• Journal of Intelligence and Information Systems
• /
• v.29 no.1
• /
• pp.249-263
• /
• 2023

Collaborative Filtering, a representative recommendation system methodology, consists of two approaches: neighbor methods and latent factor models. Among these, the latent factor model using matrix factorization decomposes the user-item interaction matrix into two lower-dimensional rectangular matrices, predicting the item's rating through the product of these matrices. Due to the factor vectors inferred from rating patterns capturing user and item characteristics, this method is superior in scalability, accuracy, and flexibility compared to neighbor-based methods. However, it has a fundamental drawback: the need to reflect the diversity of preferences of different individuals for items with no ratings. This limitation leads to repetitive and inaccurate recommendations. The Adaptive Deep Latent Factor Model (ADLFM) was developed to address this issue. This model adaptively learns the preferences for each item by using the item description, which provides a detailed summary and explanation of the item. ADLFM takes in item description as input, calculates latent vectors of the user and item, and presents a method that can reflect personal diversity using an attention score. However, due to the requirement of a dataset that includes item descriptions, the domain that can apply ADLFM is limited, resulting in generalization limitations. This study proposes a Generalized Adaptive Deep Latent Factor Recommendation Model, G-ADLFRM, to improve the limitations of ADLFM. Firstly, we use item ID, commonly used in recommendation systems, as input instead of the item description. Additionally, we apply improved deep learning model structures such as Self-Attention, Multi-head Attention, and Multi-Conv1D. We conducted experiments on various datasets with input and model structure changes. The results showed that when only the input was changed, MAE increased slightly compared to ADLFM due to accompanying information loss, resulting in decreased recommendation performance. However, the average learning speed per epoch significantly improved as the amount of information to be processed decreased. When both the input and the model structure were changed, the best-performing Multi-Conv1d structure showed similar performance to ADLFM, sufficiently counteracting the information loss caused by the input change. We conclude that G-ADLFRM is a new, lightweight, and generalizable model that maintains the performance of the existing ADLFM while enabling fast learning and inference.

• Journal of Technology Innovation
• /
• v.31 no.1
• /
• pp.105-148
• /
• 2023

This paper provides a new method of measuring the degree of technological progress which contributes to real economic growth based on Schumpeter's Trilogy. Using Microdata of Statistics Korea, the results of measuring and comparing the actual growth contribution of technological progress during the period 2003-2018 by the total factor productivity growth rate(growth accounting method), the R&D investment contribution rate, and the Schumpeterian innovation growth rate, respectively are as follows. First, the measurement of the real growth contribution of technological progress by the growth rate of total factor productivity and the growth rate of Schumpeterian innovation shows contradictory results. Second, when the growth rate of production is in a decreasing trend, the difference between the growth rate of production and the growth rate of total factor productivity increases compared to when it is in an increasing trend. Conversely, when there is an increasing trend, the difference between the growth rate of production and the growth rate of total factor productivity becomes smaller compared to when it is in a decreasing trend.. Third, the technological opportunity that affects the innovation growth rate, i.e., the contribution of R&D incentives to innovative growth is only 3.3%. The reason why this result is different from the existing perception of the contribution of technological progress to growth is that different entities are being measured while measuring the same term of technological progress. Therefore, the growth rate of total factor productivity should be used to measure macroeconomic efficiency, R&D investment should be used to measure the effectiveness of new technology supply, and the Schumpeterian innovation rate should be used to measure the economic impact of technological progress. The policy implications of the research results of this thesis are as follows: ① Transition from a policy of one-sided technology supply to a policy of convergence of technology supply and new technology demand support, ② Mission-oriented R&D policy and R&D policy that links national R&D with private R&D, ③ Reclassification of capital goods reflecting the degree of new knowledge.

• Journal of the Society of Disaster Information
• /
• v.19 no.2
• /
• pp.395-410
• /
• 2023

Purpose: The purpose of this study was to analyze the current state of drone terrorism response at such critical national facilities and derive improvements, especially to identify problems in laws and systems to effectively utilize the anti-drone system and present directions for improvement. Method: A qualitative research method was used for this study by analyzing a variety of issues not discussed in existing research papers and policy documents through in-depth interviews with subject matter experts. In-depth interviews were conducted based on 12 semi-structured interviews by selecting 16 experts in the field of anti-drone and terrorism in Korea. The interview contents were recorded with the prior consent of the study participants, transcribed back to the Korean file, and problems and improvement measures were derived through coding. For this, the threats and types were analyzed based on the cases of drone terrorism occurring abroad and measures to establish anti-drone system were researched from the perspective of laws and systems by evaluating the possibility of drone terrorism in the Republic of Korea. Result: As a result of the study, improvements to some of the problems that need to be preceded in order to effectively respond to drone terrorism at critical national facilities in the Republic of Korea, have been identified. First, terminologies related to critical national facilities and drone terrorism should be clearly defined and reflected in the Integrated Defense Act and the Terrorism Prevention Act. Second, the current concept of protection of critical national facilities should evolve from the current ground-oriented protection to a three-dimensional protection concept that considers air threats and the Integrated Defense Act should reflect a plan to effectively install the anti-drone system that can materialize the concept. Third, a special law against flying over critical national facilities should be enacted. To this end, legislation should be enacted to expand designated facilities subject to flight restrictions while minimizing the range of no fly zone, but the law should be revised so that the two wings of "drone industry development" and "protection of critical national facilities" can develop in a balanced manner. Fourth, illegal flight response system and related systems should be improved and reestablished. For example, it is necessary to prepare a unified manual for general matters, but thorough preparation should be made by customizing it according to the characteristics of each facility, expanding professional manpower, and enhancing response training. Conclusion: The focus of this study is to present directions for policy and technology development to establish an anti-drone system that can effectively respond to drone terrorism and illegal drones at critical national facilities going forward.

• Journal of Intelligence and Information Systems
• /
• v.29 no.2
• /
• pp.171-187
• /
• 2023

When a person's face is recognized through a recording device such as a low-pixel surveillance camera, it is difficult to capture the face due to low image quality. In situations where it is difficult to recognize a person's face, problems such as not being able to identify a criminal suspect or a missing person may occur. Existing studies on face recognition used refined datasets, so the performance could not be measured in various environments. Therefore, to solve the problem of poor face recognition performance in low-quality images, this paper proposes a method to generate high-quality images by performing image quality improvement on low-quality facial images considering various environments, and then improve the performance of facial feature point detection. To confirm the practical applicability of the proposed architecture, an experiment was conducted by selecting a data set in which people appear relatively small in the entire image. In addition, by choosing a facial image dataset considering the mask-wearing situation, the possibility of expanding to real problems was explored. As a result of measuring the performance of the feature point detection model by improving the image quality of the face image, it was confirmed that the face detection after improvement was enhanced by an average of 3.47 times in the case of images without a mask and 9.92 times in the case of wearing a mask. It was confirmed that the RMSE for facial feature points decreased by an average of 8.49 times when wearing a mask and by an average of 2.02 times when not wearing a mask. Therefore, it was possible to verify the applicability of the proposed method by increasing the recognition rate for facial images captured in low quality through image quality improvement.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.25 no.4
• /
• pp.337-345
• /
• 2023

Estimating the rice heading date is one of the most crucial agricultural tasks related to productivity. However, due to abnormal climates around the world, it is becoming increasingly challenging to estimate the rice heading date. Therefore, a more objective classification method for estimating the rice heading date is needed than the existing methods. This study, we aimed to classify the rice heading stage from various images using a CNN classification model. We collected top-view images taken from a drone and a phenotyping tower, as well as slanted-view images captured with a RGB camera. The collected images underwent preprocessing to prepare them as input data for the CNN model. The CNN architectures employed were ResNet50, InceptionV3, and VGG19, which are commonly used in image classification models. The accuracy of the models all showed an accuracy of 0.98 or higher regardless of each architecture and type of image. We also used Grad-CAM to visually check which features of the image the model looked at and classified. Then verified our model accurately measure the rice heading date in paddy fields. The rice heading date was estimated to be approximately one day apart on average in the four paddy fields. This method suggests that the water head can be estimated automatically and quantitatively when estimating the rice heading date from various paddy field monitoring images.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.25 no.4
• /
• pp.436-445
• /
• 2023

Methane (CH₄) emissions from rice paddies are mainly observed using the closed chamber method or the eddy covariance method. In this study, a new observation technique combining a portable gas analyzer (Model LI-7810, LI-COR, Inc., USA) and an automatic opening/closing chamber (Model Smart Chamber, LI-COR, Inc., USA) was introduced based on the strengths and weaknesses of the existing measurement methods. A cylindrical collar was manufactured according to the maximum growth height of rice and used as an auxiliary measurement tool. All types of measured data can be monitored in real time, and CH₄ flux is also calculated simultaneously during the measurement. After the measurement is completed, all the related data can be checked using the software called 'SoilFluxPro'. The biggest advantage of the new observation technique is that time-series changes in greenhouse gas concentrations can be immediately confirmed in the field. It can also be applied to small areas with various treatment conditions, and it is simpler to use and requires less effort for installation and maintenance than the eddy covariance system. However, there are also disadvantages in that the observation system is still expensive, requires specialized knowledge to operate, and requires a lot of manpower to install multiple collars in various observation areas and travel around them to take measurements. It is expected that the new observation technique can make a significant contribution to understanding the CH₄ emission pathways from rice paddies and quantifying the emissions from those pathways.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.29 no.7
• /
• pp.779-793
• /
• 2023

Intensive and long-term aquaculture activities in Korea have generated considerable amounts of organic matter, deteriorating the sedimentary environment and ecosystem. The Korean government enacted the Fishery Management Act to preserve and manage the environment of fish farms. Based on this, a fisheries environment assessment has been conducted on fish cage farms since 2014, necessitating the development of a scientific and objective evaluation method suitable for the domestic environment. Therefore, a benthic health index (BHI) was developed using the relationship between benthic polychaete communities and organic matter, a major source of pollution in fish farms. In this study, the development process and calculation method of the BHI have been introduced. The BHI was calculated by classifying 225 species of polychaetes appearing in domestic coastal and aquaculture areas into four groups by linking the concentration gradient of the total organic carbon in the sediment and the distributional characteristics of each species and assigning differential weights to each group. Using BHI, the benthic fauna communities were assigned to one of the four ecological classes (Grade 1: Normal, Grade 2: Slightly polluted, Grade 3: Moderately polluted, and Grade 4: Heavily polluted). The application of the developed index in the field enabled effective evaluation of the Korean environment, being relatively more accurate and less affected by the season compared with the existing evaluation methods like the diversity index or AZTI's Marine Biotic Index developed overseas. In addition, using BHI will be useful in the environmental management of fish farms, as the environment can be graded in quantified figures.

• Journal of Environmental Impact Assessment
• /
• v.32 no.6
• /
• pp.577-589
• /
• 2023

In Korea, the technical techniques for assessing visual impacts are standardized, but the methods for assessing the marine landscape itself are not standardized and need to be improved. In particular, in the landscape impact assessment of offshore wind power generation in Korea, it is necessary to recognize the landscape itself as a receptor and prepare a system that can evaluate the characteristics and sensitivity of the landscape. In this study, we propose an evaluation method for preparing a marine landscape quality assessment document that reflects the project characteristics of offshore wind power projects, and examine the possibility of utilization by applying it to actual project sites as an example. To evaluate the quality of marine scenery in offshore wind power projects, evaluation items of landscape characteristics, physical characteristics, and socio-cultural characteristics were evaluated based on the preliminary survey contents, and the quality of marine scenery was divided into five grades. Next, the evaluation criteria of the evaluation items were synthesized and the quality of the marine landscape was classified into preservation grade (grade 5), semi-preservation grade (grade 4), buffer grade (grade 3), semi-improvement grade (grade 2), and improvement grade (grade 1). In addition, the Sinan-Ui Offshore Wind Farm, an actual project site, was randomly selected to conduct the evaluation process and examine its utilization. This study aims to complement the existing method of visual impact assessment in offshore wind power projects and evaluate the quality of the marine landscape itself to effectively conserve marine landscape resources during offshore wind power projects. Rather than relying on mechanical and quantitative evaluation, this study is expected to be used as a basis for comprehensive understanding of the location and socio-cultural characteristics of the project site and for communication and cooperation with stakeholders.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_4
• /
• pp.1097-1109
• /
• 2023

Urban is an area where small-scale changes to individual buildings occur frequently. An existing urban building database requires periodic updating to increase its usability. However, there are limitations in data collection for building changes over a wide urban. In this study, we check the possibility of detecting building changes and updating a building database by using satellite images that can capture a wide urban region by a single image. For this purpose, building areas in a satellite image are first extracted by projecting 3D coordinates of building corners available in a building database onto the image. Building areas are then divided into roof and facade areas. By comparing textures of the roof areas projected, building changes such as height change or building removal can be detected. New height values are estimated by adjusting building heights until projected roofs align to actual roofs observed in the image. If the projected image appeared in the image while no building is observed, it corresponds to a demolished building. By checking buildings in the original image whose roofs and facades areas are not projected, new buildings are identified. Based on these results, the building database is updated by the three categories of height update, building deletion, or new building creation. This method was tested with a KOMPSAT-3A image over Incheon Metropolitan City and Incheon building database available in public. Building change detection and building database update was carried out. Updated building corners were then projected to another KOMPSAT-3 image. It was confirmed that building areas projected by updated building information agreed with actual buildings in the image very well. Through this study, the possibility of semi-automatic building change detection and building database update based on single satellite image was confirmed. In the future, follow-up research is needed on technology to enhance computational automation of the proposed method.

• Korean Chemical Engineering Research
• /
• v.61 no.4
• /
• pp.523-541
• /
• 2023

As accessibility to 3D printers increases, there is a growing frequency of exposure to chemicals associated with 3D printing. However, research on the toxicity and harmfulness of chemicals generated by 3D printing is insufficient, and the performance of toxicity prediction using in silico techniques is limited due to missing molecular structure data. In this study, quantitative structure-activity relationship (QSAR) model based on data-centric AI approach was developed to predict the toxicity of new 3D printing materials by imputing missing values in molecular descriptors. First, MissForest algorithm was utilized to impute missing values in molecular descriptors of hazardous 3D printing materials. Then, based on four different machine learning models (decision tree, random forest, XGBoost, SVM), a machine learning (ML)-based QSAR model was developed to predict the bioconcentration factor (Log BCF), octanol-air partition coefficient (Log Koa), and partition coefficient (Log P). Furthermore, the reliability of the data-centric QSAR model was validated through the Tree-SHAP (SHapley Additive exPlanations) method, which is one of explainable artificial intelligence (XAI) techniques. The proposed imputation method based on the MissForest enlarged approximately 2.5 times more molecular structure data compared to the existing data. Based on the imputed dataset of molecular descriptor, the developed data-centric QSAR model achieved approximately 73%, 76% and 92% of prediction performance for Log BCF, Log Koa, and Log P, respectively. Lastly, Tree-SHAP analysis demonstrated that the data-centric-based QSAR model achieved high prediction performance for toxicity information by identifying key molecular descriptors highly correlated with toxicity indices. Therefore, the proposed QSAR model based on the data-centric XAI approach can be extended to predict the toxicity of potential pollutants in emerging printing chemicals, chemical process, semiconductor or display process.