KR102647903B1 - Apparatus and method for predicting customized genetic skin type based on questionnaire using machine learning - Google Patents

Abstract

A questionnaire-based customized genetic skin type prediction device, comprising: a memory storing at least one instruction and at least one processor executing the at least one instruction, wherein the processor receives the customer's response to a pre-prepared questionnaire. Receive and input the received response into a pre-trained machine learning model to derive variable values for genetic indicators related to skin risk, determine the customer's skin type based on the derived variable values, and determine the skin type of the customer. The accuracy of type prediction can be improved, and the time and cost required to predict skin type can be minimized.

Description

Survey-based customized genetic skin type prediction device and method using machine learning {APPARATUS AND METHOD FOR PREDICTING CUSTOMIZED GENETIC SKIN TYPE BASED ON QUESTIONNAIRE USING MACHINE LEARNING}

Embodiments relate to a survey-based customized genetic skin type prediction device and method. More specifically, it relates to an apparatus and method for deriving an individual's genetic characteristics related to skin by applying survey data conducted for each individual to a machine learning model and predicting a customized genetic skin type based on this.

Recently, the public's interest in skin and beauty is increasing. In the past, cosmetics were mainly the exclusive property of women, but now the demand for cosmetics is increasing regardless of age and gender. Additionally, as the types of cosmetics become more diverse, personalized cosmetics are attracting attention.

In the past, skin type was determined through a survey of customers, and based on this, criteria for selecting personalized cosmetics were provided. For example, American Dr. Leslie Bauman presented the first questionnaire to determine skin type through “The Skin Type Solution.”

However, this method of determining skin type based on a simple questionnaire not only involves subjective judgments that differ from person to person, but also uses only limited data, so the objectivity and accuracy of skin type determination is low and customer satisfaction is not high. there was.

Meanwhile, research has been conducted on biomarker technology that confirms a customer's genetic information through genetic testing to determine skin type and predicts skin risk based on this. This method of determining skin type based on genetic testing has the advantage of being able to derive objective results using biological information, but since it must necessarily be accompanied by a genetic test for the customer, there is a time delay in providing the service and the cost of excessive testing is high. There is a downside to this.

Republic of Korea Registration No. 10-1713510 Republic of Korea Registration No. 10-2001337 Republic of Korea Application No. 10-2019-0173477 Republic of Korea Application No. 10-2019-0173478 Republic of Korea Application No. 10-2019-0173479 Republic of Korea Application No. 10-2019-0173480 Republic of Korea Application No. 10-2019-0173481

Embodiments apply survey response data to a machine learning model to derive genetic information related to the risk of each customer's skin, thereby predicting and providing a customized skin type based on more accurate and objective standards.

Embodiments seek to minimize the time and cost required to predict skin type by deriving the customer's skin genetic information through a survey without genetic testing for the customer.

The technical challenges that this embodiment aims to achieve are not limited to the technical challenges described above, and other technical challenges can be inferred from the following embodiments.

A questionnaire-based customized genetic skin type prediction device according to an embodiment includes a memory storing at least one instruction and at least one processor executing the at least one instruction, wherein the processor responds to a pre-prepared questionnaire. Receive the customer's response, input the received response into a pre-trained machine learning model to derive variable values for genetic indicators related to skin risk, and determine the customer's skin type based on the derived variable value. can be decided.

Additionally, the processor may process the received response into an input data format for the machine learning model.

Additionally, the processor can convert variable values related to the genetic index into binary code depending on the presence or absence of skin risk.

Additionally, the risk of the skin may include the risk of wrinkles, moisture, pigmentation, oiliness, and sensitivity.

Additionally, the genetic indicator related to the risk of the skin may be a biomarker for predicting the risk of any one of wrinkles, moisture, pigmentation, oiliness, and sensitivity.

In addition, the biomarkers have rs_ids of rs1929013, rs9873353, rs1362404, rs143938096, rs74653330, rs11685354, rs4653497, rs59784607, rs308971, rs151209785 and rs733. It may be at least one SNP among 4780.

Additionally, the skin type can be classified into 32 skin types based on variable values for the genetic index.

Additionally, the machine learning model may be updated using the received response and the derived variable value.

Additionally, the machine learning model may be learned based on an ensemble learning algorithm.

A questionnaire-based customized genetic skin type prediction method according to an embodiment includes the steps of receiving a customer's response to a pre-prepared questionnaire, inputting the received response into a pre-trained machine learning model to generate a genetic index related to the risk of the skin. It may include deriving variable values for and determining the skin type of the customer based on the derived variable values.

In addition, the step of processing the responses to the questionnaire into the input data format of the machine learning model may be further included.

Additionally, the step of deriving the variable value related to the genetic index may include converting the variable value into binary code according to the presence or absence of skin risk.

Additionally, the step of determining the customer's skin type may include determining the customer's skin type based on the risk of wrinkles, moisture, pigmentation, oiliness, and sensitivity.

Additionally, the step of updating the machine learning model using the received response and the derived variable value may be further included.

The computer program stored in the medium according to one embodiment may be combined with hardware to execute the questionnaire-based customized genetic skin type prediction method.

According to the above-described embodiments, data related to the customer's genetic information can be derived through survey data, and individual skin type can be predicted based on this. Accordingly, more accurate skin type prediction is possible, and the time and cost required to predict skin type can be minimized.

In addition, according to the above-described embodiments, the predicted skin type can be used to recommend customized procedures in the medical industry such as dermatology and the beauty industry such as aesthetics, or can be applied to the development and recommendation of customized cosmetics in the cosmetics industry.

The effects of the embodiments are not limited to the effects described above, and effects not mentioned can be clearly understood by those skilled in the art from this specification and the attached drawings.

Figure 1 is a diagram schematically showing the questionnaire-based customized genetic skin type prediction flow and application based on analysis results.
Figure 2 is a block diagram of a questionnaire-based customized genetic skin type prediction device according to an embodiment.
FIG. 3 is a diagram illustrating the learning and analysis structure of a machine learning model applied to the questionnaire-based customized genetic skin type prediction device of FIG. 2.
Figure 4a is a diagram showing input data input to a machine learning model according to one embodiment.
Figure 4b is a diagram showing output data output from a machine learning model according to one embodiment.
Figure 5 is a diagram showing 32 skin types classified through a questionnaire-based customized genetic skin type prediction device according to an embodiment.
Figure 6 is a flowchart of a questionnaire-based customized genetic skin type prediction method according to an embodiment.

Phrases such as “in some embodiments,” “according to one example,” or “in one embodiment” that appear in various places in this specification do not necessarily all refer to the same embodiment.

Since the embodiments can be subject to various changes and have various forms, some embodiments will be illustrated in the drawings and described in detail. However, this is not intended to limit the embodiments to a specific disclosed form, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the embodiments. The terms used in the specification are merely used to describe the embodiments and are not intended to limit the embodiments.

The terms used in the embodiments are general terms that are currently widely used as much as possible while considering the functions in the embodiments, but this is due to the intention or precedents of technicians working in the technical field to which the embodiments belong, the emergence of new technologies, etc. It may vary depending on In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the relevant section. Therefore, the terms used in the embodiments should be defined based on the meaning of the term and the overall content of the embodiments, rather than simply the name of the term.

Terms such as “mechanism,” “database,” “element,” “means,” and “configuration” may be used broadly and are not limited to mechanical and physical configurations. Additionally, terms such as “-unit” and “-module” used in the specification refer to a unit that processes at least one function or operation, which may be implemented as hardware or software, or as a combination of hardware and software.

Additionally, connection lines or connection members between components shown in the drawings merely exemplify functional connections and/or physical or circuit connections. In an actual device, connections between components may be represented by various replaceable or additional functional connections, physical connections, or circuit connections.

Additionally, terms including ordinal numbers such as 'first' or 'second' used in this specification may be used to describe various components, but the components should not be limited by the terms. Terms are used only to distinguish one component from another.

Additionally, some components in the drawing may be depicted with their size or proportions somewhat exaggerated. Additionally, components shown in one drawing may not be shown in other drawings.

Throughout the specification, 'examples' is an arbitrary division for easily explaining the invention in the present disclosure, and each embodiment does not need to be mutually exclusive. For example, configurations disclosed in one embodiment may be applied and/or implemented in another embodiment, and may be applied and/or implemented with changes without departing from the scope of the present disclosure.

Additionally, the terms used in this disclosure are for describing the embodiments and are not intended to limit the embodiments. In this disclosure, the singular form also includes the plural form unless otherwise specified.

Throughout the specification, the term 'skin risk' refers to the potential risk inherent in an individual's skin, and may have different risks depending on individual genetic characteristics. Skin risks include, for example, risk for wrinkles related to sagging or wrinkling of the skin, risk for moisture related to dryness, risk for pigmentation related to the amount of melanin, oiliness, risk for oil related to the amount of oil in the skin. , it can be classified by the risk of sensitivity related to the reactivity of the skin to external factors.

Specifically, the risk of wrinkles can mean the depth of wrinkles or the likelihood of wrinkles, the risk of skin can mean a lack of moisture content in the skin, and the risk of pigmentation can mean the darkness of the skin and the amount of spots or freckles. This can mean that the risk of oiliness can mean excessive oil in the skin, and the risk of sensitivity can mean that the skin is hypersensitive to lactic acid or has a high incidence of redness.

The skin risk is not limited to the above-mentioned examples, and various skin risks can be defined depending on the measurement area and measurement object. Meanwhile, the term 'skin type' refers to a person's comprehensive skin characteristics and can be determined based on the skin's risk level.

Throughout the specification, the term 'SNP (single nucleotide polymorphism)' refers to a genetic change or mutation that shows a difference of one nucleotide in the nucleic acid sequence. SNPs are known to cause diversity in traits or phenotypes between individuals. Meanwhile, in order to distinguish various SNPs from each other, each SNP has a marker called rs_id assigned by NCBI (National Center for Biotechnology Information). A person skilled in the art related to this example will be able to confirm the location and sequence of the SNP using rs_id.

Throughout the specification, the term 'machine learning' is one of the application technologies of artificial intelligence, and refers to a technology in which a computer learns through learning data and analyzes the input data based on the learning results to derive desired or useful information. do.

Below, with reference to the attached drawings, embodiments of the present disclosure will be described in detail so that those skilled in the art can easily implement them. However, embodiments of the present disclosure may be implemented in various different forms and are not limited to the embodiments described in the present disclosure.

Figure 1 is a diagram schematically showing the questionnaire-based customized genetic skin type prediction flow and application based on analysis results.

Referring to FIG. 1, skin type prediction according to the present disclosure may proceed through a survey implementation step (1), a machine learning-based analysis step (2), and a skin type prediction step (3). In addition, based on the results analyzed in the above process, the genetic information prediction step (4) and the customized product/service recommendation step (5) may be further performed. The genetic information prediction step (4) and the customized product/service recommendation step (5) may not only be performed sequentially as shown, but may also be performed separately, regardless of the order shown.

In the survey implementation step (1), customers who wish to predict their skin type can conduct a survey using a previously prepared questionnaire. The questionnaire can be prepared with a variety of questions, and each question can be related to the risk level of the skin. The skin risk can be a risk related to various characteristics of the skin. For example, skin risks may be related to wrinkles, moisture, pigmentation, oiliness, and sensitivity.

Surveys can be conducted through an app running on a terminal (e.g. computer, tablet, smartphone) or via the web. Survey results can be stored in a database included in the survey terminal or transmitted to another computing device that runs a machine learning model.

Meanwhile, the survey may be conducted through a traditional survey form, and the response data for the survey may later be transmitted or input to a computing device that runs a machine learning model.

In the machine learning-based analysis step (2), the customer's survey response results can be analyzed based on machine learning. Here, the machine learning model that analyzes the survey response results may be trained in advance to derive genetic indicators related to the risk of the customer's skin based on the survey response results. Accordingly, the learned machine learning model can receive survey data as input and derive genetic information related to the skin of the customer who provided the survey data as output. In other words, data on genetic indicators related to the risk of the customer's skin can be derived from the customer's survey response data.

In the skin type prediction step (3), individual skin type can be predicted using genetic information, which is result data derived from a machine learning model. Referring to Patent Document 0003 to Patent Document 0007 described in the prior art literature, as a genetic indicator related to the risk of the skin, biomarkers for predicting the risk of wrinkles, moisture (moisturizing), pigmentation, oil, and sensitivity are a priori. has been studied, and research results on genetic indicators related to various skin risks are being revealed. Therefore, the customer's genetic information can be derived through responses to the customer's survey, and as a result, the risk to the skin can be predicted.

For example, according to a customer's survey results, it may be predicted that the customer is at risk for wrinkles, pigmentation, and sensitivity, but not at risk for moisture and oil. In this way, customers can be provided with skin risks based on their genetic information derived from their survey responses. Furthermore, customers can be provided with their skin type classified based on the risk of the analyzed skin.

In the genetic information prediction step (4), the customer can be provided with more detailed information than the skin type derived based on the customer's survey results. Since various genetic information related to the customer's skin can be predicted from the information included in the survey results, not only the results for the five skin characteristics described above, but also genetic information for more detailed skin characteristics can be provided. For example, customers can receive additional information about genes related to subcutaneous fat, genes related to smoking sensitivity, and genes related to melanin synthesis.

In the customized product/service recommendation step (5), the customer can be recommended ingredients suitable for his or her skin based on the predicted skin type, or can be recommended products containing the ingredients available on the market. Additionally, customers may be recommended dermatological procedures or surgeries necessary to improve their skin based on their predicted skin type. In other words, information about the customer's skin derived through the embodiments of the present disclosure may be used in the cosmetics industry, beauty industry, etc.

Hereinafter, with reference to FIGS. 2 to 5, the questionnaire-based customized genetic skin type prediction device will be described in detail.

Figure 2 is a block diagram of a questionnaire-based customized genetic skin type prediction device according to an embodiment.

The questionnaire-based customized genetic skin type prediction device 100 according to an embodiment may include a processor 110, a memory 120, an input unit 130, and an output unit 140. However, in addition to the components shown in FIG. 2, general-purpose components (e.g., a communication unit for transmitting and receiving data) may be further included in the questionnaire-based customized genetic skin type prediction device 100.

The processor 110 may control the overall operation of the questionnaire-based customized genetic skin type prediction device 100. For example, the processor 110 may execute at least one instruction stored in the memory 120.

The processor 110 may be implemented as an array of multiple logic gates, or may be implemented as a combination of a general-purpose microprocessor 110 and a memory 120 storing a program that can be executed on the microprocessor 110. Additionally, those skilled in the art can understand that the present embodiment may be implemented with other types of hardware.

According to one embodiment, the processor 110 receives the customer's response to a pre-prepared questionnaire and inputs the received response into the pre-trained machine learning model 200 to generate variable values for genetic indicators related to skin risk. can be derived, and the customer's skin type can be determined based on the derived variable values.

Here, the variable value related to the genetic indicator may mean data containing information about skin risk. For example, variable values related to genetic indicators may be quantified data containing information including the presence or absence of a risk for skin wrinkles or the presence or absence of a risk for pigmentation.

Surveys can be prepared with a variety of questions to determine the customer's skin characteristics. For example, the survey included information about the customer's skin type, skin concerns, condition of the face after washing the face, condition of the face after makeup, condition of the skin in the T-zone area, skin gloss, experiences such as dermatitis or eczema, etc. It can consist of a variety of questions, such as the presence or absence of thread veins, criteria for selecting cosmetics, the amount of fatty foods consumed, the amount of sweet foods consumed, whether or not the patient smokes, whether or not he or she gets enough sleep, the level of stress, and whether he or she has experienced pregnancy and childbirth.

In one embodiment, a questionnaire may be prepared with 87 questions as shown in Table 1 below. Referring to Table 1, for each question, the question number, code name, variable name, description of the variable, and explanation of the response are listed.

No. Main Category code name variable name Description of variables Description of response
(-9=missing value) One survey Q01S1 inattention 1-1. What do you think your skin type is? 1=No
2=Yes 2 survey Q01S2 Complexity 1-1. What do you think your skin type is? 1=No
2=Yes 3 survey Q01S3 Intelligence 1-1. What do you think your skin type is? 1=No
2=Yes 4 survey Q01S4 neutrality 1-1. What do you think your skin type is? 1=No
2=Yes 5 survey Q01S5 medium to dry 1-1. What do you think your skin type is? 1=No
2=Yes 6 survey Q01S6 middle-grade 1-1. What do you think your skin type is? 1=No
2=Yes 7 survey Q02S1 Skin concern: Hyperpigmentation 1-2. What is your biggest skin concern? 1=No
2=Yes 8 survey Q02S2 Skin concern: lack of moisture 1-2. What is your biggest skin concern? 1=No
2=Yes 9 survey Q02S3 Skin concerns: Elasticity 1-2. What is your biggest skin concern? 1=No
2=Yes 10 survey Q02S4 Skin concern: Wrinkles 1-2. What is your biggest skin concern? 1=No
2=Yes 11 survey Q02S5 Skin concerns: pores 1-2. What is your biggest skin concern? 1=No
2=Yes 12 survey Q02S6 Skin concerns: dead skin cells 1-2. What is your biggest skin concern? 1=No
2=Yes 13 survey Q02S7 Skin concerns: acne, rashes 1-2. What is your biggest skin concern? 1=No
2=Yes 14 survey Q02S8 Skin concerns: blackheads 1-2. What is your biggest skin concern? 1=No
2=Yes 15 survey Q02S9 Skin concerns: Sensitivity 1-2. What is your biggest skin concern? 1=No
2=Yes 16 survey Q02S10 Skin concerns: Excessive oil 1-2. What is your biggest skin concern? 1=No
2=Yes 17 survey Q02S11 Skin concerns: redness 1-2. What is your biggest skin concern? 1=No
2=Yes 18 survey Q03S1 Pulling after washing the face 2-1. What is the condition of your face without basic makeup 30 minutes after washing your face? 1=No (comfortable without pulling or splitting feeling)
2=Yes (pulling, splitting feeling) 19 survey Q03S2 Pulling after washing 2-1. What is the condition of your face without basic makeup 30 minutes after washing your face? 1=No (comfortable without pulling or splitting feeling)
2=Yes (grey due to a lot of oil) 20 survey Q04S1 Makeup comes off easily 2-2. How does your face look and feel after 2 hours of makeup? 1=No
2=Yes 21 survey Q04S2 Makeup goes well 2-2. How does your face look and feel after 2 hours of makeup? 1=No
2=Yes 22 survey Q05S1 T-zone dryness 2-3. How does your skin feel in the T-zone area on a regular basis? 1=No
2=Yes 23 survey Q05S2 T zone oil 2-3. How does your skin feel in the T-zone area on a regular basis? 1=No
2=Yes 24 survey Q06S1 U zone drying 2-4. How does your skin feel in the U-zone area on a regular basis? 1=No
2=Yes 25 survey Q06S2 U zone oil 2-4. How does your skin feel in the U-zone area on a regular basis? 1=No
2=Yes 26 survey Q07S1 Pore condition 2-5. How is your skin texture on a regular basis? 1=Pores are barely noticeable, skin texture is delicate
2=Some pores are visible in the T-zone area, but the overall skin texture is pus.
3=Overall pores are large and skin texture is uneven. 27 survey Q08S1 no skin shine 2-6. How does your skin shine (gloss) on a regular basis? 1=No
2=Yes 28 survey Q08S2 Excessive skin shine 2-6. How does your skin shine (gloss) on a regular basis? 1=No
2=Yes 29 survey Q09S1 Cosmetic change trouble 3-1. Do skin problems occur when cosmetics, soap, or water are changed? 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 30 survey Q10S1 weather skin redness 3-2. Does your face turn red easily depending on external circumstances such as weather or wind? 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 31 survey Q11S1 allergic rash 3-3. Previous experience with atopic dermatitis, eczema, or contact dermatitis (allergic skin rash). 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 32 survey Q12S1 stress skin trouble 3-4. If you don't sleep or are stressed, your skin condition changes (pimples, redness, etc.) 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 33 survey Q13S1 thread vein 3-5. The skin is thin, so the veins are visible. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 34 survey Q14S1 Spots, moles, freckles 4-1. Spots, moles, or freckles appear on the skin. 0=Not at all. / I can't see it at all.
1=It was seen, but has disappeared / Less than 5 are visible. /slightly visible.
2=Less than 5 are visible. / More than 5 are visible. 35 survey Q15S1 Dull 4-2. The skin color of the face is dark and dull compared to the skin of other areas. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 36 survey Q16S1 even tone 4-3. The overall facial tone is uneven. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 37 survey Q17S1 sun sensitive 4-4. After normal exposure to sunlight, it does not turn red but turns black. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 38 survey Q18S1 Pigmentation 4-5. Pigmentation occurs easily in areas where skin trouble occurs and the skin is thin. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 39 survey Q19S1 fine wrinkles 5-1. There are fine wrinkles on the face. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 40 survey Q20S1 thick wrinkles 5-2. There are thick wrinkles on the face (e.g. nasolabial folds, wrinkles between the eyebrows, etc.) 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 41 survey Q21S1 neck wrinkles 5-3. There are wrinkles on the neck. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 42 survey Q22S1 Skin elasticity 5-4. If you pull on the skin, it will slowly bounce back. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 43 survey Q23S1 skin marks 5-5. Once a mark appears on the skin, it slowly disappears. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 44 survey Q24S1 skin aging 6-1. Other people usually see you as older than your actual age. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 45 survey Q25S1 skin while 6-2. How much younger do you look compared to your natural age? 0=I look more than 4 years younger than my age.
1=I look less than 3 years younger than my age. / I look 1-3 years younger than my age.
2=I look my age.
3=I look 1-3 years older than my age. / He looks less than three years older than me.
4=I look more than 4 years older than my age. 46 survey Q26S1 Pore size and shape 6-3. What is the size and shape of your pores? 0=Pores are barely visible. / Almost invisible. / Small and regular.
1=Pores are slightly visible. / Large pores begin to appear / Many of the pores are elongated.
2=The pores are large and clearly visible. / Most pores are large and stretched vertically. 47 survey Q27S1 face sagging 6-4. There is sagging face. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 48 survey Q28S1 face volume 6-5. There are places where the facial volume is turned off in an expressionless state. 0=Not at all.
1=No.
2=average.
3=Yes.
4=Always like that. 49 survey Q29S1 Cosmetic Selection: Ingredients 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes 50 survey Q29S2 Cosmetic selection: feeling of use 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes 51 survey Q29S3 Cosmetic Selection: Feeling 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes 52 survey Q29S4 Cosmetics selection: Advertising (information available in advance) 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes (advertising) 53 survey Q29S5 Cosmetics Selection: Brand 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes 54 survey Q29S6 Cosmetics Selection: Price 7-1. When choosing cosmetics, what do you consider the most important? 1=No
2=Yes 55 survey Q29S7 Cosmetic selection: Fragrance 7-1. When choosing cosmetics, what do you consider first? 1=No
2=Yes 56 survey Q29S8 Cosmetics selection: color 7-1. When choosing cosmetics, what do you consider first? 1=No
2=Yes 57 survey Q29S9 Cosmetics selection: design 7-1. When choosing cosmetics, what do you consider first? 1=No
2=Yes 58 survey Q30S1 Cosmetics Fragrance: Weak Fragrance 7-2. What are the criteria for selecting cosmetic fragrances? 1=No
2=Yes 59 survey Q30S2 Cosmetic fragrance: No fragrance 7-2. What are the criteria for selecting cosmetic fragrances? 1=No
2=Yes 60 survey Q30S3 Cosmetic scent: A unique scent like perfume 7-2. What are the criteria for selecting cosmetic fragrances? 1=No
2=Yes 61 survey Q30S4 Cosmetics scent: Soap scent 7-2. What are the criteria for selecting cosmetic fragrances? 1=No
2=Yes 62 survey Q30S5 Cosmetic fragrance: Baby product fragrance 7-2. What are the criteria for selecting cosmetic fragrances? 1=No
2=Yes 63 survey Q31S1 Feeling of applying cosmetics: moist 7-3. What are the criteria for choosing the feeling of applying cosmetics? 1=No
2=Yes 64 survey Q31S2 Feeling of applying cosmetics: Feeling of being absorbed quickly 7-3. What are the criteria for choosing the feeling of applying cosmetics? 1=No
2=Yes 65 survey Q31S3 Feeling of applying cosmetics: Average 7-3. What are the criteria for choosing the feeling of applying cosmetics? 1=No
2=Yes 66 survey Q31S4 Feeling of applying cosmetics: nutritious feeling 7-3. What are the criteria for choosing the feeling of applying cosmetics? 1=No
2=Yes 67 survey Q31S5 Feeling of applying cosmetics: dry and soft feeling 7-3. What are the criteria for choosing the feeling of applying cosmetics? 1=No
2=Yes (soft feeling / dry and soft feeling) 68 survey Q32S1 Cosmetic properties: elasticity 7-4. What are the selection criteria for cosmetic properties? 1=No
2=Yes 69 survey Q32S2 Cosmetic properties: average 7-4. What are the selection criteria for cosmetic properties? 1=No
2=Yes 70 survey Q32S3 Cosmetic properties: light 7-4. What are the selection criteria for cosmetic properties? 1=No
2=Yes 71 survey Q32S4 Cosmetic properties: sticky feeling 7-4. What are the selection criteria for cosmetic properties? 1=No
2=Yes (heavy feeling) 72 survey Q32S5 Cosmetic properties: Type as smooth as water 7-4. What are the selection criteria for cosmetic properties? 1=No
2=Yes (water-smooth type) 73 survey Q33S1 fatty and sweet food 8-1. I often eat fatty and sweet foods. 0=Every day / Always.
1=Often / Yes.
2=Often / Usually.
3=Sometimes / No.
4=Not at all / Not at all. 74 survey Q34S1 outdoor activities 8-2. I often do outdoor activities (exercise, tanning, etc.). 0=almost never
1=Sometimes / Less than 3 times a week
2=4~5 times a week / 2~3 times a week / 4 or more times a week
3=Daily 75 survey Q35S1 sunscreen 8-3. I always use sunscreen. 0=Never use / Rarely use
1=Not often / Less than 3 times a week
2=Use as much as possible / During outdoor activities / More than 4 times a week
3=Daily 76 survey Q36S1 Skin worsening (diet) 8-4. My skin got worse due to dieting. 0=Not at all / Not at all.
1=Has it gotten worse/No.
2=Not severe/average.
3=Slightly severe / Yes.
4=Very severe / Very true. 77 survey Q38S1 amount of smoking 8-6. Smoke (exposure to secondhand smoke) 0=Not at all / I don't smoke at all
1=Sometimes / Smoking less than half of the time
2=Medium / smokes half to 1 pack
3=Always / Often / Smoking more than 1 pack 78 survey Q39S1 stress 8-7. I tend to get a lot of stress. 0=Always
1=Often
2=average
3=Sometimes
4=Not at all / Almost none 79 survey Q40S1 dermatological procedures 8-8. I often receive dermatological procedures. 0=Always
1=Often
2=average
3=Sometimes
4=Not at all / Almost none 80 survey Q41S1 pregnancy and birth experience 8-9. I have experience with pregnancy and childbirth. 1=None / None
2=Yes / Yes 81 survey Q42S1 Skin worsening (pregnancy, childbirth) 8-10. My skin got worse due to pregnancy and childbirth. -9=Not applicable. (No pregnancy/childbirth experience)
0=Not at all. / Not at all
1=No. / It has gotten worse
2=average. / not severe
3=Yes. / slightly severe
4=Very much so. / very severe 82 survey Q43S1 sleep time 8-11. get enough sleep 0=Not getting a good night's sleep at all. / Sleep less than 6 hours every night
1=1 time
2=2~3 times/sleeping for more than 6 hours 3 times a week
3=4~5 times/sleeping for more than 6 hours 5 times a week
4=Get a good night’s sleep every day. 83 survey Q44S1 Dead skin care 8-12. I often treat dead skin cells with peeling and scrubs. 0=No exfoliation treatment at all. / Almost never
1=1 time / more than once a week
2=2~3 times / 3 or more times a week
3=4~5 times
4=Exfoliate every day. 84 survey Q45S1 fingernail marks 8-13. If your nails are thick, your skin will turn red and swell. 0=Not at all.
1=No.
2=average / average
3=Yes.
4=Always like that. / it really is 85 survey Q46S1 cosmetic trouble 8-14. Trouble occurs when I change my cosmetics. 0=Not at all. / nothing at all
1=No.
2=average. / often
3=Yes.
4=Always like that. / often 86 survey Q47S1 T zone U zone trouble 8-15. I have troubles centered around the T zone (forehead, bridge of nose) and U zone (around chin). 0=Not at all. / nothing at all
1=No.
2=average. / often
3=Yes. / often
4=Always like that. / always

As shown in Table 1, the questions may be related to various areas and topics, and the customer can answer each question, for example, “Yes / No” or “Not at all / Not / Usually / Yes / Always”, etc. You can respond with Specifically, if the customer considers their skin type to be dry, they can respond yes/no to the question, and such customer responses will be recorded as numbers (1=no, 2=yes), for example. You can.

In one embodiment, the processor 110 may process the received response into the input data format of the machine learning model 200. The input data format may refer to a pre-specified data format in which the machine learning model 200 can perform analysis without preprocessing the data. That is, the processor 110 can process the received response into a data format suitable for analysis by the machine learning model 200.

For example, the processor 110 receives the customer's response to the survey, converts the response to each question into a number, and processes it into a data frame or series arranged in the order of the 87 questions listed in Table 1. can do. Specifically, the input data of the machine learning model 200 can be processed in the form of a 1 by 87 matrix as shown below.

['Q01S1', 'Q01S2', 'Q01S3', 'Q01S4', 'Q01S5', 'Q01S6', 'Q02S1', 'Q02S2', 'Q02S3', 'Q02S4', 'Q02S5', 'Q02S6', ' Q02S7', 'Q02S8', 'Q02S9', 'Q02S10', 'Q02S11', 'Q03S1', 'Q03S2', 'Q04S1', 'Q04S2', 'Q05S1', 'Q05S2', 'Q06S1', 'Q06S2' , 'Q07S1', 'Q08S1', 'Q08S2', 'Q09S1', 'Q10S1', 'Q11S1', 'Q12S1', 'Q13S1', 'Q14S1', 'Q15S1', 'Q16S1', 'Q17S1', ' Q18S1', 'Q19S1', 'Q20S1', 'Q21S1', 'Q22S1', 'Q23S1', 'Q24S1', 'Q25S1', 'Q26S1', 'Q27S1', 'Q28S1', 'Q29S1', 'Q29S2' , 'Q29S3', 'Q29S4', 'Q29S5', 'Q29S6', 'Q29S7', 'Q29S8', 'Q29S9', 'Q30S1', 'Q30S2', 'Q30S3', 'Q30S4', 'Q30S5', ' Q31S1', 'Q31S2', 'Q31S3', 'Q31S4', 'Q31S5', 'Q32S1', 'Q32S2', 'Q32S3', 'Q32S4', 'Q32S5', 'Q33S1', 'Q34S1', 'Q35S1' , 'Q36S1', 'Q38S1', 'Q39S1', 'Q40S1', 'Q41S1', 'Q42S1', 'Q43S1', 'Q44S1', 'Q45S1', 'Q46S1', 'Q47S1', 'Q48S1']

Meanwhile, here, each question (e.g. Q01S1, Q01S2, Q01S3, ..., Q48S1) can be entered with a number corresponding to the answer that can be selected. For example, “I have fine wrinkles on my face.” For the question, 4 (=always), 3 (=yes), 2 (=usually), 1 (=not), 0 (=not at all), or -9 (=missing value) can be entered. . Specifically, the processed input data may be as follows.

[2, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 2, 1, 2, 1, 2, 1 , 1, 2, 1, 1, 1, 0, 1, 1, 2, 2, 2, 3, 3, 2, 2, 3, 1, 1, 1, 1, 0, 2, 2, 2, 1 , 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 2, 1, 1, 1, 1, 2, 1, 1, 1, 1, 3, 1, 2 , 2, 0, 1, 4, 2, 1, 2, 0, 2, 0, 0, 0]

Below, the machine learning model 200 using the above-described customer response will be described with reference to FIG. 3.

FIG. 3 is a diagram illustrating the learning and analysis structure of the machine learning model 200 applied to the questionnaire-based customized genetic skin type prediction device 100 of FIG. 2.

According to one embodiment, the machine learning model 200 may be trained in advance to receive a customer's response 20 to a survey and derive a variable value 30 related to a genetic index related to skin risk. The learning data set 10 for learning the machine learning model 200, for example, may be extracted from 1000 pieces of raw data obtained through a survey and genetic testing conducted in advance on approximately 1000 customers. You can.

In other words, each learning data can be extracted from survey data answered by each customer and data on genetic information obtained by analyzing biological samples collected from the customer. Here, the biological sample may be, for example, tissue, urine, mucus, saliva, blood, or lymph fluid, but is not limited thereto.

In one embodiment, the genetic indicator related to skin risk may be a biomarker for predicting the risk of any one of wrinkles, moisture, pigmentation, oiliness, and sensitivity. Biomarkers refer to indicators that can identify the biological characteristics of a target object using biological materials such as proteins, DNA, and RNA. For example, a biomarker may be a SNP.

Specifically, the SNP associated with the risk of wrinkles may be at least one of rs117381658, rs1961184, rs1929013, and rs7042102, and the SNP associated with the risk of moisture may be rs9873353, rs34567709, rs1362404, rs7853290, rs143938096, and It can be at least one of rs12955989 and , the SNP associated with the risk of pigmentation may be at least one of rs74653330, rs34466224, rs11685354, rs4653497, rs59784607, and rs76548385, and the SNP associated with the risk of oiliness may be rs308971, rs151209785, and rs957791. 9, at least one of rs147804495, rs8107564 and rs6490805 may be, and the SNP associated with the risk of sensitivity may be at least one of rs7334780 and rs41308.

The machine learning model 200 for one embodiment converts response data to a customer's survey (QnA Data) and genetic marker data (SNP Data) related to the skin risk of the customer into a learning data set 10. The machine learning model 200 can be trained by configuring it. Accordingly, the correlation between responses to the survey and genetic indicators related to skin risk can be learned.

The machine learning model 200 may be learned based on various learning algorithms. For example, the learning algorithm may be any one of a decision tree classification algorithm, a random forest classification algorithm, and a support vector machine classification algorithm, but is not limited thereto.

According to one embodiment, the machine learning model 200 may be learned based on an ensemble learning algorithm. An ensemble learning algorithm may be a learning technique that learns through multiple learning algorithms and combines them to achieve improved learning performance. For example, the ensemble learning algorithm may include boosting techniques and bagging techniques.

The boosting technique is a technique that improves the performance of a learning model by sequentially learning and predicting multiple weak learners, and assigning weights to the next learner for data whose prediction is incorrect. The bagging technique is an abbreviation for Bootstrap Aggregating and is a technique to improve the deviation of the learning model by randomly sampling a given learning data set and training them through multiple learners.

If an ensemble learning algorithm is used when learning the machine learning model 200, the possibility of overfitting of the machine learning model 200 is reduced, and the analysis/prediction accuracy of the machine learning model 200 can be improved. .

Ensemble learning algorithms may include, for example, AdaBoost, Gradient Boost, Lightbgm, Xgboost, LGBMClassifier, and random forest algorithms. Table 2 below shows the results of a GWAS (genome-wide association study) analysis using the LGBMClassifier algorithm, and describes the analysis accuracy for SNPs related to skin risk.

Phenotypes SNP Precision Acc. Recall Acc. Wrinkle rs117381658 0.238 0.147 rs1961184 0.591 0.325 rs1929013 0.891 0.983 rs7042102 0.654 0.972 Moisture rs9873353 0.992 1.000 rs34567709 0.222 0.030 rs1362404 0.938 0.960 rs7853290 0.333 0.025 rs143938096 0.996 1.000 rs12955989 0.400 0.041 Melanin rs74653330 0.996 0.996 rs34466224 0.000 0.000 rs11685354 0.811 1.000 rs4653497 0.725 0.974 rs59784607 0.957 0.968 rs76548385 0.167 0.043 Oil rs308971 0.930 0.992 rs151209785 0.996 1.000 rs9577919 0.200 0.033 rs147804495 0.646 0.608 rs8107564 0.000 0.000 rs6490805 0.451 0.256 Sensitivity rs7334780 0.905 0.988 rs41308 0.515 0.443

The machine learning model (200) learned based on the LGBMClassifier algorithm is rs1929013 for the risk of wrinkles, rs9873353, rs1362404 and rs143938096 for the risk of moisture, and rs74653330, rs11685354, rs4653497 and rs5978460 for the risk of pigmentation. 7, Risk of oil For rs308971 and rs151209785, the risk prediction accuracy of rs7334780 for sensitivity was confirmed to be high.

In other words, as can be seen from the accuracy (precision accuracy, recall accuracy) listed in Table 2, the machine learning model 200 based on the LGBMClassifier algorithm has the above-mentioned SNPs for each of the five skin risks. Accuracy was high when predicting risk.

In one embodiment, the SNP for predicting the risk of wrinkles is rs1929013, the SNPs for predicting the risk of moisture are rs9873353, rs1362404, and rs143938096, and the SNPs for predicting the risk of pigmentation are rs74653330, rs11685354, rs4653497, and rs5978. 4607 The SNPs for predicting the risk for oiliness may be rs308971 and rs151209785, and the SNP for predicting the risk for sensitivity may be rs7334780.

Prediction of risk through SNP can be predicted based on the bases of the allele. For example, for a particular SNP that predicts risk for skin, an allele pair is predicted to be at risk for skin if it contains one or more bases that increase risk for skin, and skin if it does not contain any bases that increase risk. It can be predicted that there is no risk. Here, the allele with a base that increases the risk of skin is called an effect allele.

For example, in rs74653330, which predicts the risk for pigmentation, the effective allele that increases the risk for pigmentation may be C (cytosine). At this time, customers whose bases in the allele pair were analyzed to be at least one of CC, CA, and AC are predicted to be at risk for pigmentation, and customers whose bases in the allele pair were analyzed to be AA are predicted to have pigmentation. It can be predicted that there is no risk for deposition. In other words, the risk to the customer's skin can be predicted according to the results of analysis of the customer's effective alleles.

When response data (20) to a customer's questionnaire that wants to analyze skin type is input into the previously learned machine learning model (200), the machine learning model (200) generates variable values ( 30) can be derived.

According to one embodiment, the processor 110 converts the variable value 30 related to the derived genetic indicator into a binary code according to the presence or absence of skin risk, and variables related to the genetic indicator related to skin risk. The value (30) can be set to 0 or 1.

For example, if there is a risk for wrinkles, moisture, and pigmentation based on the genetic markers for a specific customer who responded to a survey, but there is no risk for oil and sensitivity, then there is a risk for wrinkles, moisture, and pigmentation. Variable values related to oil content and sensitivity can be converted to the number 1, and variable values related to oil content and sensitivity can be converted to the number 0. In this way, the variable value 30 related to the genetic indicator can be converted into a binary number corresponding to the presence or absence of skin risk.

Below, with reference to FIGS. 4A and 4B, the input data and output data of the machine learning model 200 will be described in more detail.

FIG. 4A is a diagram showing input data input to the machine learning model 200 according to an embodiment, and FIG. 4B is a diagram showing output data output from the machine learning model 200 according to an embodiment.

Referring to FIG. 4A, input data (20a, 20b, 20c, ...) input to the machine learning model 200 according to one embodiment is shown in the form of a data frame. The data frame of the input data (20a, 20b, 20c, ...) may be arranged in order from left to right, with responses to the 87 questions (QnAs) described above coded with numbers. Here, one row means one input data sample. An explanation of the responses to each question, indicated by numbers, is provided in the rightmost column of Table 1.

Referring to FIG. 4B, output data (30a, 30b, 30c, ...) output from the machine learning model 200 according to one embodiment is shown in the form of a data frame. The data frame of the output data (30a, 30b, 30c, ...) is coded and arranged in binary code based on the presence or absence of risk for the 11 SNPs (SNP_1, SNP_2, SNP_3, ...) described above from left to right. You can. Here, one row means one output data sample. The variable value for each SNP, expressed as a binary number, may have a value of 0.0, meaning there is no risk for the skin risk associated with the SNP, and a value of 1.0 may mean there is a risk.

In one embodiment, when the first input data 20a of FIG. 4A is input to the machine learning model 200, the first output data 30a of FIG. 4B may be output, and based on the first output data 30a, the first The skin type of the customer who provided the input data 20a may be determined.

In the output data 30a of the first row of Figure 4b, a value of 1.0 is assigned to all 11 SNPs (SNP_1, SNP_2, SNP_3, ...), which means that there is a risk for all skin risks associated with all SNPs. It can mean.

In summary, there is one SNP related to the risk of wrinkles and sensitivity (rs1929013 and rs7334780, respectively), so the risk of wrinkles and sensitivity can be predicted depending on the value of one SNP variable, and the SNP related to the risk of moisture and pigmentation There are three each (rs9873353, rs1362404, rs143938096 and rs74653330, rs11685354, rs4653497, respectively), so the sensitivity of moisture and pigmentation can be determined respectively depending on the combination of the three SNP variable values, and there are two SNPs related to the risk of oiliness. (rs308971, rs151209785), the risk of oily skin can be predicted according to the values of the two SNP variables.

The customer's skin type can be determined based on variable values 30 related to genetic indicators. For example, if the variable value (30) related to the genetic index derived for a specific customer includes information that there is a risk for wrinkles and moisture, and includes information that there is no risk for pigmentation, oil, and sensitivity, The customer's skin type can be determined as sagging or dry skin type.

According to one embodiment, skin types can be classified into 32 skin types based on variable values 30 related to genetic indicators. For example, if the variable value (30) related to the genetic indicator derived from the machine learning model (200) is converted to binary code according to the risk level for five skin characteristics, the presence/absence of risk for each of the five characteristics is determined. Accordingly, 32 skin types can be derived.

Below, with reference to Figure 5, the 32 skin types will be described in detail.

Figure 5 is a diagram showing 32 skin types classified through the questionnaire-based customized genetic skin type prediction device 100 according to an embodiment.

According to one embodiment, risks to the skin are divided into excessive oil (O) and insufficient oil (C) for oil, sensitive (S) and non-sensitive (N) for sensitivity, and wrinkles. It can be classified into elasticity (E) and sagging (W), moist (M) and dry (R) for moisture, and bright (B) and dull (D) for pigmentation.

The 32 skin types are predicted based on genetic indicators derived from customer survey results, and can provide more objective and accurate results than skin types predicted based on conventional simple surveys.

Skin types are not limited to the above-described embodiments, and more diverse types of skin types can be derived based on the type of skin risk or the diversity of variable values 30 related to genetic indicators related to risk.

In general, when the data applied to the machine learning model 200 changes, the optimal machine learning model 200 for analyzing the changed data may also change, so the machine learning model 200 needs to be continuously updated.

According to one embodiment, the machine learning model 200 may be updated using the received response 20 and the derived variable value 30. For example, input data to be analyzed through machine learning (e.g., customer response 20) is input and analyzed into the previously learned machine learning model 200, and then the input data and the analyzed output data (e.g. : The derived variable value (30) can be fed back to the machine learning model (200). Accordingly, the machine learning model 200 may be updated in a direction that minimizes the loss function value.

The updated machine learning model has improved performance compared to the existing previously learned machine learning model 200, and the analysis accuracy of skin risk can be improved.

Referring again to FIG. 2, the memory 120 is hardware that stores various data processed within the questionnaire-based customized genetic skin type prediction device 100, and includes data (e.g., data for processing and control of the processor 110). : Survey information, genetic marker information related to skin risk) can be saved.

Specifically, the memory 120 contains various items necessary for predicting skin type, such as prepared questionnaires, parameters of machine learning algorithms, biomarkers related to skin risk, customer survey result data, and genetic information data according to the customer's genetic test. Can contain data. In addition, the memory 120 can store the learning data set 10 for learning the machine learning model 200, input data related to survey responses, and derived output data in a database.

The memory 120 may be a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory 120 (for example, SD or XD memory (120, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) It may include at least one type of storage medium among (Only Memory), magnetic memory 120, magnetic disk, and optical disk.

The input unit 130 may be configured to allow the questionnaire-based customized genetic skin type prediction device 100 according to embodiments to receive questionnaire information. For example, the input unit 130 may include at least one of a user input device (e.g., keyboard, touch pad, touch screen), a scanner, and a camera that can receive responses to a questionnaire from a customer who wants to predict skin type. You can.

The output unit 140 may be a device for outputting result information derived from the questionnaire-based customized genetic skin type prediction device 100 according to embodiments. For example, the output unit 140 may include a display that can output the predicted customer's skin type, the predicted customer's genetic information, and a list of recommended products and services.

The questionnaire-based customized genetic skin type prediction device 100 according to an embodiment may further include a battery (not shown). The battery can supply the power necessary for the questionnaire-based customized genetic skin type prediction device 100 to operate. For example, the battery can supply power so that the processor 110 can operate, and can supply power necessary for the input unit 130 and output unit 140 to operate.

Figure 6 is a flowchart of a questionnaire-based customized genetic skin type prediction method according to an embodiment. The questionnaire-based customized genetic skin type prediction method may consist of steps processed in time series by a processor, and additional processing steps may be added between each step.

Referring to FIG. 6, steps for performing a questionnaire-based customized genetic skin type prediction method are shown. Each step may be implemented by a system including at least one processor. For example, the system may be, but is not limited to, a personal computer (PC), laptop, smartphone, wearable device, or tablet device.

In step 610, the processor may receive the customer's response to a pre-prepared questionnaire. The survey is information used to predict the customer's skin type, and may consist of questions about the customer's skin condition, skin concerns, eating habits, and selection criteria for cosmetics. For example, a pre-prepared questionnaire may include the contents listed in Table 1 above.

The survey can be conducted through a terminal of a company that provides skin type prediction services or through an Internet website. In other words, the survey method is not limited to a specific method such as an online/offline method, and the customer's response results to the survey can be transmitted to the processor through an input device, storage medium, or Internet network.

According to one embodiment, the processor may process the received response into the input data format of a machine learning model. Since the customer's response is input into a machine learning model to derive a genetic index regarding the risk level of the customer's skin, data processing can be performed into a form that can be interpreted by the learned machine learning model.

In other words, customer response data to surveys can be processed into a form that can be applied to a machine learning model. For example, responses to various questions included in a survey may be variableized or coded and converted or processed into a data frame.

The processed data may be in an optimized form for the learned machine learning model. Accordingly, the speed and accuracy of analysis of skin risk based on customer survey responses can be improved.

In step 630, the processor inputs the received response into a pre-trained machine learning model to derive variable values for genetic indicators related to skin risk.

The machine learning model may be learned to predict what genetic information the customer has related to skin risk through the customer's response data to the questionnaire. For example, a machine learning model can be learned by survey response data on a large number of customers conducted in advance and genetic information data derived from genetic testing for those customers. The learned machine learning model can use the customer's responses to the questionnaire to derive the customer's genetic information.

According to one embodiment, the processor may convert the variable value into binary code depending on the presence or absence of skin risk. For example, variable values for genetic indicators related to skin risk can be converted into binary codes depending on the presence or absence of risk based on the entered survey data. As variable values are simplified to 0 or 1, the learning speed, analysis speed, and analysis accuracy of the machine learning model can be improved.

In step 650, the processor may determine the customer's skin type based on the derived variable values. The derived variable value may, for example, be related to whether there is a risk or no risk with respect to the risk of the skin.

As another example, the derived variable value may be a variable quantified according to the degree or level of the risk of the skin. As another example, the derived variable values may be variables that indicate the presence or absence of risk or the degree of risk for the risk of skin classified into various types.

The customer's skin type may be determined according to variable values derived from the customer's survey data. For example, if there were many negative responses to questions related to wrinkles in the survey data, the variable value of the genetic index related to the risk of wrinkles can be derived as a value indicating the risk, and accordingly, the customer's skin type can be determined to be vulnerable to wrinkles.

According to one embodiment, the processor may determine the customer's skin type based on risk for wrinkles, moisture, pigmentation, moisture, and sensitivity.

The processor can convert variable values regarding the risk of wrinkles, moisture, pigmentation, moisture, and sensitivity into binary codes according to the presence or absence of risk, and accordingly, the customer's skin type can be determined as one of 32 classifications. However, the customer's skin type is not limited to this, and may appear more diversely depending on skin characteristics related to risk and variable values related to skin risk. For example, if the risk of wrinkles is considered separately by region of the face (e.g., around the mouth, around the eyes, and forehead), various skin types may appear in each area with respect to wrinkles.

The present embodiments may also be implemented in the form of a recording medium containing instructions executable by a computer, such as program modules executed by a computer. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include both computer storage media and communication media.

Computer storage media includes both volatile and non-volatile, 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. Communication media typically includes computer readable instructions, data structures, other data such as program modules, modulated data signals, or other transmission mechanisms, and includes any information delivery medium.

Additionally, in this specification, “unit” may be a hardware component such as a processor or circuit, and/or a software component executed by the hardware component such as a processor.

The description of the present specification described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. You will be able to. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as unitary may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

Meanwhile, those skilled in the art related to the embodiments will understand that the above-described substrate can be implemented in a modified form without departing from the essential characteristics. Therefore, the disclosed methods should be considered from an explanatory rather than a restrictive perspective. The scope of the invention is indicated in the claims, not the foregoing description, and all differences within the equivalent scope should be construed as being included in the invention.

100: Survey-based customized genetic skin type prediction device
110: processor
120: memory
130: input unit
140: output unit

Claims (15)

a memory storing at least one instruction; and
At least one processor executing the at least one instruction;
The processor,
Receiving customer responses to pre-prepared questionnaires, converting the customer responses into the format of input data for a pre-trained machine learning model,
Input the customer's response converted into the machine learning model to derive variable values for genetic indicators related to skin risk,
Determining the skin type of the customer based on information about the risk of the skin included in the variable value of the genetic index,
The machine learning model is,
Survey data is input and genetic information related to the skin of the customer who provided the survey data is output. Survey data conducted in advance on multiple customers and genetic information obtained through genetic testing of the multiple customers are output. is learned as training data,
The risk of the skin is,
Risk of wrinkles, moisture, pigmentation, oiliness and sensitivity,
Genetic indicators related to the risk of the skin are,
It is a biomarker to predict the risk of any one of wrinkles, moisture, pigmentation, oiliness, and sensitivity.
Biomarkers for predicting the risk of wrinkles are:
It is a SNP with an rs_id of at least one of rs117381658, rs1961184, rs1929013, and rs7042102,
Biomarkers for predicting the risk of moisture are:
It is a SNP with an rs_id of at least one of rs9873353, rs34567709, rs1362404, rs7853290, rs143938096, and rs12955989,
Biomarkers for predicting the risk of pigmentation are:
It is a SNP with an rs_id of at least one of rs74653330, rs34466224, rs11685354, rs4653497, rs59784607, and rs76548385,
Biomarkers for predicting the risk of the oil are:
It is a SNP with an rs_id of at least one of rs308971, rs151209785, rs9577919, rs147804495, rs8107564, and rs6490805,
Biomarkers for predicting the risk of sensitivity are:
A questionnaire-based customized genetic skin type prediction device that is a SNP with an rs_id of at least one of rs7334780 and rs41308. delete According to claim 1,
The processor,
A questionnaire-based customized genetic skin type prediction device that converts variable values related to the genetic indicator into a binary code corresponding to the presence or absence of skin risk. delete delete delete According to claim 1,
The skin type is,
A questionnaire-based customized genetic skin type prediction device that is classified into 32 skin types based on the variable values of the genetic indicators. According to claim 1,
The machine learning model is,
A questionnaire-based customized genetic skin type prediction device that is updated by receiving the customer's response and variable values related to the genetic index. According to claim 1,
The machine learning model is,
A survey-based customized genetic skin type prediction device that is learned based on an ensemble learning algorithm. A questionnaire-based customized genetic skin type prediction device receiving a customer's response to a pre-prepared questionnaire and converting the customer's response into the format of input data for a pre-trained machine learning model;
Inputting the customer's response converted by the questionnaire-based customized genetic skin type prediction device into the machine learning model to derive variable values for genetic indicators related to skin risk; and
A step wherein the questionnaire-based customized genetic skin type prediction device determines the skin type of the customer based on information about the risk of the skin included in the variable value of the genetic index,
The machine learning model is,
Survey data is input and genetic information related to the skin of the customer who provided the survey data is output. Survey data conducted in advance on multiple customers and genetic information obtained through genetic testing of the multiple customers are output. is learned as training data,
The risk of the skin is,
Risk of wrinkles, moisture, pigmentation, oiliness and sensitivity,
Genetic indicators related to the risk of the skin are,
It is a biomarker to predict the risk of any one of wrinkles, moisture, pigmentation, oiliness, and sensitivity.
Biomarkers for predicting the risk of wrinkles are:
It is a SNP with an rs_id of at least one of rs117381658, rs1961184, rs1929013, and rs7042102,
Biomarkers for predicting the risk of moisture are:
It is a SNP with an rs_id of at least one of rs9873353, rs34567709, rs1362404, rs7853290, rs143938096, and rs12955989,
Biomarkers for predicting the risk of pigmentation are:
It is a SNP with an rs_id of at least one of rs74653330, rs34466224, rs11685354, rs4653497, rs59784607, and rs76548385,
Biomarkers for predicting the risk of the oil are:
It is a SNP with an rs_id of at least one of rs308971, rs151209785, rs9577919, rs147804495, rs8107564, and rs6490805,
Biomarkers for predicting the risk of sensitivity are:
A survey-based customized genetic skin type prediction method that is a SNP with an rs_id of at least one of rs7334780 and rs41308. delete According to claim 10,
The step of deriving variable values for the genetic index is,
A questionnaire-based customized genetic skin type prediction method comprising converting the variable value into a binary code corresponding to the presence or absence of skin risk. According to claim 10,
The step of determining the customer's skin type is,
A survey-based customized genetic skin type prediction method comprising determining the customer's skin type based on the risk of wrinkles, moisture, pigmentation, oiliness, and sensitivity. According to claim 10,
Further comprising: updating the machine learning model by inputting the customer's response and variable values related to the genetic index into the machine learning model. A computer program coupled with hardware and stored on a medium for executing the method according to claim 10.

Images