• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.47 no.2
• /
• pp.134-146
• /
• 2024

This study introduces and experimentally validates a novel approach that combines Instruction fine-tuning and Low-Rank Adaptation (LoRA) fine-tuning to optimize the performance of Large Language Models (LLMs). These models have become revolutionary tools in natural language processing, showing remarkable performance across diverse application areas. However, optimizing their performance for specific domains necessitates fine-tuning of the base models (FMs), which is often limited by challenges such as data complexity and resource costs. The proposed approach aims to overcome these limitations by enhancing the performance of LLMs, particularly in the analysis precision and efficiency of national Research and Development (R&D) data. The study provides theoretical foundations and technical implementations of Instruction fine-tuning and LoRA fine-tuning. Through rigorous experimental validation, it is demonstrated that the proposed method significantly improves the precision and efficiency of data analysis, outperforming traditional fine-tuning methods. This enhancement is not only beneficial for national R&D data but also suggests potential applicability in various other data-centric domains, such as medical data analysis, financial forecasting, and educational assessments. The findings highlight the method's broad utility and significant contribution to advancing data analysis techniques in specialized knowledge domains, offering new possibilities for leveraging LLMs in complex and resource-intensive tasks. This research underscores the transformative potential of combining Instruction fine-tuning with LoRA fine-tuning to achieve superior performance in diverse applications, paving the way for more efficient and effective utilization of LLMs in both academic and industrial settings.

• Health Policy and Management
• /
• v.10 no.4
• /
• pp.75-98
• /
• 2000

This study applied Data Envelopment Analysis(DEA) and Ratio Analysis and Regression Analysis to a set of Korean Public Corporation Medical Centers to evaluate their relative human resource efficiencies. The output measure used in this study was based on health insurance system which was used in both in-patient departments and out-patient departments. Inputs included working time of the doctors, nurses, technicians, and managerial department staff. Based on the data provided on the inputs and outputs, the analysis showed 23 of the 34 hospitals to be relatively inefficient. Each hospital with an efficiency rating of less than 1 was considered relatively inefficient. In addition, managerial strategies based on dual variables were constructed to indicate the manner In which inefficient hospitals may be made efficient. A subsequent analysis of t-test revealed that the bed occupancy rate, medical revenue per 100beds, value added revenue per staff, medical revenue per staff were statistically significant. The results of this study suggest the DEA is a promising tool for evaluating relative human resource efficiency in hospitals which have multiple inputs and outputs and where the efficient production function is not specifiable with any precision. But it is considered that efficiency evaluations may be most effective]y accomplished by Incorporating a combination of methodologies such as ratio analysis and regression analysis.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.18 no.6
• /
• pp.82-90
• /
• 2019

To reduce emissions of marine pollutants, regulations are being tightened around the world. In the shipbuilding and shipping industries, various countermeasures are being put forward. As there are limits to applying countermeasures to ships already in operation, however, it is necessary for these vessels to use energy efficiently. The sensors installed on ships typically gather a very large amount of data, and thus a big data platform is needed to manage and analyze the data. In this paper, we build a big data analysis platform based on Hadoop and Spark, and we present a method to detect abnormal ship operation using the platform. We also utilize real ship operation data to discuss the data analysis experiment.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.373-378
• /
• 2004

Recently the electrostatic 2-axis MEMS stages have been fabricated for the purpose of an application to PSD (Probebased Storage Device). However, most of them have low area efficiency, which is undesirable as data storage devices, since all of the components (springs, comb electrodes, anchors, platform, etc.) are placed in-plane. In this paper, we present a novel structure of electrostatic 2-axis MEMS stage that is characterized by having large area platform. For large area efficiency, the actuator part consisting of mainly comb electrodes and springs is placed right below the platform. In this article, the structures and operational principle of the MEMS stages are described, followed by design procedure, structural and modal analysis using FEM(Finite Element Method). The area efficiency of the MEMS stage was designed to be about 55％, that is very large compared with conventional ones having a few percentage.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.700-703
• /
• 2004

A reliability analysis and design procedure based on the design of experiment (DOE) is combined with the response surface method (RSM) for numerical efficiency. The procedure established is based on a 3$^n$ full factorial DOE for numerical quadrature using explicit formula of optimum levels and weights derived for general distributions. The full factorial moment method (FFMM) shows good performance in terms of accuracy and ability to treat non-normally distributed random variables. But, the FFMM becomes very inefficient because the number of function evaluation required increases exponentially as the number of random variables considered increases. To enhance the efficiency, the response surface moment method (RSMM) is proposed. In RSMM, experiments only with high probability are conducted and the rest of data are complemented by a quadratic response surface approximation without mixed terms. The response surface is updated by conducting experiments one by one until the value of failure probability is converged. It is calculated using the Pearson system and the four statistical moments obtained from the experimental data. A measure for checking the relative importance of an experimental point is proposed and named as influence index. During the update of response surface, mixed terms can be added into the formulation.

• Genomics & Informatics
• /
• v.16 no.4
• /
• pp.38.1-38.3
• /
• 2018

Gene-gene interaction (GGI) analysis is known to play an important role in explaining missing heritability. Many previous studies have already proposed software to analyze GGI, but most methods focus on a binary phenotype in a case-control design. In this study, we developed "Hierarchical structural CoMponent analysis of Gene-Gene Interactions" (HisCoM-GGI) software for GGI analysis with a continuous phenotype. The HisCoM-GGI method considers hierarchical structural relationships between genes and single nucleotide polymorphisms (SNPs), enabling both gene-level and SNP-level interaction analysis in a single model. Furthermore, this software accepts various types of genomic data and supports data management and multithreading to improve the efficiency of genome-wide association study data analysis. We expect that HisCoM-GGI software will provide advanced accessibility to researchers in genetic interaction studies and a more effective way to understand biological mechanisms of complex diseases.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2001.04a
• /
• pp.105-109
• /
• 2001

The micro optical parts such as ferrules are required to be manufactured within very small tolerances, as the slight deviation of the tolerance would give very large amount of loss in communication efficiency. For efficient optical communication, outer diameter, fiber diameter, fiber separation and eccentricity are significant parameters to be inspected., Thus we developed an automatic inspection system to evaluate shape parameters of the optical fiber connectors(ferrule) upto submicron accuracy using machine vision. new optical probe of multi fields of views has been developed and the image processing and data analysis algorithms have been complemented in real time basis. The developed system is successfully used in the practical ferrule manufacturing industry, and about 0.1$\mu\textrm{m}$ accuracy can be obtained with very fast inspection time.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2003.06a
• /
• pp.133-136
• /
• 2003

In these days, the reliability analysis and prediction are applied for many industrial products and many products require guaranteeing the quality and efficiency of their products. In this study reliability prediction for core units of machine tools has been performed in order to improve and analyze its reliability. ATC(Automatic Tool Changer) and interface Card of PC-NC that are core component of the machine tools were chosen as the target of the reliability prediction. A reliability analysis tool was used to obtain the reliability data(failure rate database) for reliability prediction. It is expected that the results of reliability prediction be applied to improve and evaluate its reliability. Failure rate, MTBF (Mean Time Between Failure) and reliability for core units of machine tools were evaluated and analyzed in this study.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2006.05a
• /
• pp.619-620
• /
• 2006

In this paper, the CAE Service System for Collaborative Engineering Environment with web services and Multi-frontal Method has been investigated and developed. The enabling technologies such as SOAP and .NET Framework play great roles in the development of integrated distributed application software. In addition to the distribution of analysis modules, numerical solution process itself is again divided into parallel processes using Multi-frontal Method for computational efficiency. We believe that the proposed approach for the analysis can be extended to the entire product development process for sharing and utilizing common product data in the distributed engineering environment.

• Design & Manufacturing
• /
• v.16 no.4
• /
• pp.40-45
• /
• 2022

The EV electric vehicle market is growing rapidly worldwide. An electric vehicle means a vehicle that uses energy charged through an electricity source as power. The precision of the press is important to mass-produce the drive motor, which is a key component of the electric vehicle. The size of the driving motor is increasing, and The size of the mold is also growing. In this study, the precision of large high-speed presses for mass production of driving motors was measured. A study was conducted on the measurement method of press and the analysis of measurement data. A drive motor is a component that transmits power by converting electrical energy into kinetic energy. EV driven motors have key material properties to improve efficiency. The material properties are the thickness of the material. As a method for improving performance, use a 0.2mm thin steel sheet. Mold is also becoming larger. As the mold grows, the size of the high-speed press for mass production of the driving motor is also increasing. Also, the precision of the press is the most important because it uses a thin iron plate material. So the importance of large press precision is being emphasized. In this study, the effect of large high-speed press structure on precision was verified