KR20210028999A - Apparatus for location-based restaurant recommendation service and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and a method for a location-based restaurant recommendation service. The method using the apparatus includes: a step of performing big data construction by accumulating choice and rejection scores in response to a user′s choice or rejection response regarding each restaurant in a restaurant list that is held; a step of performing terminal location-based restaurant search and generating an initial list when receiving a restaurant recommendation service request from a user terminal; a step of correcting the initial list by determining list removal or non-removal using the choice score accumulated at the restaurant and post-visit elapsed time in a case where the initial list has a restaurant where the user visited within N days; a step of finally determining a recommendation object by calculating the probability of list recommendation using the choice and rejection scores accumulated at the restaurant and an industry-corresponding category score with regard to each restaurant in the corrected list; and a step of recommending the final list of finally determined restaurants to the user terminal by exposing the same to the user terminal. According to the present invention, it is possible to select an optimum restaurant list for a user and recommend and provide the same to the user terminal based on the location of the user terminal, the user′s visit history, and past big data accumulated by restaurant.

Description

Location-based restaurant recommendation service apparatus and method thereof

The present invention relates to a location-based restaurant recommendation service apparatus and method thereof, and more particularly, a location-based restaurant recommendation service apparatus and method capable of providing a restaurant recommendation to a user terminal in consideration of a current location and past big data. It is about.

With the recent development of mobile communication technology, mobile devices such as smart phones are rapidly popularized, and mobile apps that can operate on mobile devices are also being actively developed.

Among them, a system for recommending products, contents, multimedia, etc. assists customers to more easily search for services that may be used or purchased, and is mainly provided based on web or mobile.

In addition, as social media activities of users who share a person's daily life increase and interest in restaurants increases, service apps that recommend restaurants on mobile devices are also diversifying.

However, most restaurant recommendation apps rely on the restaurant's location, price range, and customer ratings, and do not take into account the customer's recent visit history or past big data. In general, in the case of a restaurant that has recently been visited, it is unlikely that the user will revisit within a few days.

Therefore, it is required to develop a restaurant recommendation platform that considers a user's past visit history rather than a recommendation service based on a simple customer rating or location.

The technology behind the present invention is disclosed in Korean Patent Publication No. 2019-0057821 (published on May 29, 2019).

An object of the present invention is to provide a location-based restaurant recommendation service apparatus and method capable of providing an optimal restaurant recommendation to a user terminal in consideration of a user's recent restaurant visit history and big data constructed for each restaurant.

In the present invention, in a restaurant recommendation service method using a restaurant recommendation service apparatus, the step of accumulating selection scores and rejection scores in response to a user's selection or rejection response for each restaurant in a restaurant list, and constructing big data; , When a restaurant recommendation service is requested from the user terminal, searching for a restaurant based on the location of the terminal to create an initial list, and if there is a restaurant with the user's visit history within the last N days in the initial list, it is accumulated in the corresponding restaurant. Modifying the initial list by determining whether to remove the list using the selected selected score and the elapsed time after the visit, and targeting each restaurant in the remaining list after the modification, to the selection score, rejection score, and business type accumulated in the corresponding restaurant. A restaurant recommendation service method comprising the step of determining a recommendation target by calculating a list recommendation probability using the corresponding category score, and providing a recommendation by exposing a final list composed of the finally determined restaurants to the user terminal. to provide.

In addition, the restaurant recommendation service method may further include receiving a selection response or a rejection response for a restaurant in the final list from the user terminal, and updating a selection score or a rejection score corresponding to the restaurant.

In addition, in the providing of the recommendation, priorities of restaurants in the final list are determined, and recommendations are provided to the user terminal by sequentially exposing them one by one from a higher ranking, but a selection response for the recommended restaurant is received from the user terminal. In one case, after the selection score corresponding to the restaurant is updated, the restaurant is set as a destination to perform navigation, and when a rejection response is received, the rejection score corresponding to the restaurant is updated, and then the next ranking restaurant can be recommended. .

In addition, in the providing of the recommendation, the priority may be determined based on at least one of a selection score corresponding to each restaurant in the final list, a rejection score, and a ratio of a selection score to a rejection score.

In addition, in the step of modifying the initial list, the following equation is applied in response to the i-th section in which the elapsed time after the visit of the visiting restaurant belongs among the L time sections obtained by dividing the last N days (N>1) into L equally. If the list removal probability of the visited restaurant is obtained, and the list removal probability is greater than 0, the visited restaurant may be removed from the initial list, but if it is less than 0, the visited restaurant may be maintained.

Here, i is the time interval index (i=1,2,...,L), and the smaller i is, the closer it is to the latest time.When i=1, the i-1 section removal probability is applied with a value of 100, and the default rate of decline Is the same for each time section, the selection score is the accumulated selection score of the corresponding restaurant, and the selection score weight is between 0 and 1, and the smaller i is, the larger the value.

In addition, in the step of modifying the initial list, if a predetermined restaurant that has been visited within the last 1 day exists in the initial list and the corresponding restaurant corresponds to a user's previously registered favorite restaurant, the probability of removing the list of the favorite restaurants However, the list of the favorite restaurants by applying the above equation corresponding to the i-th section in which the elapsed time after the visit of the favorite restaurant belongs among the L time sections obtained by dividing 24 hours corresponding to the 1st by L equally. You can calculate the probability of removal.

In addition, the final determination of the recommendation target includes checking the j-th section in which the cumulative rejection number of the corresponding restaurant belongs among a plurality of preset cumulative rejection number sections for each restaurant in the remaining list, and then using the following equation. Thus, the list recommendation probability of each restaurant is calculated, and the top K restaurants having a high list recommendation probability or restaurants having a recommendation probability greater than or equal to a reference value may be finally determined as a recommendation target.

Here, the cumulative number of rejections of the restaurant is determined corresponding to the cumulative number of rejection points, j is an index of the cumulative rejection number, the larger j is, the higher the cumulative rejection number, and the rejection weight of the j section decreases as j increases. The selection weight of the j section increases as j increases, the selection score is the accumulated selection score of the corresponding restaurant, and the category weight represents the weight applied to the business type of the corresponding restaurant.

In addition, the present invention provides a storage unit that accumulates a selection score and a rejection score in response to a user's selection or rejection response for each restaurant in the restaurant list to build big data, and when a restaurant recommendation service is requested from the user terminal. A generator that searches for restaurants based on the location of the terminal and generates an initial list, and if there is a restaurant in the initial list with a history of the user's visit within the last N days, the selection score accumulated in the restaurant and the elapsed time after the visit are calculated. The revising unit, which determines whether to remove the list by using the list, and modifies the initial list, and for each restaurant in the remaining list after the modification, a list using the selection score accumulated in the restaurant, rejection score, and category score corresponding to the business type. It provides a restaurant recommendation service apparatus including a determination unit that determines a recommendation target by calculating a recommendation probability, and a recommendation unit that exposes and recommends a final list consisting of the finally determined restaurants to the user terminal.

According to the present invention, it is possible to provide a recommendation to the user terminal by selecting an optimal restaurant list required for the user based on the location of the user terminal, the user's past visit history, and the accumulated past big data for each restaurant.

According to the present invention, after initial search for a restaurant list based on the location of the user terminal, the user's past visit history and the accumulated past big data for each restaurant are additionally considered, and the unnecessary list is removed, and then the recommended restaurant is selected based on the recommendation probability. It can be finally selected and recommended to users.

1 is a diagram showing a location-based restaurant recommendation service system according to an embodiment of the present invention.
2 is a diagram showing the configuration of the restaurant recommendation apparatus shown in FIG. 1.
3 is a diagram illustrating a restaurant recommendation method using FIG. 2.
4 is a diagram showing in more detail a method for recommending a restaurant according to an embodiment of the present invention.

Then, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art can easily implement the present invention.

1 is a diagram showing a location-based restaurant recommendation service system according to an embodiment of the present invention.

As shown in FIG. 1, the restaurant recommendation service apparatus 100 according to an embodiment of the present invention is connected to a wired/wireless network through a communication network with a plurality of user terminals 200 to transmit and receive mutual information, and to be located at the user terminal 200. Provides a restaurant recommendation service based on.

The restaurant recommendation service device 100 may correspond to a restaurant recommendation service server for providing a restaurant recommendation service to a user terminal 200 connected to an online platform implemented as a web or a mobile app. have.

The user terminal 200 may correspond to terminals such as a smart phone, a pad, a desktop, a PC, a notebook, and a tablet, and the restaurant recommendation service device 100 through a wired or wireless communication network. ) Can be connected to the network. The user terminal 200 may correspond to a user terminal registered or authenticated as a member through a web or an app. Of course, the user terminal 200 can join and authenticate the service platform by linking with accounts such as Google, Facebook, and KakaoTalk.

2 is a diagram showing the configuration of the restaurant recommendation apparatus shown in FIG. 1.

As shown in Figure 2, the restaurant recommendation apparatus 100 according to an embodiment of the present invention includes a storage unit 110, a generation unit 120, a modification unit 130, a determination unit 140, and a recommendation unit 150. Includes.

The storage unit 110 basically stores a list of restaurants held, basic information of each restaurant (location, address, contact information, provided menu, business hours, prices, etc.), personal information of each user, and subscription status, etc., and recommends restaurants. It stores and manages various data to provide services in an integrated manner.

In addition, the storage unit 110 accumulates a selection score and a rejection score in response to a user's selection or rejection response for each restaurant in the currently owned restaurant list to build big data. Here, each cumulative score is proportional to or corresponds to the cumulative number of responses for the corresponding response.

Such big data is accumulated in real time and updated in real time. Of course, the past big data constructed in this way is combined with the location of the user terminal and the user's past restaurant visit history to be used to recommend the best restaurant list to the user.

The restaurant's visit history may be previously stored in association with a user account within the recommended service app. For example, information about a restaurant visited by an actual user among restaurants recommended after the user searches in the app (eg, restaurant name, restaurant identification code, visit time, etc.) may be separately stored and managed in the storage unit 110 .

Of course, the restaurant visit information includes whether to receive a selection response to the previously recommended restaurant, the time of arrival of the recommended restaurant through the navigation function after the selection response, the time of entering the visit confirmation through the pop-up screen provided after arrival, and the credit card of the user terminal 200. It can be collected through various methods such as checking the usage history, and the information of the restaurant visited by the user can be directly input.

When receiving a request for a restaurant recommendation service from the user terminal 200, the generation unit 120 searches for and selects a restaurant based on the location of the terminal to generate an initial list.

If the user terminal 200 receives a restaurant recommendation or search button after executing the restaurant recommendation service app, the generation unit 120 searches for restaurants within a predetermined radius based on the location of the current user terminal and initializes the list. Create a list. Since the initial list created in this way corresponds to a kind of raw data, it is not directly exposed to the user terminal 200.

The revising unit 130 checks whether a restaurant with a user's visit history within the last N days exists in the initial list provided by the generation unit 120, and if there is, the previously accumulated selection scores and visits for the restaurant. After that, the initial list is modified by deciding whether to remove the list using the elapsed time.

If there are no restaurants in the initial list that have been visited within the last 5 days, all restaurants are candidates for exposure, but if there are restaurants that have been visited within the last 5 days, the list removal probability is calculated, and if the calculated probability is greater than 0, it is removed from the initial list. Modify the initial list.

The determination unit 140 determines a recommendation target by calculating a list recommendation probability using a selection score, a rejection score, and a category score corresponding to a business type for each of the restaurants in the revised remaining list. Here, of course, the recommended target refers to a list of restaurants to be exposed on the screen of the actual user terminal 200.

The recommendation unit 150 provides a recommendation by exposing a final list composed of the finally determined restaurants to the user terminal. In this case, the recommendation unit 150 may provide recommendations to the user terminal 200 by exposing restaurants in the final list one by one according to priority.

At this time, when the selection response is fed back from the user terminal 200 in response to the recommended restaurant, the storage unit 110 updates the cumulative selection score of the restaurant, and when the rejection response is fed back, the cumulative rejection of the restaurant Save updated score. Of course, for this purpose, a selection button and a rejection button for a user's positive and negative feedback on a corresponding restaurant may be individually provided on a screen for recommending a restaurant.

As a simple example, assume that the current cumulative selection score for restaurant "A" is 5 points, the cumulative rejection score is 10 points, and 1 point is added for each response. In this state, if the user presses the selection button when recommending the restaurant "A", the cumulative selection score is updated from 5 points to 6 points, and if the reject button is pressed, the cumulative rejection score is updated from 10 points to 11 points. do. Of course, the examples of the addition score and the cumulative score for each reaction in the embodiment of the present invention are not necessarily limited to those described above.

When receiving the selection response, the recommendation unit 150 may switch to a navigation screen to provide a route guidance service for the restaurant "A". Of course, when the rejection response is received, it is immediately switched from the recommendation screen for the restaurant "A" to the recommendation screen for the next-ranked restaurant "B", and the same operation as described above can be repeated.

Hereinafter, a restaurant recommendation service method according to an embodiment of the present invention will be described in more detail.

3 is a diagram illustrating a restaurant recommendation method using FIG. 2.

First, the storage unit 110 builds big data by accumulating the selection score and the rejection score in response to the user's selection or rejection response for each restaurant in the restaurant list (S310). The built big data is used to recommend restaurants to network-connected user terminals.

When receiving a request for a restaurant recommendation service from the user terminal 200, the generator 120 searches for a restaurant based on the current terminal location and generates an initial list (S320).

For example, a list of restaurants located within a set radius (eg, 0.3km, 1km, 2km, 5km) around the location of the current user terminal 200 is inquired, and an initial list is created from the inquired list. At this time, a search may be attempted from a preset initial radius (basic radius) (eg, 0.3 km), and if the number of searches does not satisfy a criterion or more, re-search may be performed while gradually expanding the search radius until it is satisfied.

Here, in the embodiment of the present invention, when there is a restaurant with a visitation history of a user within the last N days (eg, 5 days) in the initial list, the process of determining whether to delete the visited restaurant from the initial list is performed. It's rough.

In the case of a restaurant (general restaurant) with a history of recent visits by the user, since it is highly likely that the intention or frequency of revisiting is low, the initial restaurant list is modified by calculating the removal probability based on the elapsed time after the visit to the corresponding restaurant.

To this end, in the case of a restaurant with the user's visit history within the last N days (for example, 5 days) within the initial list, the revision government 130 accumulates the selection score accumulated in the restaurant (hereinafter, visited restaurant) and after the visit. Whether to remove the list is determined using the elapsed time, and the initial list is modified based on this (S330).

Here, of course, if the restaurant that has been visited within the last N days is not on the list, step S340 may be immediately performed. The following embodiments of the present invention focus on a case where a restaurant visited within the last N days exists in the initial list. Here, N has a value of 2 or more, which is greater than 1, and for convenience of description, it is assumed that N=5.

Specifically, in step S330, the revision unit 130 calculates Equation 1 below in response to the i-th section in which the elapsed time after the visit of the visiting restaurant belongs among the L time sections obtained by dividing the last N days (e.g., 5 days) into L equally. By applying, the probability of removing the list of restaurants visited (the probability of removing the i section) is obtained. At this time, if the calculated list removal probability is greater than 0, the visited restaurant is removed from the initial list, but if it is less than 0, it is maintained in the list.

Here, i is a time interval index (i=1,2,...,L), and the smaller i is, the closer it is to the latest time, and when i=1, a value of 100 is applied to the i-1 section removal probability.

In addition, in Equation 1, the basic deterioration rate is the same for each time interval, the selection score is the accumulated selection score of the restaurant, and the selection score weight is between 0 and 1, and the smaller i is, the larger the value.

If, if N=5, L=6, six time intervals are determined by dividing the five days' mealtime into six. At this time, each time section may be divided equally, but may be divided into a form in which the width of the section becomes wider as the past age increases, as shown in Table 1 to be described later.

It can be seen from Equation 1 that in order to obtain the removal probability of the visited restaurant belonging to the predetermined-th section, it is necessary to accumulate and obtain the removal probability up to the previous section.

For better understanding, the formula for each section in Equation 1 may be summarized as in Equation 2 below.

Equation 2 assumes the case of N=5 and L=6, and a separate formula is not applied to the removal probability of section 7 which is a section greater than i=6. The larger the i, the farther the past time, so in other words, if any restaurant on the initial list is a restaurant that has been visited in the past (for example, a restaurant that has visited within 6 days), rather than within the last 5 days, the list is subject to removal. It does not mean that it is kept on the initial list regardless of the formula.

Table 1 shows an example of calculating a list removal probability for a visited restaurant for a restaurant visited by the user within the last 5 days (eg, restaurant "C") among the restaurants on the initial list generated by the generator 120.

At this time, it can be seen that the time for the past 5 days (total 120 hours) from the present is divided into 6 time intervals (N=5, L=6).

It is assumed that the selection score accumulated so far in the visiting restaurant (restaurant "C") = 5 points, and the default rate of decline = 10% is preset. The selection score weights exist from 0.1 to 0.6, and the closer to the latest time (the smaller i) is set to increase by 0.1.

The selection score reduction rate (%) in Table 1 corresponds to the product of the parenthesis part of Equation 1, that is, the selection score and the selection score weight, and the total reduction rate (%) in Table 1 is the bracket part of Equation 1, that is, the basic decrease rate ( %) and the rate of decline in the selection score (%).

If the elapsed time after the user's visit to the restaurant "C" = 20 hours, since 20 hours belongs to the second interval (i = 2), the removal probability of the two intervals i = 2 in Equation 1 may be obtained.

To this end, first, by substituting i=1 in Equation 1, the probability of removing one section is obtained. In this case, when i = 1, it is obtained using the default value of 100 as the i-1 section removal probability value present in the right hand side of Equation 1. As a result, it can be seen that 87% (=100-[10 + (0.6 × 5)]) was derived as shown in Table 1 for the probability of removing section 1.

Next, substituting i=2 to find the removal probability of section 2, and substituting the removal probability of section 1 obtained earlier with the i-1 section removal probability value is 75% (=87-[10 + (0.5 × 5)]) Becomes.

From this, it can be seen that the probability decreases by 10% by default starting from 100% at the beginning for each section, and at this time, the probability decreases by additionally reflecting the selection score decrease rate.

As a result, the calculated list removal probability for the restaurant "C" visited within the last 5 days is 75%, and since this is a value greater than 0, the initial list is modified by removing the restaurant "C" from the initial list. In the case of Table 1, since the probability is 0 or higher even for section 6, it can be seen that the restaurant "C" is subject to list removal for 120 hours after the visit. However, after 120 hours, it is not included in the list removal.

Of course, if 0 or less is derived as a result of calculating the removal probability of section 2 of Equation 1 for the visited restaurant “C”, the restaurant “C” is immediately removed from the initial list. For example, if the elapsed time after visit of the user of the restaurant "C" = 20 hours (i = 2), the basic decline rate = 25%, and the accumulated selection score of the restaurant "C" = 50 points, i in Equation 1 Since the probability of removing two sections with =2 is less than 0, restaurant "C" is removed from the initial list.

This process is performed by applying to all restaurants that the user has confirmed to have visited within the last 5 days among the restaurants in the initial list.

In this way, in the case of a general restaurant visited by the user within the last 5 days, the probability of removing the list for a maximum of 120 hours is calculated.

However, among restaurants with a history of recent visits, since there is a possibility that the user's intention to revisit and the frequency of visits is high, the probability of removing the list for a maximum of 24 hours is calculated as follows.

That is, if there is a restaurant that has been visited within the last 1 day among the restaurants in the initial list and the corresponding restaurant corresponds to the user's previously registered favorite restaurant, the revising unit 130 calculates the probability of removing the list for the corresponding favorite restaurant. , By applying N=1 instead of N=5, the probability of list removal is calculated by dividing (subdivide) L equally for 1 day, that is, 24 hours a day.

More specifically, the revision unit 130 applies Equation 1 in response to the i-th section in which the elapsed time after the visit of the favorite restaurant belongs among L time sections obtained by dividing 24 hours corresponding to one day into L equals. Calculate the probability of removing the list of restaurants.

Table 2 below shows an example of a list removal probability calculation for favorite restaurants visited within the last 1 day.

In the case of favorite restaurants that have been visited within the last 1 day, unlike general restaurants that are not registered as favorites, it can be seen that the intention to revisit or repeat visits within a short time is high, so the probability is calculated by subdividing 24 hours into 6 sections as follows. .

Table 2 shows a case where the selection score accumulated so far in a favorite restaurant (eg, restaurant “D”) = 10 points.

In this way, in the case of restaurants that the user has visited within the last 1 day and registered as favorites, the probability of removing the list for up to 24 hours is calculated. In the case of Table 2, since the probability is 0 or higher even for section 6, it can be seen that the restaurant "D" is subject to list removal for 24 hours after the visit.

In addition, according to this, in the case of favorite restaurants, after 24 hours have passed, they are not included in the list removal target. Therefore, if any restaurant on the initial list is a favorite restaurant (for example, a favorite restaurant visited within 2 days) 24 hours after the user's visit, the formula is not applied because it is not subject to the list removal. maintain.

In this way, after modifying the initial list, the determination unit 140 targets each restaurant in the remaining list after the modification, and recommends the list by using the selection score, rejection score, and category score corresponding to the business type accumulated in the corresponding restaurant. A recommendation target is finally determined by calculating the probability (S340).

Specifically, for each restaurant in the remaining list, the determination unit 140 checks the j-th section in which the cumulative rejection number of the corresponding restaurant belongs among a plurality of predetermined cumulative rejection number sections, and then, by using Equation 3 below, Calculate the probability of recommending a list of.

Here, of course, the determination unit 140 may finally determine the top K restaurants having a high list recommendation probability or restaurants having a recommendation probability greater than or equal to a reference value as recommendation targets.

Here, j is the cumulative rejection number section index (j=1,2,...,M), and the larger j represents a section with a higher cumulative rejection number. At this time, it is obvious that the cumulative number of rejections of the restaurant is determined corresponding to the cumulative number of rejection points. That is, if the rejection score is 1 point per accumulation, 1 point is accumulated corresponding to one rejection response, and if the cumulative number of rejection scores is 10, the cumulative number of geometries is 10 and a total of 10 points are accumulated.

In Equation 3, the representative value of the j section can be determined by averaging the random number values of the number of settings selected within the j section, and the rejection weight of the j section decreases as j increases, and the selection weight of the j section increases as j increases. Is set. And, as described above, the selection score represents the accumulated selection score of the corresponding restaurant. The category weight represents the weight applied to the business type of the restaurant. The business type of a restaurant can be varied, such as buffet, Korean, Chinese, dessert, cafe, and bakery, and weights for each business type can be preset.

Tables 3 and 4 show specific examples of obtaining the recommendation probability of Equation 3.

First, Table 3 shows the process of obtaining the [1-A] part described in the first line of Equation 3. Here, A = (representative value of section j × rejection weight of section j).

Table 3 is expressed in percentage units by multiplying each of the A value and the [1-A] value by 100 for convenience of explanation.

The cumulative rejection number section was divided into a total of 14 sections (j=1~14) according to the cumulative rejection number. In Table 3, it can be seen that the representative value of each section is determined by rounding up after averaging the five random number values selected within the section. Here, the five random number values selected within a section may be determined according to a minimum/maximum distribution value of the cumulative rejection number within a corresponding section.

For example, if the cumulative rejection number for a predetermined restaurant (eg, restaurant "D") in the remaining list is 8, it is calculated according to the calculation formula of section 3 (cumulative rejection number: 6-10 times) with j=3. It can be seen that the random number values are determined from 6 to 10.

At this time, if 8, which is the average (rounded) of the random number, is multiplied by 0.015, which is the rejection weight of section 3, it becomes 0.12 (A=0.12), and subtracting A from 1 yields 0.88 (1-A=0.88). That is, in section 3 where j=3, [1-A]=0.88, and multiplying this by 100 results in 88%.

Table 4 below shows the process of obtaining the [B] part (additional recommendation probability) described in the second line of Equation 3. Here, B = [(selection score + category score) × selection weight for j section].

As shown in Table 4, it is assumed that the cumulative selection score for the restaurant "D" = 5 points, and the category score of the corresponding restaurant industry = 5 points. Since the selection weight in section 3 where j = 3 in Table 4 is 0.002, 15, which is the sum of the selection score and the category score, is multiplied by the selection weight of 0.002 to obtain 0.03. If you multiply this by 0.03 by 100, you can see that it becomes 3%.

That is, when the 88% value obtained in Table 3 is added to the 3% value obtained in Table 4, the total recommendation probability of 91% is derived. Of course, as can be seen from Equation 3, the total recommendation probability does not exceed 100%.

In this way, the determination unit 140 obtains a recommendation probability for all restaurants in the remaining list, and finally determines a recommendation target through this.

Next, the recommendation unit 150 provides a recommendation by exposing a final list composed of the finally determined restaurants to the user terminal (S350).

Here, the recommendation unit 150 may determine the priorities of restaurants in the final list and provide recommendations by sequentially exposing them to the user terminal one by one from the higher order.

Here, the priority may simply be determined in the order of the highest probability of recommending the list, but at least one of the cumulative selection scores corresponding to each of the restaurants in the final list, the cumulative rejection score, and the cumulative selection score compared to the accumulated rejection score. You can also prioritize based on one.

In addition, the recommendation unit 150 recommends restaurants from the highest ranking, and receives the selection response or rejection response for the recommended restaurant from the user terminal 200 to update the selection score or rejection score corresponding to the restaurant again. Can be (S360). That is, the score corresponding to the current user's selection or rejection reaction is added to the accumulated selection or rejection score and updated.

In addition, when the recommendation unit 150 receives the selection response for the restaurant recommended to the user from the user terminal, updates the selection score corresponding to the restaurant, sets the restaurant as a destination, executes a navigation function, and links with a map. And follow the directions.

However, when the rejection response is received, the rejection score corresponding to the corresponding restaurant may be updated, and then the next ranked restaurant may be recommended again. Here, the navigation function may be an external app that works with the recommended service app or may be embedded in the recommended service app.

As described above, the embodiment of the present invention basically searches for a restaurant list based on the location of the user terminal, and then considers the user's recent restaurant visit history, favorite history, and past big data built for each restaurant. After removing unnecessary lists first, a list with a high recommendation probability among the remaining lists can be finally selected and recommended to the user.

As described above, in the present embodiment, a list of restaurants to be recommended may be finally selected and recommended to the user terminal in consideration of the location of the user terminal, the user's past visit history, and the accumulated past big data for each restaurant.

In addition, according to an embodiment of the present invention, a desired search condition (category, business type, etc.) may be first input from a user before a restaurant list search, and restaurants in which a large number of selection scores are accumulated are organized into a higher list so that they are frequently recommended and exposed. It can also be adjusted. Of course, restaurants with a large number of rejection points may be organized into a sub-list so that fewer recommendations and exposures may be made.

In addition, it is also possible to recommend a restaurant by analyzing the user's preferences and dispositions of non-preferred stores by using detailed information of the restaurant. For example, by judging the business type of the restaurant frequently selected by the user, recommending a restaurant with the same business type in the future. And it is possible to further increase the exposure probability. In the case of rejection scores, the same principle is applicable.

4 is a diagram showing in more detail a method for recommending a restaurant according to an embodiment of the present invention.

First, when a restaurant inquiry is requested from a user through a restaurant recommendation app executed on the user terminal 200, the restaurant recommendation device 100 (server) checks the current location of the user terminal 200, Restaurants existing within a set radius (distance) from the current location are searched and inquired (S410).

Thereafter, an initial list consisting of the inquired restaurants is generated (S402). Of course, the initial list is not exposed to the user terminal 200.

Then, the restaurant recommendation apparatus 100 determines whether there is a restaurant in the initial list that the user has visited within the last 5 days (S403).

If there is, the probability of removing the list for the visited restaurant is calculated using Equation 1 (S404). In this case, N=5 in Equation 1 is applied and the list removal probability for the visited restaurant is calculated using the section information of Table 1 in which a total of 120 hours (5 days) is divided into several time sections.

When calculating the list removal probability, as shown in Equation 1, the selection score accumulated to the present restaurant, the weight applied thereto, the elapsed time after the visit, and the basic deterioration rate are considered, respectively.

Then, the initial list is modified based on the calculated probability (S405). If the calculated probability is greater than 0, the restaurant to be visited is determined as an unnecessary list and deleted from the initial list. If the calculated probability is equal to or negative, it is maintained in the initial list.

In addition, if there is no restaurant visited within the last 5 days in step S403, it is determined whether there is a favorite restaurant visited within the last 1 day (S406). Here, of course, if there are no restaurants visited within the last 5 days in the initial list and no favorite restaurants visited within the last 1 day, step S411 may be immediately performed.

As a result of the determination in step S406, if there is a favorite restaurant that has been visited within the last 1 day, the probability of removing the list of the favorite restaurants is calculated using Equation 1 (S404). However, in this case, N=1 in Equation 1 is applied and the probability of removing the list of favorite restaurants is calculated using the section information of Table 2 in which a total of 24 hours (1 day) is divided into several time sections (S404).

This process is repeatedly applied to all visited restaurants in the initial list to modify and update the initial list (S406). The modified initial list is referred to as the residual list.

Thereafter, the restaurant recommendation apparatus 100 calculates each list recommendation probability by applying Equation 3 to the restaurants in the remaining list (S411).

When calculating the list recommendation probability, the cumulative number of rejections for the corresponding restaurant, the representative value of the corresponding section to which the cumulative rejection number belongs, the rejection weight of the section, the cumulative selection score and category score of the corresponding restaurant, and the selection weight of the section are considered.

Thereafter, according to the calculated list recommendation probability, a recommendation (exposure) target list to be exposed and provided on the screen of the user terminal 200 is determined and added to the final list (S412). In addition, restaurants are recommended and provided in order from the highest ranking according to the recommendation order determined in the final list (S413).

A selection or rejection response (positive or negative feedback) is received from the user for the restaurant recommended on the screen of the user terminal 200 (S414). When the selection response is received, the selection score for the corresponding restaurant is increased by 1 and the corresponding restaurant is set as a destination to perform directions (S415, S416). When the rejection response is received, the rejection score is increased by 1 for the corresponding restaurant, and the next ranked restaurant is exposed on the screen through screen flipping to provide recommendations (S417).

According to the present invention as described above, after initial search for the restaurant list based on the location of the user terminal, the user's past visit history and the accumulated past big data for each restaurant are additionally considered, and the unnecessary list is removed, and then based on the recommendation probability. It is possible to provide recommendations to users by final selection of restaurants to be recommended.

The present invention has been described with reference to the embodiments shown in the drawings, but these are merely exemplary, and those of ordinary skill in the art will appreciate that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

100: restaurant recommendation service device 110: storage unit
120: generation unit 130: revision government
140: decision unit 150: recommendation unit
200: user terminal

Claims (14)

In the restaurant recommendation service method using a restaurant recommendation service device,
Constructing big data by accumulating a selection score and a rejection score in response to a user's selection or rejection response for each restaurant in the restaurant list being held;
Upon receiving a request for a restaurant recommendation service from a user terminal, searching for a restaurant based on the location of the terminal and generating an initial list;
Modifying the initial list by determining whether to remove the list using the selection score accumulated in the restaurant and the elapsed time after the visit, if there is a restaurant in the initial list with a history of the user's visit within the last N days;
Determining a recommendation target by calculating a list recommendation probability using a selection score, a rejection score, and a category score corresponding to a business type for each of the restaurants in the revised remaining list; And
And providing a recommendation by exposing a final list of the finally determined restaurants to the user terminal. The method according to claim 1,
Receiving a selection response or a rejection response to the restaurant in the final list from the user terminal, and updating the selection score or rejection score corresponding to the restaurant. The method according to claim 1,
The step of providing the recommendation,
Priority of restaurants in the final list is determined, and recommendations are provided to the user terminal by sequentially exposing them one by one from the highest priority,
When the selection response for the recommended restaurant is received from the user terminal, the selection score corresponding to the restaurant is updated, and then the restaurant is set as a destination to execute the navigation, and when a rejection response is received, the corresponding restaurant is A restaurant recommendation service method that recommends the next-ranked restaurant after updating the rejection score. The method of claim 3,
The step of providing the recommendation,
A restaurant recommendation service method for determining the priority based on at least one of a selection score corresponding to each of the restaurants in the final list, a rejection score, and a ratio of a selection score to a rejection score. The method according to claim 1,
The step of modifying the initial list,
Of the L time intervals obtained by dividing the last N days (N>1) into L equally, the following equation is applied to the i-th interval in which the elapsed time after the visit of the visited restaurant belongs to obtain the probability of removing the list of the visited restaurants, and the list If the removal probability is greater than 0, the restaurant is removed from the initial list, but if it is less than O, the restaurant recommendation service method:

Here, i is the time interval index (i=1,2,...,L), and the smaller i is, the closer it is to the latest time.When i=1, the i-1 section removal probability is applied with a value of 100, and the default rate of decline Is the same for each time interval, the selection score is the accumulated selection score of the corresponding restaurant, and the selection score weight is between 0 and 1, and the smaller i is, the larger the value. The method of claim 5,
The step of modifying the initial list,
If a predetermined restaurant that has been visited within the last 1 day exists in the initial list and the corresponding restaurant corresponds to the user's previously registered favorite restaurant, the probability of removing the list of the favorite restaurants is calculated,
The probability of removing the list of the favorite restaurants is calculated by applying the equation in response to the i-th section in which the elapsed time after the visit of the favorite restaurant belongs among L time sections obtained by dividing the 24 hours corresponding to the 1st by L. Recommended restaurant service method. The method according to claim 1,
The step of finally determining the recommendation target,
For each restaurant in the remaining list, after checking the j-th section in which the cumulative rejection number of the corresponding restaurant belongs among a plurality of predetermined cumulative rejection section sections, the list recommendation probability of each restaurant is calculated using the following equation,
A restaurant recommendation service method of finally determining the top K restaurants with a high probability of recommending the list or restaurants having a recommendation probability of more than a reference value as a recommendation target:

Here, the cumulative number of rejections of the restaurant is determined corresponding to the cumulative number of rejection points, j is an index of the cumulative rejection number, the larger j is, the higher the cumulative rejection number, and the rejection weight of the j section decreases as j increases. The selection weight of the j section increases as j increases, the selection score is the accumulated selection score of the corresponding restaurant, and the category weight represents the weight applied to the business type of the corresponding restaurant. A storage unit configured to build big data by accumulating selection scores and rejection scores in response to a user's selection or rejection response for each restaurant in the restaurant list held;
A generator for generating an initial list by searching for a restaurant based on a terminal location when a restaurant recommendation service is requested from a user terminal;
If a restaurant with a user's visit history within the last N days is present in the initial list, a modification unit modifies the initial list by determining whether to remove the list using the selection score accumulated in the restaurant and the elapsed time after the visit;
A determination unit for final determination of a recommendation target by calculating a list recommendation probability using a selection score, a rejection score, and a category score corresponding to a business type for each of the restaurants in the revised remaining list; And
A restaurant recommendation service apparatus comprising a recommendation unit for exposing and recommending a final list of the finally determined restaurants to the user terminal. The method of claim 8,
The storage unit,
A restaurant recommendation service device that receives a selection response or a rejection response for a restaurant in the final list from the user terminal and updates a selection score or a rejection score corresponding to the restaurant. The method of claim 8,
The recommendation unit,
Priority of restaurants in the final list is determined, and recommendations are provided to the user terminal by sequentially exposing them one by one from the highest priority,
When the selection response for the recommended restaurant is received from the user terminal, the selection score corresponding to the restaurant is updated, and then the restaurant is set as a destination to execute the navigation, and when a rejection response is received, the corresponding restaurant is A restaurant recommendation service device that recommends the next-ranked restaurant after updating the rejection score. The method of claim 10,
The recommendation unit,
A restaurant recommendation service device that determines the priority based on at least one of a selection score corresponding to each of the restaurants in the final list, a rejection score, and a ratio of a selection score to a rejection score. The method of claim 8,
The correction unit,
Of the L time intervals obtained by dividing the last N days (N>1) into L, the following equation is applied to the i-th interval in which the elapsed time after the visit of the visited restaurant belongs to obtain the probability of removing the list of the visited restaurants, and the list A restaurant recommendation service device that removes visited restaurants from the initial list if the removal probability is greater than 0, but maintains if it is less than O:

Here, i is the time interval index (i=1,2,...,L), and the smaller i is, the closer it is to the latest time.When i=1, the i-1 section removal probability is applied with a value of 100, and the default rate of decline Is the same for each time interval, the selection score is the accumulated selection score of the corresponding restaurant, and the selection score weight is between 0 and 1, and the smaller i is, the larger the value. The method of claim 12,
The correction unit,
If a predetermined restaurant that has been visited within the last 1 day exists in the initial list and the corresponding restaurant corresponds to the user's previously registered favorite restaurant, the probability of removing the list of the favorite restaurants is calculated,
The probability of removing the list of the favorite restaurants is calculated by applying the equation in response to the i-th section in which the elapsed time after the visit of the favorite restaurant belongs among L time sections obtained by dividing the 24 hours corresponding to the 1st by L. Restaurant recommendation service device. The method of claim 8,
The determination unit,
For each restaurant in the remaining list, after checking the j-th section in which the cumulative rejection number of the corresponding restaurant belongs among a plurality of predetermined cumulative rejection section sections, the list recommendation probability of each restaurant is calculated using the following equation,
A restaurant recommendation service device that finally determines the top K restaurants having a high probability of recommending the list or restaurants having a recommendation probability of greater than or equal to a reference value as a recommendation target:

Here, the cumulative number of rejections of the restaurant is determined corresponding to the cumulative number of rejection points, j is an index of the cumulative rejection number, the larger j is, the higher the cumulative rejection number, and the rejection weight of the j section decreases as j increases. The selection weight of the j section increases as j increases, the selection score is the accumulated selection score of the corresponding restaurant, and the category weight represents the weight applied to the business type of the corresponding restaurant.

Images