• Journal of Digital Convergence
• /
• v.20 no.4
• /
• pp.185-192
• /
• 2022

This study investigates the effect of the semi-supervised learning(SSL) method on predicting default risk of peer-to-peer(P2P) loans. Despite its proven performance, the supervised learning(SL) method requires labeled data, which may require a lot of effort and resources to collect. With the rapid growth of P2P platforms, the number of loans issued annually that have no clear final resolution is continuously increasing leading to abundance in unlabeled data. The research data of P2P loans used in this study were collected on the LendingClub platform. This is why an SSL model is needed to predict the default risk by using not only information from labeled loans(fully paid or defaulted) but also information from unlabeled loans. The results showed that in terms of default risk prediction and despite the use of a small number of labeled data, the SSL method achieved a much better default risk prediction performance than the SL method trained using a much larger set of labeled data.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.307-310
• /
• 2021

지도 학습을 하기 위해선 레이블이 부착된 데이터셋이 필요하다. 크라우드소싱 서비스를 통해 데이터셋을 구축하는데 다수의 주석자(Annotator)가 관여한다. 다수의 주석자가 레이블을 할당하고 과반수인 레이블을 최종 정답으로 결정한다. 이 과정에서 최종 정답과 다른 후보 레이블의 정보가 누락된다. 이를 완화하고 목표 작업에 대한 성능을 높이기 위해 후보 레이블에 대한 정보를 반영하는 멀티 디코더 모델을 제안한다. KLUE-TC, SNLI, MNLI 데이터셋으로 정량적 성능 평가를 수행하였으며 실험한 데이터셋 모두 일괄적인 성능 향상을 보였다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.5
• /
• pp.482-488
• /
• 2014

This paper aims to analyze user's emotion automatically by analyzing Twitter, a representative social network service (SNS). In order to create sentiment analysis models by using machine learning techniques, sentiment labels that represent positive/negative emotions are required. However it is very expensive to obtain sentiment labels of tweets. So, in this paper, we propose a sentiment analysis model by using self-training technique in order to utilize "data without sentiment labels" as well as "data with sentiment labels". Self-training technique is that labels of "data without sentiment labels" is determined by utilizing "data with sentiment labels", and then updates models using together with "data with sentiment labels" and newly labeled data. This technique improves the sentiment analysis performance gradually. However, it has a problem that misclassifications of unlabeled data in an early stage affect the model updating through the whole learning process because labels of unlabeled data never changes once those are determined. Thus, labels of "data without sentiment labels" needs to be carefully determined. In this paper, in order to get high performance using self-training technique, we propose 3 policies for updating "data with sentiment labels" and conduct a comparative analysis. The first policy is to select data of which confidence is higher than a given threshold among newly labeled data. The second policy is to choose the same number of the positive and negative data in the newly labeled data in order to avoid the imbalanced class learning problem. The third policy is to choose newly labeled data less than a given maximum number in order to avoid the updates of large amount of data at a time for gradual model updates. Experiments are conducted using Stanford data set and the data set is classified into positive and negative. As a result, the learned model has a high performance than the learned models by using "data with sentiment labels" only and the self-training with a regular model update policy.

• Information Systems Review
• /
• v.25 no.1
• /
• pp.47-73
• /
• 2023

Management decision-making based on artificial intelligence(AI) plays an important role in helping decision-makers. Business decision-making centered on AI is evaluated as a driving force for corporate growth. AI-based on accurate analysis techniques could support decision-makers in making high-quality decisions. This study proposes an effective decision-making method with the application of multi-label feature selection. In this regard, We present a CFS-BR (Correlation-based Feature Selection based on Binary Relevance approach) that reduces data sets in high-dimensional space. As a result of analyzing sample data and empirical data, CFS-BR can support efficient decision-making by selecting the best combination of meaningful attributes based on the Best-First algorithm. In addition, compared to the previous multi-label feature selection method, CFS-BR is useful for increasing the effectiveness of decision-making, as its accuracy is higher.

• Journal of Internet Computing and Services
• /
• v.18 no.1
• /
• pp.65-75
• /
• 2017

The Detection Model is the model to find the result of a certain purpose using artificial intelligent, data mining, intelligent algorithms In Cyber Security, it usually uses to detect intrusion, malwares, cyber incident, and attacks etc. There are an amount of unlabeled data that are collected in a real environment such as security data. Since the most of data are not defined the class labels, it is difficult to know type of data. Therefore, the label determination process is required to detect and analysis with accuracy. In this paper, we proposed a KDFL(K-means and D-S Fusion based Labeling) method using D-S inference and k-means(unsupervised) algorithms to decide label of data records by fusion, and a detection model architecture using a proposed labeling method. A proposed method has shown better performance on detection rate, accuracy, F1-measure index than other methods. In addition, since it has shown the improved results in error rate, we have verified good performance of our proposed method.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2023.05a
• /
• pp.204-205
• /
• 2023

This study studied a prediction model for sending collision alarms for small fishing boats based on semi-supervised learning(SSL). The supervised learning (SL) method requires a large number of labeled data, but the labeling process takes a lot of resources and time. This study used service data collected through a data pipeline linked to 'intelligent maritime traffic information service' and data collected from real-sea experiment. The model accuracy was improved as a result of learning not only real-sea experiment data with labeling determined based on actual user satisfaction but also service data without label determined together.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.4
• /
• pp.169-177
• /
• 2011

As demand for efficiency in handling dynamic XML data grows, new dynamic XML labeling schemes have been researched. The key idea of the dynamic XML labeling scheme is to find ancestor-descendent-sibling relationships and to minimize memory space to store total label, response time and range of relabeling incurred by update operations. The prime number labeling scheme is a representative scheme which supports dynamic XML documents. It determines the ancestor-descendant relationships between two elements by a simple divisibility test of labels. When a new element is inserted into the XML data using this scheme, it does not change the label values of existing nodes. However, since each prime number must be used exclusively, labels can become significantly large. Therefore, in this paper, we introduce a novel technique to effectively reduce the problem of label overflow. The suggested idea is based on tree decomposition. When label overflow occurs, the full tree is divided into several sub-trees, and nodes in each sub-tree are separately labeled. Through experiments, we show the effectiveness of our scheme.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10a
• /
• pp.163-165
• /
• 2003

도메인 지식은 텍스트의 포맷과 의미 정보를 이용하여 웹에 존재하는 텍스트의 다양한 의미를 이해할 수 있도록 도와준다. 그러나 도메인 지식은 텍스트에 데이터의 의미를 표현하는 레이블이 존재하지 알을 경우에 텍스트 인식을 제대로 수행할 수 없기 때문에 무용지물이 되고 만다. 이러한 문제를 해결하기 위해 본 논문에서는 레이블이 존재하지 않는 텍스트의 의미를 효과적으로 추론할 수 있는 엔티티 인식 모델을 제안한다 엔티티 인식 모델은 베이지언 모델과 컨텍스트 정보를 결합한 방법으로서, 구조 분석을 수행한 HTML 문서의 텍스트 토큰에 대해서 어떤 엔티티에 속할 것인가를 결정하는 기능을 수행한다. 실험 결과 본 모델을 사용할 경우 기존에는 레이블이 없어서 인식되지 않았던 텍스트들을 효과적으로 인식하는 것을 확인할 수 있었다.

• Journal of KIISE
• /
• v.44 no.7
• /
• pp.686-691
• /
• 2017

Telemarketing has become the center of marketing action of the industry in the information society. Recently, machine learning has emerged in many areas, especially, financial prediction. Financial data consists of lots of unlabeled data in most parts, and therefore, it is difficult for humans to perform their labeling. In this paper, we propose a fusion method of semi-supervised learning for automatic labeling of unlabeled data to predict telemarketing. Specifically, we integrate labeling results of label propagation and co-training with a decision tree. The data with lower reliabilities are removed, and the data are extracted that have consistent label from two labeling methods. After adding them to the training set, a decision tree is learned with all of them. To confirm the usefulness of the proposed method, we conduct the experiments with a real telemarketing dataset in a Portugal bank. Accuracy of the proposed method is 83.39%, which is 1.82% higher than that of the conventional method, and precision of the proposed method is 19.37%, which is 2.67% higher than that of the conventional method. As a result, we have shown that the proposed method has a better performance as assessed by the t-test.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.170-175
• /
• 2022

일반적으로 대규모 언어 모델들은 다량의 데이터를 오랜시간 사전학습하면서 레이블을 예측하기 위한 성능을 높여왔다. 최근 언어 모델의 레이블 예측에 대한 정확도가 높아지면서, 언어 모델이 왜 해당 결정을 내렸는지 이해하기 위한 신뢰도 높은 Natural Language Explanation(NLE) 을 생성하는 것이 시간이 지남에 따라 주요 요소로 자리잡고 있다. 본 논문에서는 높은 레이블 정확도를 유지하면서 동시에 언어 모델의 예측에 대한 신뢰도 높은 explanation 을 생성하는 참신한 자연어 추론 시스템을 제시한 Natural-language Inference over Label-specific Explanations(NILE)[1] 을 소개하고 한국어 데이터셋을 이용해 NILE 과 NLE 를 활용하지 않는 일반적인 자연어 추론 태스크의 성능을 비교한다.