Seonghun Lee

Dr. Seonghun Lee

orcid

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 August 2020 in Animals
Verified
Journal Article
Open access
PDF file
MLMingyung Lee
Mingyung Lee
SLSeonghun Lee
Seonghun Lee
JPJaehwa Park
Jaehwa Park
SSSeongwon Seo
Seongwon Seo
The aim of the study was to group the lactation curve (LC) of Holstein cows in several clusters based on their milking characteristics and to investigate physiological differences among the clusters. Milking data of 330 lactations which have a milk yield per day during entire lactation period were used. The data were obtained by refinement from 1332 lactations from 724 cows collected from commercial farms. Based on the similarity measures, clustering was performed using the k-medoids algorithm; the number of clusters was determined to be six, following the elbow method. Significant differences on parity, peak milk yield, DIM at peak milk yield, and average and total milk yield (p < 0.01) were observed among the clusters. Four clusters, which include 82% of data, show typical LC patterns. The other two clusters represent atypical patterns. Comparing to the LCs generated from the previous models, Wood, Wilmink and Dijsktra, it is observed that the prediction errors in the atypical patterns of the two clusters are much larger than those of the other four cases of typical patterns. The presented model can be used as a tool to refine characterization on the typical LC patterns, excluding atypical patterns as exceptional cases.

7 October 2009 in Journal of the Science of Food and Agriculture
Verified
Journal Article
Closed access
PDF file
SLSeonghun Lee
Seonghun Lee
BYBu‐Young Yi
Bu‐Young Yi
Su-Jeong KimSu-Jeong Kim
Su-Jeong Kim

7 April 2019 in Concurrency and Computation: Practice and Experience
Verified
Journal Article
Closed access
PDF file
SLSeonghun Lee
Seonghun Lee
JPJaehwa Park
Jaehwa Park

19 July 2026 in Agriengineering
Verified
Journal Article
Open access
PDF file
SLSeonghun Lee
Seonghun Lee
WJWonjin Jung
Wonjin Jung
SCSubin Cho
Subin Cho
MKMyeong-Kwan Kim
Myeong-Kwan Kim
DPDonghee Park
Donghee Park
JPJaehwa Park
Jaehwa Park
Modern dairy farms continuously record individual cow milk yields, yet turning these records into on-demand forecasts requires a multi-step data-to-model pipeline that natural language alone cannot drive. We study large language model (LLM) orchestration over a verified forecasting pipeline, asking not whether it predicts yield but how faithfully it converts requests into correct tool-execution conditions. Against a deterministic Oracle pipeline, we compare a single-LLM function-calling framework and a multi-agent framework that distributes the same workflow across specialized agents. Using Gemini 2.5 Flash, both answer 1085 queries (4340 forecast points) varying in phrasing, horizon, and missingness, drawn from 13 months of commercial-farm records. On execution-matched points (those reproducing the Oracle’s cow, dates, input window, and horizon), the LLM output is statistically equivalent to the Oracle (paired TOST within ±0.5 kg, pTOST<0.001), so agreement is largely fixed by the deterministic tools. This equivalence is therefore confined to execution-matched points and does not describe overall system reliability. The frameworks differ instead in fidelity: the single-LLM framework reaches only 94.31% completion (zero-filled RMSE 9.70 kg), whereas the multi-agent framework reaches 99.88% at 35% added runtime; the single-LLM losses reflect systematic date-anchoring errors and silent false-accepts, not random noise. In this case study, LLM frameworks can orchestrate verified forecasting tools, and separating responsibilities improves workflow fidelity more than accuracy.

1 August 2017
Verified
Conference Paper
Closed access
SLSeonghun Lee
Seonghun Lee
MOMessaoudi Oumaima
Messaoudi Oumaima
SRSu-Hyeon Ryu
Su-Hyeon Ryu
JPJaehwa Park
Jaehwa Park
A real-time data acquisition system for precision agricultural is presented. Using the system, the spatiotemporal data for the specific local area can be collected at small expense. A hardware frame is designed to take synchronized multi-view images of the target plants as well as environmental data. To monitor the growth of plants or crops, set of images are taken periodically and a multi-view stereo technique is adopted to build a 3-D model to represent the target plant. For real-time monitoring an application running on the Android platform is developed also.

11 March 2022 in Agriculture
Verified
Journal Article
Open access
PDF file
SLSeonghun Lee
Seonghun Lee
JPJaehwa Park
Jaehwa Park
Machine learning techniques provide efficient data analysis tools without mathematical derivations. Data-centric LC representations are highly demanded to use these tools for LC-related research. A novel data-oriented LC representation model using piecewise linear regression (PWLR) is presented. This representation is intended to be used directly as data for machine learning along with other associated data at an individual base. An LC is represented in vector form as a series of connected line segments and the location and number of segments are determined by the maximum residual. The critical points are determined at the rapid transit point in the LC. The Bayesian information criterion was used to choose the proper number of line segments to avoid the overfitting problem. To demonstrate the validity of the PWLR model as an LC descriptor, its approximation accuracy and representation generality were tested experimentally. The results revealed that the PWLR model is advantageous for representing the LCs of an individual or a large herd that are directly applicable to data-driven approaches.