Earticle

본 연구는 형태학적 특징 정량화와 딥러닝 기반 다중 시점 분석을 결합한 송이버섯 품질 등급 자동 분류 시스템을 제안한다. 송이버섯은 형태적 특징에 따라 등급이 결정되나, 전통적 육안 검사는 주관성과 일관성 문제가 있으며, 기존 단일 시점 딥러닝 방법은 3차원 정보 손실로 인한 분류 한계를 보였다. 본 연구는 이중 카메라 시스템을 활용하여 466쌍의 정면-측면 이미지 데이터셋을 구축하고, 규칙 기반의 YOLOv8n 형태 분석 모듈과 딥러닝 기반의 Dual-ConvNeXt 분류 모듈, Dual-EfficientNet 분류 모듈의 3가지 모듈을 Majority Voting 방식으로 앙상블하여 5등급 분류를 수행하였다. 제안 시스템은 94.51%의 정확도를 달성하여 단일 시점 최고 성능(87.91%) 대비 6.6%p, 형태 지표 기반 방식(74.73%) 대비 19.78%p 향상되었다. 본 연구는 다중 시점 정보 융합의 효과를 입증하였으며, 송이버섯 자동 선별의 실용화와 다른 농산물 품질 평가로의 확장 가능성을 제시한다.

This study proposes an automated matsutake mushroom quality grading system that combines rule-based morphological feature quantification with multi-view deep learning analysis using a dual-camera setup. While matsutake mushrooms are graded based on morphological characteristics, traditional visual inspection suffers from subjectivity and inconsistency, and existing single-view deep learning methods encounter classification limitations due to three-dimensional information loss. A dataset of 466 frontal-lateral image pairs was constructed, and five-grade classification was performed by integrating three modules — a rule-based YOLOv8n morphological analysis module, a Dual-ConvNeXt classification module, and a Dual-EfficientNet classification module — via Majority Voting ensemble. The proposed system achieves 94.51% accuracy, improving 6.6 percentage points over the best single-view method (87.91%) and 19.78 percentage points over the morphological approach (74.73%). This research demonstrates the effectiveness of multi-view information fusion and presents the potential for practical implementation of automated matsutake sorting and extension to quality assessment of other agricultural products.

본 연구는 송이버섯의 자루(줄기)의 길이와 갓 펴짐 정도를 수치화하고, 이를 바탕으로 등급을 자동으로 분류하는 딥러닝 기반 시스템을 제안한다. 송이버섯의 형태적 특성인 자루 길이, 갓 펴짐 정도를 수치화하여 선별하는 알고리 즘을 개발하였으며, 이 시스템은 Dino-SAM2 모델을 활용한 인스턴스 세그멘테이션 네트워크와 ResNet-50 기반 분류 네트워크로 구성된다. 인스턴스 세그멘테이션 네트워크를 통해 송이버섯의 자루 길이 측정과 갓 펴짐 정도를 정량적으로 평가하며, Dino-SAM2가 생성한 마스크에 특정 문자열이 포함되지 않은 경우에는 자루 길이와 갓 펴짐 정도를 수치화할 수 없다. 이러한 마스크 생성 실패 상황에 대응하기 위해 ResNet-50 기반 분류 네트워크를 보조 시스템으로 활용하여 송이버섯의 전체적 형태를 기반으로 품질을 판단한다. 두 가지 접근법의 결과를 통합하여 1등 급, 2등급, 3등급(개산), 3등급(성장 정지), 등외 5가지 등급으로 분류한다. 시스템의 성능 평가 결과, Dino-SAM2 가 마스크 생성 성공 시 90%의 높은 정확도를 보였으며, 마스크 생성 실패 시에도 ResNet-50을 통한 방어 메커니 즘을 통해 전체 시스템의 성능 79%를 달성하였다. 이는 단일 모델 의존성을 줄이고 실패 상황에 대한 강건성을 제 공하는 효과적인 방어 전략임을 입증한다. 본 연구는 송이버섯 선별 작업의 효율성과 객관성을 높이는 데 기여할 것으로 기대한다.

This study proposes a deep learning-based system that quantifies the stalk length and cap opening degree of pine mushrooms and automatically classifies them based on these measurements. We developed an algorithm that quantifies and sorts the morphological characteristics of pine mushrooms, including stalk length and cap opening degree. The system consists of an instance segmentation network utilizing the Dino-SAM2 model and a classification network based on ResNet-50. The instance segmentation network quantitatively evaluates the stalk length and cap opening degree of pine mushrooms, but when the mask generated by Dino-SAM2 does not include specific strings, the stalk length and cap opening degree cannot be quantified. To address these mask generation failures, a ResNet-50-based classification network is used as an auxiliary system to assess the quality based on the overall shape of the mushroom. By integrating the results from both approaches, mushrooms are classified into five grades: Grade 1, Grade 2, Grade 3 (open cap), Grade 3 (growth arrest), and Non-standard. Performance evaluation shows that the system achieved 90% accuracy when Dino-SAM2 successfully generated masks, and even in cases of mask generation failure, the defensive mechanism through ResNet-50 helped achieve an overall system performance of 79%. This demonstrates an effective defensive strategy that reduces dependence on a single model and provides robustness in failure situations. This research is expected to contribute to improving the efficiency and objectivity of pine mushroom sorting operations.

In this paper, an alternative approach to select base classifiers forming a parallel Heterogeneous ensemble is proposed. The fundamental concept is to trim poorly performing classifiers; thus, a more effective Heterogeneous ensemble can be generated. More specifically, the proposed trimming approach finds an optimal subset of classifiers to form the desired heterogeneous ensemble. The main challenge is how to detect poor performance classifiers. To address this issue, the differences in effectiveness between base classifiers forming the ensemble are utilized to spot weak classifiers. For evaluating the proposed approach, eighteen benchmark datasets are used for generating the heterogeneous ensemble classification and comparisons with the state-of-the-art methods are conducted. The experimental analysis demonstrated the effectiveness and superiority of the proposed approach when compared to other state-of-the-art approaches.

In this paper, a novel method is proposed to build an ensemble of classifiers by using a feature selection schema. The feature selection schema identifies the best feature sets that affect the arrhythmia classification. Firstly, a number of feature subsets are extracted by applying the feature selection schema to the original dataset. Then classification models are built by using the each feature subset. Finally, we combine the classification models by adopting a voting approach to form a classification ensemble. The voting approach in our method involves both classification error rate and feature selection rate to calculate the score of the each classifier in the ensemble. In our method, the feature selection rate depends on the extracting order of the feature subsets. In the experiment, we applied our method to arrhythmia dataset and generated three top disjointed feature sets. We then built three classifiers based on the top-three feature subsets and formed the classifier ensemble by using the voting approach. Our method can improve the classification accuracy in high dimensional dataset. The performance of each classifier and the performance of their ensemble were higher than the performance of the classifier that was based on whole feature space of the dataset. The classification performance was improved and a more stable classification model could be constructed with the proposed approach.

겨울철 도로 결빙으로 인한 사고는 대부분 큰 사고로 이어진다. 이는 운전자가 도로의 결빙을 사전에 자각하기 어렵기 때문이다. 본 연구에서는 AutoML과 CNN의 앙상블 모델을 이용하여 도로교통 이머징 리스크 를 정확하게 탐지하는 방법을 연구한다. 비정형 데이터인 이미지를 이용한 CNN 이미지 특징 추출 기반 도로교 통 이머징 리스크 분류 모델과 정형 데이터인 기상 데이터를 이용한 AutoML 기반 도로교통 이머징 리스크 분류 모델을 각각 학습시킨다. 그 후 모델들에서 도출된 확률값을 입력하여 CNN 기반 분류 모델을 보완하도록 앙상 블 모델을 설계한다. 이를 통해 도로교통 이머징 리스크 분류 성능을 향상하고 더 정확하고 빠르게 운전자에게 경고하여 안전한 주행이 가능하도록 한다.

Most accidents caused by road icing in winter lead to major accidents. Because it is difficult for the driver to detect the road icing in advance. In this work, we study how to accurately detect road traffic emerging risk using AutoML and CNN's ensemble model that use both structured and unstructured data. We train CNN-based road traffic emerging risk classification model using images that are unstructured data and AutoML-based road traffic emerging risk classification model using weather data that is structured data, respectively. After that the ensemble model is designed to complement the CNN-based classification model by inputting probability values derived from of each models. Through this, improves road traffic emerging risk classification performance and alerts drivers more accurately and quickly to enable safe driving.

IoT 시장의 성장과 더불어 linux 아키텍쳐를 사용하는 디바이스들에 대해 악성코드 보안 위협이 꾸준히 증가하고 있다. 하지만, Mirai 등의 심각한 보안피해를 야기한 주요 악성코드들을 제외하면 linux 악성코드에 대한 보안 커뮤니티의 관련 기술이나 연구는 전무한 수준이다. 또한, IoT 환경의 디바이스, 벤더, 아키텍쳐 등의 다양성이 더욱 심화됨에 따라 linux 악성코드 대응 난이도 또한 심화되고 있다. 따라서, 본 논문에서는 linux 아키텍쳐의 주요 포맷인 ELF를 분석하고 이를 기반으로 한 분석 시스템과, IoT 환경을 고려한 바이너리 기반의 분석 시스템을 제안한다. ELF 기반의 분석 시스템은 상대적으로 고속으로 다수의 악성코드에 대해 전처리 분류 할 수 있으며 상대적으로 저속의 바이너리 기반의 분석 시스템은 전처리 하지 못한 데이터에 대해 모두 분류 가능하다. 이러한 두 개의 프로세스는 서로 상호보완되어 효과적으로 linux 기반의 악성코드를 분류할 수 있을 것이라 기대한다.

With the growth of the IoT market, malware security threats are steadily increasing for devices that use the linux architecture. However, except for the major malware causing serious security damage such as Mirai, there is no related technology or research of security community about linux malware. In addition, the diversity of devices, vendors, and architectures in the IoT environment is further intensifying, and the difficulty in handling linux malware is also increasing. Therefore, in this paper, we propose an analysis system based on ELF which is the main format of linux architecture, and a binary based analysis system considering IoT environment. The ELF-based analysis system can be pre-classified for a large number of malicious codes at a relatively high speed and a relatively low-speed binary-based analysis system can classify all the data that are not preprocessed. These two processes are supposed to complement each other and effectively classify linux-based malware.

최근 기업의 보안 시스템으로부터 수집되는 보안 인텔리전스 수는 악성코드의 확산으로 인해 기하급수적으로 증가하고 있다. 빅 데이터 환경이 도래하면서 기업들은 침해사고에 대한 다양한 정보를 이용할 수 있게 되면서 기업이 수집할 수 있는 침해사고 정보가 다양해지고 있다. 이에 따라 보안 인텔리전스를 구성하고 있는 침해사고의 다양한 속성을 사용하여 보다 정확하게 유사침해사고를 그룹별로 분류할 필요성이 요구되고 있다. 본 연구에서는 유사도 비교 분석 이론에 근거하여 침해사고를 공격유형과 침해자원을 고려한 다중 프로파일을 개발하고, 이를 활용하여 보안 인텔리전스를 구성하고 있는 침해사고 유형 분류의 정확성을 개선하는 다중 프로파일 기반 앙상블 모델을 제안한다. 제안 모델은 침입탐지시스템에서 수집된 계층적 침해자원에 대한 유사도 분석을 통해 새로운 침해사고를 효과적으로 분석할 수 있다. 사실적이고 의미 있는 침해사고의 구성을 통한 유형 분류는 새로운 침해사고에 대한 유사 침해사고를 정확하게 분류 제공함으로써 분석의 실용성을 향상시킨다.

Threat intelligences collected from cyber incident sharing system and security events collected from Security Information & Event Management system are analyzed and coped with expanding malicious code rapidly with the advent of big data. Analytical classification of the threat intelligence in cyber incidents requires various features of cyber observable. Therefore it is necessary to improve classification accuracy of the similarity by using multi-profile which is classified as the same features of cyber observables. We propose a multi-profile ensemble model performed similarity analysis on cyber incident of threat intelligence based on both attack types and cyber observables that can enhance the accuracy of the classification. We see a potential improvement of the cyber incident analysis system, which enhance the accuracy of the classification. Implementation of our suggested technique in a computer network offers the ability to classify and detect similar cyber incident of those not detected by other mechanisms.

감성 분석에서 정확한 감성 분류는 중요한 연구 주제이다. 본 연구는 최근 많은 연구가 이루어지는 word2vec과 앙상블 방법을 이용하여 효과적으로 한국어 리뷰를 감성 분류하는 방법을 제시한다. 연구는 20 만 개의 한국 영화 리뷰 텍스트에 대해, 품사 기반 BOW 자질과 word2vec를 사용한 자질을 생성하고, 두 개의 자질 표현을 결합한 통합 자질을 생성했다. 감성 분류를 위해 Logistic Regression, Decision Tree, Naive Bayes, Support Vector Machine의 단일 분류기와 Adaptive Boost, Bagging, Gradient Boosting, Random Forest의 앙상블 분류기를 사용하였다. 연구 결과로 형용사와 부사를 포함한 BOW자질과 word2vec자질로 구성된 통합 자질 표현이 가장 높은 감성 분류 정확도를 보였다. 실증결과, 단일 분류기인 SVM이 가장 높은 성능을 나타내었지만, 앙상블 분류기는 단일 분류기와 비슷하거나 약간 낮은 성능을 보였다.

Accurate sentiment classification is an important research topic in sentiment analysis. This study suggests an efficient classification method of Korean sentiment using word2vec and ensemble methods which have been recently studied variously. For the 200,000 Korean movie review texts, we generate a POS-based BOW feature and a feature using word2vec, and integrated features of two feature representation. We used a single classifier of Logistic Regression, Decision Tree, Naive Bayes, and Support Vector Machine and an ensemble classifier of Adaptive Boost, Bagging, Gradient Boosting, and Random Forest for sentiment classification. As a result of this study, the integrated feature representation composed of BOW feature including adjective and adverb and word2vec feature showed the highest sentiment classification accuracy. Empirical results show that SVM, a single classifier, has the highest performance but ensemble classifiers show similar or slightly lower performance than the single classifier.

본 연구는 최신 딥러닝 아키텍처를 활용하여 도배 하자의 자동 분류 시스템을 제안한다. 특히 데이터 불 균형과 하자 유형 간 유사성이라는 기술적 과제를 효과적으로 해결하기 위한 새로운 방법론을 제시한다. 데이터 불균형과 하자 유형 간 유사성 문제를 해결하기 위해 ConvNeXt 아키텍처를 기반으로 하되, 하자 유형별 특성을 고려한 맞춤형 데이터 증강 기법과 앙상블 학습을 도입하였다. 총 19개 하자 유형에 대해 3,457개의 이미지로 구 성된 데이터셋을 활용하였으며, 데이터 불균형 해소를 위해 Focal Loss와 Class-Balanced Loss를 결합한 하이브 리드 손실 함수를 적용하였다. 실험 결과, 제안된 모델은 Weighted F1 Score 0.889를 달성하여 기존 ResNet50 기반 모델 대비 8.8%의 성능 향상을 보였으며, 45.2ms의 평균 추론 시간은 실제 하자 검수 업무에 적용 가능한 수준임을 검증하였다. 본 연구의 결과는 건설 현장에서의 도배 하자 검수 자동화 가능성을 제시하며, 품질 관리 프로세스의 효율성 향상에 기여할 것으로 기대된다.

This study proposes an automated classification system for wallpaper defects using state-of-the-art deep learning architectures. In particular, it introduces novel methodologies to effectively address technical challenges such as data imbalance and similarities between defect types. To address challenges such as data imbalance and similarity between defect types, the model is based on the ConvNeXt architecture and incorporates customized data augmentation methods and ensemble learning tailored to the characteristics of each defect type. A dataset of 3,457 images across 19 defect types was utilized, and a hybrid loss function combining Focal Loss and Class-Balanced Loss was applied to mitigate data imbalance. Experimental results demonstrate that the proposed model achieves a weighted F1 score of 0.889, representing an 8.8% performance improvement over existing ResNet50-based models, with an average inference time of 45.2ms validating its feasibility for practical defect inspection tasks. These findings suggest the potential for automating wallpapering defect inspection in construction sites, thereby contributing to enhanced efficiency in quality control processes.

We propose ensemble-based modeling for classifying streaming data with concept drift. The concept drift is a phenomenon in which the distribution of streaming data changes. In this paper, the types of the concept drift are categorized into the change of data distribution and the change of class distribution. The proposed ensemble modeling generates a meta-ensemble which consists of ensembles of classifiers. Whenever a change of class distribution occurs in streaming data, our modeling builds a new classifier of an existing ensemble and whenever a change of data distribution occurs, it builds a new ensemble which consists of an only one classifier. In our approach, new classifiers of a meta-ensemble on streaming data will be generated dynamically according to the estimated distribution of streaming data. We compared the results of our approach and of the chunk-based ensemble approach, which builds new classifiers of an ensemble periodically. In experiments with 13 benchmark data sets, our approach produced an average of 21.95% higher classification accuracy generating an average of 61.7% fewer new classifiers of an ensemble than the chunk-based ensemble method using partially labeled samples. We also examine that the time points when our approach builds new classifiers are appropriate for maintaining performance of an ensemble.

Aiming at the characteristics of high dimension and small samples in microarray data, this paper proposes a selective ensemble method to classify microarray data. Firstly, kruskal-wallis test is used to filter irrelevant genes with classification task and to obtain a set of genes, and then a reduced training set is produced from original training set according to gene subset obtained. Secondly, multiple gene subsets are generated by using neighborhood rough set model with different radius and used to construct training subsets on above reduced training set. Thirdly, every constructed training subset is used to train a classifier by using SVM algorithm, and then multiple classifiers are produced as base classifiers. Finally, a set of base classifiers are selected by using teaching-learning-based optimization and build an ensemble classifier by weighted voting. Five benchmarks tumor microarray datasets are applied to evaluate performance of our proposed method. Experimental results indicate our proposed method is very effective and efficient for classifying microarray data, and it improves not only classification accuracy, but also decrease memory costs and computation times.

In order to lower the classification cost and improve the performance of the classifier, this paper proposes the approach of the dynamic cost-sensitive ensemble classification based on extreme learning machine for imbalanced massive data streams (DCECIMDS). Firstly, this paper gives the method of concept drifts detection by extracting the attributive characters of imbalanced massive data streams. If the change of attributive characters exceeds threshold value, the concept drift occurs. Secondly, we give Cost-sensitive extreme learning machine algorithm, and the optimal cost function is defined by the dynamic cost matrix. Build the cost-sensitive classifiers model for imbalanced massive data streams under MapReduce, and the data streams are processed in parallel. At last, the weighted cost-sensitive ensemble classifier is constructed, and the dynamic cost-sensitive ensemble classification based on extreme learning machine classification is given. The experiments demonstrate that the proposed ensemble classifier under the MapReduce framework can reduce the average misclassification cost and can make the classification results more reliable. DCECIMDS has high performance by comparing to the other classification algorithms for imbalanced data streams and can effectively deal with the concept drift.

The idea of ensemble learning is to train multiple models, each with the objective to predict or classify a set of results. Most of the errors from a model’s learning are from three main factors: variance, noise, and bias. By using ensemble methods, we’re able to increase the stability of the final model and reduce the errors mentioned previously. By combining many models, we’re able to reduce the variance, even when they are individually not great. In this paper we propose an ensemble model and applied it to classification problem. In iris, Pima indian diabeit and semiconductor fault detection problem, proposed model classifies well compared to traditional single classifier that is logistic regression, SVM and random forest .

의료 데이터 마이닝의 목적은 효율적인 알고리즘 및 기법을 통하여 각종 질병을 예측 분류하고 신뢰도를 높이는데 있다. 기존의 연구로 단일모델을 기반으로 하는 알고리즘이 존재하며 나아가 모델의 더 좋은 예측과 분류 정확도를 위하여 다중모델을 기반으로 하는 앙상블 기법을 적용한 연구도 진행되고 있다. 본 논문에서는 의료데이터의 보다 높은 예측의 신뢰도를 위하여 기존의 앙상블 기법에 사분위간 범위를 적용한 I-ENSEMBLE을 제안한다. 갑상선 기능 저하증 진단을 위한 데이터를 통해 실험 적용한 결과 앙상블의 대표적인 기법인 Bagging, Boosting, Stacking기법 모두 기존에 비해 현저하게 향상된 정확도를 나타내었다. 또한 기존 단일모델 기법과 비교하여 다중모델인 앙상블 기법에 사분위간 범위를 적용했을 때 더 뚜렷한 효과를 나타냄을 확인하였다.

The purpose of medical data mining for efficient algorithms and techniques throughout the various diseases is to increase the reliability of estimates to classify. Previous studies, an algorithm based on a single model, and even the existence of the model to better predict the classification accuracy of multi-model ensemble-based research techniques are being applied. In this paper, the higher the medical data to predict the reliability of the existing scope of the ensemble technique applied to the I-ENSEMBLE offers. Data for the diagnosis of hypothyroidism is the result of applying the experimental technique, a representative ensemble Bagging, Boosting, Stacking technique significantly improved accuracy compared to all existing, respectively. In addition, compared to traditional single-model techniques and ensemble techniques Multi modeling when applied to represent the effects were more pronounced.

현재 의료 현장에서 초음파 진단은 과거 청진기와 같다고 할 수 있다. 그러나 초음파의 특성상 검사자의 숙련도에 따라 결과 예측이 불확실하다는 단점을 가진다. 따라서 본 논문에서는 이런 문제를 해결하기 위해 딥러닝 기술을 기반으 로 초음파 검사 중 간병변 탐지의 정확도를 높이고자 한다. 제안 논문에서는 CNN 모델과 앙상블 모델을 이용해 병변 분류의 정확도 비교 실험하였다. 실험결과 CNN 모델에서 분류 정확도는 평균 82.33%에서 앙상블모델의 경우 평균 89.9%로 약 7% 높은 것을 확인하였다. 또한 앙상블 모델이 평균 ROC커브에서도 0.97로 CNN모델보다 약 0.4정도 높은 것을 확인하였다.

In the current medical field, ultrasound diagnosis can be said to be the same as a stethoscope in the past. However, due to the nature of ultrasound, it has the disadvantage that the prediction of results is uncertain depending on the skill level of the examiner. Therefore, this paper aims to improve the accuracy of liver lesion detection during ultrasound examination based on deep learning technology to solve this problem. In the proposed paper, we compared the accuracy of lesion classification using a CNN model and an ensemble model. As a result of the experiment, it was confirmed that the classification accuracy in the CNN model averaged 82.33% and the ensemble model averaged 89.9%, about 7% higher. Also, it was confirmed that the ensemble model was 0.97 in the average ROC curve, which is about 0.4 higher than the CNN model.

Classification is a predictive modeling for a categorical target variable. Various classification ensemble methods, which predict with better accuracy by combining multiple classifiers, became a powerful machine learning and data mining paradigm. Well-known methodologies of classification ensemble are boosting, bagging and random forest. In this article, we assume that decision trees are used as classifiers in the ensemble. Further, we hypothesized that tree size affects classification accuracy. To study how the tree size in uences accuracy, we performed experiments using twenty-eight data sets. Then we compare the performances of ensemble algorithms; bagging, double-bagging, boosting and random forest, with different tree sizes in the experiment.

In this paper, we propose the ensemble-based classification model which extracts just new data patterns from the streaming-data by using clustering and generates new classification models to be added to the ensemble in order to reduce the number of data labeling while it keeps the accuracy of the existing system. The proposed technique performs clustering of similar patterned data from streaming data. It performs the data labeling to each cluster at the point when a certain amount of data has been gathered. The proposed technique applies the K-NN technique to the classification model unit in order to keep the accuracy of the existing system while it uses a small amount of data. The proposed technique is efficient as using about 3% less data comparing with the existing technique as shown the simulation results for benchmarks, thereby using clustering.

본 논문에서는 멀티 센서가 장착된 스마트 인솔로 측정한 보행 데이터에 대해 앙상블 딥러닝 네트워크를 이용하여 보행의 타입을 분류하는 시스템을 제안한다. 보행 타입 분류 시스템은 인솔에 의해 측정된 데이터를 정규화하는 부분과 딥러닝 네트워크를 이용하여 보행의 특징을 추출하는 부분, 그리고 추출된 특징을 입력으로 보행의 타입을 분류하는 부분으로 구성되어 있다. 서로 다른 특성을 가지는 CNN과 LSTM을 기반으로 하는 네트워크를 독립적으로 학습하여 두 종류의 보행 특징 맵을 추출하였으며, 각각의 분류 결과를 결합하여 최종적인 앙상블 네트워크의 분류 결과를 도출하였다. 20~30대 성인의 걷기, 뛰기, 빠르게 걷기, 계단 오르기와 내려가기, 언덕 오르기와 내려가기의 7종류의 보행에 대해, 스마트 인솔을 이용하여 실측한 멀티 센서 데이터를 제안한 앙상블 네트워크로 분류해 본 결과 90% 이상의 높은 분류율을 보이는 것을 확인하였다.

This paper proposes a system for classifying gait types using an ensemble deep learning network for gait data measured by a smart insole equipped with multi-sensors. The gait type classification system consists of a part for normalizing the data measured by the insole, a part for extracting gait features using a deep learning network, and a part for classifying the gait type by inputting the extracted features. Two kinds of gait feature maps were extracted by independently learning networks based on CNNs and LSTMs with different characteristics. The final ensemble network classification results were obtained by combining the classification results. For the seven types of gait for adults in their 20s and 30s: walking, running, fast walking, going up and down stairs, and going up and down hills, multi-sensor data was classified into a proposed ensemble network. As a result, it was confirmed that the classification rate was higher than 90%.

Phishing website has become a crucial concern in cyber security applications. It is performed by fraudulently deceiving users with the aim of obtaining their sensitive information such as bank account information, credit card, username, and password. The threat has led to huge losses to online retailers, e-business platform, financial institutions, and to name but a few. One way to build anti-phishing detection mechanism is to construct classification algorithm based on machine learning techniques. The objective of this paper is to compare different classifier ensemble approaches, i.e. random forest, rotation forest, gradient boosted machine, and extreme gradient boosting against single classifiers, i.e. decision tree, classification and regression tree, and credal decision tree in the case of website phishing. Area under ROC curve (AUC) is employed as a performance metric, whilst statistical tests are used as baseline indicator of significance evaluation among classifiers. The paper contributes the existing literature on making a benchmark of classifier ensembles for web phishing detection.

최근 코로나(Covid-19)의 영향으로 스마트 헬스케어 관련 산업과 비대면 방식의 원격 진단을 통한 질환 분류 예측 연구의 필요성이 증가하고 있다. 일반적으로 호흡기 질환의 진단은 비용이 많이 들고 숙련된 의료 전문가를 필요로 하여 현실적으로 조기 진단 및 모니터링에 한계가 있다. 따라서, 간단하고 편리한 청진기로부터 수집된 호흡음을 딥러닝 기반 모델을 활용하여 높은 정확도로 분류하고 조기 진단이 필요하다. 본 연구에서는 청진을 통해 수집된 폐음 데이터를 이용하여 이상 호흡음 분류모델을 제안한다. 데이터 전처리로는 대역통과필터(BandPassFilter)방법론을 적용하고 로그 멜 스펙트로그램(Log-Mel Spectrogram)과 Mel Frequency Cepstral Coefficient(MFCC)을 이용하여 폐음의 특징적인 정보를 추출하였다. 추출된 폐음의 특징에 대해서 효과적으로 분류할 수 있는 병렬 합성곱 신경망 네트워크(Parallel CNN network)모델을 제안하고 다양한 머신러닝 분류기(Classifiers)와 결합한 스태킹 앙상블(Stacking Ensemble) 방법론을 이용하여 이상 호흡음을 높은 정확도로 분류하였다. 본 논문에서 제안한 방법은 96.9%의 정확도로 이상 호흡음을 분류하였으며, 기본모델의 결과 대비 정확도가 약 6.1% 향상되었다.

As the COVID-19 pandemic rapidly changes healthcare around the globe, the need for smart healthcare that allows for remote diagnosis is increasing. The current classification of respiratory diseases cost high and requires a face-to-face visit with a skilled medical professional, thus the pandemic significantly hinders monitoring and early diagnosis. Therefore, the ability to accurately classify and diagnose respiratory sound using deep learning-based AI models is essential to modern medicine as a remote alternative to the current stethoscope. In this study, we propose a deep learning-based respiratory sound classification model using data collected from medical experts. The sound data were preprocessed with BandPassFilter, and the relevant respiratory audio features were extracted with Log-Mel Spectrogram and Mel Frequency Cepstral Coefficient (MFCC). Subsequently, a Parallel CNN network model was trained on these two inputs using stacking ensemble techniques combined with various machine learning classifiers to efficiently classify and detect abnormal respiratory sounds with high accuracy. The model proposed in this paper classified abnormal respiratory sounds with an accuracy of 96.9%, which is approximately 6.1% higher than the classification accuracy of baseline model.