• Journal of Intelligence and Information Systems
• /
• v.22 no.4
• /
• pp.1-18
• /
• 2016

Collaborative Filtering (CF) predicts the focal user's preference for particular item based on user's preference rating data and recommends items for the similar users by using them. It is a popular technique for the personalization in e-commerce to reduce information overload. However, it has some limitations including sparsity and scalability problems. In this paper, we use a method to integrate social network information into collaborative filtering in order to mitigate the sparsity and scalability problems which are major limitations of typical collaborative filtering and reflect the user's qualitative and emotional information in recommendation process. In this paper, we use a novel recommendation algorithm which is integrated with collaborative filtering by using Social SVD++ algorithm which considers social network information in SVD++, an extension algorithm that can reflect implicit information in singular value decomposition (SVD). In particular, this study will evaluate the performance of the model by reflecting the real-world user's social network information in the recommendation process.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.6
• /
• pp.19-28
• /
• 2021

Nowadays, interest in health care is increasing due to Coronavirus (COVID-19), and a lot of people are doing home training as there are more difficulties in using fitness centers and public facilities that are used together. In this paper, we propose a personalized exercise recommendation algorithm using personalized propensity information to provide more accurate and meaningful exercise recommendation to home training users. Thus, we classify the data according to the criteria for obesity with a k-nearest neighbor algorithm using personal information that can represent individuals, such as eating habits information and physical conditions. Furthermore, we differentiate the exercise dataset by the level of exercise activities. Based on the neighborhood information of each dataset, we provide personalized exercise recommendations to users through a dimensionality reduction algorithm (SVD) among model-based collaborative filtering methods. Therefore, we can solve the problem of data sparsity and scalability of memory-based collaborative filtering recommendation techniques and we verify the accuracy and performance of the proposed algorithms.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.11a
• /
• pp.1294-1297
• /
• 2013

본 논문은 기계학습을 통하여 특허문서를 국제 특허 분류(IPC) 기준에 따라 자동으로 분류하는 시스템에 관한 연구로 잠재 의미 색인 기법을 이용하여 분류의 성능을 높일 수 있는 방법을 제안하기 위한 연구이다. 종래 특허문서에 관한 IPC 자동 분류에 관한 연구가 단어 매칭 방식의 색인 기법에 의존해서 이루어진바가 있으나, 현대 기술용어의 발생 속도와 다양성 등을 고려할 때 특허문서들 간의 관련성을 분석하는데 있어서는 단어 자체의 빈도 보다는 용어의 개념에 의한 접근이 보다 효과적일 것이라 판단하여 잠재 의미 색인(LSI) 기법에 의한 분류에 관한 연구를 하게 된 것이다. 실험은 단어 매칭 방식의 색인 기법의 대표적인 자질선택 방법인 정보획득량(IG)과 카이제곱 통계량(CHI)을 이용했을 때의 성능과 잠재 의미 색인 방법을 이용했을 때의 성능을 SVM, kNN 및 Naive Bayes 분류기를 사용하여 분석하고, 그중 가장 성능이 우수하게 나오는 SVM을 사용하여 잠재 의미 색인에서 명사가 해당 용어의 개념적 의미 구조를 구축하는데 기여하는 정도가 어느 정도인지 평가함과 아울러, LSI 기법 이용시 최적의 성능을 나타내는 특이값의 범위를 실험을 통해 비교 분석 하였다. 분석결과 LSI 기법이 단어 매칭 기법(IG, CHI)에 비해 우수한 성능을 보였으며, SVM, Naive Bayes 분류기는 단어 매칭 기법에서는 비슷한 수준을 보였으나, LSI 기법에서는 SVM의 성능이 월등이 우수한 것으로 나왔다. 또한, SVM은 LSI 기법에서 약 3%의 성능 향상을 보였지만 Naive Bayes는 오히려 20%의 성능 저하를 보였다. LSI 기법에서 명사가 잠재적 의미 구조에 미치는 영향은 모든 단어들을 내용어로 한 경우 보다 약 10% 더 향상된 결과를 보여주었고, 특이값의 범위에 따른 성능 분석에 있어서는 30% 수준에 Rank 되는 범위에서 가장 높은 성능의 결과가 나왔다.

• Journal of fish pathology
• /
• v.22 no.3
• /
• pp.335-342
• /
• 2009

The stability of immunoglobulin M (IgM) on different serum storage conditions and specific antibody response were tested using the serum collected from sevenband grouper Epinephelus septemfasciatus by enzyme-linked immunosorbent assay (ELISA). To test the effect of storage temperature and duration, sevenband grouper antiserum against bovine serum albumin (BSA) was stored at -80, -20 or 4${^{\circ}C}$ for 1, 34, 61 or 119 days. In addition, to test the effect of repeated freeze-thawing condition, the anti-BSA fish serum was frozen at -20 and -80${^{\circ}C}$ and then thawn and frozen for 1, 5 or 10 times repeatedly. Consequently, no significant difference was found in ELISA optical density (O.D.) values of sera for the above mentioned storage conditions: different temperatures (-80, -20 and 4${^{\circ}C}$), durations of storage (1, 34, 61 and 119 days), and repeated thaw-freeze cycles (1, 5, and 10 times), indicating that IgMs of test fish were stable. The specific antibody response of sevenband grouper was observed after BSA-immunization of the test fish reared at 20 ${^{\circ}C}$ or 25${^{\circ}C}$. At the rearing temperature of 20${^{\circ}C}$, the specific antibody against BSA first appeared at 14 days and maximum antibody titer was observed between 21 and 28 days, while at the rearing temperature of 25 ${^{\circ}C}$, specific antibody appeared at 7 days and maximum antibody titer was observed between 14 and 21 days. In conclusion, the rearing temperature at 25${^{\circ}C}$ gave a faster and higher specific antibody response than at 20${^{\circ}C}$ and the specific antibody response maintained for approximately 2 months at 20℃ and 25${^{\circ}C}$.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.309-312
• /
• 2019

본 논문에서는 계산의 복잡성을 줄이고 피크 신호 대 잡음 비율(PSNR) 성능을 개선하기 위한 새로운 프레임 상향 변환 (Frame Rate Up Conversion) 알고리즘을 제안한다. 제안된 알고리즘을 사용하기 위한 모션 추정 과정(Motion Estimation) 은 이전 프레임과 현재 프레임에서 마크로블록(Macroblock) 값의 최소 차이값(Sum of absolute differences) 을 이용하여 보간된 프레임(Interpolated Frame) 의 마크로블록이 가지게 되는 모션 벡터 값을 추출한다. 이 때 반복된 배경 패턴 및 여러 움직임들 때문에 모션 추정 과정에서 출력되는 벡터값이 비정상적으로 출력되는 경우가 있다. 여기서 제안된 알고리즘을 통해 모션 벡터값들의 특이치(Outlier) 를 검출하고 이를 교정하기 위한 분산값(Variance) 을 이용하여 모션 벡터 평활화 작업(Motion Vector Smoothing) 을 거친다. 이와 같이 제안된 알고리즘을 이용하여 실험한 결과값으로 프레임률 상향 변환 과정을 통해 상대적으로 계산의 복잡성은 낮으면서 양호한 PSNR 값이 출력됨을 확인할 수 있다.

• Journal of the Institute of Convergence Signal Processing
• /
• v.14 no.3
• /
• pp.154-161
• /
• 2013

In this paper, Gabor cosine and sine transform considered as human visual filter is applied to watermarking methods for digital images. Four algorithms by using singular values or principal components of SVD in the frequency domain are proposed for watermark embedding and extraction. Two dimensional image is used as an embedded watermark. To measure the similarity between the embedded watermark image and the extracted one, a normalized correlation value is computed for the comparison of the four proposed methods with various attacks. Extracted watermark images are also provided for visual inspection. The proposed GCST-SVD method which embeds a watermark image into the lowest vertical or horizontal ac frequency band can provide useful watermarking algorithm with high correlation values and visual watermark features from experimental results for various attacks.

• Journal of the Korean Society of Radiology
• /
• v.12 no.2
• /
• pp.217-223
• /
• 2018

This study to search the diagnostic performance of shear wave elastography(SWE) in breast mass and to compare the biopsy result and stiffness obtained from shear wave elastography. Diagnostic breast ultrasonography and SWE were targeted for 157 patients who had breast ultrasonography was diagnosed mass from June 2017 to September 2017. Pathology results of 157 patients showed a benign 92 patients(Age, $44.54{\pm}11.84$) and a malignancy 65 patients(Age, $51.55{\pm}10.54$). Final evaluation, biopsy result, and quantitative SWE result were obtained and compared with each other according to Breast Imaging Reporting and Data System(BI-RADS) of diagnostic breast ultrasonography. Quantitative SWE value and pathologic result showed the highest diagnostic specificity of 83.70% in Emean and sensitivity of 89.23% in Emin. Quantitative SWE result and biopsy result is statistically significant.(p=0.000). The optimal cut-off value for malignant lesions was 66.3 kPa and 63.7 kPa, respectively, for the sensitivity, specificity, high maximum mean elasticity value(Emax) and mean elasticity value(Emean) and this showed the highest diagnostic area under the ROC curve(Az) value compared to other SWE measurement(p=0.000). The addition of SWE to conventional US in breast mass make a increase diagnostic specificity and reduce unnecessary biopsy. Therefore, it is expected that it will be helpful to analyze the breast mass using the above analysis and apparatus.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.4
• /
• pp.561-567
• /
• 2013

Some of the feature vectors for speech recognition are rare and unusual. These patterns lead to overfitting for the parameters of the speech recognition system and, as a result, cause structural risks in the system that hinder the good performance in recognition. In this paper, as a method of removing these unusual patterns, we try to exclude vectors whose norms are larger than a specified cutoff value and then train the speech recognition system. The objective of this study is to exclude as many unusual feature vectors under the condition of no significant degradation in the speech recognition error rate. For this purpose, we introduce a cutoff parameter and investigate the resultant effect on the speaker-independent speech recognition of isolated words by using FVQ(Fuzzy Vector Quantization)/HMM(Hidden Markov Model). Experimental results showed that roughly 3%~6% of the feature vectors might be considered as unusual, and therefore be excluded without deteriorating the speech recognition accuracy.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.41 no.4
• /
• pp.29-37
• /
• 2004

This paper provides a reliable H$_{\infty}$ state feedback controller design method for delayed singular systems with actuator failures occurred within the prescribed subset. The sufficient condition for the existence of a reliable H$_{\infty}$ controller and the controller design method are presented by linear matrix inequality(LMI), singular value decomposition, Schur complements, and changes of variables. The proposed controller guarantees not only asymptotic stability but also H$_{\infty}$ norm bound in spite of existence of actuator failures. Since the obtained sufficient condition can be expressed as an LMI fen all variables can be calculated simultaneously. Moreover, the controller design method can be extended to the problem of robust reliable H$_{\infty}$ controller design method for singular systems with parameter uncertainties, time-varying delay, and actuator failures. A numerical example is given to illustrate the validity of the result.

• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.125-140
• /
• 2013

We call a data set in which the number of records belonging to a certain class far outnumbers the number of records belonging to the other class, 'imbalanced data set'. Most of the classification techniques perform poorly on imbalanced data sets. When we evaluate the performance of a certain classification technique, we need to measure not only 'accuracy' but also 'sensitivity' and 'specificity'. In a customer churn prediction problem, 'retention' records account for the majority class, and 'churn' records account for the minority class. Sensitivity measures the proportion of actual retentions which are correctly identified as such. Specificity measures the proportion of churns which are correctly identified as such. The poor performance of the classification techniques on imbalanced data sets is due to the low value of specificity. Many previous researches on imbalanced data sets employed 'oversampling' technique where members of the minority class are sampled more than those of the majority class in order to make a relatively balanced data set. When a classification model is constructed using this oversampled balanced data set, specificity can be improved but sensitivity will be decreased. In this research, we developed a hybrid model of support vector machine (SVM), artificial neural network (ANN) and decision tree, that improves specificity while maintaining sensitivity. We named this hybrid model 'hybrid SVM model.' The process of construction and prediction of our hybrid SVM model is as follows. By oversampling from the original imbalanced data set, a balanced data set is prepared. SVM_I model and ANN_I model are constructed using the imbalanced data set, and SVM_B model is constructed using the balanced data set. SVM_I model is superior in sensitivity and SVM_B model is superior in specificity. For a record on which both SVM_I model and SVM_B model make the same prediction, that prediction becomes the final solution. If they make different prediction, the final solution is determined by the discrimination rules obtained by ANN and decision tree. For a record on which SVM_I model and SVM_B model make different predictions, a decision tree model is constructed using ANN_I output value as input and actual retention or churn as target. We obtained the following two discrimination rules: 'IF ANN_I output value <0.285, THEN Final Solution = Retention' and 'IF ANN_I output value ${\geq}0.285$, THEN Final Solution = Churn.' The threshold 0.285 is the value optimized for the data used in this research. The result we present in this research is the structure or framework of our hybrid SVM model, not a specific threshold value such as 0.285. Therefore, the threshold value in the above discrimination rules can be changed to any value depending on the data. In order to evaluate the performance of our hybrid SVM model, we used the 'churn data set' in UCI Machine Learning Repository, that consists of 85% retention customers and 15% churn customers. Accuracy of the hybrid SVM model is 91.08% that is better than that of SVM_I model or SVM_B model. The points worth noticing here are its sensitivity, 95.02%, and specificity, 69.24%. The sensitivity of SVM_I model is 94.65%, and the specificity of SVM_B model is 67.00%. Therefore the hybrid SVM model developed in this research improves the specificity of SVM_B model while maintaining the sensitivity of SVM_I model.