KR102436134B1 - Trip-planning method considering weather condition and stay-time - Google Patents

Abstract

The present invention relates to a travel itinerary planning method taking into account weather conditions and stay time.
In addition to information such as travel blogs, travel-related websites and communities, the present invention provides information on weather conditions and travel time, user characteristic information (gender, age group, occupation, etc.) It is about the methodology of using artificial intelligence to plan travel itineraries. That is, in the present invention, the AI system can plan a travel schedule that maximizes user satisfaction for places of interest while satisfying the travel schedule planning constraints (residence time, expected precipitation, temperature and humidity index, etc.).

Description

How to plan a travel itinerary taking into account weather conditions and length of stay

The present invention relates to a method of providing travel planning information to a user.

Before visiting a destination, travelers determine points of interest and determine the order in which they visit the points of interest. The problem of determining the places of interest and the order of visits is called the Trip-Planning Problem. Conventionally, as a way to solve the travel itinerary planning problem, research on solving the problem has been actively conducted after defining it as the Orienteering Problem.

The orienteering problem is a routing problem that determines the path that maximizes the score without exceeding a specific budget from the starting point to the ending point in the graph structure. The orienteering problem-solving method for the itinerary planning problem showed a problem in effectiveness because it did not consider the time it takes to stay or visit the place of interest. Moreover, although one of the biggest concerns for travelers is weather information, systematic consideration for it was lacking.

The inventors of the present invention have completed the present invention after long research and efforts to solve these problems.

When a traveler plans a travel itinerary and actually travels, satisfaction with a place of interest is greatly influenced by time and weather factors. For example, in rainy weather, people are more satisfied with indoor activities than outdoor activities. It's a good idea to plan to visit places with good night views after sunset. The time component and the weather component are dynamic information. An object of the present invention in consideration of such dynamic information is to consider a time element and a weather element in generating and providing a travel itinerary planning method. The former relates to residence time, the latter to weather conditions. It also comprehensively considers user satisfaction with stay time and weather conditions.

On the other hand, other objects not specified in the present invention will be additionally considered within the range that can be easily inferred from the following detailed description and effects thereof.

As a travel itinerary planning method in consideration of weather conditions and stay time of the present invention for achieving the above object:

A traveler terminal accessing the travel schedule planning artificial intelligence server and requesting a travel schedule planning data generation service by inputting information about the travel date and place of interest;

The step of the travel schedule planning artificial intelligence server executing a travel schedule generator;

The travel itinerary generator collects data on user characteristics and travel location characteristics from a multi-layered artificial neural network, and plans travel schedules using user satisfaction with temporal constraints on stay time and weather constraints as random variables. It characterized in that it comprises the step of generating data and providing it to the traveler terminal.

In addition, in the travel itinerary planning method in consideration of weather conditions and stay time according to an embodiment of the present invention, the weather constraint is that x _i is the place of interest to be visited i-th, and the stay time t _i is the interest. When representing the time of staying at place x _i , it includes the expected precipitation (Rain _i ), temperature and humidity index (THI _i ), and wind speed (Wind _i ). to plan,

(One)

(2)

(3)

In addition, in the travel itinerary planning method in consideration of weather conditions and residence time according to an embodiment of the present invention, the user satisfaction is calculated using neural collaboration filtering using the multi-layered artificial neural network, wherein the input layer is the user characteristic It may be divided into an input layer and a place of interest input layer.

In addition, in the travel itinerary planning method in consideration of weather conditions and residence time according to an embodiment of the present invention, it is preferable to optimize the user satisfaction calculated through the neural collaborative filtering using a priority-based genetic algorithm.

According to the present invention, there is an advantage in that the travel itinerary planning problem that the user is worried about can be solved through an artificial neural network. The system can present a travel itinerary that takes into account dynamic information such as stay time, weather information, travel purpose, individual age, and places of interest of the user, and furthermore, it can provide information on hot places that people frequently seek during the travel itinerary. have.

On the other hand, even if it is an effect not explicitly mentioned herein, it is added that the effects described in the following specification expected by the technical features of the present invention and their potential effects are treated as described in the specification of the present invention.

1 shows a system configuration according to a preferred embodiment of the present invention.
2 schematically shows the overall process of a method for generating travel itinerary planning data according to a preferred embodiment of the present invention.
3 is a model of a method for obtaining user satisfaction using neural collaboration filtering according to a preferred embodiment of the present invention.
4 shows a method of applying the POI node and priority using the genetic algorithm of the present invention.
※ It is revealed that the accompanying drawings are exemplified as a reference for understanding the technical idea of the present invention, and the scope of the present invention is not limited thereby.

In the first aspect of the present invention, temporal uncertainty is modeled using continuous random variables, and constraints related to weather and residence time are determined.

In the second aspect of the present invention, in order to infer user satisfaction for a place of interest, a multi-layered artificial neural network will be used to model using neural collaborative filtering.

In the third aspect of the present invention, optimization based on a genetic algorithm will be attempted to find a travel itinerary that satisfies the constraint conditions and maximizes user satisfaction with the place of interest.

Hereinafter, the configuration of the present invention guided by various embodiments of the present invention and effects resulting from the configuration will be described with reference to the drawings. In the description of the present invention, if it is determined that the subject matter of the present invention may be unnecessarily obscured as it is obvious to those skilled in the art with respect to related known functions, the detailed description thereof will be omitted.

The method of the present invention for generating and providing an optimized travel itinerary using user satisfaction may be provided through a schematic system configuration example of FIG. 1 .

In a preferred embodiment of the present invention, the traveler terminal 1 accessing the travel itinerary planning artificial intelligence server 100 using the Internet and requesting the travel itinerary planning data generation service is a desktop computer terminal, a notebook computer, or a smart phone. It may be a mobile device such as a tablet PC or the like. In another preferred embodiment, the traveler terminal 1 may be a server computer that manages or passes through another network.

The travel itinerary planning artificial intelligence server 100 receives the request of the traveler terminal 1 and generates travel itinerary planning data. This travel itinerary planning artificial intelligence server 100 includes one or more server devices, and includes a plurality of hardware/software equipment. In particular, in the present invention, the travel schedule planning artificial intelligence server 100 includes a travel schedule generator 110 that generates travel schedule planning data using an artificial intelligence algorithm, which will be described later.

The travel schedule planning artificial intelligence server 100 builds the travel information learning database 120 . Information such as travel destination information, user satisfaction related to travel destination information, and travel schedule may be learned and stored through the travel information learning database 120 . Although not shown, other information related to travel itinerary planning data and a database related to user information may be further included.

The travel schedule planning artificial intelligence server 100 may collect weather information by accessing the external server 200 providing weather information through a communication network and the weather database 210 . Weather includes data such as temperature and humidity, precipitation, and wind speed.

In addition, the travel schedule planning artificial intelligence server 100 may manage the multi-layered artificial neural network 300 through the Internet. The multi-layered artificial neural network 300 collects data on various events and user characteristics generated by users in a network configuration connected to users in the network. The user characteristics may include the users' gender, age group, number of travelers, occupation, and the like. In addition, data such as season, day of the week, and time of visit are collected. In addition, data such as categories, activities, ratings, indoor/outdoor, etc. are collected as characteristics of travel places.

2 schematically shows a process for generating travel itinerary planning data according to a preferred embodiment of the present invention.

The traveler terminal requests a travel itinerary planning data generation service to the server by inputting information about the travel itinerary including information on the date and place (S100). Then, the travel schedule planning artificial intelligence server starts the process of generating the travel schedule planning data in consideration of the weather and the residence time by executing the travel schedule generator (S110).

First, constraints related to the residence time and weather are verified (S120). This verification process uses the formulas predefined in the itinerary generator.

The travel planning problem defines user satisfaction while satisfying constraints. This can be defined as the problem of finding an itinerary that maximizes user satisfaction. If the order of visits is determined by selecting m ( m is an integer greater than or equal to 1) from among n ( n is an integer greater than or equal to 1) candidates of interest, the journey I is <x ₀ , x ₁ , ... , x _m >. where x _i is the i-th place of interest to visit. Then, when the user stays at each point of interest x _i for the duration t _i , the user's satisfaction is defined as s _i (x _i , t _i ) .

Therefore, the problem of finding the journey I that maximizes satisfaction in all visited places can be defined as the following Equation 1.

(식 1)

(Equation 1)

Now, let's verify the constraint conditions in earnest according to the present invention. When solving the travel planning problem, first of all, certain constraints must be satisfied. In the present invention, two kinds of constraint conditions are verified. First, there is a temporal constraint on the place of interest. In these constraints, the arrival time (A _i ), the departure time from the point of interest (D _i ), and the return time (B) after the trip are verified. Next is the constraint regarding the weather. As these constraints, the expected precipitation (Rain _i ), temperature and humidity index (THI _i ), and wind speed (Wind _i ) can be used as constraints. In addition, time-related constraints A _i , D _i , and B are defined as continuous probabilities, and a constraint that a place of interest can only be visited once can be added.

In the embodiment of the present invention, the time-related constraint is defined as the continuous probability of Equations 2 and 3 as follows.

(식2)

(Formula 2)

(식 3)

(Equation 3)

That is, it is possible for a user to visit a place of interest only within a specific time interval [ Open _{i ,} Close _i ]. The time T _m at the end of the trip must be within the time limit B. Therefore, the continuous probability of T _m can be expressed as Equation 4.

(식 4)

(Equation 4)

In Equation 4, ε is a variable representing the reliability of the constraint. If this variable must satisfy the constraint with 95% confidence, ε = 0.95.

In an embodiment of the present invention, the constraints on the weather are in the form of a function having the determining variables of the place of interest and the residence time, respectively, as follows for the expected precipitation (Rain _i ), the temperature and humidity index (THI _i ), and the wind speed (Wind _i ). It is defined by Equation 5 to Equation 7.

(식 5)

(Equation 5)

(식 6)

(Equation 6)

(식 7)

(Equation 7)

It is defined as visiting the place of interest only when the conditions defined by Equations 5, 6, and 7 are satisfied at the residence time t _i staying at the place of interest x _i . That is, Rain _i ^max , THI _i ^max , and Wind _i ^max represent the values of the weather constraint at the POI xi.

After verification of the time constraint and weather constraint as above is completed, the user's satisfaction s _i (x _i , t _i ) is calculated when the user stays at the place of interest x _i for the dwell time t _i ( S130).

In a preferred embodiment of the present invention, random variables related to residence time are modeled as conditional Gaussian Distributions. The residence time is calculated as in Equation 8.

(식 8)

(Equation 8)

In a preferred embodiment of the present invention, a Neural Collaborative Filtering technique was used to calculate the user's satisfaction s _i (x _i , t _i ). Neural collaborative filtering is a known technique for expressing the correlation between a user vector and a target vector. In particular, the multi-layered artificial neural network 300 of FIG. 1 is used.

The neural collaborative filtering algorithm of the present invention may be diagrammed as shown in FIG. 3 . As shown, between the input layers 310 and 311 and the output layer 340, the hidden layer is composed of four layers including the embedding layers 320 and 321 that make the data input from the input layer into a vector. can

The input layer is divided into a user characteristic input layer 310 and a place of interest input layer 311 .

For the user characteristic input layer 310, gender, age group, number of people traveling, occupation, season, and day of the week were used. Factors related to user characteristics were divided into gender, age group, number of people traveling, occupation, etc., and factors related to travel days and times were classified by season, day of the week, and time of visit, and used as an input layer.

In addition, the place of interest input layer 311 inputs categories, activities that can be performed in a place, a place rating in a social network service, and indoor/outdoor divisions.

The user characteristic input layer 310 is generated as an embedding vector that is a user-time latent variable in the embedding layer 320 . The POI input layer 311 is generated as an embedding vector that is a POI latent variable in the embedding layer 321 .

Then, the user satisfaction of the output layer 340 is calculated through sequential layers of the neural collaborative filtering layer 330 composed of the layers of the pre-designed artificial neural network. After learning a neural network using user preference data defined for each situation such as places of interest and weather input in the input layer, input data such as users, weather, and places of interest to the learned network, a multi-layered neural collaboration filtering layer User satisfaction is calculated and output through the network of 330 .

Next, the travel schedule is optimized using a genetic algorithm using the optimized user satisfaction (S140). This genetic algorithm is known as a method of finding an approximate optimal solution within a fixed time. It is a representative technique of evolutionary computation that mimics the evolution of living things.

In such a genetic algorithm, the decision variables of an optimization problem are expressed in an arrangement called a chromosome. And it is evaluated through a fitness function. Depending on the result, the process of changing the chromosome using a genetic operator is repeated. As the number of iterations increases, the chromosome changes in the direction of increasing the fit to the fit function. And in order to express the decision variable as a chromosome, a priority-based genetic algorithm was used.

It is difficult to apply directly to a genetic algorithm that uses a fixed length because the number of travel places varies according to the constraint of the travel itinerary. Therefore, it is preferable to express the visit priorities for the m POI candidate nodes in an array. Therefore, it is necessary to convert the variable-length visited node array representing the travel itinerary into a fixed-length array. When the priority of nodes is displayed in an array, the travel schedule can be expressed as a fixed-length array. Then, the travel itinerary is expressed as a node, and the chromosome to be used in the chromosome algorithm is expressed. 4 conceptually illustrates the process of this embodiment.

Fig. 4(a) shows a connection relationship with a node, Fig. 4(b) shows a priority value of a node, and Fig. 4(c) shows a generation path.

As shown in Fig. 4(a), it is assumed that there are 5 POIs. When creating a travel itinerary that starts at node 1 (place of interest) and ends at node 5, the visit priority array of each node can be used as a chromosome.

2, 1, 3, 5, and 4 of FIG. 4(b) indicate priority values in selecting a node. If you start from node 1, the nodes you can go to are 2, 3, and 4. The priority value of node 2 is 1, the priority value of node 3 is 3, and the priority value of node 4 is 5. The node with the highest priority value, 4, is selected. After moving to node 4, the nodes that can go from node 4 are 1, 3, and 5. The priority value of node 1 is 2, the priority value of node 3 is 3, and the priority value of node 5 is 4. Since the priority value of node 5 is the largest, eventually, as shown in FIG. 4(c) , node 1, node 4, and node 5 move in the order, and the itinerary ends with node 5 to the end position.

The fit function can be used as a score when choosing a path. Optimization is performed so that a path with a high score has a higher selection probability.

Preferably, the fitting function using the user satisfaction function uses a method of adding a penalty to the user satisfaction when the constraint is violated as in Equation 9 in order to express the constraint.

(식 9)

(Equation 9)

Φ ( I , t _o ) is a function that checks whether a constraint violation occurs when path I is selected at time t ₀ . If no constraint violation occurs (when Φ ( I , t _o ) = 0), the calculated user satisfaction is used as the value of the objective function. When a constraint violation occurs, it is used as a penalty by adding the number of constraint violations to the low value defined in f _worst , and the path that created such a condition should be selected as much as possible. On the other hand, E[X] means the average considering the expected value of the random variable X, that is, the probability. S _i |I, t _o means satisfaction S _i at location i when the itinerary (travel schedule) I and departure time t _o are given.

In conclusion, if optimization is performed through such a genetic algorithm, an optimized travel itinerary can be created in consideration of user satisfaction. That is, through the method proposed in the present invention, a travel schedule that maximizes user satisfaction while satisfying the constraint conditions is derived as a result, and travel schedule planning data is generated ( S150 ). And this result data is provided to the traveler terminal.

For reference, the travel itinerary planning method in consideration of weather conditions and residence time according to an embodiment of the present invention may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software.

Examples of computer-readable media include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floppy disks, and ROM, RAM, A hardware device specifically configured to store and execute program instructions, such as flash memory, may be included. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter as well as machine codes such as those generated by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

The protection scope of the present invention is not limited to the description and expression of the embodiments explicitly described above. In addition, it is added once again that the protection scope of the present invention cannot be limited due to obvious changes or substitutions in the technical field to which the present invention pertains.

Claims (4)

A method of generating travel itinerary planning data by deriving a travel itinerary that maximizes user satisfaction while satisfying constraints as a result:
A traveler terminal accessing the travel itinerary planning artificial intelligence server and requesting a travel itinerary planning data generation service by inputting information on the travel date and place of interest;
The step of the travel schedule planning artificial intelligence server executing a travel schedule generator;
The travel itinerary generator collects data on user characteristics and travel location characteristics from a multi-layered artificial neural network, and plans travel schedules using user satisfaction with temporal constraints on stay time and weather constraints as random variables. Generating data and providing it to the traveler terminal,
When the user's satisfaction is defined as s _i ( x _i , t _i ) when staying at a place of interest x _i for a stay time t _i , the journey (I) that maximizes user satisfaction is calculated by the following equation,

The user satisfaction is calculated using neural collaboration filtering using the multi-layer artificial neural network, wherein the input layer is divided into a user characteristic input layer and a place of interest input layer,
To generate travel schedule planning data by expressing the constraint using the user satisfaction and then deriving a travel schedule that maximizes user satisfaction while satisfying the constraint using the fit function of the equation

Φ ( I , t _o ) is a function that checks whether a constraint violation occurs when path I is chosen at time t ₀ . If no constraint violation occurs (when Φ ( I , t _o ) = 0), the calculated user satisfaction is used as the value of the objective function. If a constraint violation occurs, it is used as a penalty by adding the number of constraint violations to the low value defined in f _worst . E[X] means the average considering the expected value of the random variable X. S _i |I, t _o means satisfaction S _i at location i given the itinerary I and departure time t _o .
A method of generating travel itinerary planning data using a multi-layered artificial neural network.
According to claim 1,
The weather constraint is, when x _i is the i-th place of interest to visit, and the residence time t _i represents the time to stay in the place of interest x _i , the expected precipitation (Rain _i ), temperature and humidity index (THI _i ) , including wind speed (Wind _i ), each of which is to plan a travel itinerary to visit a place of interest only when the following equations are satisfied,
(One)

(2)

(3)

A method of generating travel itinerary planning data using a multi-layered artificial neural network.
delete According to claim 1,
The method of generating travel itinerary planning data using a multi-layered artificial neural network, wherein the user satisfaction calculated through the neural collaborative filtering is optimized using a priority-based genetic algorithm.

Images