선택한 단어 수는 입니다.
최소 단어 이상 선택하여야 합니다.
최소 단어 이상 선택하여야 합니다.
선택한 단어 수는 30입니다.
최대 10 단어까지만 선택 가능합니다.
최대 10 단어까지만 선택 가능합니다.
문제 정의
- Logistic regression analysis was adopted in this study in order to determine the presence of gender differences in the manifestation of skin characteristics. We also used a Gaussian mixture model to group the subjects according to the measured skin characteristics, and the decision tree model was used for lifestyle assessment for each group.
- This study aims to explore and determine the relationship between skin type and lifestyle, by applying a data mining technique yet not attempted in the dermatology sector. As mentioned above, skin types are currently being classified based on the independent examination of the individual skin characteristics and the setting of threshold values.
- Specifically, our study is distinctive in that the decision tree method was employed to derive the correlations between skin type and life style, as well as the rules of classification pertaining to each skin type, based on the demographic elements and the selfreported life style and skincare status. This study is also important for researching the specific skin characteristics of Koreans.
제안 방법
- , the skin type, as well as the regular patterns associated with the same. All the results of quantitative analysis were integrated and arranged according to the possible similarities between the skin-type characteristics of the subjects grouped together in each cluster and their association with the life styles of the subjects.
- Future research can be directed to the observation and analysis of the changes in skin type, by applying a study design where the lifestyle-related conditions of the subjects of the same skin type are artificially varied, in order to derive more accurate relationships between lifestyle and skin type. Another limitation of this study is that the changes in skin conditions with respect to the seasonal factors in Korea, i.e., in a temperate climate zone with four seasons, were not analyzed. We believe that the careful analysis of these aspects in future research could assist in a more comprehensive and accurate determination of the skin characteristics.
- As briefly mentioned above, we used GMM clustering, in order to group the subjects, and attempted a skin type classification by analyzing the 8 skin-related features in each group. In order to determine the optimal number of clusters, we observed the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) while increasing the number of clusters.
- Based on the data collected from both a questionnaire survey and skin measurement of 1,246 subjects, we conducted analyses by using the multivariate statistical analysis and data mining techniques, in order to determine the relationship between skin type and life style. Because of the gender-dependent significant differences in skin type as determined by logistic regression, we conducted individual analyses for male and female subjects.
- Because of the gender-dependent significant differences in skin type as determined by logistic regression, we conducted individual analyses for male and female subjects. Based on the data measuring the skin conditions, we defined two clusters for male subjects and four clusters for female subjects, by using the GMM clustering technique. We then subjected the variables to ANOVA and C4.
- In contrast, this study adopted a cluster analysis, where the subjects were grouped into clusters, with respect to the interrelated skin-specific measurement values. Based on the results, we performed a classification modeling, where we represented the identified clusters as a response variable, and the lifestylerelated variables as independent variables. The data on current lifestyle trends were collected from the subjects by means of a questionnaire.
- As mentioned above, skin types are currently being classified based on the independent examination of the individual skin characteristics and the setting of threshold values. In contrast, this study adopted a cluster analysis, where the subjects were grouped into clusters, with respect to the interrelated skin-specific measurement values. Based on the results, we performed a classification modeling, where we represented the identified clusters as a response variable, and the lifestylerelated variables as independent variables.
- As briefly mentioned above, we used GMM clustering, in order to group the subjects, and attempted a skin type classification by analyzing the 8 skin-related features in each group. In order to determine the optimal number of clusters, we observed the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) while increasing the number of clusters. In general, the optimal number is determined as the one representing the smallest AIC and BIC values.
- Prior to cluster analysis by using the 8 measured variables pertaining to skin conditions, a pre-testing process was conducted to determine whether each measured variable shows a gender-dependent difference in distribution. In order to implement the test, we performed a logistic regression analysis by setting each of the 8 measured variables as an independent variable (X), and the gender as the response variable. Specifically, 8 models were generated by setting the likelihood of the subject belonging to the male class set to be p.
- GMM can also be applied efficiently during the pre-treatment stage, as it can be used simultaneously for clustering and the detection of outliers (Roberts, 1999). In this study, we have attempted to determine the optimal number of clusters by applying GMM to the data (representative of the measurement of skin characteristics), and simultaneously classify the skin type of the subjects (by grouping them into several clusters according to their skin characteristics).
- In this section, we will provide an overview of the effects of lifestyle including cosmetic usage on the skin condition based on the results of analysis presented in the previous sections. The subjects were classified into clusters by cluster analysis, associating the general skin conditions quantitatively measured by a device. In addition, the life styles of the different clusters were analyzed by using the ANOVA and the decision tree analysis technique.
- After preprocessing the data, 1,214 subjects were included in the data mining analysis. The variables of the questionnaire were converted to values between 0-1 for scale adjustment, and a binary derived variable was added for the female subjects, in order to separately examine the effects of menstruation on the status of the skin. This assisted in the addition of the factor, ‘irregular menstruation’ to the existing ‘menopausal status’ and ‘regular menstruation’ factors.
- Most existing studies have been conducted by using a univariate approach, where the skin types were classified by independent examination of the individual skin-specific values, and the setting of threshold values. Therefore, this study is significantly different in that the skin types were determined by adopting a multivariate approach, where the measured skin-specific values for all subjects, and their correlations were simultaneously taken into account. To achieve this, we used data mining techniques.
- Under the assumption that the clusters determined in our cluster analysis represented the skin types of the individual subjects, we performed ANOVA and a decision tree analysis, in order to determine the relationship between skin type and life style.
대상 데이터
- We deleted 32 subjects either having missing values or outlying. After preprocessing the data, 1,214 subjects were included in the data mining analysis. The variables of the questionnaire were converted to values between 0-1 for scale adjustment, and a binary derived variable was added for the female subjects, in order to separately examine the effects of menstruation on the status of the skin.
- In this survey, we included 1,246 adults aged 16-75 (mean 41±16 standard deviation (SD); female: 934, male:312).
데이터처리
- The subjects were classified into clusters by cluster analysis, associating the general skin conditions quantitatively measured by a device. In addition, the life styles of the different clusters were analyzed by using the ANOVA and the decision tree analysis technique. Table 8 summarizes the cluster-specific characteristics, integrating the results of all the analyses.
- As observed in the two figures, the female tree intended to classify four clusters has more complicated tree structure than the male tree classifying two clusters. Therefore, more various independent variables influencing the classification of each cluster were selected in the female tree, which is consistent with the obtained results of the ANOVA. As seen in Figure 4, the use of treatment product was revealed to occupy the top root node of the male tree, as the factor with the largest influence on the classification of the two clusters.
이론/모형
- , skin type. As mentioned above, we used the C4.5 algorithm, and the final model was obtained by using the reduced error pruning method. As a result, the final trees were found to have a training error of approximately 26% in both male and female datasets.
- 5 uses a pruning algorithm to prevent the side-effect of over fitting of the final model with the data. The reduced error pruning method was used in this study.
- Based on the data measuring the skin conditions, we defined two clusters for male subjects and four clusters for female subjects, by using the GMM clustering technique. We then subjected the variables to ANOVA and C4.5 technique (a decision tree analysis technique), in order to determine the relationships between lifestyle of the subjects and the characteristics of each cluster, i.e., the skin type, as well as the regular patterns associated with the same. All the results of quantitative analysis were integrated and arranged according to the possible similarities between the skin-type characteristics of the subjects grouped together in each cluster and their association with the life styles of the subjects.
성능/효과
- 5 algorithm, and the final model was obtained by using the reduced error pruning method. As a result, the final trees were found to have a training error of approximately 26% in both male and female datasets. This means that approximately 74% of subjects in the training datasets are correctly classified by the trees.
- (2014) has studied the influence of gender differences on skin elasticity, focusing on the agedependent changes in Caucasian men and women. It was discovered that skin elasticity shows different changing patterns in males and females, with menopause triggering a rapid reduction in skin elasticity. From this, it could be inferred that skin properties demonstrate genderrelated differences.
후속연구
- A limitation of this study is that we did not conduct a longitudinal study to determine the intra-subject changes of skin-specific values over time. Future research can be directed to the observation and analysis of the changes in skin type, by applying a study design where the lifestyle-related conditions of the subjects of the same skin type are artificially varied, in order to derive more accurate relationships between lifestyle and skin type. Another limitation of this study is that the changes in skin conditions with respect to the seasonal factors in Korea, i.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.