Yongjun Kim

We present a comprehensive empirical characterization of sensor performance across 350 stations in Korea’s national air quality monitoring network (AirKorea), analyzing 748 792 hourly measurements of six pollutants over a 89-day period (6 December 2025–6 March 2026). Our analysis reveals statistically significant drift signals, consistent with possible sensor drift, in 33.8% of PM 2.5 stations and up to 61.3% of SO 2 stations (linear trend p < 0.05 ), with drift rates varying substantially across pollutant types. Spatial cross-calibration of 279 adjacent station pairs ( < 5 km) yields a median Pearson correlation of 0.936 for PM 2.5 , but 23.7% of pairs exhibit systematic biases exceeding 5 μ g m − 3 . We also document pronounced heteroscedastic noise characteristics: daily coefficient of variation (CV) ranges from 0.115 ( SO 2 ) to 0.365 ( NO 2 / O 3 ), with CV inversely related to concentration level. Station failure analysis reveals 36.3% of stations experienced data gaps exceeding 24 h, with mean time between failures of 192.8 d for urban stations. These findings provide actionable benchmarks for calibration scheduling, quality control, and sensor network design. We propose a sensor health index that integrates drift, noise, and data availability into a single composite metric for prioritizing station maintenance.

Climate change is accelerating the introduction of new fruit varieties worldwide, yet machine learning models for agricultural quality prediction require years of historical data to be reliable. This bottleneck severely limits AI deployment in smart farming: a new variety planted today may lack sufficient quality records for a decade. Transfer learning from established crops offers a promising shortcut, but our investigation reveals a critical pitfall: naively combining data from multiple established crops is highly unreliable—improving results in some target domains (+23.7%) while causing catastrophic performance degradation in others (R² drops from 0.553 to −0.494). We term this the source overwhelming problem, where abundant source data drowns the signal from scarce target observations. We present Balanced Multi-Source Transfer Learning (BMTL), a principled framework that controls source-target balance to enable effective cross-crop knowledge transfer, operating in a few-shot regime where a small target validation set (n^val_T ≥ 10) is available. We establish a counterintuitive finding: source-target balance matters more than domain similarity for successful transfer. On fruit quality prediction spanning 4 crops (apple, citrus, grape, pear; 3,721 quality measurements with 20,896 integrated weather records from the Korea Meteorological Administration), BMTL improves R2 by +0.143 points (from 0.393 to 0.536, a +36.4% relative gain) over single-crop training in 7 of 8 tested scenarios, and up to +117.8% when target data is extremely scarce (n_T =50). On UCI benchmarks, BMTL is the only transfer-based method avoiding catastrophic negative transfer on high-shift domains. The proposed method reduces the data collection burden for new crop varieties, enabling agricultural AI systems to become operational years earlier—with direct benefits for food security and supply chain management.

Despite the increasing attention on smart farms as a solution to rural issues such as aging agricultural populations, a shortage of young farmers, decreased production area, and reduced investment leading to stagnant income, exports, and growth rates, many farms still rely on traditional methods like cultivating tangerines in open fields. Despite this, increasing farm income requires producing high-quality tangerines and selling them at premium prices, with fruit sweetness being a crucial factor. Therefore, there is a need to examine the close correlation between tangerine quality and sweetness. In this paper, we use deep learning with the PyCaret library to predict and analyze tangerine sweetness using data from seven regions in Jeju and 13 comprehensive factors influencing sweetness, including terrain, temperature, humidity, precipitation, sunlight, wind speed, acidity, sugar-acid ratio, and others. Although applying all 13 factors could achieve over 90% accuracy, our study, limited to seven factors, still achieves a respectable 82.4% prediction accuracy, demonstrating the significant impact of weather data on sweetness. Moreover, these optimistic predictions enable the estimation of tangerine quality and price formation in the market for the coming year, allowing tangerine farmers and related agencies to respond to market conditions proactively. Furthermore, by applying these data to smart farms to control factors influencing tangerine sweetness, it is anticipated that high-quality tangerine production and increased farm income can be achieved.

The rapid development of information and communication technology has fostered a natural integration of technology and design. As a result, there is increasing interest in Augmented Reality (AR) business card systems that leverage digital media. This research aims to advance the design of an AR-based participatory business card information system in line with contemporary trends. Key aspects of this study include applying technology to acquire contextual information from paper business cards, transmitting it to a server, and delivering it to mobile devices; facilitating interactivity between users and content through a screen interface; providing multimedia business content (video, image, text, 3D elements) via image markers recognized by users on mobile devices, while also adapting the type and method of content delivery. The AR business card system designed in this research enhances traditional paper business cards by incorporating visual information and interactive elements and automatically generating buttons linked to phone numbers, location information, and homepages. This innovative approach enables users to interact and enriches their overall experience while adhering to strict quality control measures.

We live in a flood of big data and information through computers, communications, social media, and mass media. In other words, we can get the information we want quickly and easily, but we have many questions about the accuracy and reliability of this information. That is, there are many problems in trying to obtain accurate knowledge of such reckless details, and in particular, advertisement articles provided by online newspapers need to be clearer and more manageable when individuals try to find precise information and reports. Such experiences are threatened even to the foundation of existence due to distrust of Internet newspapers and advertisement evasion. To solve this problem, this study used emotion analysis of natural language processing to classify general and advertisement articles. Getting going Existing similar studies have mainly been undertaken to classify such advertisement articles, such as spam mail classification, and most of these studies used general natural language processing. However, this paper is a study that analyzes text data to understand further the meaning of the words, sentences, and phrases and adds steps to explore emotions to provide more accurate information that individuals want.

This study uses the API of Upbit, one of Korea’s cryptocurrency exchanges, to predict continuous time series for a limited period and cryptocurrencies using LSTM, a machine learning technique. The trading (buying and selling) point algorithm presented in this study was used to conduct experimental research on efficient profit creation for cryptocurrency investment. Several related studies have shown the results of time series prediction for long-term forecasts, such as a week or several months. Still, they have not attempted to make an ultra-short-term prediction in units of one minute. This paper attempts such a 1 min prediction. This is an experiment to create efficient profits by setting efficient trading (buying and selling) points using machine learning techniques and repeating these operations by an algorithm. Applying it to cryptocurrency shows the possibility of time series prediction.