• Journal of Information Technology Services
• /
• 제18권2호
• /
• pp.143-159
• /
• 2019

Machine learning (ML) is a method of fitting given data to a mathematical model to derive insights or to predict. In the age of big data, where the amount of available data increases exponentially due to the development of information technology and smart devices, ML shows high prediction performance due to pattern detection without bias. The feature engineering that generates the features that can explain the problem to be solved in the ML process has a great influence on the performance and its importance is continuously emphasized. Despite this importance, however, it is still considered a difficult task as it requires a thorough understanding of the domain characteristics as well as an understanding of source data and the iterative procedure. Therefore, we propose methods to apply deep learning for solving the complexity and difficulty of feature extraction and improving the performance of ML model. Unlike other techniques, the most common reason for the superior performance of deep learning techniques in complex unstructured data processing is that it is possible to extract features from the source data itself. In order to apply these advantages to the business problems, we propose deep learning based methods that can automatically extract features from transaction data or directly predict and classify target variables. In particular, we applied techniques that show high performance in existing text processing based on the structural similarity between transaction data and text data. And we also verified the suitability of each method according to the characteristics of transaction data. Through our study, it is possible not only to search for the possibility of automated feature extraction but also to obtain a benchmark model that shows a certain level of performance before performing the feature extraction task by a human. In addition, it is expected that it will be able to provide guidelines for choosing a suitable deep learning model based on the business problem and the data characteristics.

• Speech Sciences
• /
• 제4권2호
• /
• pp.55-68
• /
• 1998

In this paper, a novel method of noise reduction of speech based on a complex adaptive noise canceler and method of optimal feature extraction are proposed. This complex adaptive noise canceler needs simply the noise detection, and LMS algorithm used to calculate the adaptive filter coefficient. The method of optimal feature extraction requires the variance of noise. The experimental results have shown that the proposed method effectively reduced noise in noisy speech. Optimal feature extraction has shown similar characteristics in noise-free speech.

• Biotechnology and Bioprocess Engineering:BBE
• /
• 제9권3호
• /
• pp.207-211
• /
• 2004

Predispersed solvent extraction (PDSE) of succinic acid with Tri-n-octylamine (TOA) dissolved in 1-octanol from aqueous solutions of 50 g/L succinic acid was examined. It was found that the equilibrium data in PDSE was equal to that in conventional solvent extraction in spite of the lack of mechanical mixing in PDSE. The influence of salts on succinic acid extraction and the stability of colloidal liquid aphrons (CLAs) were also investigated. Results indicated that in the presence of sodium chloride, less succinic acid was extracted by CLAs and the stability of CLAs decreased. However, the stability of CLAs was sufficient to make PDSE practically applicable to real fermentation broth, considering the concentration range of salts in the fermentation process for succinic acid.

• Journal of Advanced Information Technology and Convergence
• /
• 제10권1호
• /
• pp.45-56
• /
• 2020

Despite extensive research, performance enhancement of keyphrase (KP) extraction remains a challenging problem in modern informatics. Recently, deep learning-based supervised approaches have exhibited state-of-the-art accuracies with respect to this problem, and several of the previously proposed methods utilize Bidirectional Encoder Representations from Transformers (BERT)-based language models. However, few studies have investigated the effective application of BERT-based fine-tuning techniques to the problem of KP extraction. In this paper, we consider the aforementioned problem in the context of scientific articles by investigating the fine-tuning characteristics of two distinct BERT models - BERT (i.e., base BERT model by Google) and SciBERT (i.e., a BERT model trained on scientific text). Three different datasets (WWW, KDD, and Inspec) comprising data obtained from the computer science domain are used to compare the results obtained by fine-tuning BERT and SciBERT in terms of KP extraction.

• Proceedings of the IEEK Conference
• /
• 대한전자공학회 2000년도 추계종합학술대회 논문집(2)
• /
• pp.47-50
• /
• 2000

DC Gummel-Poon SPICE model parameter extraction program has been implemented. This program extracts the parameters from measured data using Levenberg-Marquardt algorithm. Measured data consist of forward and reverse Gummel plot, forward and reverse output characteristics and RE and RC measurements.

• Elastomers and Composites
• /
• 제48권4호
• /
• pp.288-293
• /
• 2013

Crosslink characteristics of maleic anhydride-grafted EPDM (MAH-g-EPDM)/zinc oxide composite were investigated by weight losses after dichloroacetic acid (DCA)/toluene cosolvent extraction at different temperatures and by measurement of crosslink densities. The chemical changes were analyzed using attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR). The weight losses by extraction at high temperature ($90^{\circ}C$) were remarkably greater than those at room temperature and those by DCA/toluene cosolvent extraction were greater than those by toluene one by more than 5 times. The crosslink densities were measured after the solvent extraction, and the second crosslink densities were higher than the first ones. The first crosslink density was lower when the extraction temperature was high, and it was much lower for the toluene extraction than for the DCA/toluene cosolvent extraction. The second crosslink density of the sample extracted with DCA/toluene cosolvent was greater than that extracted with toluene. The extracted components were depending on the extraction solvents and temperatures, for example; only strong crosslinked networks were remained when extracting with DCA/toluene cosolvent at high temperature, while only uncrosslinked polymer chains were extracted when extracting with toluene at room temperature. Therefore, crosslink characteristics of the MAH-g-EPDM/zinc oxide composite can be analyzed by comparison of the extracted components according to the extraction solvents and temperatures and by measurement of successive crosslink densities.

• Culinary science and hospitality research
• /
• 제20권3호
• /
• pp.13-21
• /
• 2014

We developed a method for improving the antioxidant activities of Citrus extracts through cyclic low pressure drying(CLPD) and a high hydrostatic pressure extraction (HPE) process. Citrus fruits were prepared for water extraction at $60^{\circ}C$ and 300 MPa for 5 min (high hydrostatic pressure extraction, HPE5) and 15 min (high hydrostatic pressure extraction, HPE15) after cyclic low pressure drying method. Extraction yields obtained by cyclic low pressure drying and high hydrostatic pressure extraction process were 20.41, 23.47, and 28.19%, respectively. Total polyphenol contents were increased by combined process. Generally, CLPD and HPE resulted in higher yields than the conventional extraction process. Further, HPE15 showed 48.21% DPPH radical scavenging activity (EDA, %) at $1,000{\mu}g/mL$. In general, antioxidant activities of Citrus increased by CLPD and HPE. Therfore, CLPD and HPE of Citrus resulted in higher antioxidant activity than conventional water extraction.

• Proceedings of the KSRS Conference
• /
• 대한원격탐사학회 2008년도 International Symposium on Remote Sensing
• /
• pp.147-150
• /
• 2008

This study refers to develop a semi-automatic extraction of agricultural land use and vegetation information using high resolution satellite images. Data of IKONOS satellite image (May 25 of 2001) and QuickBird satellite image (May 1 of 2006) which resembles with the spatial resolution and spectral characteristics of KOMPSAT3. The precise agricultural land use classification was tried using ISODATA unsupervised classification technique and the result was compared with on-screen digitizing land use accompanying with field investigation. For the extraction of vegetation information, three crops of paddy, com and red pepper were selected and the spectral characteristics were collected during each growing period using ground spectroradiometer. The vegetation indices viz. RVI, NDVI, ARVI, and SAVI for the crops were evaluated. The evaluation process is under development using the ERDAS IMAGINE Spatial Modeler Tool.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• 제14권6호
• /
• pp.503-511
• /
• 2002

Heat flow characteristics during hot water extraction process was studied experimentally. Data were taken at various outlet port type for the fixed inlet port type, inlet-outlet temperature differences and mass flow rates. In this study, the temperature distribution in a storage tank and an outlet temperature were measured to predict a degree of stratification in the storage tank, and a useful rate of hot water was analysed with respect to the variables dominating a extraction process. Experimental results show that the degree of stratification and useful rate of hot water are all high in a low flow rate in case of using modified distributor I (MDI) as the outlet port type.

• ETRI Journal
• /
• 제33권4호
• /
• pp.547-557
• /
• 2011

This study presents an approach for extracting boundaries of various buildings, which have concave boundaries, inner yards, non-right-angled corners, and nonlinear edges. The approach comprises four steps: building point segmentation, boundary tracing, boundary grouping, and regularization. In the second and third steps, conventional algorithms are improved for more accurate boundary extraction, and in the final step, a new algorithm is presented to extract nonlinear edges. The unique characteristics of airborne light detection and ranging (LIDAR) data are considered in some steps. The performance and practicality of the presented algorithm were evaluated for buildings of various shapes, and the average omission and commission error of building polygon areas were 0.038 and 0.033, respectively.