• Journal of Multimedia Information System
• /
• 제6권4호
• /
• pp.225-234
• /
• 2019

Deep learning has become increasingly popular in both academic and industrial areas nowadays. Various domains including pattern recognition, Computer vision have witnessed the great power of deep neural networks. However, current studies on deep learning mainly focus on quality data sets with balanced class labels, while training on bad and imbalanced data set have been providing great challenges for classification tasks. We propose in this paper a method of data analysis-based data reduction techniques for selecting good and diversity data samples from a large dataset for a deep learning model. Furthermore, data sampling techniques could be applied to decrease the large size of raw data by retrieving its useful knowledge as representatives. Therefore, instead of dealing with large size of raw data, we can use some data reduction techniques to sample data without losing important information. We group PCB characters in classes and train deep learning on the ResNet56 v2 and SENet model in order to improve the classification performance of optical character recognition (OCR) character classifier.

• Journal of the Korean Data and Information Science Society
• /
• 제7권2호
• /
• pp.203-210
• /
• 1996

A linear neural unit with a modified anti-Hebbian learning rule has been shown to be able to optimally fit curves, surfaces, and hypersurfaces by adaptively extracting the minor component of the input data set. In this paper, we study how to use the robust version of this neural fitting method for linear regression analysis. Furthermore, we compare this method with other methods when data set is contaminated by outliers.

• Tuberculosis and Respiratory Diseases
• /
• 제86권3호
• /
• pp.203-215
• /
• 2023

Background: Surgical resection is the standard treatment for early-stage lung cancer. Since postoperative lung function is related to mortality, predicted postoperative lung function is used to determine the treatment modality. The aim of this study was to evaluate the predictive performance of linear regression and machine learning models. Methods: We extracted data from the Clinical Data Warehouse and developed three sets: set I, the linear regression model; set II, machine learning models omitting the missing data: and set III, machine learning models imputing the missing data. Six machine learning models, the least absolute shrinkage and selection operator (LASSO), Ridge regression, ElasticNet, Random Forest, eXtreme gradient boosting (XGBoost), and the light gradient boosting machine (LightGBM) were implemented. The forced expiratory volume in 1 second measured 6 months after surgery was defined as the outcome. Five-fold cross-validation was performed for hyperparameter tuning of the machine learning models. The dataset was split into training and test datasets at a 70:30 ratio. Implementation was done after dataset splitting in set III. Predictive performance was evaluated by R² and mean squared error (MSE) in the three sets. Results: A total of 1,487 patients were included in sets I and III and 896 patients were included in set II. In set I, the R² value was 0.27 and in set II, LightGBM was the best model with the highest R² value of 0.5 and the lowest MSE of 154.95. In set III, LightGBM was the best model with the highest R² value of 0.56 and the lowest MSE of 174.07. Conclusion: The LightGBM model showed the best performance in predicting postoperative lung function.

• International Journal of Advanced Culture Technology
• /
• 제5권2호
• /
• pp.74-81
• /
• 2017

Recently, there are many companies that use systems based on artificial intelligence. The accuracy of artificial intelligence depends on the amount of learning data and the appropriate algorithm. However, it is not easy to obtain learning data with a large number of entity. Less data set have large generalization errors due to overfitting. In order to minimize this generalization error, this study proposed DGA(Dropout Genetic Algorithm) which can expect relatively high accuracy even though data with a less data set is applied to machine learning based genetic algorithm to deep learning based dropout. The idea of this paper is to determine the active state of the nodes. Using Gradient about loss function, A new fitness function is defined. Proposed Algorithm DGA is supplementing stochastic inconsistency about Dropout. Also DGA solved problem by the complexity of the fitness function and expression range of the model about Genetic Algorithm As a result of experiments using MNIST data proposed algorithm accuracy is 75.3%. Using only Dropout algorithm accuracy is 41.4%. It is shown that DGA is better than using only dropout.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제2권10호
• /
• pp.705-712
• /
• 2013

능동적 학습(active learning)은 소수의 라벨 데이터로 구성된 훈련 집합이 주어진 경우에 분류기 학습에 가장 도움이 될 만한 언라벨드 데이터를 선택하여 전문가에 의한 라벨링을 통해 훈련 집합에 포함시키는 과정을 반복함으로써 분류기의 성능을 향상시키는 것을 목적으로 한다. 본 논문에서는 워드 연결(ward's linkage)을 이용한 계층적 군집화(hierarchical clustering)를 바탕으로 한 능동적 학습 방법을 제안한다. 제안된 방법은 각 군집에서 적어도 하나의 샘플을 포함하도록 초기 훈련 집합을 능동적으로 구성하거나 또는 기존의 훈련 집합을 확장함으로써 전체 데이터 분포를 반영할 수 있게 한다. 기존의 능동적 학습 방법들 중 대부분은 초기 훈련 집합이 주어져 있을 경우를 가정하는 반면에 제안하는 방법은 초기 클래스 정보를 가진 훈련 데이터가 주어지지 않은 경우와 주어진 경우에 모두 적용 가능하다. 실험을 통하여 제안하는 방법이 비교 방법들에 비해 분류기 성능을 크게 향상시킬 수 있는 효과적인 데이터 선택을 수행함을 보인다.

• 한국컴퓨터정보학회논문지
• /
• 제23권12호
• /
• pp.131-136
• /
• 2018

This paper proposes a computer-aided diagnostic algorithm in a non-invasive way. Currently, clinical diagnosis of jaundice is performed through blood sampling. Unlike the old methods, the non-invasive method will enable parents to measure newborns' jaundice by only using their mobile phones. The proposed algorithm enables high accuracy and quick diagnosis through machine learning. In here, we used the SVM model of machine learning that learned the feature extracted through image preprocessing and we used the international jaundice research data as the test data set. As a result of applying our developed algorithm, it took about 5 seconds to diagnose jaundice and it showed a 93.4% prediction accuracy. The software is real-time diagnosed and it minimizes the infant's pain by non-invasive method and parents can easily and temporarily diagnose newborns' jaundice. In the future, we aim to use the jaundice photograph of the newborn babies' data as our test data set for more accurate results.

• Asia pacific journal of information systems
• /
• 제31권2호
• /
• pp.185-196
• /
• 2021

The paper aims to analyze and predict sales of liquor in the state of Iowa by applying machine learning algorithms to models built for prediction. We have taken recourse of Azure ML and Spark ML for our predictive analysis, which is legacy machine learning (ML) systems and Big Data ML, respectively. We have worked on the Iowa liquor sales dataset comprising of records from 2012 to 2019 in 24 columns and approximately 1.8 million rows. We have concluded by comparing the models with different algorithms applied and their accuracy in predicting the sales using both Azure ML and Spark ML. We find that the Linear Regression model has the highest precision and Decision Forest Regression has the fastest computing time with the sample data set using the legacy Azure ML systems. Decision Tree Regression model in Spark ML has the highest accuracy with the quickest computing time for the entire data set using the Big Data Spark systems.

• 정보보호학회논문지
• /
• 제29권4호
• /
• pp.795-805
• /
• 2019

최근 알려지지 않은 공격에 대처하기 위한 네트워크 비정상(anomaly) 탐지 기술에 대한 관심이 한층 높아지고 있다. 이러한 기술 개발을 위해 데이터 마이닝(data mining), 기계학습(machine learning), 그리고 딥러닝(deep learning)등을 활용한 다양한 연구가 진행되고 있다. 본 논문에서는 분류(classification) 문제를 다루는 데이터 마이닝 기술 중 가장 전통적인 방법 중 하나인 의사결정나무(decision tree)를 이용하여 NSL-KDD 데이터 셋을 대상으로 네트워크 비정상 탐지 가능성을 보여준다. 의사결정나무의 과대적합(over-fitting) 단점을 해소하기 위해 카이-제곱(chi-square) 테스트를 통해 최적의 속성 선택(feature selection)을 수행하고, 선택된 13개의 속성을 사용한 의사결정나무 모델 환경에서 NSL-KDD 시험 데이터 셋 KDDTest+에 대해 84% 그리고 KDDTest-21에 대해 70%의 네트워크 비정상 검출 정확도를 보였다. 제시된 정확도는 기존 의사결정나무 모델 적용 시 이들 시험 데이터 셋을 대상으로 알려진 정확도 81% 그리고 64% 수준과 비교해 약 3% 그리고 6% 각각 향상된 결과다.

• 한국정보통신학회논문지
• /
• 제23권1호
• /
• pp.13-19
• /
• 2019

기계학습의 정확도는 학습용 데이터의 양과 데이터의 품질에 많은 영향을 받는다. 기존의 웹을 기반으로 학습용 데이터를 수집하는 것은 실제 학습과 무관한 데이터가 수집 될 수 있는 위험성이 있으며 데이터의 투명성을 보장할 수가 없다. 본 논문에서는 블록체인구조에서 블록들이 직접 병렬적으로 데이터를 수집하게 하고 각 블록들이 수집한 데이터를 타 블록의 데이터와 비교하여 양질의 데이터만을 선별하는 방안을 제안한다. 제안하는 시스템은 각 블록들은 데이터를 서로 블록체인을 통해 공유하며 All-reduce 구조의 Parallel-SGD를 활용하여 다른 블록들의 데이터와 비교를 통해 양질의 데이터만을 선별하여 학습용 데이터셋을 구성할 수가 있다. 또한 본 논문에서는 제안한 구조의 성능을 확인하기 위해 실험을 통해 기존의 벤치마크용 데이터셋의 이미지를 활용하여 변조된 이미지 사이에서 원본 이미지만을 양질의 데이터로 판별함을 확인하였다.

• 한국건축시공학회:학술대회논문집
• /
• 한국건축시공학회 2021년도 가을 학술논문 발표대회
• /
• pp.30-31
• /
• 2021

This study aims to build a deep learning model that can predict the value of concrete mixing properties according to a given concrete strength value. A model was created for a total of 1,291 concrete data, including 8 characteristics related to concrete mixing elements and environment, and the compressive strength of concrete. As the deep learning model, DNN-3L-256N, which showed the best performance on the prior study, was used. The average value for each characteristic of the data set was used as the initial input value. In results, in the case of 'curing temperature', which had a narrow range of values in the existing data set, showed the lowest error rate with less than 1% error based on MAE. The highest error rate with an error of 12 to 14% for fly and bfs.