KR20230015555A - Food Recommendation Apparatus and Method - Google Patents

Abstract

The present invention provides a food recommendation device and method. The food recommendation device includes: an information collection unit that collects and obtains food intake data including user meal information, which is information about food consumed during a meal during a previous predetermined period of a user and information on food eaten by a large number of people and the ingredients of each meal; a relationship analysis unit that obtains a food-ingredient bipartite graph for analyzing the relationship between food and ingredients based on the food intake data and a food-meal bipartite graph for analyzing the relationship between food and meals; a food selection unit that searches for related food in the food-ingredient bipartite graph based on food obtained from the user meal information to select a predetermined number of representative food with high correlation, searches for related food in a food-meal bipartite graph based on each representative food to select a predetermined number of combination food with high correlation corresponding to each representative food, and obtains a food set consisting of the representative food and the combination food; and a food recommendation unit that optimizes the serving amount of each food based on the correlation of the food included in the food set and selects and recommends a predetermined number of food sets with a high overall correlation calculated based on the optimal serving amount of each food. Therefore, food combinations can be recommended to ensure an appropriate amount of nutrients based on individual tastes and health conditions.

Description

Food Recommendation Apparatus and Method {Food Recommendation Apparatus and Method}

The present invention relates to a food recommendation apparatus and method, and more particularly to a food recommendation apparatus and method capable of recommending optimally combined foods based on personal fitness and nutrition.

As the quality of life has improved, interest in health has increased, and among them, social interest in food is greatly increasing. In particular, in modern times, a wide variety of foods are being supplied compared to before, and personal nutritional imbalance is gradually becoming more serious due to people's biased diet. In addition, as the number of people with various adult diseases increases with aging, there is a problem that it is difficult for individuals to recognize the ingredients contained in each food or the nutritional components according to the amount of intake, even though there are many foods to avoid. Moreover, in the case of Korea, since various foods must be combined at each meal due to the side dish culture, there is a problem in that each individual's diet selection is more complicated and difficult.

Korean Registered Patent No. 10-2132975 (registered on July 6, 2020)

An object of the present invention is to provide a food recommendation device and method capable of recommending a food combination so that an appropriate amount of nutritional components can be supplemented.

Another object of the present invention is to provide a food recommendation apparatus and method capable of recommending a food combination according to personal tastes and health conditions.

In order to achieve the above object, a food recommendation device according to an embodiment of the present invention provides user meal information, which is information about food consumed by a user during a previous predetermined period of time, food consumed by a plurality of people, and each food. Information collection unit for collecting and obtaining food intake data containing information on the ingredients of the; a relationship analysis unit that acquires a food-ingredient bipartite graph for analyzing a relationship between food and ingredients and a food-meal bipartite graph for analyzing a relationship between food and meal based on the food intake data; Based on the food obtained from the user's meal information, search for related foods in the food-ingredient bipartite graph to select a predetermined number of representative foods with high correlation, and based on each of the representative foods, in the food-meal bipartite graph. a food selection unit for obtaining a food set consisting of the representative food and the combination food by searching for related foods and selecting a predetermined number of combination foods having a high correlation corresponding to each of the representative foods; And a food recommendation unit that optimizes the serving amount of each food based on the correlation of the foods included in the food set, and selects and recommends a predetermined number of food sets having a high total correlation calculated based on the optimal serving amount of each food. do.

The food selection unit searches for related foods using common ingredients in the food-ingredient dichotomy graph based on the food acquired from the user meal information, and determines the relationship between the standard food and the related food as a proportion of ingredients included in common. a representative food selection unit for selecting a representative food from a predetermined number of related foods having a high modified AAI by calculating a modified Adamic-Adar Index (hereinafter referred to as AAI) in consideration of; And based on each of the representative foods, search for related foods that are foods served at the same meal in the food-meal dichotomy graph, calculate the association between each representative food and the related food by AAI, and calculate a predetermined number of associations with a high AAI. It may include a combination food selection unit that selects a food combination and obtains a food set composed of the representative food and the combination food.

The relationship analyzer classifies a plurality of foods and a plurality of food ingredients in the food intake data into two different categories, connects ingredients included in each of the plurality of foods, and obtains the food-ingredient bipartite graph. a graph acquisition unit; and a food-meal relationship graph obtaining unit configured to classify meals and foods into two different categories from the food intake data and connect the foods consumed at each meal to obtain the food-meal dichotomy graph.

The food recommendation unit optimizes the serving amount of each food so that the total correlation calculated as the sum of the correlations of the foods included in the food set is maximized, and applies the optimal serving amount for each food calculated for each predetermined nutritional component to the optimal serving amount for each food. The nutritional component amount of each food set is calculated according to , and a predetermined number of food sets having the highest total correlation may be selected and recommended while satisfying the predetermined nutritional component-specific condition.

The food recommendation unit calculates the total correlation to which the condition relaxation penalty, which is calculated according to the difference from the condition for each predetermined nutritional component, is applied, optimizes the serving amount of each food so that the total correlation to which the condition relaxation penalty is applied is maximized, and optimizes the serving amount It is possible to select and recommend a pre-specified number of food sets having the highest overall correlation according to .

In order to achieve the above object, a nutrition-based food recommendation method according to another embodiment of the present invention provides user meal information, which is information about food consumed during a meal for a predetermined period of time by a user, food consumed during a meal of a plurality of people, and Collecting and obtaining food intake data including information on ingredients of each food; Obtaining a bipartite graph including a food-ingredient bipartite graph for analyzing a relationship between food and ingredients and a food-meal bipartite graph for analyzing a relationship between food and meal based on the food intake data; Based on the food obtained from the user's meal information, search for related foods in the food-ingredient bipartite graph to select a predetermined number of representative foods with high correlation, and based on each of the representative foods, in the food-meal bipartite graph. obtaining a food set consisting of the representative food and the combination food by searching for related foods and selecting a predetermined number of combination foods having a high correlation corresponding to each of the representative foods; And optimizing the serving amount of each food based on the correlation of the foods included in the food set, and selecting and recommending a predetermined number of food sets having a high total correlation calculated based on the optimal serving amount of each food. .

Accordingly, the food recommendation apparatus and method according to an embodiment of the present invention may recommend a food combination so that an appropriate amount of nutritional components according to personal tastes and health conditions may be supplemented.

1 shows a nutrition-based food recommendation system according to an embodiment of the present invention.
FIG. 2 shows an example of a schematic structure of the food recommendation device of FIG. 1 .
FIG. 3 shows an example of a bipartite graph obtained by the graph acquisition unit of FIG. 2 .
Figure 4 shows an example of food classification and unit amount set according to the food classification unit of FIG.
5 shows an example of personal taste setting of the food classification unit of FIG. 2 .
6 shows a nutrition-based food recommendation method according to an embodiment of the present invention.

In order to fully understand the present invention and its operational advantages and objectives achieved by the practice of the present invention, reference should be made to the accompanying drawings illustrating preferred embodiments of the present invention and the contents described in the accompanying drawings.

Hereinafter, the present invention will be described in detail by describing preferred embodiments of the present invention with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the described embodiments. And, in order to clearly describe the present invention, parts irrelevant to the description are omitted, and the same reference numerals in the drawings indicate the same members.

Throughout the specification, when a part "includes" a certain component, it means that it may further include other components, not excluding other components unless otherwise stated. In addition, terms such as "... unit", "... unit", "module", and "block" described in the specification mean a unit that processes at least one function or operation, which is hardware, software, or hardware. And it can be implemented as a combination of software.

1 shows a nutrition-based food recommendation system according to an embodiment of the present invention.

Referring to FIG. 1 , the nutrition-based food recommendation system according to the present embodiment may include a plurality of user terminals 10, medical terminals 20, a database server 30, and a food recommendation server 40.

The plurality of user terminals 10 may be implemented not only in portable user terminals such as smartphones, but also in stationary user terminals such as PCs and laptops, and food recommendation apps are installed, respectively.

The user may input the type and amount of each food he/she has consumed into the user terminal 10 in which the food recommendation app is installed, and the user terminal 10 may transmit the input type and amount of food to the food recommendation server 40 through the network. forward to In addition, the user terminal 10 may transmit preferred and non-preferred food information input by the user according to the user's personal taste, or prohibited food and recommended food information according to disease to the food recommendation server 40 .

The user terminal 10 may transmit a food recommendation request to the food recommendation server 40 according to the user's command, and display the recommended food to the user when the recommended food is transmitted from the food recommendation server 40 .

The medical terminal 20 is a terminal used in medical institutions such as hospitals, and is preferably used under user approval. Similar to the user terminal 10, the medical terminal 20 may transmit prohibited food, recommended food information, etc. according to the user's disease to the food recommendation server 40, and in a special case such as a state in which the user is hospitalized The type and amount of food provided to the user may be transmitted to the food recommendation server 40 . That is, the medical terminal 20 may be used as a means to replace or assist the user terminal 10 when a medical institution provides food information to be consumed or prohibited according to the user's health condition.

The database server 30 stores food intake data and transmits the stored food intake data according to a request of the food recommendation server 40 . In this embodiment, a meal means an action by which a user consumes at least one food at a time, and it may include snacks as well as one meal such as breakfast, lunch, and dinner. In addition, food refers to all foods that can be individually classified and provided, such as steamed rice, various kinds of grilled meat, stew, kimchi, hamburger, and the like.

In some cases, the database server 30 may receive and store the type and amount of food consumed by each user from a plurality of user terminals 10 or medical terminals.

In this embodiment, for example, it is assumed that the food intake survey (24-hour recall survey by individual) is used as food intake data among the public health and nutrition survey details provided by the Korea Centers for Disease Control and Prevention. The food intake survey (individual 24-hour recall survey) item of the National Health and Nutrition Examination Survey is the data recorded by surveying the meal details of the subjects designated in advance for 24 hours. Including the weight of ingredients used in For example, in the case of the 2018 data, the experiment was conducted on 7064 people, and 1783 types of food and 3079 types of ingredients included in the food were included. In addition, the food intake survey items include nutritional component information for each food according to the standard food composition table announced by the Rural Development Administration.

The food recommendation server 40 is the food recommendation device of the present embodiment, and analyzes food intake data to analyze the relationship between meals consumed by a large number of people and food and between food and food ingredients. In addition, meals and food information consumed by the user are obtained and stored from each of the plurality of user terminals 10, and when a food recommendation request is transmitted from the user terminal 10, the analyzed meal and food and relationship information between the food and food ingredients are stored. Based on this, suitable food combinations are recommended for the user's meal.

At this time, the food recommendation server 40 according to the present embodiment recommends food by referring to food combination patterns of a plurality of people, but may recommend food combinations by considering nutritional components and amounts suitable for the user. In addition, food combinations and amounts are obtained and recommended so that an appropriate amount of nutritional components can be supplemented according to user conditions and health conditions set by each user.

Also, the food recommendation server 40 may be configured to directly acquire food intake data. In this case, the database server 30 may be integrated into the food recommendation server 40 and may be omitted.

FIG. 2 shows an example of a schematic structure of the food recommendation device of FIG. 1 .

Referring to FIG. 2 , the food recommendation device 40 according to the present embodiment includes an information collection unit 100, a relationship analysis unit 200, a food selection unit 300, a food classification unit 400, a condition setting unit ( 500) and a food recommendation unit 600.

The information collection unit 100 collects various types of information for recommending food to users. The information collection unit 100 may include a user meal information acquisition unit 110 and a food intake information collection unit 120 . The user meal information acquisition unit 110 collects user meal information including the type and amount of food consumed by each user from each of the plurality of user terminals 10 . The user meal information acquisition unit 110 is a type and amount of food consumed by each of the plurality of users from the plurality of user terminals 10 or the medical terminal 20 during a recent pre-determined period (for example, one week). Can be obtained as user meal information. In addition, the food intake information collecting unit 120 collects and obtains the food intake data.

It is very rare for people to consume only one food at a meal, and usually consumes a combination of 4 to 5 foods. Among the foods consumed together in a meal as described above, one or two main types of main foods and the remaining sub foods may be classified as core foods, and additional foods such as fruits or beverages may also exist depending on the meal. Food categories such as main food, sub food, and additional food may be pre-specified and acquired together with the collected food intake data.

As described above, the food intake data includes the ingredients included in each food and the corresponding nutritional component information, and the amount and nutritional components of the ingredients included here are set based on one serving. Accordingly, the unit amount for one serving may also be pre-specified in the food intake data. That is, the food intake data may include an average amount of the corresponding food consumed by one person in one meal and may be acquired together.

Here, it is assumed that the database server 30 is included in the food recommendation device 40, and the information collection unit 100 collects user meal information and food intake data by replacing the database server 30, but the database server If 30 is configured separately, the information collection unit 100 may be omitted from the food recommendation device 40 .

The relationship analyzer 200 analyzes relationships between the user and food, between food and meal, and between food and food ingredients, based on user meal information and food intake data collected by the information collection unit 100 . The relationship analysis unit 200 may analyze the relationship between the collected information according to various data analysis techniques, but here, as an example, a bipartite graph is used between the user and food, food and meal, and food and food ingredients. analyze the relationship between

FIG. 3 shows an example of a bipartite graph obtained by the graph acquisition unit of FIG. 2 .

As shown in FIG. 3, the bipartite graph includes data ((u ₁ , u ₂ , …, u _i ), (v ₁ , v ₂ , …, v _j ) divided into two categories (U and V). As a graph obtained by connecting each other, among the data ((u ₁ , u ₂ , …, u _i ), (v ₁ , v ₂ , …, v _j )) included in different categories (U, V) are related to each other It is used to simply represent data.

Referring to the example of FIG. 3 , the data (u ₁ ) of the first category (U) is related only to the data (v ₁ ) of the second category (V), but the data (v ₁ ) of the second category (V) It can be seen that the data (u ₁ ) of the first category (U) is associated with the other two data (u ₂ , u _i ), and the data (u ₂ ) of the first category (U) is associated with the second category ( It can be easily seen that V) is related to two data (v ₁ , v ₂ ). That is, the bipartite graph is a data analysis means that can very simply set and check the correlation between data of different categories.

The relationship analysis unit 200 may include a user food relationship graph acquisition unit 210, a food ingredient relationship graph acquisition unit 220, and a food meal relationship graph acquisition unit 230.

The user food relationship graph acquisition unit 210 obtains a relationship between each user and the food consumed by the user in a bipartite graph using the user meal information. For example, the user food relationship acquisition unit 210 may set the user as the first category (U) and the food as the second category (V) to obtain a food-ingredient bipartite graph. However, as another example, food may be set as the first category (U) and users may be set as the second category (V).

The food ingredient relationship graph acquisition unit 220 obtains a relationship between each food and ingredient as a bipartite graph using food intake data. The food-ingredient relationship graph acquisition unit 220 may obtain a food-ingredient bipartite graph by setting food as the first category (U) and food as the second category (V).

On the other hand, the food meal relationship graph acquisition unit 230, based on the food intake data collected by the food intake information collection unit 120, the relationship between the meals of a number of people and the food consumed by the people at the time of eating a bipartite graph. Acquire The food-meal relationship graph obtaining unit 230 may obtain a meal-food dichotomy graph by setting a first category (U) to food consumed by a plurality of people and setting food consumed by a plurality of people to a second category (V). there is.

The food selection unit 300 selects food candidates to be recommended to the user based on the bipartite graph obtained by the relationship analysis unit 200 . In this embodiment, the food selection unit 300 utilizes the degree of a common neighbor to use the Adamic-Adar Index (AAI), which is an index for comparing intimacy between two points. Food candidates to be recommended to the user can be sorted based on the

AAI is an algorithm proposed to numerically express the importance according to the amount of shared links between two nodes in a social network. When applied to the bipartite graph of FIG. 3, a plurality of first data of the first category U ( The number of second data (v _m ) to which the two first data (u _p , u _q ) are commonly connected among u ₁ , u ₂ , …, u _i , that is, the number of commonly associated second data (v _m ) It is calculated to have a higher value as the number (N(v)) of the first data (u) connected to the second data (v _m ) connected in common is large. As the number of second data (v _m ) commonly connected to the two first data (up _p , u _q ) increases, the correlation increases, but as the number of first data connected to the second data (v _m ) becomes rarer, This is because the importance of the correlation between the two first data u _p and u _q is high.

For example, AAI (AAI(u ₁ , u ₂ )) for the _two first data u ₁ and u 2 of the first category U may be calculated as in Equation 1.

The food selection unit 300 may include a representative food selection unit 310 and a combination food selection unit 320 .

The representative food selection unit 310 corresponds to each of a plurality of foods consumed during a meal for a predetermined period in the past (for example, one week) from the user meal information acquired by the user meal information acquisition unit 110, for example. The data to be checked in the first category (U) of the food-ingredient bipartite graph obtained by the food ingredient relationship graph acquisition unit 220. Then, for each of the identified plurality of food data (up ), food data (u _q ) of the first category (U) is searched for through the _medium of the food material of the second category (V). That is, food data using common ingredients may be searched based on the ingredients of the food consumed by the user.

However, when food data is searched based only on food consumed by an individual user, the type of food to be searched may be limited. In order to overcome this problem, the representative food selection unit 310 includes a second category (V) connected to the current user (up) in the first category (U) of the user-food _bipartite graph obtained by the user food relationship graph acquisition unit 210. ) may be searched for, and a related user (u _q ), who is another user who has commonly consumed the searched food, may be searched again in the first category (U). In addition, based on the food consumed by the searched related user (u _q ), related food may be searched again using a food-ingredient dichotomy graph. In this case, since the related food data is searched by additionally considering the food consumed by the related user who consumed common food with the current user, the type of food to be searched for can be expanded. At this time, the association between the current user (up _p ) and the associated user (u _q ) is calculated based on the AAI so that the number of associated users does not increase excessively, and the calculated AAI is calculated based on a predetermined number of associated users (u You can also use a food-ingredient bipartite graph based on the food eaten by _q ).

On the other hand, when related food data (u _q ), which is other food data related to a plurality of food data (u _p ) consumed by the user in the first category (U) of the food-ingredient bipartite graph, is searched for, the consumed food data (u _p ), the association between each associated food data (u _q ) is calculated based on AAI, and the searched related food data (u _q ) is sorted according to the calculated AAI. This is to confirm the user's preferred food ingredients based on the food history of the user or other users who have consumed common food in the past, and to prioritize recommending foods according to the preferred food ingredients.

However, food includes various ingredients, and among them, there are important ingredients that are the core of the food, while additional ingredients such as seasonings also exist. Therefore, it is difficult to see the relationship between food ingredients as an equal relationship. Although AAI (AAI (up _p , u _q )) for the two first data (up _p , u _q ) in the food-food dichotomous graph has a large number of common ingredients (v) that are commonly connected, and the common ingredients are The smaller the number of foods (u) used, the higher the value, and the food (u _q ) related to important ingredients is more likely to have a high AAI. The importance of commonly included ingredients is not properly reflected in the relationship between foods.

Accordingly, in this embodiment, the representative food selection unit 310 corrects the AAI of Equation 1 as in Equation 2 so that the proportion according to the weight of the food ingredient is additionally reflected in the association between foods, thereby determining the association between foods. It is calculated using the modified AAI (mAAI) of Equation 2, and the highly related foods (u _q ) are sequentially sorted according to the modified AAI (mAAI).

Here, Am represents the average weight of the ingredient (v) within all foods containing the ingredient (v), and m _p , m _q are the weight of the ingredient (v) included in the food (up _p , u _q ), respectively. indicate

Then, the representative food selection unit 310 selects a predetermined number (eg, 10) of foods having a large modified AAI from the aligned foods (u _q ) and delivers them to the combination food selection unit 320.

The combination food selection unit 320 searches for and arranges combination foods that people have combined and consumed for each of a predetermined number of representative foods selected and delivered by the representative food selection unit 310 .

The food u _q arranged according to the representative food selector 310 includes important ingredients of the food consumed by the previous user or related user, and a predetermined number of high-ranking foods may be referred to as the representative food. Here, the representative food is simply food selected according to importance, and may be a main food, a sub food, or an additional food.

The combination food selection unit 320 receives a predetermined number of representative foods selected by the representative food selection unit 310, and the authorized representative food is obtained by the food meal relationship graph acquisition unit 230. The food-meal dichotomy graph Check in the first category (U). Then, for each of the identified plurality of food data ( _{up), food data (u q )} of the first category (U), which is related to the meal of the second category (V), is searched for. In other words, it searches for food data that people consumed along with representative foods during meals.

When the associated food data (u _q ) related to the food data (up _p ) for the representative food in the first category (U) of the food-meal dichotomy graph is searched for, the combination food selection unit 320 displays the representative food data (u _p ) The association between each associated food data (u _q ) is calculated based on AAI, and the searched related food data (u _q ) is sorted according to the calculated AAI, and the top criteria with high AAI for each representative food A specified number of associated food data (u _q ) is selected as a combination food to be combined with the corresponding representative food.

In other words, the foods that people consume with the representative foods are sequentially arranged. Since the relationship between foods consumed together during a meal can be regarded as an equal relationship, the combination food selection unit 320 may arrange the foods according to the AAI calculated according to Equation 1. Then, the combination food selection unit 320 selects a predetermined number of foods from among a plurality of foods arranged according to AAI as combination foods to be combined with corresponding representative foods.

In this embodiment, in the food selection unit 300, the representative food selection unit 310 uses a food-ingredient dichotomy graph to search for and sort food data using common ingredients with the food previously consumed by the user, so that the combined food selection unit (320) searches and sorts food data consumed by people using a food-meal dichotomy graph, so unrelated foods are not searched and AAI is not calculated. Therefore, representative food and combination food can be sorted with low computational cost.

On the other hand, the food classification unit 400 classifies the type and amount of food collected in the information collection unit (100). The food classification unit 400 may include a food category setting unit 410 and a unit food amount setting unit 420 .

Figure 4 shows an example of food classification and unit amount set according to the food classification unit of FIG. As shown in the fourth column (main) of FIG. 4 , the food category setting unit 410 may assign a separator according to the category of food collected by the information collecting unit 100 . That is, the food category setting unit 410 may assign different delimiters depending on whether the main food, the sub food, or the side food is different. In FIG. 4, a delimiter of 1 is assigned to the main food and a delimiter of 0 is assigned to the sub food. The case where is given is shown.

Here, the food category setting unit 410 assigns separators according to food categories so that foods of each food category classified according to the separators can be evenly included in meals at a predetermined ratio or more.

Although the representative food and the combination food to be consumed together in the food selection unit 300 are sorted based on the relationship between food ingredients and meals, this is a sorting order that considers only the relationship between foods. Therefore, when foods are selected and recommended according to the order in which they are arranged in the food selector 300, the proportion of main foods or the proportion of sub foods may be too high. In order to prevent this problem, the food category setting unit 410 may preliminarily assign a separator according to the food category so that the condition setting unit 500 to be described later may include various foods according to the food category.

Meanwhile, the unit food amount setting unit 420 discretizes the food amount of each food. As shown in the third column (serving_size) in the table of FIG. 4, the amount of each food served during a meal is very diverse and different for each type. However, it can be discretized and set as a unit amount according to one serving for each food. Accordingly, the unit food amount setting unit 420 sets the weight per serving according to the type of food so that each food increases or decreases per serving.

Here, in the food classification unit 400, a manager who manages the food recommendation device may set and store each food category and unit food amount. However, if the main food, sub food, and additional food are already assigned to the collected food intake data, the food category setting unit 410 may be omitted, and if the unit amount for one serving is set in advance in the food information, Unit food amount setting unit 420 may be omitted.

The condition setting unit 500 sets various constraints to be considered when the food recommendation device recommends food to the user. Here, user conditions according to the user's taste may be set together with restrictions according to nutritional components.

The condition setting unit 500 may include a basic condition setting unit 510 and a user condition setting unit 520 .

The basic condition setting unit 510 sets conditions considering nutritional components of food. Here, as an example, the basic condition setting unit 510 basically sets sodium intake (na), calories (en), saturated fat ratio (sfa), fat ratio (fat), carbohydrate ratio (carb), and vitamin C content (vitc). assumed to be taken into account. In this case, the sodium intake (na) is set to be 2000 mg or less, the calorie (en) is 2000 kcal or less, the saturated fat percentage (sfa) is 7% or less, the fat percentage (fat) is between 15 and 30%, and the carbohydrate percentage ( carb) is between 55 and 65% and the vitamin C content (vitc) is 50 to 1000 mg. However, this is a simple example and the type and number of nutritional ingredients may be variously changed.

In addition, the basic condition setting unit 510 may set additional conditions so that the composition of food can maintain the form of a normal meal. For example, the basic condition setting unit 510 may set food classified as main food in the food category to include a predetermined weight (eg, 700 g) or more. This is to prevent a case in which the user does not trust the recommended food because only sub food such as side dishes or vegetables are excessively provided and main food is not provided. In addition, the serving amount of a particular type of food, such as fruit, may be pre-determined.

The user condition setting unit 520 may set conditions for foods to be necessarily included or excluded according to preferred and non-preferred food information and forbidden food information according to disease, etc., applied from the user terminal 10 . For example, the user may set kimchi to be necessarily included in the recommended food or set not to be included.

5 shows an example of personal taste setting of the food classification unit of FIG. 2 .

5 shows an example in which conditions are set so that kimchi must be included. In this case, as shown in the right column, at least one of various kimchi classified as kimchi in the food category is included, rather than simply including cabbage kimchi. It can be seen that all foods classified as kimchi are activated as 1.

Here, when an ingredient other than a food to be necessarily included or excluded is authorized, the food including the corresponding ingredient can be identified using the food-ingredient bipartite graph acquired by the food ingredient relationship graph acquisition unit 220 .

In addition, the user condition setting unit 520 may additionally set personal nutritional conditions according to the user's health condition, and the conditions set by the user condition setting unit 520 overlap with the conditions set by the basic condition setting unit 510. In this case, the conditions set in the user condition setting unit 520 may be prioritized.

The food recommendation unit 600 has a predetermined number of representative foods extracted from the representative food selection unit 310 of the food selection unit 300 and a predetermined number of foods extracted according to each representative food from the combination food selection unit 320. A plurality of combined food sets are authorized, and a food set meeting a condition set in the condition setting unit 500 and an amount of food included in the food set are determined and transmitted to the user terminal 10 as recommended food. Here, the number of food sets may be the same as the number of selected representative foods.

The food recommendation unit 600 arranges the food set based on the linear programming method based on the AAI and serving amounts of the foods included in the food set, and satisfies the condition set in the condition setting unit 500 among the food sets arranged on the top. A food set of a predetermined number (for example, three) is determined as a recommended food.

The food recommendation unit 600 designates the AAI of each food included in each of a plurality of food sets as a score (s), configures a linear function by setting the serving amount of each food as a variable (x), and constructs a linear function The food serving amount (x) is varied within a specified range so that the calculated value is maximized to optimize the food serving amount in each food set. That is, the food recommendation unit 600 optimizes the serving amount of each food by using the correlation between the foods included in the food set.

If the number of foods included in the food set is n, the serving amount of each food in the food set can be optimized according to Equation 3.

Here, s ₁ to s _n represent the scores of each of the n foods included in the food set, and x ₁ to x _n represent the serving amounts of each of the n foods.

At this time, the amount of food to be provided (x) may be limited to be provided in discrete units such as 0 servings, 1 serving, and 2 servings (0, 1, 2) based on the unit amount. This is because, when the amount of food provided (x) is not limited, the calculated value of the calculation function of Equation 3 may exceed the amount within a common sense range that the user can consume during a meal. That is, certain foods can be restricted so that they are not provided in excess of two servings per meal.

However, in the case of some foods, since the serving amount is determined based on the fact that many people share the food and consume it, the food serving amount (x) is 0, not 0 servings, 1 serving, and 2 servings (0, 1, 2). It can also be set to be served in discrete units of servings, 0.5 servings, 1 serving, 1.5 servings (0, 0.5, 1, 1.5).

When the amount of food provided for each of the plurality of food sets is optimized according to Equation 3, the plurality of food sets are sorted according to the optimized calculation value. That is, the food sets are arranged in an order of high correlation among the food sets calculated based on the optimized serving amount.

In addition, it is checked whether the food combination of the food set in which the food serving amount is determined according to the optimized calculation value satisfies the condition set by the condition setting unit 500. At this time, whether or not the nutritional conditions set in the basic condition setting unit 510 are satisfied may be calculated according to Equation 4.

Here, a _m1 to a _mn represent specific nutritional components (m) in each unit amount of n foods, and b _mu and b _ml represent upper and lower limit conditions set for specific nutritional components (m), respectively.

That is, Equation 4 determines whether a specific nutritional component (m) according to the optimal serving amount (x ₁ to x _n ) of foods included in the food set exceeds the upper and lower limit conditions (b _mu , b _ml ).

However, in general, the nutritional components that users should consume are set per day, not per meal. Therefore, the food recommendation unit 600 may adjust the upper and lower limit conditions (b _mu , b _ml ) by additionally considering nutritional components provided by food consumed by the user at the previous meal of the corresponding day. That is, the upper and lower limit conditions (b _mu , b _ml ) may be changed so that nutritional components are not insufficient or exceeded on a daily basis.

Thereafter, the food recommendation unit 600 selects a food set that satisfies the condition set by the condition setting unit 500 from the food set for which the amount of food to be provided according to the optimized calculation value is determined, and transmits the selected food set to the user terminal as recommended food. At this time, if there are a plurality of food sets satisfying the condition set by the condition setting unit 500, a predetermined number of food sets may be selected in the sorted order, that is, in the order of food sets having a large optimized calculation value, and transmitted to the user terminal. .

Additionally, in this embodiment, the food recommendation unit 600 may additionally apply a condition relaxation penalty to prevent a case in which a set of recommended foods is very limited or not selected according to set conditions.

The condition relaxation penalty can be set by weighting the pre-specified penalty coefficient (p _m ) for each nutritional component (m ) over the condition (e _m ) of each nutritional component (m). In this case, the linear function of Equation 3 is It can be transformed into Equation 5 by subtracting the penalty according to

Also, the food recommendation unit 600 may determine an optimal serving amount of food in each food set according to Equation 5, and arrange the food sets based on calculation values according to the determined optimal serving amount of food to determine recommended foods.

As a result, the food recommendation device according to the present embodiment configures a food set by extracting a representative food with a high correlation from the food consumed by the user at a previous meal, and extracting a plurality of combination foods that are good to be consumed together with the extracted representative food. It is possible to determine the serving amount of each food from the configured food set and recommend it to the user. In addition, it is possible to recommend food considering the user's taste and nutrition by allowing the user to select a food set that satisfies basic conditions such as nutritional components according to the food set and user conditions set by the individual user.

6 shows a nutrition-based food recommendation method according to an embodiment of the present invention.

Referring to the food recommendation method of FIG. 6 with reference to FIGS. 1 to 5 , first, food intake data is collected and obtained (S11). Here, the food intake data includes information on the food consumed by a large number of people and the ingredients used in the food, for example, the meal time during the day, the type of food consumed at the time of meal, the weight of the food material used for each food, and Nutrient component information for each food may be included.

Then, when information on food intake data is collected, user meal information including information on the type and amount of food consumed during each user's meal for a predetermined period of time and during the meal is obtained (S12). User meal information may be transmitted and obtained from a plurality of user terminals 10 .

When the food intake data and user meal information are acquired, a bipartite graph is obtained to analyze the relationship between the user and food, food and meal, and food and ingredients based on the obtained user meal information and food intake data (S13). At this time, it can be obtained by dividing the food-food bipartite graph representing the relationship between food and food and the food-meal bipartite graph representing the relationship between food and food, and the user-food bipartite graph representing the relationship between the user and food can be obtained additionally.

When the food-ingredient bipartite graph and food-meal bipartite graph are obtained, based on the user's meal information, based on the food consumed by the user for a pre-determined period, search for related foods having common ingredients using the food-ingredient bipartite graph. And, the correlation between the two foods is calculated by the modified AAI of Equation 2 considering the specific gravity of the ingredients, and a predetermined number of foods having a high modified AAI value are selected as representative foods (S14).

Based on each of a predetermined number of representative foods, a food-meal dichotomy graph is used to search for related foods that people consume with the corresponding food, and the relationship between the two foods is calculated with AAI to determine a predetermined number of high AAI values. The food of each is selected as a combination food for each representative food (S15).

When a predetermined number of representative foods and a predetermined number of combination foods for each representative food are selected, the representative food and the corresponding combination food are set as a food set (S16).

Meanwhile, separately from this, when food intake data and user meal information are obtained, food categories are classified into main food, sub food, and additional food, and a separator according to the category is assigned (S21). A unit amount is set based on the serving amount of each food (S22).

However, if food categories such as main food, sub food, and additional food are pre-classified in the collected food intake data, the category classification step (S21) may be omitted. In addition, if the unit amount for each food is set in advance in the food intake data, the food unit amount setting step (S22) may also be omitted.

In addition, basic conditions for recommended food sets may be set (S23). Here, as the basic condition, various nutritional components to be provided as a food set, amount of food according to category, and essential food may be set.

In addition, user conditions according to the user's personal taste or health condition are additionally set (S24). Here, the user conditions may include the user's preferred and non-preferred food and the required amount of each nutritional component according to the user's health condition. If the basic condition and the user condition are set as conditions for the same content, the user condition may be set first.

Here, the basic conditions may be set in advance by the administrator, and the user conditions may also be set in advance by the user. In this case, the basic condition setting step and the user condition setting step may be omitted.

When the food set is set, the serving amount of each food is optimized based on the AAI value and the serving amount of the food included in each set food set (S17). At this time, the serving amount can be optimized by setting the serving amount to a discrete variable according to the unit amount and calculating the serving amount at which the calculated value of the linear function represented by Equation 3 is maximized while varying within a predetermined range.

Then, based on the optimized amount of food provided, it is determined whether each of the plurality of food sets satisfies the set basic condition and user condition, and a food set satisfying the condition is selected (S18).

If there is a food set that satisfies the condition, the food set that satisfies the condition is recommended to the user as recommended food (S19). In this case, among the food sets satisfying the condition, food sets of a predetermined reference number may be recommended to the user in the order of increasing calculated values of the linear function. However, if there is no food set that satisfies the condition or the number of food sets that _satisfy the condition is less than the standard number, the penalty coefficient ( The condition relaxation penalty accumulated by weighting p _m ) is additionally reflected in the linear function of Equation 3, and the modified linear function is used as shown in Equation 5. This is to allow the user to be recommended a food set with a high computational value even though a penalty is applied because the amount of food provided in the food set does not satisfy the required condition, so that the user can resolve the complaint about the recommended food to a certain level. am.

The method according to the present invention may be implemented as a computer program stored in a medium for execution on a computer. Here, computer readable media may be any available media that can be accessed by a computer, and may also include all computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, including read-only memory (ROM) dedicated memory), random access memory (RAM), compact disk (CD)-ROM, digital video disk (DVD)-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

Although the present invention has been described with reference to the embodiments shown in the drawings, this is only exemplary, and those skilled in the art will understand 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.

10: user terminal 20: medical terminal
30: database server 40: food recommendation server
100: information collection unit 110: user meal information acquisition unit
120: food intake information collection unit 200: relationship analysis unit
210: user food relationship graph acquisition unit
220: food ingredient relationship graph acquisition unit
230: food meal relationship graph acquisition unit
300: food selection unit 310: representative food selection unit
320: combination food selection unit 400: food classification unit
410: food category setting unit 420: unit food amount setting unit
500: condition setting unit 510: basic condition setting unit
520: user condition setting unit 600: food recommendation unit

Claims (18)

Information collection obtained by collecting user meal information, which is information about the user's previous food consumption during a pre-specified period, and food intake data, which includes information about the food consumed by a large number of people and the ingredients of each meal. wealth;
a relationship analysis unit that acquires a food-ingredient bipartite graph for analyzing a relationship between food and ingredients and a food-meal bipartite graph for analyzing a relationship between food and meal based on the food intake data;
Based on the food obtained from the user's meal information, search for related foods in the food-ingredient bipartite graph to select a predetermined number of representative foods with high correlation, and based on each of the representative foods, in the food-meal bipartite graph. a food selection unit for obtaining a food set consisting of the representative food and the combination food by searching for related foods and selecting a predetermined number of combination foods having a high correlation corresponding to each of the representative foods; and
A food recommendation unit that optimizes the serving amount of each food based on the relevance of the foods included in the food set and selects and recommends a predetermined number of food sets having a high total correlation calculated based on the optimal serving amount of each food Food recommendation device. The method of claim 1, wherein the food selection unit
Based on the food obtained from the user's meal information, search for related foods using common ingredients in the food-ingredient bipartite graph, and modify the relationship between the standard food and the related food in consideration of the proportion of commonly included ingredients. a representative food selection unit for selecting a representative food from a predetermined number of related foods having a high modified AAI calculated by a modified Adamic-Adar Index (hereinafter referred to as AAI); and
Based on each of the representative foods, a related food that is a food served at the same meal is searched for in the food-meal dichotomy graph, and the association between each representative food and the related food is calculated by AAI, and a predetermined number of related foods having a high AAI are calculated. A food recommendation device including a combination food selection unit for selecting a combination food to obtain a food set consisting of the representative food and the combination food. The method of claim 2, wherein the relationship analysis unit
a food ingredient relationship graph acquisition unit configured to classify a plurality of foods and a plurality of ingredients into two different categories in the food intake data, and acquire the food-ingredient bipartite graph by connecting ingredients included in each of the plurality of foods; and
and a food-meal relationship graph acquisition unit configured to classify meals and foods into two different categories from the food intake data and to obtain the food-meal dichotomy graph by connecting foods consumed at each meal. The method of claim 2, wherein the representative food selection unit
The modified AAI (mAAI) is expressed as

(Where u _p represents the standard food and related food in the food-ingredient bipartite graph, v is the food ingredient, Am represents the average weight of the food ingredient (v) within all foods including the food ingredient (v), m _p , m _q represent the weight of ingredients (v) included in food (u _p , u _q ), respectively.)
Food recommendation device that calculates according to. The method of claim 2, wherein the combination food selection unit
Equation above AAI

(Where u _p represents the standard food and related food in the food-meal dichotomy graph, and v represents the meal.)
Food recommendation device that calculates according to. The method of claim 3, wherein the food recommendation unit
The serving amount of each food is optimized so that the total correlation calculated as the sum of the correlations of the foods included in the food set is maximized, and the optimal serving amount for each food calculated for each predetermined nutritional component is applied to each food according to the optimal serving amount for each food. A food recommendation device that calculates the amount of nutritional components of a set, selects and recommends a predetermined number of food sets having the highest total correlation while satisfying the predetermined nutritional component-specific conditions. The method of claim 6, wherein the food recommendation unit
According to the linear programming method, the AAI between each of the representative foods and the combination food is designated as a score (s), and the serving amount of each food is set as a variable (x) according to the unit amount provided for each food. Equation so that the calculated value of the linear function representing the total correlation is maximized

A food recommendation device that calculates an optimal serving amount for each food by searching for a variable (x) that satisfies within a predetermined range. The method of claim 3, wherein the food recommendation unit
Calculate the total relevance to which the condition relief penalty, which is calculated according to the difference from the condition for each predetermined nutritional component, is applied, optimize the serving amount of each food so that the total relevance to which the condition relief penalty is applied is maximized, and the total relevance according to the optimal serving amount A food recommendation device that selects and recommends the highest predetermined number of food sets. The method of claim 8, wherein the food recommendation unit
According to the linear programming method, the AAI between each of the representative foods and the combination food is designated as score (s), the serving amount of each food is set as a variable (x) according to the unit amount provided, and each nutritional component ( m) of the condition excess amount (e _m ), which is composed of a condition relaxation penalty set by weighting the penalty coefficient (p _m ) for each nutritional component specified in advance, so that the calculated value of the linear function representing the overall association of the food set is maximized math formula

A food recommendation device that calculates an optimal serving amount for each food by searching for a variable (x) that satisfies within a predetermined range. Collecting and obtaining user meal information, which is information about food consumed by a user during a pre-specified period of time, and food intake data including information about food eaten by a plurality of people and ingredients of each meal;
Obtaining a bipartite graph including a food-ingredient bipartite graph for analyzing a relationship between food and ingredients and a food-meal bipartite graph for analyzing a relationship between food and meal based on the food intake data;
Based on the food obtained from the user's meal information, search for related foods in the food-ingredient bipartite graph to select a predetermined number of representative foods with high correlation, and based on each of the representative foods, in the food-meal bipartite graph. obtaining a food set consisting of the representative food and the combination food by searching for related foods and selecting a predetermined number of combination foods having a high correlation corresponding to each of the representative foods; and
Optimizing the serving amount of each food based on the correlation of the foods included in the food set, and selecting and recommending a predetermined number of food sets having a high total correlation calculated based on the optimal serving amount of each food How to recommend. 11. The method of claim 10, wherein obtaining the set of food comprises:
Based on the food obtained from the user's meal information, search for related foods using common ingredients in the food-ingredient bipartite graph, and modify the relationship between the standard food and the related food in consideration of the proportion of commonly included ingredients. selecting a representative food from a predetermined number of related foods having a high modified AAI by calculating with a modified Adamic-Adar Index (hereinafter referred to as AAI); and
Based on each of the representative foods, a related food that is a food served at the same meal is searched for in the food-meal dichotomy graph, and the association between each representative food and the related food is calculated by AAI, and a predetermined number of related foods having a high AAI are calculated. A food recommendation method comprising selecting a combination food and configuring a food set with representative food and combination food. 12. The method of claim 11, wherein the step of obtaining the bipartite graph
obtaining the food-ingredient dichotomy graph by classifying a plurality of foods and a plurality of ingredients into two different categories in the food intake data and connecting ingredients included in each of the plurality of foods; and
and classifying meals and foods into two different categories from the food intake data, and obtaining the food-meal dichotomy graph by connecting the foods consumed at each meal to each other. The method of claim 11, wherein the step of selecting the representative food
The modified AAI (mAAI) is expressed as

(Where u _p represents the standard food and related food in the food-ingredient bipartite graph, v is the food ingredient, Am represents the average weight of the food ingredient (v) within all foods including the food ingredient (v), m _p , m _q represent the weight of ingredients (v) included in food (u _p , u _q ), respectively.)
Food recommendation method to calculate according to. 12. The method of claim 11, wherein constructing the food set comprises:
Equation above AAI

(Where u _p represents the standard food and related food in the food-meal dichotomy graph, and v represents the meal.)
Food recommendation method to calculate according to. The method of claim 12, wherein the step of selecting and recommending the food set
optimizing the serving amount of each food so that the total relevance calculated as the sum of the relevance of the foods included in the food set is maximized;
Calculating the amount of nutritional components of each food set according to the optimal serving amount for each food by applying the optimal serving amount for each food calculated for each predetermined nutritional component; and
A food recommendation method comprising the step of recommending a predetermined number of food sets having the highest total correlation while satisfying predetermined conditions for each nutritional component with calculated nutritional component amounts. 16. The method of claim 15, wherein optimizing the serving amount of food
According to the linear programming method, the AAI between each of the representative foods and the combination food is designated as a score (s), and the serving amount of each food is set as a variable (x) according to the unit amount provided for each food. Equation so that the calculated value of the linear function representing the total correlation is maximized

A food recommendation method that calculates the optimal serving amount for each food by searching for a variable (x) that satisfies within a predetermined range. The method of claim 12, wherein the step of selecting and recommending the food set
setting a condition mitigation penalty calculated according to a difference from a condition for each predetermined nutritional component;
optimizing the serving amount of each food by applying the relevance of the foods included in the food set and the condition relaxation penalty to maximize the overall relevance; and
A food recommendation method for selecting and recommending a predetermined number of food sets having the highest overall correlation to which the condition relaxation penalty is applied. 18. The method of claim 17, wherein the step of optimizing the serving amount of food
According to the linear programming method, the AAI between each of the representative foods and the combination food is designated as score (s), the serving amount of each food is set as a variable (x) according to the unit amount provided, and each nutritional component ( m) of the condition excess amount (e _m ), which is composed of a condition relaxation penalty set by weighting the penalty coefficient (p _m ) for each nutritional component specified in advance, so that the calculated value of the linear function representing the overall association of the food set is maximized math formula

A food recommendation method that calculates the optimal serving amount for each food by searching for a variable (x) that satisfies within a predetermined range.

Images