in which a non-reference base serves as a departure base and a reference base serves as an arrival base.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンとしてコンピュータを機能させるプログラムであって、
拠点毎に、各搭乗機の到着時刻及び搭乗情報が規定された到着時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積手段と、
通信ログ蓄積手段を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出するトリップログ抽出手段と
して機能させ、
機械学習エンジンは、教師データとして、到着拠点となる「後方通信ログ」と、当該後方通信ログの時刻が、到着時刻表の到着時刻の所定後時間範囲に含まれる「搭乗情報」とを対応付けて学習する
ようにコンピュータを機能させることも好ましい。 According to the present invention, it is a program that inputs a communication log of a mobile terminal possessed by a user and causes a computer to function as a machine learning engine that estimates boarding information of the user's boarding machine.
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, the pair of the destination communication log and the backward communication log whose distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length is made to function as a trip log extraction means for extracting as a trip log.
The machine learning engine associates the "rear communication log", which is the arrival base, with the "boarding information" in which the time of the rear communication log is included in the predetermined time range after the arrival time of the arrival timetable as teacher data. It is also preferable to make the computer function so that it can learn by learning.

本発明のプログラムにおける他の実施形態によれば、
拠点毎に、機械学習エンジンによって通信ログ及び搭乗情報を対応付けて既に学習済みとなった「基準拠点」と、機械学習エンジンに対して学習対象となる「非基準拠点」とに区分した際に、
トリップログ抽出手段は、基準拠点が出発拠点となり且つ非基準拠点が到着拠点となるトリップログを抽出する
ようにコンピュータを機能させることも好ましい。 According to other embodiments in the program of the present invention
When each base is divided into a "reference base" that has already been learned by associating communication logs and boarding information with the machine learning engine and a "non-standard base" that is the target of learning for the machine learning engine. ,
It is also preferable that the trip log extraction means has a computer function so as to extract trip logs in which the reference base serves as the departure base and the non-reference base serves as the arrival base.

本発明のプログラムにおける他の実施形態によれば、
搭乗情報は、ゲート番号及び／又は搭乗機便名を含む
ようにコンピュータを機能させることも好ましい。 According to other embodiments in the program of the present invention
It is also preferred to have the computer function so that the boarding information includes the gate number and / or the flight number of the boarding aircraft.

本発明のプログラムにおける他の実施形態によれば、
通信ログにおける位置は、当該携帯端末の緯度経度、及び／又は、当該携帯端末が接続する基地局ＩＤ及び電波状態である
ようにコンピュータを機能させることも好ましい。 According to other embodiments in the program of the present invention
It is also preferable to make the computer function so that the position in the communication log is the latitude and longitude of the mobile terminal and / or the base station ID and radio wave condition to which the mobile terminal is connected.

本発明のプログラムにおける他の実施形態によれば、
非基準拠点は、基準拠点との間のトリップログを用いて搭乗情報と通信ログとを対応付けて学習することによって、新規に基準拠点となる
ようにコンピュータを機能させることも好ましい。 According to other embodiments in the program of the present invention
It is also preferable that the non-reference base makes the computer function so as to become a new reference base by learning the boarding information and the communication log in association with each other by using the trip log with the reference base.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンを有する学習装置であって、
拠点毎に、各搭乗機の出発時刻及び搭乗情報が規定された出発時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積手段と、
通信ログ蓄積手段を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出するトリップログ抽出手段と
を有し、
機械学習エンジンは、教師データとして、出発拠点となる「先方通信ログ」と、当該先方通信ログの時刻が、出発時刻表の出発時刻の所定前時間範囲に含まれる「搭乗情報」とを対応付けて学習する
ことを特徴とする。 According to the present invention, it is a learning device having a machine learning engine that inputs a communication log of a mobile terminal owned by a user and estimates boarding information of the user's boarding machine.
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
It has a trip log extraction means that extracts a set of a destination communication log and a backward communication log whose distance between positions of communication logs that are continuous in time series is equal to or longer than a predetermined length by using a communication log storage means as a trip log. ,
The machine learning engine associates the "destination communication log", which is the departure base, with the "boarding information" in which the time of the destination communication log is included in the predetermined time range before the departure time in the departure timetable as teacher data. It is characterized by learning.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンを有する学習装置であって、
拠点毎に、各搭乗機の到着時刻及び搭乗情報が規定された到着時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積手段と、
通信ログ蓄積手段を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出するトリップログ抽出手段と
を有し、
機械学習エンジンは、教師データとして、到着拠点となる「後方通信ログ」と、当該後方通信ログの時刻が、到着時刻表の到着時刻の所定後時間範囲に含まれる「搭乗情報」とを対応付けて学習する
ことを特徴とする。 According to the present invention, it is a learning device having a machine learning engine that inputs a communication log of a mobile terminal owned by a user and estimates boarding information of the user's boarding machine.
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
It has a trip log extraction means that extracts a set of a destination communication log and a backward communication log whose distance between positions of communication logs that are continuous in time series is equal to or longer than a predetermined length by using a communication log storage means as a trip log. ,
The machine learning engine associates the "rear communication log", which is the arrival base, with the "boarding information" in which the time of the rear communication log is included in the predetermined time range after the arrival time of the arrival timetable as teacher data. It is characterized by learning.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンを有する装置の学習方法であって、
装置は、
拠点毎に、各搭乗機の出発時刻及び搭乗情報が規定された出発時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積部と
を有し、
装置は、
通信ログ蓄積部を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出する第１のステップと、
機械学習エンジンに対して、教師データとして、出発拠点となる「先方通信ログ」と、当該先方通信ログの時刻が、出発時刻表の出発時刻の所定前時間範囲に含まれる「搭乗情報」とを対応付けて学習させる第２のステップと
を実行することを特徴とする。 According to the present invention, it is a learning method of a device having a machine learning engine that inputs a communication log of a mobile terminal possessed by a user and estimates boarding information of the user's boarding machine.
The device is
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Each base has a communication log storage unit that stores communication logs including the time and position of each mobile terminal.
The device is
The first step of extracting a set of the destination communication log and the backward communication log in which the distance between the positions of the communication logs continuous in chronological order is equal to or longer than a predetermined length by using the communication log storage unit as a trip log.
For the machine learning engine, as teacher data, the "destination communication log" that serves as the departure base and the "boarding information" in which the time of the other party communication log is included in the predetermined time range before the departure time of the departure timetable are provided. It is characterized in that the second step of associating and learning is executed.

本発明によれば、ユーザによって所持される携帯端末の通信ログを入力し、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンを有する装置の学習方法であって、
装置は、
拠点毎に、各搭乗機の到着時刻及び搭乗情報が規定された到着時刻表と、
拠点毎に、各携帯端末の時刻及び位置を含む通信ログを蓄積した通信ログ蓄積部と
を有し、
装置は、
通信ログ蓄積部を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる先方通信ログ及び後方通信ログの組を、トリップログとして抽出する第１のステップと、
機械学習エンジンに対して、教師データとして、到着拠点となる「後方通信ログ」と、当該後方通信ログの時刻が、到着時刻表の到着時刻の所定後時間範囲に含まれる「搭乗情報」とを対応付けて学習する第２のステップと
を実行することを特徴とする。 According to the present invention, it is a learning method of a device having a machine learning engine that inputs a communication log of a mobile terminal possessed by a user and estimates boarding information of the user's boarding machine.
The device is
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Each base has a communication log storage unit that stores communication logs including the time and position of each mobile terminal.
The device is
The first step of extracting a set of the destination communication log and the backward communication log in which the distance between the positions of the communication logs continuous in chronological order is equal to or longer than a predetermined length by using the communication log storage unit as a trip log.
For the machine learning engine, as teacher data, the "backward communication log" that serves as the arrival base and the "boarding information" in which the time of the rearward communication log is included in the predetermined time range after the arrival time of the arrival timetable are provided. It is characterized in that the second step of learning in association with each other is executed.

本発明のプログラム、装置及び方法によれば、ユーザによって所持される携帯端末の通信ログを用いて、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンの学習モデルを構築する。 According to the program, device, and method of the present invention, a learning model of a machine learning engine that estimates boarding information of a user's boarding machine is constructed by using a communication log of a mobile terminal possessed by the user.

本発明におけるシステム環境の構成図である。It is a block diagram of the system environment in this invention. 本発明における学習装置の機能構成図である。It is a functional block diagram of the learning apparatus in this invention. 通信ログ蓄積部及びトリップログ抽出部を表す説明図である。It is explanatory drawing which shows the communication log storage part and trip log extraction part. 出発拠点における機械学習エンジンへ入力する教師データを表す説明図である。It is explanatory drawing which shows the teacher data to input to the machine learning engine at the departure base. 到着拠点における機械学習エンジンへ入力する教師データを表す説明図である。It is explanatory drawing which shows the teacher data to input to the machine learning engine at the arrival base. 空港に基づく基準拠点及び非基準拠点を表す説明図である。It is explanatory drawing which shows the reference base and non-standard base based on an airport. 全国の基準空港と非基準空港とを表す説明図である。It is explanatory drawing which shows the standard airport and the non-standard airport of the whole country. 非基準拠点を出発拠点とし、基準空港を到着拠点とした搭乗機のトリップログを用いて、通信ログと搭乗情報とを対応付けて学習する説明図である。It is explanatory drawing which learns by associating a communication log with a boarding information by using a trip log of a boarding machine which has a non-reference base as a departure base and a reference airport as an arrival base. 基準拠点を出発拠点とし、非基準空港を到着拠点とした搭乗機のトリップログを用いて、通信ログと搭乗情報とを対応付けて学習する説明図である。It is explanatory drawing which learns by associating a communication log and a boarding information with a trip log of a boarding machine which has a reference base as a departure base and a non-standard airport as an arrival base.

以下、本発明の実施の形態について、図面を用いて詳細に説明する。 Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

図１は、本発明におけるシステム環境の構成図である。 FIG. 1 is a configuration diagram of a system environment according to the present invention.

本発明の主な実施形態として、ユーザが、空港（拠点）で航空機（搭乗機）に搭乗することを想定している。
図１によれば、空港内では、ユーザが所持する携帯端末２は、通信事業設備として設置された基地局（又はアクセスポイント）と常時通信可能となっている。通信事業設備は、携帯端末２から定期的に「通信ログ」を収集している。 As a main embodiment of the present invention, it is assumed that a user board an aircraft (boarding machine) at an airport (base).
According to FIG. 1, in the airport, the mobile terminal 2 owned by the user can always communicate with the base station (or access point) installed as the communication business equipment. The telecommunications business equipment regularly collects "communication logs" from the mobile terminal 2.

通信ログには、少なくとも「時刻」及び「位置」が記録されている。
「位置」とは、当該携帯端末の緯度経度、及び／又は、当該携帯端末が接続する基地局ＩＤ及び電波状態であってもよい。
「緯度経度」とは、例えばＧＰＳ(Global Positioning System)測位であってもよいし、複数の基地局との通信に基づく基地局測位であってもよい。
また、緯度経度のような地理座標に拘わらず、単に、携帯端末２が接続した１つ以上の基地局の「基地局ＩＤ及び電波状態」であってもよい。基地局ＩＤや電波状態は、位置に基づいて異なるものとなるためである。
ここで、電波状態とは、受信信号強度（ＲＳＳＩ(Received Signal Strength Indicator)）や信号対干渉雑音比（ＳＩＮＲ(Signal to Interference Noise Ratio)、ＲＴＴ(Round Trip Time)のような種々の無線測定項目であってもよい。 At least "time" and "position" are recorded in the communication log.
The "position" may be the latitude and longitude of the mobile terminal and / or the base station ID and radio wave condition to which the mobile terminal is connected.
The "latitude / longitude" may be, for example, GPS (Global Positioning System) positioning, or base station positioning based on communication with a plurality of base stations.
Further, regardless of the geographic coordinates such as latitude and longitude, it may simply be the "base station ID and radio wave condition" of one or more base stations to which the mobile terminal 2 is connected. This is because the base station ID and the radio wave condition are different depending on the position.
Here, the radio wave condition refers to various wireless measurement items such as received signal strength (RSSI (Received Signal Strength Indicator)), signal-to-interference noise ratio (SINR), and RTT (Round Trip Time). May be.

図１のように、その空港間では、基地局と通信不可能であるために、移動中の通信ログを取得することはできない。例えば、携帯端末２の通信ログは、出発拠点となる空港ａの搭乗ゲート付近で発生した後、一定時間、通信ログが発生しない状態が継続する。その後、遠方の到着拠点となる空港ｂの到着ゲート付近で、突然、次の通信ログが発生する。
また、同一の出発空港ａと到着空港ｂとの間では、同一の航空路線に、異なる航空会社の複数の航空便が運航する場合もある。 As shown in FIG. 1, since it is impossible to communicate with the base station between the airports, it is not possible to acquire the communication log during movement. For example, the communication log of the mobile terminal 2 is generated near the boarding gate of the airport a, which is the departure base, and then the communication log is not generated for a certain period of time. After that, the next communication log suddenly occurs near the arrival gate of airport b, which is a distant arrival base.
Further, between the same departure airport a and arrival airport b, a plurality of flights of different airlines may operate on the same air route.

各空港では、「出発時刻表」が予め規定されていると共に、航空便のフライト履歴として「到着時刻表」も公開されている。 At each airport, a "departure timetable" is prescribed in advance, and an "arrival timetable" is also published as a flight history of flights.

出発時刻表は、拠点（空港）毎に、各搭乗機の出発時刻及び搭乗情報が規定されたものである。
図１によれば、例えば以下のような「出発時刻表」が規定されている。
空港ａ：［時刻］［航空便名（到着先）］［ゲート番号］
空港ｂでは、その「時刻」に、その「航空便名（到着先）」の搭乗機が、その「ゲート番号」のゲートに接続されたことを表す。 The departure timetable defines the departure time and boarding information of each boarding aircraft for each base (airport).
According to FIG. 1, for example, the following "departure timetable" is defined.
Airport a: [Time] [Flight number (destination)] [Gate number]
At airport b, it means that the boarding machine of the "airline flight number (destination)" was connected to the gate of the "gate number" at the "time".

図２は、本発明における学習装置の機能構成図である。 FIG. 2 is a functional configuration diagram of the learning device according to the present invention.

図２によれば、本発明の学習装置１は、＜学習段階＞と＜推定段階＞とに区分される。学習段階として、通信ログ蓄積部１０１と、出発時刻表１０２と、到着時刻表１０３と、トリップログ抽出部１１と、機械学習エンジン１２とを有する。これら機能構成部は、装置に搭載されたコンピュータを機能させるプログラムを実行することによって実現される。また、これら機能構成部の処理の流れは、装置の推定方法としても理解できる。 According to FIG. 2, the learning device 1 of the present invention is divided into a <learning stage> and a <estimation stage>. As a learning stage, it has a communication log storage unit 101, a departure timetable 102, an arrival timetable 103, a trip log extraction unit 11, and a machine learning engine 12. These functional components are realized by executing a program that makes the computer mounted on the device function. Further, the processing flow of these functional components can be understood as an estimation method of the device.

＜学習段階＞
図３は、通信ログ蓄積部及びトリップログ抽出部を表す説明図である。 <Learning stage>
FIG. 3 is an explanatory diagram showing a communication log storage unit and a trip log extraction unit.

［通信ログ蓄積部１０１］
通信ログ蓄積部１０１は、拠点（空港）毎に、各携帯端末の時刻及び位置を含む「通信ログ」を蓄積したものである。
図３によれば、空港毎に、端末ＩＤ(IDentifier)に対して、時刻及び位置が対応付けて記録されている。 [Communication log storage unit 101]
The communication log storage unit 101 stores a "communication log" including the time and position of each mobile terminal for each base (airport).
According to FIG. 3, the time and position are recorded in association with the terminal ID (IDentifier) for each airport.

［トリップログ抽出部１１］
トリップログ抽出部１１は、通信ログ蓄積部１０１を用いて、時系列に連続する通信ログの位置間の距離が所定長以上となる「先方通信ログ」及び「後方通信ログ」の組を、トリップログとして抽出する。
「位置間の距離」とは、航空機を利用したであろうと推定される所定長の距離（例えば200km以上）である。携帯端末２の通信ログが、そのような長距離の間も途切れるということは、航空機を利用していると想定することができる。
特に、ユーザは、航空機に搭乗する際に、出発ゲート付近で携帯端末２の電源をOFFにしたり、フライトモードに変更したりする場合が多い。また、航空機から降機した際に、到着ゲート付近で、携帯端末２の電源をONにしたり、通常モードに変更したりする場合が多い。 [Trip log extraction unit 11]
The trip log extraction unit 11 uses the communication log storage unit 101 to trip a set of a “destination communication log” and a “backward communication log” in which the distance between the positions of the communication logs that are continuous in time series is equal to or longer than a predetermined length. Extract as a log.
The "distance between positions" is a predetermined length of distance (for example, 200 km or more) that is estimated to have used an aircraft. The fact that the communication log of the mobile terminal 2 is interrupted even during such a long distance can be assumed to be using an aircraft.
In particular, when boarding an aircraft, the user often turns off the power of the mobile terminal 2 or changes to the flight mode near the departure gate. Further, when the aircraft disembarks, the power of the mobile terminal 2 is often turned on or changed to the normal mode near the arrival gate.

［機械学習エンジン１２］
機械学習エンジン１２は、教師あり学習(Supervised learning)に基づくものである。教師データｘ及びｙ（二値分類問題）を対応付けて入力し、ｘ及びｙのデータ間の写像関数を近似的に学習する。これによって、未知データｘを入力することよって予測データｙを得ることができる。具体的には、サポートベクタマシンやニューラルネットワークなどを用いることができる。 [Machine learning engine 12]
The machine learning engine 12 is based on supervised learning. The teacher data x and y (binary classification problem) are input in association with each other, and the mapping function between the x and y data is approximately learned. Thereby, the prediction data y can be obtained by inputting the unknown data x. Specifically, a support vector machine, a neural network, or the like can be used.

機械学習エンジン１２は、教師データとして、出発拠点と到着拠点と別々に、通信ログと搭乗情報とを対応付けて学習する。 The machine learning engine 12 learns the departure base and the arrival base separately by associating the communication log with the boarding information as teacher data.

図４は、出発拠点における機械学習エンジンへ入力する教師データを表す説明図である。 FIG. 4 is an explanatory diagram showing teacher data input to the machine learning engine at the departure base.

（出発拠点）「先方通信ログ」と、当該先方通信ログの時刻が、出発時刻表の出発時刻の所定前時間範囲に含まれる搭乗機の「搭乗情報」とを対応付けて学習する。
「所定前時間範囲」とは、出発時刻の前の一定時間範囲であって、その範囲に、先方通信ログの時刻が含まれているとする。例えば出発時刻10:00である場合、所定前時間範囲を30分間とすれば、9:30〜10:00の時間範囲に、先方通信ログの時刻が含まれていることを要する。 (Departure base) The "destination communication log" and the "boarding information" of the boarding machine whose time in the destination communication log is included in the predetermined time range before the departure time in the departure timetable are learned in association with each other.
The "predetermined time range" is a fixed time range before the departure time, and it is assumed that the range includes the time of the communication log of the other party. For example, when the departure time is 10:00, if the predetermined time range is 30 minutes, it is necessary that the time of the other party's communication log is included in the time range from 9:30 to 10:00.

機械学習エンジン１２は、位置のみ入力して学習するものであってもよいし、時刻及び位置を入力して学習するものであってもよい。具体的には、以下のいずれかの情報を対応付けて入力して学習する。
［先方通信ログ］ <-> ［搭乗情報］
位置 <-> ゲート番号
位置 <-> 航空便名（到着先）
時刻及び位置 <-> 出発時刻及びゲート番号
時刻及び位置 <-> 出発時刻及び航空便名（到着先）
先方通信ログの時刻及び位置は、搭乗機へ搭乗する前に通過するゲート付近のものである。搭乗するゲート番号が特定できれば、出発時刻表から、航空便名（到着先）も特定することができる。 The machine learning engine 12 may be one that learns by inputting only a position, or may be one that learns by inputting a time and a position. Specifically, learning is performed by inputting any of the following information in association with each other.
[Recipient communication log] <-> [Boarding information]
Position <-> Gate number
Location <-> Flight number (destination)
Time and position <-> Departure time and gate number
Time and location <-> Departure time and flight number (arrival destination)
The time and position of the communication log of the other party is near the gate that passes before boarding the boarding aircraft. If the boarding gate number can be specified, the flight number (arrival destination) can also be specified from the departure timetable.

図５は、到着拠点における機械学習エンジンへ入力する教師データを表す説明図である。 FIG. 5 is an explanatory diagram showing teacher data input to the machine learning engine at the arrival base.

（到着拠点）「後方通信ログ」と、当該後方通信ログの時刻が、到着時刻表の到着時刻の所定後時間範囲に含まれる搭乗機の「搭乗情報」とを対応付けて学習する。
「所定後時間範囲」とは、到着時刻の後の一定時間範囲であって、その範囲に、後方通信ログの時刻が含まれているとする。例えば到着時刻10:00である場合、所定後時間範囲を30分間とすれば、10:00〜10:30の時間範囲に、後方通信ログの時刻が含まれていることを要する。 (Arrival base) The "rear communication log" and the "boarding information" of the boarding machine whose time in the rear communication log is included in the predetermined time range after the arrival time in the arrival timetable are learned in association with each other.
The "predetermined time range" is a fixed time range after the arrival time, and it is assumed that the range includes the time of the backward communication log. For example, when the arrival time is 10:00, if the predetermined time range is 30 minutes, it is necessary that the time of the backward communication log is included in the time range of 10:00 to 10:30.

この場合も、機械学習エンジン１２は、位置のみを学習するものであってもよいし、時刻及び位置を含めて学習するものであってもよい。具体的には、以下のいずれかの情報を対応付けて入力して学習する。
［後方通信ログ］ <-> ［搭乗情報］
位置 <-> ゲート番号
位置 <-> 航空便名（出発元）
時刻及び位置 <-> 到着時刻及びゲート番号
時刻及び位置 <-> 到着時刻及び航空便名（出発元）
後方通信ログの時刻及び位置は、搭乗機から降機した後に通過するゲート付近のものである。降機したゲート番号が特定できれば、到着時刻表から、航空便名（出発元）も特定することができる。 In this case as well, the machine learning engine 12 may learn only the position, or may learn including the time and the position. Specifically, learning is performed by inputting any of the following information in association with each other.
[Backward communication log] <-> [Boarding information]
Position <-> Gate number
Location <-> Flight number (departure source)
Time and location <-> Arrival time and gate number
Time and location <-> Arrival time and flight number (departure source)
The time and position of the rear communication log are near the gate that passes after getting off the boarding aircraft. If the gate number at which the aircraft disembarked can be specified, the flight number (departure source) can also be specified from the arrival timetable.

図６は、空港に基づく基準拠点及び非基準拠点を表す説明図である。 FIG. 6 is an explanatory diagram showing a reference base and a non-reference base based on the airport.

図６によれば、拠点毎に、以下のように区分されている。
［基準拠点］：機械学習エンジン１２によって通信ログ及び搭乗情報を対応付け
て既に学習済みとなった拠点（図６（ａ）参照）
［非基準拠点］：機械学習エンジン１２に対して学習対象となる拠点（図６（ｂ）
参照） According to FIG. 6, each base is classified as follows.
[Reference base]: Communication log and boarding information are associated with each other by the machine learning engine 12.
Bases that have already been learned (see Fig. 6 (a))
[Non-reference base]: A base to be learned for the machine learning engine 12 (FIG. 6 (b))
reference)

図７は、全国の基準空港と非基準空港とを表す説明図である。 FIG. 7 is an explanatory diagram showing standard airports and non-standard airports nationwide.

図７によれば、例えば羽田空港の場合、ゲートの位置が、航空会社によって十分に離れており、ゲート毎に通信ログの測位結果が全く異なっているために、通信ログとゲートとを対応付けて学習させることができる。これによって、羽田空港を「基準空港」とすることができる。
また、地方空港は、通信ログとゲートとを対応付けて学習していない場合、「非基準空港」となる。地方空港は、羽田空港との間で航空機の発着便を用いて、基準空港となるように学習していく。 According to FIG. 7, for example, in the case of Haneda Airport, the gate positions are sufficiently separated depending on the airline, and the positioning result of the communication log is completely different for each gate. Therefore, the communication log and the gate are associated with each other. Can be learned. As a result, Haneda Airport can be designated as a "reference airport."
In addition, a regional airport is a "non-standard airport" if the communication log and the gate are not learned in association with each other. Regional airports will learn to become standard airports by using flights to and from Haneda Airport.

本発明によれば、第１の非基準空港は、第１の基準拠点との間のトリップログを用いて搭乗情報と通信ログとを対応付けて学習することによって、新規に第２の基準拠点となる。
これができると、第２の非基準拠点は、第２の基準拠点との間のトリップログを用いて搭乗情報と通信ログとを対応付けて学習することによって、新規に第３の基準拠点となる。
このように、最初の基準空港から順に、芋づる式に、非基準空港が基準空港として学習されていく。 According to the present invention, the first non-reference airport newly learns the boarding information and the communication log in association with each other by using the trip log between the first reference base and the second reference base. It becomes.
If this is possible, the second non-reference base will become a new third reference base by learning the boarding information and the communication log in association with each other using the trip log between the second reference base and the second reference base. ..
In this way, from the first reference airport, the non-standard airport is learned as the reference airport in the Imozuru formula.

図８は、非基準拠点を出発拠点とし、基準空港を到着拠点とした搭乗機のトリップログを用いて、通信ログと搭乗情報とを対応付けて学習する説明図である。 FIG. 8 is an explanatory diagram for learning by associating a communication log with boarding information using a trip log of a boarding machine having a non-reference base as a departure base and a reference airport as an arrival base.

拠点毎に、基準拠点又は非基準拠点が明確であるとする。
（Ｓ１）基準空港を到着拠点とし、非基準空港を出発拠点とするトリップログを抽出する。
（Ｓ２）非基準空港の出発時刻表によって、非基準空港の搭乗情報（出発ゲート番号、航空便名（到着空港））を特定する。
（Ｓ３）非基準空港の先方通信ログと搭乗情報とを対応付けて学習する。 It is assumed that the reference base or the non-standard base is clear for each base.
(S1) Extract trip logs with the reference airport as the arrival base and the non-standard airport as the departure base.
(S2) The boarding information (departure gate number, flight number (arrival airport)) of the non-standard airport is specified by the departure timetable of the non-standard airport.
(S3) Learn by associating the destination communication log of the non-standard airport with the boarding information.

図９は、基準拠点を出発拠点とし、非基準空港を到着拠点とした搭乗機のトリップログを用いて、通信ログと搭乗情報とを対応付けて学習する説明図である。 FIG. 9 is an explanatory diagram for learning by associating a communication log with boarding information using a trip log of a boarding aircraft whose departure base is a reference base and whose arrival base is a non-standard airport.

（Ｓ１）基準空港を出発拠点とし、非基準空港を到着拠点とするトリップログを抽出する。
（Ｓ２）非基準空港の到着履歴表によって、非基準空港の搭乗情報（到着ゲート番号、航空便名（出発空港））を特定する。
（Ｓ３）非基準空港の後方通信ログと搭乗情報とを対応付けて学習する。 (S1) Extract trip logs with the reference airport as the departure base and the non-standard airport as the arrival base.
(S2) The boarding information (arrival gate number, flight number (departure airport)) of the non-standard airport is specified by the arrival history table of the non-standard airport.
(S3) Learn by associating the rear communication log of the non-standard airport with the boarding information.

＜推定段階＞
機械学習エンジン１２は、ユーザによって所持される携帯端末の「通信ログ」を入力し、当該ユーザの搭乗機の「搭乗情報」（ゲート番号や航空便名）を推定することができる。 <Estimation stage>
The machine learning engine 12 can input the "communication log" of the mobile terminal possessed by the user and estimate the "boarding information" (gate number and flight number) of the user's boarding machine.

以上、詳細に説明したように、本発明のプログラム、装置及び方法によれば、ユーザによって所持される携帯端末の通信ログを用いて、当該ユーザの搭乗機の搭乗情報を推定する機械学習エンジンの学習モデルを構築することができる。 As described in detail above, according to the program, device and method of the present invention, the machine learning engine that estimates the boarding information of the user's boarding machine using the communication log of the mobile terminal owned by the user. You can build a learning model.

前述した本発明の種々の実施形態について、本発明の技術思想及び見地の範囲の種々の変更、修正及び省略は、当業者によれば容易に行うことができる。前述の説明はあくまで例であって、何ら制約しようとするものではない。本発明は、特許請求の範囲及びその均等物として限定するものにのみ制約される。 With respect to the various embodiments of the present invention described above, various changes, modifications and omissions within the scope of the technical idea and viewpoint of the present invention can be easily made by those skilled in the art. The above explanation is just an example and does not attempt to restrict anything. The present invention is limited only to the scope of claims and their equivalents.

１学習装置
１０１通信ログ蓄積部
１０２出発時刻表
１０３到着時刻表
１１トリップログ抽出部
１２機械学習エンジン
２携帯端末 1 Learning device 101 Communication log storage unit 102 Departure timetable 103 Arrival timetable 11 Trip log extraction unit 12 Machine learning engine 2 Mobile terminal

Claims

A program that allows a computer to function as a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, the pair of the destination communication log and the backward communication log whose distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length is made to function as a trip log extraction means for extracting as a trip log. ,
The machine learning engine corresponds to the "destination communication log" which is the departure base and the "boarding information" in which the time of the other party communication log is included in the predetermined time range before the departure time of the departure timetable as teacher data. A program characterized by making a computer function so that it can be attached and learned.

Each of the bases is divided into a "reference base" that has already been learned by associating communication logs and boarding information with the machine learning engine and a "non-reference base" that is a learning target for the machine learning engine. When you do
The program according to claim 1, wherein the trip log extracting means causes a computer to function so as to extract a trip log in which a non-reference base serves as a departure base and a reference base serves as an arrival base.

A program that allows a computer to function as a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, the pair of the destination communication log and the backward communication log whose distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length is made to function as a trip log extraction means for extracting as a trip log. ,
The machine learning engine corresponds to the "rear communication log" which is the arrival base and the "boarding information" in which the time of the rear communication log is included in the predetermined time range after the arrival time of the arrival timetable as teacher data. A program characterized by making a computer function so that it can be attached and learned.

Each of the bases is divided into a "reference base" that has already been learned by associating communication logs and boarding information with the machine learning engine and a "non-reference base" that is a learning target for the machine learning engine. When you do
The program according to claim 3, wherein the trip log extracting means causes a computer to function so as to extract a trip log in which a reference base serves as a departure base and a non-reference base serves as an arrival base.

The program according to any one of claims 1 to 4, wherein the boarding information includes a gate number and / or a flight number of a boarding machine to function a computer.

The position in the communication log is the latitude and longitude of the mobile terminal, and / or the base station ID to which the mobile terminal is connected and the radio wave state. The program described in any one of the items.

A claim characterized in that a non-reference base makes a computer function so as to become a new reference base by learning by associating boarding information with a communication log using a trip log to and from the reference base. The program according to any one of 1 to 6.

A learning device having a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, there is a trip log extraction means for extracting a set of a destination communication log and a backward communication log in which the distance between the positions of communication logs continuous in time series is equal to or longer than a predetermined length as a trip log. And
The machine learning engine corresponds to the "destination communication log" which is the departure base and the "boarding information" in which the time of the other party communication log is included in the predetermined time range before the departure time of the departure timetable as teacher data. A learning device characterized by attaching and learning.

A learning device having a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Communication log storage means that stores communication logs including the time and position of each mobile terminal for each base,
Using the communication log storage means, there is a trip log extraction means for extracting a set of a destination communication log and a backward communication log in which the distance between the positions of communication logs continuous in time series is equal to or longer than a predetermined length as a trip log. And
The machine learning engine corresponds to the "rear communication log" which is the arrival base and the "boarding information" in which the time of the rear communication log is included in the predetermined time range after the arrival time of the arrival timetable as teacher data. A learning device characterized by attaching and learning.

It is a learning method of a device having a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
The device is
Departure timetables that specify the departure time and boarding information of each boarding aircraft for each base,
Each base has a communication log storage unit that stores communication logs including the time and position of each mobile terminal.
The device is
Using the communication log storage unit, the first step of extracting as a trip log a set of a destination communication log and a backward communication log in which the distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length,
For the machine learning engine, as teacher data, the "destination communication log" which is the departure base and the "boarding information" in which the time of the other party communication log is included in the predetermined time range before the departure time of the departure timetable. A learning method of an apparatus, which comprises executing a second step of associating and learning.

It is a learning method of a device having a machine learning engine that inputs the communication log of a mobile terminal owned by a user and estimates the boarding information of the user's boarding machine.
The device is
An arrival timetable that defines the arrival time and boarding information of each boarding aircraft for each base,
Each base has a communication log storage unit that stores communication logs including the time and position of each mobile terminal.
The device is
Using the communication log storage unit, the first step of extracting as a trip log a set of a destination communication log and a backward communication log in which the distance between the positions of the communication logs continuous in time series is equal to or longer than a predetermined length,
For the machine learning engine, as teacher data, "rear communication log" which is an arrival base and "boarding information" in which the time of the rear communication log is included in a predetermined time range after the arrival time of the arrival timetable. A learning method of an apparatus, which comprises executing a second step of learning in association with each other.