XGBoost와 SHAP 기법을 활용한 근로자 이직 예측에 관한 연구

이재준 ( Lee Jae Jun ),이유린 ( Lee Yu Rin ),임도현 ( Lim Do Hyun ),안현철 ( Ahn Hyun Chul ) 한국정보시스템학회 2021 情報시스템硏究 Vol.30 No.4

Purpose In order for companies to continue to grow, they should properly manage human resources, which are the core of corporate competitiveness. Employee turnover means the loss of talent in the workforce. When an employee voluntarily leaves his or her company, it will lose hiring and training cost and lead to the withdrawal of key personnel and new costs to train a new employee. From an employee's viewpoint, moving to another company is also risky because it can be time consuming and costly. Therefore, in order to reduce the social and economic costs caused by employee turnover, it is necessary to accurately predict employee turnover intention, identify the factors affecting employee turnover, and manage them appropriately in the company. Design/methodology/approach Prior studies have mainly used logistic regression and decision trees, which have explanatory power but poor predictive accuracy. In order to develop a more accurate prediction model, XGBoost is proposed as the classification technique. Then, to compensate for the lack of explainability, SHAP, one of the XAI techniques, is applied. As a result, the prediction accuracy of the proposed model is improved compared to the conventional methods such as LOGIT and Decision Trees. By applying SHAP to the proposed model, the factors affecting the overall employee turnover intention as well as a specific sample’s turnover intention are identified. Findings Experimental results show that the prediction accuracy of XGBoost is superior to that of logistic regression and decision trees. Using SHAP, we find that jobseeking, annuity, eng_test, comm_temp, seti_dev, seti_money, equl_ablt, and sati_safe significantly affect overall employee turnover intention. In addition, it is confirmed that the factors affecting an individual’s turnover intention are more diverse. Our research findings imply that companies should adopt a personalized approach for each employee in order to effectively prevent his or her turnover.

RVDT용 DSP 기반 위상 자동보정 디지털 신호처리기 FPGA 구현

김성미,서연호,진유린,이민웅,조성익,이종열,Kim, Sung-mi,Seo, Yeon-ho,Jin, Yu-rin,Lee, Min-woong,Cho, Seong-ik,Lee, Jong-yeol 한국정보통신학회 2017 한국정보통신학회논문지 Vol.21 No.6

RVDT(Rotary Variable Differential Transformer)는 각도 변위를 측정하는 센서로써 출력 신호는 DSBSC-AM(Double SideBand Suppressed Carrier AM) 신호이기 때문에 출력 신호로부터 각도 변위를 알아내기 위하여 DSBSC-AM 복조 과정이 필요하다. 본 논문에서는 DSBSC-AM 신호의 복조기인 코스타스 루프를 수정하여 RVDT 출력 신호로부터 각도 변위를 추출하는 DADC(Digital Angle to DC)를 FPGA(Field Programmable Gate Array)로 구현하였다. 본 논문에서 설계된 DADC는 4선식과 5선식 RVDT에 적용가능하며, 코스타스 루프의 사용으로 기존의 아날로그 신호처리기와는 달리 외부의 소자를 사용하지 않고 RVDT 입력여기신호와 출력신호 사이의 위상 차이를 정확하게 보정할 수 있다. 또한 선형성 향상을 위하여 디지털 신호처리 기법이 적용되어 DADC는 기존의 아날로그 신호처리기의 선형성 오차 0.05%보다 적은 0.035%의 선형성 오차를 보였다. 구현된 DADC의 기능과 성능 테스트는 상용 RVDT 센서와 ADC(Analog to Digital Converter), 아날로그 출력단으로 구성된 통합 실험환경을 구성하여 진행하였다. A RVDT is a sensor that measures angular displacement and the output signal of RVDT is a DSBSC-AM signal. For this reason, a DSBSC-AM demodulation processor is required to determine the angular displacement from the output signal. In this paper, DADC(Digital Angle to DC) which extracts the angular displacement from the output signal of a RVDT is implemented based-on modified Costas Loop usually used in the demodulation of DSBSC-AM signal by using FPGA. DADC can used with both 4-wire and 5-wire RVDTs and can exactly compensate the phase difference between the input excitation and output signals of a RVDT unlike the conventional analog RVDT signal conditioners which require external components. Since digital signal processing technique that can enhance the linearity is exploited, DADC shows 0.035% linearity error, which is smaller than 0.005% that of a conventional analog signal conditioner. The DADC are tested in an integrated experimental environment which includes a commercial RVDT sensor, ADC and an analog output block.

AI를 활용한 실시간 수도권 약수터 수질 예측 모델 개발

엄지연(Ji-Yeon Eom),이유린(Yu-Rin Lee),임도현(Do-Hyeon Lim),임영우(Yeong-Woo Lim),안현철(Hyun-Chul Ahn),곽기영(Kee-Young Kwahk) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11

코로나19 팬데믹 전까지 수많은 사람들이 이용해온 약수터는 수도권에만 약 6백여 개가 위치해 있다. 불규칙적이고 수작업으로 수행되는 수질검사로 인해 사람들은 실시간으로 검사 결과를 알 수 없다. 본 연구에서는 약수터 수질에 영향을 미치는 요인을 탐색하고 다양한 곳에 흩어져 있는 데이터를 수집하여 실시간으로 약수터 수질을 예측할 수 있는 모델을 개발하고자 한다. There are about 600 mineral springs in the metropolitan area alone that have been used by countless people before the COVID-19 pandemic. Due to irregular and manual water quality tests, people cannot know the test results in real time. In this study, we intend to develop a model that can predict the mineral spring water quality in real time by exploring factors affecting the mineral spring water quality and collecting data scattered in various places.

직렬 커패시터 D/A 변환기를 갖는 저전력 축차 비교형 A/D 변환기

이정현(Jeong-Hyeon Lee),진유린(Yu-Rin Jin),조성익(Seong-Ik Cho) 대한전기학회 2019 전기학회논문지 Vol.68 No.1

The charge redistribution digital-to-analog converter(CR-DAC) is often used for successive approximation register analog-to-digital converter(SAR ADC) that requiring low power consumption and small circuit area. However, CR-DAC is required 2 to the power of N unit capacitors to generate reference voltage for successive approximation of the N-bit SAR ADC, and many unit capacitors occupy large circuit area and consume more power. In order to improve this problem, this paper proposes SAR ADC using series capacitor DAC. The series capacitor DAC is required 2(1+N) unit capacitors to generate reference voltage for successive approximation and charges only two capacitors of the reference generation block. Because of these structural characteristics, the SAR ADC using series capacitor DAC can reduce the power consumption and circuit area. Proposed SAR ADC was designed in CMOS 180nm process, and at 1.8V supply voltage and 500kS/s sampling rate, proposed 6-bit SAR ADC have signal-to-noise and distortion ratio(SNDR) of 36.49dB, effective number of bits(ENOB) of 5.77-bit, power consumption of 294uW.

LightGBM 기반의 학업중단 예측과 SHAP을 활용한 해석

임도현(Do Hyeon Lim),이유린(Yu-rin Lee),이재준(Jaejun Lee),곽기영(Kee-Young Kwahk),안현철(Hyunchul Ahn) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11

중, 고등학교에서의 교육은 미래 사회의 구성원을 성장시키는데 있어 중요한 역할을 하지만 지속적으로 적지 않은 수의 학생들이 학업 중단을 결정하고 있다. 학업 중단은 사회 · 경제적 손실을 야기하고, 이를 줄이기 위해 청소년의 학업 중단을 예측하는 것의 필요성이 대두되고 있다. 또한 예측에서 그치지 않고 학업 중단에 영향을 미치는 요인들을 파악해 이를 개선할 수 있는 대책을 마련하는 것이 필요하다 이에 본 논문은 LightGBM을 이용하여 학업 중단의 예측 정확도를 제고하고 기법을 통해 학업 중단에 영향을 미치는 요인에 대한 해석을 시도해 보고자 하고자 한다. Education in middle and high schools have an important role in fostering members of our future society. Thus, drop-out may cause socio-economic losses. To mitigate these losses, it is needed to predict school dropout. In addition, it is necessary to identify the factors that affect dropout, and prepare for the solution to improve it. In this paper, we aim to improve the accuracy of the school dropout prediction model using LightGBM, and provide an interpretation of the factors affecting dropout through SHAP.

악관절 장애 증상이 구강건강 상태와 정신건강 상태, 삶의 질에 미치는 영향 : 국민건강영양조사 2012년도 자료를 활용하여

강현경 ( Hyun-kyung Kang ),이지영 ( Ji-young Lee ),김유린 ( Yu-rin Kim ) 한국치위생학회 2021 한국치위생학회지 Vol.21 No.6

Objectives: The purpose of this study was to examine the effects of oral and mental health status on temporomandibular joint disorders (TMDs) and the effects of TMD symptoms on quality of life using data from the 5th 2012 National Health and Nutrition Examination Survey. Methods: A total of 1,154 people were selected as subjects for the study. Temporomandibular joint (TMJ) sounds (672 persons), TMJ pain (227 persons), and TMJ dislocation (255 persons) were the factors studied. The complex sample chi-square test was performed to compare demographic characteristics according to the three groups of TMD symptoms. Complex sample logistic regression analysis was performed to confirm the effect of oral and mental health status on TMDs, and complex sample linear regression analysis was performed to check the effect of TMDs on the quality of life (EQ-5D). Results: After adjusting for demographic characteristics, those without TMJ sound symptoms had lower pain/discomfort (OR: 0.055, CI: -0.095 to -0.016) and anxiety/depression (OR: 0.053, CI: -0.092 to -0.014). Those without TMJ pain had lower pain/discomfort (OR: 0.119, CI: -0.192 to -0.046) and anxiety/depression (OR: 0.071, CI: -0.137 to -0.004). Pain/discomfort (OR: 0.063, CI: -0.125 to -0.001) was lower in those without TMJ dislocation symptoms. After adjusting for mental health status, pain/discomfort (OR: 0.088, CI: -0.161 to -0.014) was found to be lower in those without TMJ pain symptoms (p<0.05). Conclusions: Based on the results of this study, the treatment of TMDs, oral health, and also mental health, is needed to improve the quality of life.

치위생학 전공 학생의 환경의식 관련 교육이 의료폐기물 지식과 태도에 미치는 영향

박미정(Mi-Jung Park),강현경(Hyun-Kyung Kang),김유린(Yu-Rin Kim),이지영(Ji-Young Lee) 한국구강보건과학회 2023 한국구강보건과학회지 Vol.11 No.1

Objectives: The purpose of this study was to identify differences in knowledge and attitudes about medical waste according to environmental education among dental hygiene students. Methods: In this study, 129 students were analyzed through a Google questionnaire from October to December 2022, targeting university students enrolled in dental hygiene departments. Results: The effect of environmental education on medical waste knowledge was 0.234, and the effect on medical waste attitude was 0.042. Only the effect of environmental education on medical waste knowledge was statistically significant (p<0.05). Conclusions: The study findings indicated a difference in knowledge about medical waste among students who received environmental education, but no difference was noted in attitude. Therefore, methods are needed that will further raise the level of attitude toward medical waste.

부산 일부 대학교 학생들의 니코틴 의존도와 구취의 연관성

김유린(외 5명) 신라대학교 자연과학연구소 2012 自然科學論文集 Vol.22 No.-

The purpose of this study is to examine the association of bad breath and nicotine dependency. By use of the findings from this study, This study conducted the survey of 53 smoker who visited dental hygiene lab for delivery from October to November 2012 with structured questionnaires(Fagerstrom Tolerance Questionnaire FTQ) and equipment(CO measuring instrument, Oral Chroma etc). The data were analyzed by using SPSS 20.0 program for χ2-test, ANOVA. Smokers' average age was 23 years old. The more higher nicotine dependence the more the average age was higher. The more higher nicotine dependence the less to nque brushing. No association of nicotine dependence and self-bad breath level. CO State was Significant differences(ND1(M=2.58), ND2(M=3.98), ND3(M=7.41), ND4(M=4.19), ND5(M=13.14))(F=2.616, p=.047). Dimethyl Sulfide was Significant differences(F=3.026, p=.026).