KR102330689B1 - Artificial intelligence based hotel recommendation device and method - Google Patents

Abstract

The present invention relates to an apparatus and method for generating recommended hotel information based on artificial intelligence. The user client receives an accommodation reservation request, which is the user's input for the input image information, which is the travel image information, which the user has received, and collects the input image information and accommodation property information, which is information about each accommodation product. a conversion possibility generation module including an artificial neural network module pre-learned by using as input information conversion possibility information, which is a probability that the user will purchase a specific accommodation product, as output information; and a recommended accommodation information generation module for generating recommended accommodation information based on the conversion possibility information, and outputting the generated recommended accommodation information to the accommodation reservation application module of the user client.

Description

Artificial intelligence based hotel recommendation device and method

The present invention relates to an apparatus and method for generating recommended hotel information based on artificial intelligence.

The travel industry has changed rapidly with the advent of smartphones and is one of the industries that are continuously growing every year. The travel industry can be largely divided into airlines, lodging, rental cars, activity tickets, tour guides, or package tours. Among them, the lodging reservation industry is an essential element in the travel industry, with the largest demand and supply and the most intense competition. Representative players in the lodging reservation industry are Hotels.com, Expedia, Booking.com, Agoda, Hotel Enjoy for hotel reservations, Daily Hotel, Sale Tonight, Hotel Time (how about here), Hotel Now (How about here), Hotel Time commerce for same-day hotel reservations? Yanolja), motel reservations Yanolja, how about here, pension reservations Yanolja accommodation, Woori Pension, Leave.com, Guesthouse or B&B reservations Airbnb, Kozaja, Allstay, Hotel meta search Tripadvisor, Hotels Combine, Skyscanner, Trivago, etc.

With the advent of these various travel services, travelers are starting to want to design their own travel experience and itinerary, unlike the method of simply purchasing package tours in the past. In addition, travel services are focusing on providing a new consumption experience in travel design by providing products in a variety of ways, away from pushing products and promotions that customers do not want.

Republic of Korea Patent Registration 10-1979764, Hotel reservation method and system that can compensate the difference according to the lowest price hotel reservation, Tripbitoz Co., Ltd. US Patent No. 10346402 B2, Optimized system and method for finding best fares, Expedia, Inc. US Registered Patent US 7783506 B2, System and method for managing reservation requests for one or more inventory items, Expedia, Inc. US Patent US 6826543 B1, System and method for conducting transactions involving generically identified items, Hotels.com US Patent Publication US 2016-0078374 A1, GRAPHICAL USER INTERFACE FOR HOTEL SEARCH SYSTEMS, GOOGLE INC. US Patent Publication US 2013-0031506 A1, HOTEL RESULTS INTERFACE, GOOGLE INC.

One of the ways to provide a new consumption experience in travel design is to merge travel services with social networks such as Instagram or Facebook to create a service that naturally leads to the purchase of travel products from images or videos uploaded by other users. When making a travel reservation, there is a limitation in printing nearby hotels based on location information stored together with images or videos uploaded by other users when booking a trip, so there is a problem in that it is not possible to customize the user's new consumption experience.

Accordingly, an object of the present invention is to provide an apparatus and method for generating recommended hotel information based on artificial intelligence that recommends a customized hotel product to a user based on images or videos uploaded by other users.

Hereinafter, specific means for achieving the object of the present invention will be described.

An object of the present invention, the travel image feed output module for outputting the travel image information uploaded by other users to the accommodation reservation application module of the user client, which is the user's client; The user client receives an accommodation reservation request, which is the user's input for the input image information, which is the travel image information, which the user has received, and collects the input image information and accommodation property information, which is information about each accommodation product. a conversion possibility generation module including an artificial neural network module pre-learned by using as input information conversion possibility information, which is a probability that the user will purchase a specific accommodation product, as output information; and a recommended accommodation information generation module for generating recommended accommodation information based on the conversion possibility information, and outputting the generated recommended accommodation information to the accommodation reservation application module of the user client; The facility product includes products for at least one of a hotel, a motel, a hostel, a bed and breakfast, time commerce of a hotel, an activity, a package tour, and a guide, and the artificial neural network module of the convertibility generating module is the specific accommodation product It can be achieved by providing an apparatus for generating recommended hotel information based on artificial intelligence, characterized in that learning is performed based on the difference between the actual conversion and the conversion possibility information.

Another object of the present invention is a memory module for storing a program code for generating recommended accommodation information in connection with a user client, which is a client of a user in which an accommodation reservation application module is installed; and a processing module for processing the program code, wherein the program code includes: a travel image feed output step of outputting travel image information uploaded by other users to the accommodation reservation application module of the user client, which is the user's client; The user client receives an accommodation reservation request, which is the user's input for the input image information, which is the travel image information, which the user has received, and collects the input image information and accommodation property information, which is information about each accommodation product. a conversion possibility generation step of generating the conversion possibility information for each accommodation facility product through a pre-learned artificial neural network module using the conversion possibility information that is the probability that the user will purchase a specific accommodation facility product as input information; and a recommended accommodation information generation step of generating recommended accommodation information based on the conversion possibility information, and outputting the generated recommended accommodation information to the accommodation reservation application module of the user client. The accommodation product includes a product for at least one of a hotel, a motel, a hostel, a guest house, a time commerce of a hotel, an activity, a package tour, and a guide, and the artificial neural network module of the convertibility generating module is configured to , It can be achieved by providing an artificial intelligence-based recommended hotel information generating device, characterized in that the learning based on the difference between the actual conversion for the specific accommodation product and the conversion possibility information.

Another object of the present invention, the travel image feed output module, the travel image feed output step of outputting the travel image information uploaded by other users to the accommodation reservation application module of the user client, the client of the user; The conversion possibility generation module receives, from the user client, an accommodation reservation request that is the user's input for the incoming image information that is the travel image information that the user has imported, and the incoming image information and information on each accommodation product Using the lodging facility attribute information as input information and conversion possibility information, which is the probability that the user will purchase a specific lodging facility product, as output information, the conversion possibility information for each accommodation facility product is generated through the pre-learned artificial neural network module. conversion potential generation step; and the recommended accommodation information generation module generates recommended accommodation information based on the conversion possibility information, and generates recommended accommodation information for outputting the generated recommended accommodation information to the accommodation reservation application module of the user client Including; step; wherein the accommodation product includes a product for at least one of a hotel, a motel, a hostel, a guest house, a time commerce of a hotel, an activity, a package tour, and a guide, and the artificial neural network of the conversion possibility generation module The module can be achieved by providing a method for generating recommended hotel information based on artificial intelligence, characterized in that learning is performed based on the difference between the actual conversion for the specific accommodation product and the conversion possibility information.

Another object of the present invention, configured in the user client, which is the user's client, a memory module for storing the accommodation reservation application module installed in the user client; and a processing module configured in the user client and configured to process the program code of the accommodation reservation application module; Including, wherein the program code of the accommodation reservation application module, the travel image information receiving step of receiving the travel image information from the artificial intelligence-based recommended hotel information generating device according to claim 1; a reservation request step of transmitting an accommodation reservation request for the travel image information according to the user's input; and a recommended accommodation information receiving step of receiving recommended accommodation information generated based on the travel image information from the artificial intelligence-based recommended hotel information generating device. It can be achieved by providing an artificial intelligence-based hotel reservation device, which is configured to be performed on a computer, including.

Another object of the present invention is that the accommodation reservation application module configured in the user client, which is the user's client, and installed in the user client, receives travel image information from the artificial intelligence-based recommended hotel information generating device according to claim 1 . Travel image information receiving step; a reservation request step in which the accommodation reservation application module transmits an accommodation reservation request for the travel image information according to the user's input; and a recommended accommodation information receiving step in which the accommodation reservation application module receives recommended accommodation information generated based on the travel image information from the artificial intelligence-based recommended hotel information generating device; This can be achieved by providing a hotel reservation method of

As described above, according to the present invention, there are the following effects.

First, according to an embodiment of the present invention, the effect of recommending a hotel product that has a high probability of converting to a hotel product purchase while being related to the travel image information that contributed to the inflow of the user (related to the user's hidden purchase intention) is generated

Second, according to an embodiment of the present invention, although the information for each hotel is insufficient to train the artificial neural network, there is an effect that the artificial neural network can be applied to the hotel product recommendation.

Third, according to an embodiment of the present invention, there is an effect that an artificial neural network can be applied to hotel product recommendation without separately classifying and labeling travel image information by inputting human resources.

The following drawings attached to the present specification illustrate preferred embodiments of the present invention, and serve to further understand the technical spirit of the present invention together with the detailed description of the present invention, so that the present invention is limited only to the matters described in those drawings should not be interpreted as
1 is a schematic diagram showing an artificial intelligence-based hotel list providing system according to an embodiment of the present invention;
2 is a schematic diagram showing an artificial intelligence-based switchability generating apparatus according to an embodiment of the present invention;
3 is a schematic diagram showing a hotel information collection module 10 and an image information encoding module 11 according to an embodiment of the present invention;
4 is a schematic diagram showing an embedding module 12 according to an embodiment of the present invention;
5 is a schematic diagram showing an artificial neural network of the switchability generating module 13 according to an embodiment of the present invention;
6 is a schematic diagram showing an image information encoding module 11 and a hotel information collection module 10 according to a modification of the present invention;
7 is a schematic diagram showing an embedding module 12 according to a modified example of the present invention;
8 is a schematic diagram showing a switchability generating module 13 according to a modification of the present invention;
9 is a schematic diagram showing an example of a ConvNet encoder included in the video information encoding module 11 according to an embodiment of the present invention;
10 is a schematic diagram illustrating an apparatus for generating recommended hotel information based on artificial intelligence according to an embodiment of the present invention;
11 is a schematic diagram showing a list of recommended hotels according to an embodiment of the present invention;
12 is a schematic diagram illustrating a similarity calculation according to an embodiment of the present invention;
13 is a schematic diagram illustrating an apparatus for generating an artificial intelligence-based travel image feed according to an embodiment of the present invention;
14 is a schematic diagram showing reinforcement learning of the artificial intelligence-based travel image feed generating apparatus 3 according to a modification of the present invention;
15 is a flowchart showing an example of reinforcement learning operation of the artificial intelligence-based travel image feed generating apparatus 3 according to an embodiment of the present invention;
16 is a flowchart illustrating a method for providing a hotel list based on artificial intelligence according to an embodiment of the present invention;
17 is a schematic diagram illustrating an embodiment in a hotel reservation application module of a user client of an artificial intelligence-based hotel list providing method according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings, a person of ordinary skill in the art to which the present invention pertains will be described in detail an embodiment in which the present invention can be easily carried out. However, in the detailed description of the principle of operation of the preferred embodiment of the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, the same reference numerals are used throughout the drawings for parts having similar functions and functions. Throughout the specification, when it is said that a specific part is connected to another part, this includes not only a case in which it is directly connected, but also a case in which it is indirectly connected with another element interposed therebetween. In addition, the inclusion of specific components does not exclude other components unless otherwise stated, but means that other components may be further included.

In the following description of the invention, a convolutional neural network that is a neural network using a convolutional product may be described as a CNN, a ConvNet, or the like.

In the following description of the invention, it has been described based on hotel reservations for convenience of explanation, but the scope of the present invention is not limited to hotel reservations, and time commerce of guest houses, hostels, motels, and hotels, activities, package tours, guides, rental cars, etc. You can include coverage for all travel products.

1 is a schematic diagram illustrating an artificial intelligence-based hotel list providing system according to an embodiment of the present invention. As shown in FIG. 1 , the artificial intelligence-based hotel list providing system 100 according to an embodiment of the present invention includes an artificial intelligence-based switchability generating device 1 and an artificial intelligence-based hotel information generating device. (2), an artificial intelligence-based travel image feed generating device 3 may be included. As shown in Figure 1, the artificial intelligence-based hotel list providing system 100 according to an embodiment of the present invention is included in the memory module of the user client and various users to the hotel reservation application module processed by the processing module. may be configured to transmit travel image information (including image format or video format) generated by have.

Artificial intelligence-based conversion possibility generation device

2 is a schematic diagram illustrating an artificial intelligence-based switchability generating apparatus according to an embodiment of the present invention. As shown in FIG. 2 , the artificial intelligence-based switchability generating apparatus 1 according to an embodiment of the present invention includes a hotel information collection module 10 , an image information encoding module 11 , and an embedding module 12 . , a convertibility generating module 13 . Artificial intelligence-based switchability generating device 1 according to an embodiment of the present invention is processed by a processing module of a computing device such as a specific web server, a virtual server such as a cloud server, a smartphone, a tablet PC, a desktop PC, and , may be configured to be stored in the memory module of each device.

The hotel information collection module 10 collects and encodes hotel property information (hotel name, location information, review information, convenience facility information, room product information, etc.) for a plurality of hotels by crawling from a hotel information web server such as TripAdvisor. and is a module configured to generate hotel attribute encoding information. The hotel attribute encoding information is generated for each of a plurality of hotels, respectively.

The image information encoding module 11 is a module for generating incoming image encoding information by encoding incoming image information, which is travel image information that is a basis for a hotel reservation request from a user client. 3 is a schematic diagram illustrating a hotel information collection module 10 and an image information encoding module 11 according to an embodiment of the present invention. As shown in FIG. 3 , the image information encoding module 11 receives the input image information and generates input image encoding information, and the hotel information collection module 10 crawls the hotel attribute information to generate hotel attribute encoding information. configured to do

The embedding module 12 converts the incoming image encoding information received from the image information encoding module 11 and the hotel attribute encoding information received from the hotel information collection module 10 into a continuous vector form and merges them, and a two-dimensional or more vector It is a module that generates a conversion probability matrix. 4 is a schematic diagram illustrating an embedding module 12 according to an embodiment of the present invention. As shown in FIG. 4 , the embedding module 12 converts the incoming image encoding information received from the image information encoding module 11 and the hotel attribute encoding information received from the hotel information collection module 10 into a continuous vector form. is configured to create a switchability matrix, which is a vector of two or more dimensions. In the generation of the convertibility matrix, the incoming video encoding information and the hotel attribute encoding information are matched for each hotel, and the convertibility matrix is generated for each of the plurality of hotels.

The conversion possibility generation module 13 is a module for generating conversion possibility information, which is a possibility that a user who has flowed into the corresponding travel image information (import image information) will convert to a hotel product purchase, based on the conversion probability matrix generated for each hotel. . The switchability generating module 13 according to an embodiment of the present invention may include an artificial neural network using the switchability matrix as an input vector and the switchability information as output information. 5 is a schematic diagram illustrating an artificial neural network of the switchability generating module 13 according to an embodiment of the present invention. As shown in FIG. 5 , the convertibility generating module 13 uses a convertibility matrix for each hotel as input information, and uses the convertibility information for the corresponding hotel as output information to include a pre-learned artificial neural network. can be configured. In particular, the artificial neural network module of the switchability generating module 13 may update the weight of each node by a method such as Back Propagation based on a difference (error, error) between the actual switchability and the switchability information.

According to an embodiment of the present invention, when a specific user flows into a hotel product purchase due to various travel image information generated by users, the hotel product is based on conversion possibility information, which is a possibility that the user will lead to a hotel product purchase There is an effect that can improve the conversion rate of the user through methods such as recommending or outputting the next travel video information. In addition, according to the switchability matrix, which is input information of the artificial neural network according to an embodiment of the present invention, the existing problem that is too insufficient to train the artificial neural network with only the conversion information of each individual hotel is solved.

With respect to a modification of the switchability generating module 13, FIG. 6 is a schematic diagram illustrating an image information encoding module 11 and a hotel information collection module 10 according to a modification of the present invention. As shown in FIG. 6 , the image information encoding module 11 receives the input image information and the hotel image information to generate the input image encoding information and the hotel image encoding information, and the hotel information collection module 10 receives the hotel attribute information is configured to crawl to generate hotel attribute encoding information. 7 is a schematic diagram illustrating an embedding module 12 according to a modified example of the present invention. As shown in FIG. 7 , the embedding module 12 continuously stores the incoming video encoding information received from the video information encoding module 11 , the hotel image encoding information and the hotel property encoding information received from the hotel information collection module 10 . It is configured to convert and merge into a form vector form, and to generate a convertibility matrix that is a vector of three or more dimensions. 8 is a schematic diagram illustrating a switchability generating module 13 according to a modified example of the present invention. As shown in FIG. 8 , the convertibility generation module 13 uses a convertibility matrix, which is a three-dimensional or higher vector for each hotel, as input information, and uses the convertibility information for the corresponding hotel as output information. It may be configured to include an artificial neural network. According to a modified example of the present invention, hotel image encoding information is embedded together and applied as input information to an artificial neural network, so that switchability information including calculation of the relationship between incoming image information and hotel image information is generated. .

9 is a schematic diagram illustrating an example of a ConvNet encoder included in the video information encoding module 11 according to an embodiment of the present invention. As shown in FIG. 9 , the ConvNet encoder, which is the video information encoding module 11, may be constructed as [INPUT-CONV-RELU-POOL-FC]. Incoming image information or hotel image information, which is input information, may be composed of a matrix having a channel of width 32, length 32, and height n, and the size of the input is [32x32xn]. For example, when the incoming image information is a video format composed of a plurality of sequential images, n may be composed of the number of corresponding sequential images. The CONV layer (Conv. Filter, 101) is connected to a partial region of incoming image information, and a dot product of the connected region and its own weight is calculated. The resulting volume will have a size equal to [32x32x12]. The RELU layer is an activation function applied to each element, such as max(0,x). The RELU layer does not change the size of the volume ([32x32x12]). As a result, activation map 1 (102) is generated. The POOL layer (pooling, 103) outputs a reduced volume (Activation map 2, 104) such as [16x16x12] by performing downsampling on the "horizontal, vertical" dimensions. A fully-connected (FC) layer 106 connected to the nth activation map n 105 calculates class scores and outputs a volume (output layer, 107) having a size of [mxmx1]. The output layer 107 outputs the input image information or hotel image encoding information, which is a reduced dimension of the input image information or hotel image information. The FC layer is connected to all elements of the previous volume.

The parameters of the CONV layer consist of a set of learnable filters. Each filter is small in the horizontal/vertical dimension, but covers the entire depth in the depth dimension. In the forward pass, each filter is slid in the horizontal/vertical dimension of the input volume (convolve precisely), and an n-dimensional activation map is generated. When sliding the filter over the input, a dot product is made between the filter and the input volume. Through this process, ConvNet learns a filter that activates on a specific pattern at a specific location of input information. Stacking these activation maps in the depth dimension becomes the output volume. Therefore, each element of the output volume treats only a small area of the input, and neurons in the same activation map can share the same parameters because they are the result of applying the same filter.

Artificial intelligence-based recommended hotel information generating device

10 is a schematic diagram illustrating an apparatus for generating recommended hotel information based on artificial intelligence according to an embodiment of the present invention. As shown in FIG. 10 , the apparatus 2 for generating recommended hotel information based on artificial intelligence according to an embodiment of the present invention includes a recommended hotel list generation module 20 , a similarity calculation module 21 , and a detailed adjustment list generation module 22 . The artificial intelligence-based hotel information generation device 2 according to an embodiment of the present invention is processed by a processing module of a computing device such as a specific web server, a virtual server such as a cloud server, a smartphone, a tablet PC, and a desktop PC and may be configured to be stored in a memory module of each device.

The recommended hotel list generation module 20 is a module for generating a recommended hotel list by including hotels with the conversion possibility information equal to or greater than a specific value in the recommended hotel list, based on the conversion possibility information generated by the conversion possibility generation module 13. . 11 is a schematic diagram illustrating a recommended hotel list according to an embodiment of the present invention. 11 , the recommended hotel list according to an embodiment of the present invention may refer to a list of hotels having conversion possibility information greater than or equal to a specific value.

The similarity calculation module 21 is a module for calculating the similarity between the incoming video encoding information and the hotel surrounding video encoding information for each hotel in the recommended hotel list. The hotel surrounding image encoding information according to an embodiment of the present invention inputs at least one or more travel image information generated by users within a specific distance from the location of the hotel, the hotel surrounding image information, to the image information encoding module 11 . This is encoded information. The similarity according to an embodiment of the present invention may be calculated by various similarity calculation methods such as KL-divergence, cross entropy, cosine similarity, Jaccard Similarity, and various cluttering. 12 is a schematic diagram illustrating a similarity calculation according to an embodiment of the present invention. As shown in FIG. 12 , according to an embodiment of the present invention, the similarity calculation module 21 receives video encoding information and incoming video encoding information around the hotel, which encodes the hotel surrounding video information for each hotel in the recommended hotel list, and , may be configured to calculate a similarity between the incoming video encoding information and the hotel surrounding video encoding information. That is, the image encoding information around a hotel according to an embodiment of the present invention may mean an identity for a user experience in a specific hotel.

According to a modified example of the present invention, when generating the image encoding information around the hotel input to the similarity calculation module 21, the image information encoding module 11 sets the distance between the hotel and travel image information around the hotel (image information around the hotel) It may be configured to generate (encoding by assigning a distance-based weight to each hotel surrounding image information) based on the hotel surrounding image encoding information for the hotel. According to this, there is an effect that the identity of the user experience in a specific hotel is weighted based on the distance by the image encoding information around the hotel. For example, if travel video information about swimming in the sea is generated 100m away from the hotel and travel video information about historical site visits is generated 1km away from the hotel, video encoding information around the hotel may be generated closer to the encoding information of the travel image information for sea swimming than the encoding information of the travel image information for the historical site exploration.

The refinement list generation module 22 is a module for generating a detailed adjustment list by fine-tuning the recommended hotel list based on the similarity between the incoming video encoding information calculated by the similarity calculation module 21 and the hotel surrounding video encoding information. According to an embodiment of the present invention, the recommended hotel list may be fine-tuned by using the similarity between the incoming video encoding information and the hotel surrounding video encoding information as a weight for the conversion possibility information of each hotel.

According to an embodiment of the present invention, it is possible to generate a recommended hotel list in which the user's intention to purchase hotel products is reflected through the inflow image information, which is travel image information uploaded by a specific user. do. For example, a user who is introduced by watching a travel video swimming in an infinity pool has an effect of preferentially recommending a hotel with a high conversion potential while having a travel experience identity similar to that of the incoming video information. In the existing method, in order to recommend hotels by reflecting users' inflow intentions based on experiences such as infinity pools, sea swimming, local market visits, specific food, club culture, and historical site visits, the service subject must label all hotels. There was a problem that had to be done and could not be solved. In addition, according to an embodiment of the present invention, since the recommended hotel list is first selected by the conversion possibility information and the detailed adjustment list is generated by the similarity calculation, the load of the similarity calculation module 21 that requires a lot of computing power ) is reduced.

The detailed adjustment list generation module 22 according to another modified example of the present invention automatically filters the recommended hotel list based on the property (hotel property information) of the hotel used by the user, which is the hotel previously used by the user. can be configured to adjust. In particular, while the user uses a plurality of existing hotels, if there is no change in a specific attribute, the attribute is preferentially filtered, and when there is a change in other specific attributes, the attribute may not be filtered. . For example, if a specific user has previously used Hotel A, Hotel B, and Hotel C, it searches for properties that do not change among the properties of Hotel A, Hotel B, and Hotel C. When the property of is searched, it can be configured to filter the recommended hotel list by the specific property value of the property (pet availability=1, review rating>=3.5, completion year>=2015) and then fine-tune the similarity. According to this, while a customized hotel can be provided to the user, the calculation load of the similarity calculation module 21 is further reduced.

AI-based travel video feed generation device

13 is a schematic diagram illustrating an apparatus for generating an artificial intelligence-based travel image feed according to an embodiment of the present invention. As shown in FIG. 13 , the artificial intelligence-based travel image feed generating apparatus 3 according to an embodiment of the present invention may be configured to update a specific travel image output in the travel image feed by reinforcement learning. From the point of view of reinforcement learning, the objective of the artificial intelligence-based travel image feed generating apparatus 3 according to an embodiment of the present invention is to improve the possibility that the user is converted through the output travel image information, and state and Environment may mean conversion possibility information of each hotel, hotel property information of each hotel, information on existing hotels used by the corresponding user, and image information around the hotel of each hotel, and Action is AI-based It may refer to the output of travel image information of the travel image feed generating device 3 , and the reward may be configured to be generated when the user is converted (purchasing a hotel product) through the output travel image information. According to an embodiment of the present invention, when the output of travel image information of the artificial intelligence-based travel image feed generating apparatus 3 is learned by reinforcement learning, a plurality of users may use the artificial intelligence-based travel image feed generating apparatus 3 The more it is introduced by the travel image information output by , the more the user's conversion possibility can be improved. That is, in order to improve the possibility of user conversion by the travel image feed (series of travel image information) output by the artificial intelligence-based travel image feed generating device 3, a lot of information about one user is not required, but several Even when an inflow occurs for a user, there is an effect that the artificial intelligence-based travel image feed generating device 3 can be reinforced with learning. In addition, since the switchability information is included in the environment, there is an effect that is advantageous in outputting travel image information in a direction to improve the user's switchability.

The artificial intelligence-based travel image feed generating apparatus 3 according to a modified example of the present invention may be configured to update the travel image information output by more effective reinforcement learning by the following configuration. 14 is a schematic diagram illustrating reinforcement learning of the artificial intelligence-based travel image feed generating apparatus 3 according to a modification of the present invention. 14, the artificial intelligence-based travel image feed generating apparatus 3 according to a modified example of the present invention includes a value network 211 that is an artificial neural network that learns a value function that outputs a value in a specific state, and It may include a policy network 210 for learning a policy function for outputting travel image information, and the policy network 210 and the value network 211 according to a modified example of the present invention are an artificial intelligence-based travel image feed generating device. (3) may be included. In addition, the policy network 210 and the value network 211 may be connected to the optimal travel image search module to output selected optimal travel image information.

The policy network 210 is an artificial neural network that determines the probability of the selected travel image information in each state of the artificial intelligence-based travel image feed generating device 3, and outputs the selected travel image information by learning the policy function. . The cost function of the policy network may be a function obtained by calculating a cross entropy by multiplying the policy function and the cost function of the value network and then taking a policy gradient, for example, it may be configured as in Equation 1 below. The policy network can be back propagated based on the product of cross entropy and time difference error, which is a cost function of the value network.

In Equation 1, π is a policy function, θ is a policy network parameter, π _θ (a _{i |s i} ) is the possibility of performing a specific action (output specific travel video information) in the current episode, V is a value function, and w is Value chain parameter, s _i is the state information of the current episode i, S _i+1 is the state information of the next episode i+1, r _i+1 is the reward expected to be obtained in the next episode, V _w (s _i ) may mean compensation potential in the current episode, V _w (s _i+1 ) may mean compensation potential in the next episode, and γ may mean a depreciation rate. In this case, r _i+1 may be configured to be received by an artificial neural network module that predicts switchability information. As a result, the policy network 210 according to an embodiment of the present invention initially outputs travel image information simulating the user-introduced travel image information through the policy gradient.

The policy network 210 according to an embodiment of the present invention is supervised learning based on the user's imported travel image information and performance information (conversion possibility information) prior to reinforcement learning. We can learn the basics of the policy by updating the weights of That is, the weight of the policy network can be set by supervised learning based on the existing incoming video information and performance information. According to this, there is an effect that the policy network can be learned very quickly by recording the existing incoming video information.

In addition, according to an embodiment of the present invention, in supervised learning of the policy network 210, it may be configured to be supervised based on the output of the existing travel image information including a random vector and performance information accordingly. The random vector may use, for example, a Gaussian distribution. According to this, there is an effect that the policy network can output an output policy of challenging travel image information with random probability. When supervised learning of the policy network 210 is configured to be supervised learning based on the existing inflow image information and the resulting performance information, the selection of travel image information is optimized within the existing policy. However, if a random vector is included in the supervised learning of the policy network according to an embodiment of the present invention, as reinforcement learning proceeds, the policy network can learn the output of travel image information more effectively than the existing policy. occurs

The value network 211 is an artificial neural network that derives the possibility of achieving a reward in each state that the artificial intelligence-based travel image feed generating device 3 may have, and learns a value function. The value network 211 presents the direction in which the AI-based travel image feed generating device 3, which is an agent, will be updated. To this end, the input variable of the value network 211 is set as state information, which is information about the state of the artificial intelligence-based travel image feed generating device 3 , and the output variable of the value network 211 is artificial intelligence-based travel. It may be set as compensability information (conversion possibility information) that is a possibility that the image feed generating device 3 will achieve compensation. Compensation possibility information according to an embodiment of the present invention may be calculated as a Q-function as shown in the following equation.

In Equation 2 above, Q _π may mean total compensation potential information expected in the future in the case of a state s and action a in a specific policy π, R may mean compensation for a specific period, and gamma may mean a depreciation rate. S _t may mean a state at time t, A _t may mean an action at time t, and E may mean an expected value. Compensation possibility information (Q value) according to an embodiment of the present invention defines the update direction and size of the policy network 210 .

In this case, the cost function of the value network may be a mean square error (MSE) function for the value function, and may be configured as, for example, Equation 3 below. The value network 211 may be back propagated based on a time difference error that is a cost function of the value network.

In Equation 3, V is a value function, w is a value network parameter, s _i is state information of i, which is the current episode, S _i+1 is state information of _{i+1, which} is the next episode, r i+1 is in the next episode. A reward expected to be obtained, V _w (s _i ) may mean a reward possibility in the current episode, V _w (s _i+1 ) may mean a reward possibility in the next episode, and γ may mean a depreciation rate. In this case, r _i+1 may be configured to be received by an artificial neural network module that predicts the possibility of switching.

Accordingly, when the state of the artificial intelligence-based travel image feed generating device 3 is changed, the value network may be updated in the direction of gradient descent of the cost function of Equation 3 .

According to an embodiment of the present invention, while learning the value network separately from the policy network, the Q value of the value network is supervised instead of starting at random, thereby enabling rapid learning. According to this, there is an effect that the burden of exploration can be greatly reduced in the action of selecting very complicated travel image information.

According to the artificial intelligence-based travel image feed generating apparatus 3 according to an embodiment of the present invention, when the policy network 210 that has completed supervised learning selects travel image information to be output in the current episode i, the value network 211 ) is learned to predict a reward (conversion possibility information) when outputting selected travel image information. The policy network 210 and the value network 211 of the AI-based travel image feed generating device 3 that have completed learning are combined with a simulation using an optimal travel image search module to select travel image information to be finally output. is utilized

In addition, according to the value network 211 according to an embodiment of the present invention, the update of the policy network for outputting selected travel image information can be performed for every episode. In the existing reinforcement learning, there is a problem that the update of the reinforcement learning model proceeds after all episodes are finished, so it is applied to the artificial intelligence-based travel image feed generating device 3 that must output travel image information with high conversion potential every time. It was difficult to do.

The optimal travel image search module is a configuration that searches for optimal travel image information by performing a plurality of simulations for various states and various actions based on a plurality of agents calculated in the policy network and the value network. The optimal travel image search module according to an embodiment of the present invention may utilize, for example, a Monte Carlo tree search, each node of the tree represents a state, and each edge represents a specific action for the corresponding state. It indicates the value expected according to In the optimal travel image search module according to an embodiment of the present invention, when Monte Carlo tree search is utilized, the optimal travel image search may be processed in four steps: Selection, Expansion, Evaluation, and Backup.

The selection step of the optimal travel image search module is a step in which an action with the highest value among selectable actions is selected and proceeded from the current state until a leaf node emerges. At this time, the value of the value function stored in the edge and the visit frequency value to balance the inquiry-use are used. The equation for action selection in the Selection step is as follows.

In Equation 4 above, a _t is the action (travel image information output) at time t, Q(s _t ,a) is the value of the value function stored in the tree, and u(s _t ,a) is the corresponding state-action It is a value inversely proportional to the number of visits in a pair and is used to balance exploration and use.

The expansion step of the optimal travel image search module is a step of adding a new node as a leaf node by acting according to the probability of the policy network learned through supervised learning when the simulation proceeds to the leaf node.

The evaluation stage of the optimal travel image search module is based on the value determined using the value network from the newly added leaf node (conversion potential) and the reward obtained from the leaf node using the policy network until the end of the travel video information output episode. This is the stage to evaluate the value of the node. The following equation is an example of evaluating the value of a new leaf node.

In Equation 5 above, V(s _L ) is the value of the leaf node, λ is the mixing parameter, v _θ (s _L ) is the value obtained through the value network, and z _L may mean the reward obtained by continuing the simulation.

The backup stage of the optimal travel image search module is a stage of re-evaluating the value of the nodes visited during the simulation and updating the visit frequency by reflecting the value of the newly added leaf node. The equation below is an example of node value revaluation and visit frequency update.

In Equation 6 above, s _L ⁱ represents the leaf node in the i-th simulation, 1(s,a,i) represents whether the connection (s,a) is visited in the i-th simulation, and when the tree search is completed, the algorithm is It can be configured to select the most visited connection (s,a) from the root node. According to the optimal travel image search module according to an embodiment of the present invention, it is possible to select the optimal travel image information by conducting a simulation multiple times based on the value network for a plurality of travel image information selected by the policy network. is generated

According to an embodiment of the present invention, the artificial intelligence-based travel image feed generating device 3 may be configured such that a plurality of agents are configured. When a plurality of agents are configured, the travel image information selected by the artificial intelligence-based travel image feed generating device 3 for each specific state and specific travel image information competes with each other to select the most optimal travel image information. effect is generated.

15 is a flowchart illustrating an operation example of reinforcement learning of the apparatus 3 for generating an artificial intelligence-based travel image feed according to an embodiment of the present invention. As shown in FIG. 15 , when the state s(t) is input by the artificial intelligence-based switchability generating device 1 that tracks the state information in the hotel reservation application module of a plurality of user clients, it is added to the value network 211. Various travel image information is input to the optimal travel image search module by a plurality of agents of the policy network 210, and selected travel image information is output as an action output by the optimal travel image search module As the travel video feed is progressed by a(t), the episode t ends and the episode t+1 begins. In episode t+1, again s(t+1), which is a state change due to a(t), is input by the artificial intelligence-based switchability generating device 1, and r(t+) as a reward according to a(t) 1) is directly input to update the value network 211 and the policy network 210 .

How to provide a hotel list based on artificial intelligence

In relation to the artificial intelligence-based hotel list providing method according to an embodiment of the present invention, FIG. 16 is a flowchart illustrating an artificial intelligence-based hotel list providing method according to an embodiment of the present invention. As shown in Fig. 16, the artificial intelligence-based hotel list providing method according to an embodiment of the present invention includes a travel image feed output step (S10), a hotel reservation request reception step (S11), and a conversion possibility information generation step (S12). ), a recommended hotel list generation step (S13), and a detailed adjustment list generation step (S14).

The travel image feed output step S10 is a step in which the artificial intelligence-based travel image feed generating device 3 outputs a travel image feed, which is a series of travel image information, to the hotel reservation application module of the user client.

The hotel reservation request reception step (S11) is a step in which the artificial intelligence-based switchability generating device 1 inputs a hotel reservation based on the input image information, which is specific travel image information that is the basis of the user's hotel reservation request.

The conversion possibility information generation step ( S12 ) is a step in which the artificial intelligence-based conversion possibility generation device 1 generates conversion possibility information for a plurality of hotels.

The recommended hotel list generation step (S13) is a step in which the artificial intelligence-based recommended hotel information generating device 2 generates a recommended hotel list based on the switchability information.

In the detailed adjustment list generation step (S14), the artificial intelligence-based recommended hotel information generating device 2 fine-tunes the recommended hotel list based on the similarity between the inflow image information and the travel image information within a specific distance of each hotel. This is a step in which the list is created, and the AI-based travel video feed generating device 3 updates the travel video feed based on the conversion information.

17 is a schematic diagram illustrating an embodiment in a hotel reservation application module of a user client of an artificial intelligence-based hotel list providing method according to an embodiment of the present invention. As shown in FIG. 17, when a user inputs a hotel reservation request in ① another user's travel image information (incoming image information) output to the user client, ② generated by the artificial intelligence-based hotel list providing system 100 The recommended hotel information of the detailed adjustment list is output through the hotel reservation application module. According to an embodiment of the present invention, as shown in FIG. 17 , the location-based hotel search result generated based on location information stored together with the incoming image information may be configured to be output after the recommended hotel information.

According to the present invention, there is an effect that the user introduced through the travel image information can recommend a hotel product that is highly related to the user's hidden purchase intention, while the possibility of conversion to the purchase of the hotel product is high.

As described above, those skilled in the art to which the present invention pertains will be able to understand that the present invention may be embodied in other specific forms without changing the technical spirit or essential features thereof. Therefore, the above-described embodiments are to be understood in all respects as illustrative and not restrictive. The scope of the present invention is indicated by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts should be construed as being included in the scope of the present invention.

The features and advantages described herein are not all inclusive, and many additional features and advantages will become apparent to those skilled in the art, particularly upon consideration of the drawings, the specification, and the claims. Moreover, it should be noted that the language used herein has been principally selected for readability and teaching purposes, and may not be chosen to delineate or limit the subject matter of the present invention.

The foregoing description of embodiments of the present invention has been presented for purposes of illustration. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Those skilled in the art will appreciate that many modifications and variations are possible in light of the above disclosure.

Therefore, the scope of the present invention is not limited by the detailed description, but by any claims of the application based thereon. Accordingly, the disclosure of the embodiments of the present invention is illustrative and not intended to limit the scope of the present invention as set forth in the following claims.

1: Artificial intelligence-based conversion possibility generation device
2: Artificial intelligence-based hotel information generating device
3: AI-based AI-based travel video feed generation device
10: Hotel information collection module
11: Embedding module
12: Convertible Creation Module
13: Diagnostic problem output module
20: Recommended hotel list creation module
21: Similarity Calculation Module
22: Reconciliation list generation module
100: AI-based hotel list providing system
101: Conv. Filter
102: Activation map 1
103: Pooling layer
104: Activation map 2
105: Activation map n
106: Fully connected layer
107: output layer
210: policy network
211: value chain

Claims (5)

a travel image feed output module for outputting travel image information uploaded by other users to the accommodation reservation application module of the user client, which is the user's client;
In the travel image information, the user client receives an accommodation reservation request, which is the user's input for the input image information, which is the travel image information, which is the travel image information imported by the user, from the user client, and encodes the input image information to generate the input image encoding information and an image information encoding module for generating image encoding information around the accommodation, which is encoding information of the travel image information generated within a specific distance from the location of each accommodation;
an embedding module for generating a conversion possibility matrix by merging the incoming video encoding information and accommodation property information, which is information about each accommodation product;
a conversion possibility generation module including a pre-learned artificial neural network module using the conversion probability matrix as input information and conversion probability information, which is a probability that the user purchases each accommodation product, as output information;
a recommended accommodation information generating module for generating recommended accommodation information based on the conversion possibility information;
a similarity calculation module for calculating a similarity between the incoming video encoding information and the video encoding information around the accommodation for each accommodation in the recommended accommodation information; and
a detailed adjustment list generation module for generating a detailed adjustment list by fine-tuning the recommended accommodation information based on the similarity;
including,
The accommodation products include products for at least one of hotels, motels, hostels, guest houses, time commerce of hotels, activities, package tours, and guides,
When the similarity calculation module generates the video encoding information around the accommodation, the video encoding information around the accommodation is encoded with the distance from the location of each accommodation as a weight,
The artificial neural network module of the conversion possibility generation module is characterized in that it is learned based on the difference between the actual conversion for each accommodation product and the conversion possibility information,
A device for generating recommended hotel information based on artificial intelligence.
a memory module for storing a program code for generating recommended accommodation information in connection with a user client, which is a client of a user in which the accommodation reservation application module is installed; and
a processing module for processing the program code;
including,
The program code is
A travel image feed output step of outputting the travel image information uploaded by other users to the accommodation reservation application module of the user client, which is the user's client;
In the travel image information, the user client receives an accommodation reservation request, which is the user's input for the input image information, which is the travel image information, which is the travel image information imported by the user, from the user client, and encodes the input image information to generate the input image encoding information and an image information encoding step of generating image encoding information around the accommodation, which is encoding information of the travel image information generated within a specific distance from the location of each accommodation;
an embedding step of generating a conversion possibility matrix by merging the incoming video encoding information and accommodation property information, which is information about each accommodation product;
Using the conversion possibility matrix as input information and conversion probability information, which is the probability that the user will purchase each accommodation facility product, as output information, the conversion probability matrix is received from the pre-learned artificial neural network module to receive the conversion probability matrix for each accommodation facility product. a switchability generating step of outputting the switchability information;
a recommended accommodation information generation step of generating recommended accommodation facility information based on the conversion possibility information;
a similarity calculation step of calculating a similarity between the incoming video encoding information and the video encoding information around the accommodation for each accommodation in the recommended accommodation information; and
a detailed adjustment list generation step of generating a detailed adjustment list by fine-tuning the recommended accommodation information based on the similarity;
configured to perform on a computer, including
The accommodation products include products for at least one of hotels, motels, hostels, guest houses, time commerce of hotels, activities, package tours, and guides,
When generating the video encoding information around the accommodation facility in the similarity calculation step, the video encoding information around the accommodation facility is encoded using the distance from the location of each accommodation as a weight,
The artificial neural network module, characterized in that the learning based on the difference between the actual conversion and the conversion possibility information for each accommodation product,
A device for generating recommended hotel information based on artificial intelligence.
Travel image feed output step of outputting, by the travel image feed output module, the travel image information uploaded by other users to the accommodation reservation application module of the user client, which is the user's client;
The image information encoding module receives, from the user client, an accommodation reservation request that is the user's input for the input image information that is the travel image information that the user has imported from the travel image information, and encodes the input image information. An image information encoding step of generating incoming image encoding information and generating image encoding information around accommodations, which is encoding information of the travel image information generated within a specific distance from the location of each accommodation;
an embedding step in which the embedding module merges the incoming video encoding information and accommodation property information, which is information about each accommodation facility product, to generate a conversion possibility matrix;
a conversion possibility generation step, in which the conversion possibility generation module includes an artificial neural network module previously trained using the conversion probability matrix as input information and conversion probability information, which is a probability that the user purchases each accommodation product, as output information;
a recommended accommodation information generation step in which the recommended accommodation information generation module generates recommended accommodation facility information based on the conversion possibility information;
a similarity calculation step of calculating, by a similarity calculation module, a degree of similarity between the incoming video encoding information and the video encoding information around the accommodation for each accommodation in the recommended accommodation information; and
a detailed adjustment list generation step in which the detailed adjustment list generation module generates a detailed adjustment list by fine-tuning the recommended accommodation information based on the similarity;
including,
The accommodation products include products for at least one of hotels, motels, hostels, guest houses, time commerce of hotels, activities, package tours, and guides,
When the similarity calculation module generates the video encoding information around the accommodation, the video encoding information around the accommodation is encoded with the distance from the location of each accommodation as a weight,
The artificial neural network module of the conversion possibility generation module is characterized in that it is learned based on the difference between the actual conversion for each accommodation product and the conversion possibility information,
A method of generating recommended hotel information based on artificial intelligence.
In the hotel reservation device for receiving recommended accommodation information from the artificial intelligence-based hotel information generating device according to claim 1,
a memory module configured to the user client, which is the user's client, and configured to store an accommodation reservation application module installed in the user client; and
a processing module configured in the user client and configured to process the program code of the accommodation reservation application module;
including,
The program code of the accommodation reservation application module,
a travel image information receiving step of receiving travel image information from the artificial intelligence-based recommended hotel information generating device;
a reservation request step of transmitting an accommodation reservation request for the travel image information according to the user's input; and
a recommended accommodation information receiving step of receiving the recommended accommodation information generated based on the travel image information in the artificial intelligence-based recommended hotel information generating device;
configured to perform on a computer comprising:
A hotel reservation device based on artificial intelligence.
In the hotel reservation method for receiving recommended accommodation information from the artificial intelligence-based recommended hotel information generating device according to claim 1,
A travel image information receiving step of receiving, by the accommodation reservation application module configured in the user client, which is the user's client, and installed in the user client, the travel image information from the artificial intelligence-based recommended hotel information generating device;
a reservation request step in which the accommodation reservation application module transmits an accommodation reservation request for the travel image information according to the user's input; and
a recommended accommodation information receiving step of the accommodation reservation application module receiving the recommended accommodation information generated based on the travel image information from the artificial intelligence-based recommended hotel information generating device;
containing,
A hotel reservation method based on artificial intelligence.

Images