• Title/Summary/Keyword: DNN 모델

Search Result 192, Processing Time 0.032 seconds

의료 수요 예측을 위한 딥러닝: 천식 발병을 중심으로

  • 김용선;강현욱;김정연;윤현식
    • Annual Conference of KIPS
    • /
    • 2024.10a
    • /
    • pp.384-388
    • /
    • 2024
  • 질병은 인간의 삶과 사회에 큰 영향을 미치며, 천식은 주변 사람들이 쉽게 그 심각성을 인지할 수 있는 질환이다. 천식의 유병률에도 불구하고 광범위한 데이터를 사용하여 천식 발병을 예측하는 연구가 부족하다. 그리하여 본 연구는 딥러닝 모델을 활용하여 기상데이터, 대기 오염 물질 수준, 천식 환자 수를 딥러닝 모델로 학습하고자 한다. 날짜, 위도, 경도 좌표를 기준으로 데이터가 병합되며 생성된 통합 데이터세트에서는 천식 환자 수를 결과변수로 사용하고, DNN과 LSTM 모델을 사용해 지도학습을 수행하며 모델의 성능은 MSE 및 MAE를 지표로 사용하여 평가한다. 본 연구는 질병 발생을 사전에 예측함으로써 잠재적인 질병 발생에 대한 사전 정보를 제공하고 의료인력 배치를 최적화하며 의료 서비스의 전반적인 효율성을 향상시켜 사회적 이익을 향상시키고자 한다.

Predicting the Number of Confirmed COVID-19 Cases Using Deep Learning Models with Search Term Frequency Data (검색어 빈도 데이터를 반영한 코로나 19 확진자수 예측 딥러닝 모델)

  • Sungwook Jung
    • KIPS Transactions on Software and Data Engineering
    • /
    • v.12 no.9
    • /
    • pp.387-398
    • /
    • 2023
  • The COVID-19 outbreak has significantly impacted human lifestyles and patterns. It was recommended to avoid face-to-face contact and over-crowded indoor places as much as possible as COVID-19 spreads through air, as well as through droplets or aerosols. Therefore, if a person who has contacted a COVID-19 patient or was at the place where the COVID-19 patient occurred is concerned that he/she may have been infected with COVID-19, it can be fully expected that he/she will search for COVID-19 symptoms on Google. In this study, an exploratory data analysis using deep learning models(DNN & LSTM) was conducted to see if we could predict the number of confirmed COVID-19 cases by summoning Google Trends, which played a major role in surveillance and management of influenza, again and combining it with data on the number of confirmed COVID-19 cases. In particular, search term frequency data used in this study are available publicly and do not invade privacy. When the deep neural network model was applied, Seoul (9.6 million) with the largest population in South Korea and Busan (3.4 million) with the second largest population recorded lower error rates when forecasting including search term frequency data. These analysis results demonstrate that search term frequency data plays an important role in cities with a population above a certain size. We also hope that these predictions can be used as evidentiary materials to decide policies, such as the deregulation or implementation of stronger preventive measures.

Multiple Discriminative DNNs for I-Vector Based Open-Set Language Recognition (I-벡터 기반 오픈세트 언어 인식을 위한 다중 판별 DNN)

  • Kang, Woo Hyun;Cho, Won Ik;Kang, Tae Gyoon;Kim, Nam Soo
    • The Journal of Korean Institute of Communications and Information Sciences
    • /
    • v.41 no.8
    • /
    • pp.958-964
    • /
    • 2016
  • In this paper, we propose an i-vector based language recognition system to identify the spoken language of the speaker, which uses multiple discriminative deep neural network (DNN) models analogous to the multi-class support vector machine (SVM) classification system. The proposed model was trained and tested using the i-vectors included in the NIST 2015 i-vector Machine Learning Challenge database, and shown to outperform the conventional language recognition methods such as cosine distance, SVM and softmax NN classifier in open-set experiments.

Deep Learning-based Antenna Selection Scheme for Millimeter-wave Systems in Urban Micro Cell Scenario (도심 Micro 셀 시나리오에서 밀리미터파 시스템을 위한 딥러닝 기반 안테나 선택 기법)

  • Ju, Sang-Lim;Kim, Nam-Il;Kim, Kyung-Seok
    • The Journal of the Institute of Internet, Broadcasting and Communication
    • /
    • v.20 no.5
    • /
    • pp.57-62
    • /
    • 2020
  • The millimeter wave that uses the spectrum in the 30GHz~300GHz band has a shorter wavelength due to its high carrier frequency, so it is suitable for Massive MIMO systems because more antennas can be equipped in the base station. However, since an RF chain is required per antenna, hardware cost and power consumption increase as the number of antennas increases. Therefore, in this paper, we investigate antenna selection schemes to solve this problem. In order to solve the problem of high computational complexity in the exhaustive search based antenna selection scheme, we propose a approach of applying deep learning technology. An best antenna combination is predicted using a DNN model capable of classifying multi-classes. By simulation tests, we compare and evaluate the existing antenna selection schemes and the proposed deep learning-based antenna selection scheme.

Comparison of Solar Power Generation Forecasting Performance in Daejeon and Busan Based on Preprocessing Methods and Artificial Intelligence Techniques: Using Meteorological Observation and Forecast Data (전처리 방법과 인공지능 모델 차이에 따른 대전과 부산의 태양광 발전량 예측성능 비교: 기상관측자료와 예보자료를 이용하여)

  • Chae-Yeon Shim;Gyeong-Min Baek;Hyun-Su Park;Jong-Yeon Park
    • Atmosphere
    • /
    • v.34 no.2
    • /
    • pp.177-185
    • /
    • 2024
  • As increasing global interest in renewable energy due to the ongoing climate crisis, there is a growing need for efficient technologies to manage such resources. This study focuses on the predictive skill of daily solar power generation using weather observation and forecast data. Meteorological data from the Korea Meteorological Administration and solar power generation data from the Korea Power Exchange were utilized for the period from January 2017 to May 2023, considering both inland (Daejeon) and coastal (Busan) regions. Temperature, wind speed, relative humidity, and precipitation were selected as relevant meteorological variables for solar power prediction. All data was preprocessed by removing their systematic components to use only their residuals and the residual of solar data were further processed with weighted adjustments for homoscedasticity. Four models, MLR (Multiple Linear Regression), RF (Random Forest), DNN (Deep Neural Network), and RNN (Recurrent Neural Network), were employed for solar power prediction and their performances were evaluated based on predicted values utilizing observed meteorological data (used as a reference), 1-day-ahead forecast data (referred to as fore1), and 2-day-ahead forecast data (fore2). DNN-based prediction model exhibits superior performance in both regions, with RNN performing the least effectively. However, MLR and RF demonstrate competitive performance comparable to DNN. The disparities in the performance of the four different models are less pronounced than anticipated, underscoring the pivotal role of fitting models using residuals. This emphasizes that the utilized preprocessing approach, specifically leveraging residuals, is poised to play a crucial role in the future of solar power generation forecasting.

Development of a Predictive Model forOccupational Disability Grades Using Workers'Compensation Insurance Data (산재보험 빅데이터를 활용한 장해등급 예측 모델 개발)

  • Choi, Keunho;Kim, Min Jeong;Lee, Jeonghwa
    • The Journal of Information Systems
    • /
    • v.33 no.3
    • /
    • pp.187-205
    • /
    • 2024
  • Purpose A prediction model for occupational injuries can support more proactive, efficient, and effective policy-making. This study aims to develop a model that predicts the severity of occupational injuries, classified into 15 disability grades in South Korea, using machine learning techniques applied to COMWEL data. The primary goal is to improve prediction accuracy, offering an advanced tool for early intervention and evidence-based policy implementation. Design/methodology/approach The data analyzed in this study consists of 290,157 administrative records of occupational injury cases collected between 2018 and 2020 by the Korea Workers' Compensation & Welfare Service, based on the 'Workers' Compensation Insurance Application Form' submitted for occupational injury treatment. Four machine learning models - Decision Tree, DNN, XGBoost, and LightGBM - were developed and their performances compared to identify the optimal model. Additionally, the Permutation Feature Importance (PFI) method was used to assess the relative contribution of each variable to the model's performance, helping to identify key variables. Findings The DNN algorithm achieved the lowest Mean Absolute Error (MAE) of 0.7276. Key variables for predicting disability grades included the severity index, primary disease code, primary disease site, age at the time of the injury, and industry type. These findings highlight the importance of early policy intervention and emphasize the role of both medical and socioeconomic factors in model predictions. The academic and policy implications of these results were also discussed.

Application of Machine Learning Techniques for the Classification of Source Code Vulnerability (소스코드 취약성 분류를 위한 기계학습 기법의 적용)

  • Lee, Won-Kyung;Lee, Min-Ju;Seo, DongSu
    • Journal of the Korea Institute of Information Security & Cryptology
    • /
    • v.30 no.4
    • /
    • pp.735-743
    • /
    • 2020
  • Secure coding is a technique that detects malicious attack or unexpected errors to make software systems resilient against such circumstances. In many cases secure coding relies on static analysis tools to find vulnerable patterns and contaminated data in advance. However, secure coding has the disadvantage of being dependent on rule-sets, and accurate diagnosis is difficult as the complexity of static analysis tools increases. In order to support secure coding, we apply machine learning techniques, such as DNN, CNN and RNN to investigate into finding major weakness patterns shown in secure development coding guides and present machine learning models and experimental results. We believe that machine learning techniques can support detecting security weakness along with static analysis techniques.

Arc Detection using Logistic Regression (로지스틱 회기를 이용한 아크 검출)

  • Kim, Manbae
    • Journal of Broadcast Engineering
    • /
    • v.26 no.5
    • /
    • pp.566-574
    • /
    • 2021
  • The arc is one of factors causing electrical fires. Over past decades, various researches have been carried out to detect arc occurrences. Even though frequency analysis, wavelet and statistical features have been used, arc detection performance is degraded due to diverse arc waveforms. On the contray, Deep neural network (DNN) direcly utilizes raw data without feature extraction, based on end-to-end learning. However, a disadvantage of the DNN is processing complexity, posing the difficulty of being migrated into a termnial device. To solve this, this paper proposes an arc detection method using a logistic regression that is one of simple machine learning methods.

Prediction of KBO playoff Using the Deep Neural Network (DNN을 활용한 'KBO' 플레이오프진출 팀 예측)

  • Ju-Hyeok Park;Yang-Jae Lee;Hee-Chang Han;Yoo-Lim Jun;Yoo-Jin Moon
    • Proceedings of the Korean Society of Computer Information Conference
    • /
    • 2023.01a
    • /
    • pp.315-316
    • /
    • 2023
  • 본 논문에서는 딥러닝을 활용하여 KBO (Korea Baseball Organization)의 다음 시즌 플레이오프 진출 확률을 예측하는 Deep Neural Network (DNN) 시스템을 설계하고 구현하는 방법을 제안한다. 연구 방법으로 KBO 각 시즌별 데이터를 1999년도 데이터부터 수집하여 분석한 결과, 각 시즌 데이터 중 경기당 평균 득점, 타자 OPS, 투수 WHIP 등이 시즌 결과에 유의미한 영향을 끼치는 것을 확인하였다. 모델 설계는 linear, softmax 함수를 사용하는 것보다 relu, tanh, sigmoid 함수를 사용했을 때 더 높은 정확도를 얻을 수 있었다. 실제 2022 시즌 결과를 예측한 결과 88%의 정확도를 도출했다. 폭투의 수, 피홈런 등 가중치가 높은 변수의 값이 우수할 경우 시즌 결과가 좋게 나온다는 것이 증명되었다. 본 논문에서 설계한 이 시스템은 KBO 구단만이 아닌 모든 야구단에서 선수단을 구성하는데 활용 가능하다고 사료된다.

  • PDF

AI Accelerator Design for Edge Devices (엣지 디바이스를 위한 AI 가속기 설계 방법)

  • Whoi Ree, Ha;Hyunjun Kim;Yunheung Paek
    • Annual Conference of KIPS
    • /
    • 2024.05a
    • /
    • pp.723-726
    • /
    • 2024
  • 단일 dataflow 를 지원하는 DNN 가속기는 자원 효율적인 성능을 보이지만, 여러 DNN 모델에 대해서 가속 효과가 제한적입니다. 반면에 모든 dataflow 를 지원하여 매 레이어마다 최적의 dataflow를 사용하여 가속하는 reconfigurable dataflow accelerator (RDA)는 굉장한 가속 효과를 보이지만 여러 dataflow 를 지원하는 과정에서 필요한 추가 하드웨어로 인하여 효율적이지 못합니다. 따라서 본 연구는 제한된 dataflow 만을 지원하여 추가 하드웨어 요구사항을 감소시키고, 중복되는 하드웨어의 재사용을 통해 최적화하는 새로운 가속기 설계를 제안합니다. 이 방식은 자원적 한계가 뚜렷한 엣지 디바이스에 RDA 방식을 적용하는데 필수적이며, 기존 RDA 의 단점을 최소화하여 성능과 자원 효율성의 최적점을 달성합니다. 실험 결과, 제안된 가속기는 기존 RDA 대비 32% 더 높은 에너지 효율을 보이며, latency 는 불과 1%의 차이를 보였습니다.