• Journal of KIISE
• /
• v.43 no.7
• /
• pp.781-785
• /
• 2016

Learning from human behaviors in the real world is essential for human-aware intelligent systems such as smart assistants and autonomous robots. Most of research focuses on correlations between sensory patterns and a label for each activity. However, human activity is a combination of several event contexts and is a narrative story in and of itself. We propose a novel approach of human activity prediction based on event cognition. Egocentric multi-sensor data are collected from an individual's daily life by using a wearable device and smartphone. Event contexts about location, scene and activities are then recognized, and finally the users" daily activities are predicted from a decision rule based on the event contexts. The proposed method has been evaluated on a wearable sensor data collected from the real world over 2 weeks by 2 people. Experimental results showed improved recognition accuracies when using the proposed method comparing to results directly using sensory features.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.11
• /
• pp.41-49
• /
• 2019

We are to build an unsupervised machine learning-based language model which can estimate the amount of information that are in need to process words consisting of subword-level morphemes and syllables. We are then to investigate whether the reading times of words reflecting their morphemic and syllabic structures are predicted by an information-theoretic measure such as surprisal. Specifically, the proposed Morfessor-based unsupervised machine learning model is first to be trained on the large dataset of sentences on Sejong Corpus and is then to be applied to estimate the information-theoretic measure on each word in the test data of Korean words. The reading times of the words in the test data are to be recruited from Korean Lexicon Project (KLP) Database. A comparison between the information-theoretic measures of the words in point and the corresponding reading times by using a linear mixed effect model reveals a reliable correlation between surprisal and reading time. We conclude that surprisal is positively related to the processing effort (i.e. reading time), confirming the surprisal hypothesis.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.22 no.2
• /
• pp.225-238
• /
• 2012

Among the various security threats in online games, the use of game bots is the most serious problem. In this paper, we propose a framework for user behavior analysis for bot detection in online games. Specifically, we focus on party play that reflects the social activities of gamers: In a Massively Multi-user Online Role Playing Game (MMORPG), party play log includes a distinguished information that can classify game users under normal-user and abnormal-user. That is because the bot users' main activities target on the acquisition of cyber assets. Through a statistical analysis of user behaviors in game activity logs, we establish the threshold levels of the activities that allow us to identify game bots. Also, we build a knowledge base of detection rules based on this statistical analysis. We apply these rule reasoner to the sixth most popular online game in the world. As a result, we can detect game bot users with a high accuracy rate of 95.92%.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.7
• /
• pp.299-306
• /
• 2022

In recent, sensors embedded in robots, equipment, and circuits have become common, and research for diagnosing device failures by learning measured sensor data is being actively conducted. This failure diagnosis study is divided into a classification model for predicting failure situations or types and a regression model for numerically predicting failure conditions. In the case of a classification model, it simply checks the presence or absence of a failure or defect (Class), whereas a regression model has a higher learning difficulty because it has to predict one value among countless numbers. So, the reason that regression modeling is more difficult is that there are many irregular situations in which it is difficult to determine one output from a similar input when predicting by matching input and output. Therefore, in this paper, we focus on input and output data with periodicity, analyze the input/output relationship, and secure regularity between input and output data by performing sliding window-based input data patterning. In order to apply the proposed method, in this study, current and temperature data with periodicity were collected from MMC(Modular Multilevel Converter) circuit system and learning was carried out using ANN. As a result of the experiment, it was confirmed that when a window of 2% or more of one cycle was applied, performance of 97% or more of fit could be secured.

• Journal of forensic and investigative science
• /
• v.1 no.2
• /
• pp.5-19
• /
• 2006

Data mining is an information extraction activity to discover hidden facts contained in databases. Using a combination of machine learning, statistical analysis, modeling techniques and database technology, data mining finds patterns and subtle relationships in data and infers rules that allow the prediction of future results. Typical applications include market segmentation, customer profiling, fraud detection, evaluation of retail promotions, and credit risk analysis. Law enforcement agencies deal with mass data to investigate the crime and its amount is increasing due to the development of processing the data by using computer. Now new challenge to discover knowledge in that data is confronted to us. It can be applied in criminal investigation to find offenders by analysis of complex and relational data structures and free texts using their criminal records or statement texts. This study was aimed to evaluate possibile application of data mining and its limitation in practical criminal investigation. Clustering of the criminal cases will be possible in habitual crimes such as fraud and burglary when using data mining to identify the crime pattern. Neural network modelling, one of tools in data mining, can be applied to differentiating suspect's photograph or handwriting with that of convict or criminal profiling. A case study of in practical insurance fraud showed that data mining was useful in organized crimes such as gang, terrorism and money laundering. But the products of data mining in criminal investigation should be cautious for evaluating because data mining just offer a clue instead of conclusion. The legal regulation is needed to control the abuse of law enforcement agencies and to protect personal privacy or human rights.

• Journal of the Korean Society of Food Science and Nutrition
• /
• v.38 no.10
• /
• pp.1362-1372
• /
• 2009

This study was carried out to broadly profile the dietary pattern of elderly who consumed a higher amount of sodium compared to lesser sodium consumers. Na index 14, an instant food frequency questionnaire (IFFQ) of 40 items characterizing a high amount of sodium and a food frequency questionnaire of 63 food items that were used in nationwide surveys were adopted for the assessment of dietary characteristics of the subjects. The Na index 14 consisted of food items such as kimchi stew, bean paste stew, cooked spinach dish, seaweed soup, fish stew, roasted anchovy, and seasoned vegetables. Also, the survey constructs included nutritional knowledge and anthropometric measurements of the subjects. Out a total of 135 participants, 58 elderly were determined as the high sodium intake group (HSIG) and the other 77 were the control group, via a formerly validated Na index and sodium-associated dietary habits scale (SDH). Dietary habits of "add salt or soy sauce to foods", "drink up the broth of soups and stews", and "brined fishes and vegetables daily" were among the most significant differences between HSIG and control group in SDH assessment. In addition, the HSIG were less likely to have regular meals, adequate amount of meals, and nutritionally balanced meals than the control, with the differences manifesting more in females than males. Additional findings included that the HSIG possessed a poorer nutritional knowledge and obtained much higher scores on the IFFQ. Taken together, the study urges the needs of nutrition education for the elderly who habitually salt their foods and maintain a less desirable dietary style.

• Journal of the Korea Convergence Society
• /
• v.11 no.9
• /
• pp.1-6
• /
• 2020

In this study, we propose an association inference model based on data bias optimization for road hazard detection. This is a mining model based on association analysis to collect user's personal characteristics and surrounding environment data and provide traffic accident prevention services. This creates transaction data composed of various context variables. Based on the generated information, a meaningful correlation of variables in each transaction is derived through correlation pattern analysis. Considering the bias of classified categorical data, pruning is performed with optimized support and reliability values. Based on the extracted high-level association rules, a risk detection model for personal characteristics and driving road conditions is provided to users. This enables traffic services that overcome the data bias problem and prevent potential road accidents by considering the association between data. In the performance evaluation, the proposed method is excellently evaluated as 0.778 in accuracy and 0.743 in the Kappa coefficient.

• The Journal of the Korea Contents Association
• /
• v.21 no.4
• /
• pp.658-667
• /
• 2021

Our study aimed to analyze and reinterpret, by an acoustic approach, the descriptions of the 18th century prosody and introduce the figurative pronunciation system, which is the International Phonetic Alphabet pioneer. Our methodology compares and analyzes grammars and documents on the transcription system and restructures the prosodic structure. It is certain that the 18th century grammarians widely accepted the prosody theories made by Arnauld & Lancelot of the seventeenth century. In particular, grammar scholars accepted the dichotomous classification of the accent structures as prosodic and oratorical accents. The prosodic accent has a relation to intonation, and the oratorical accent has as its key elements intonation and intensity. Regarding the temporal structure, the lengthening of the final syllable was observed systematically by grammarians of the 18th century. This time structure is similar to that of today. Therefore, we can conclude that the final elongation, an essential characteristic of the modern French accent, has already played an imbued role in 18th century prosody. Despite this, the 18th century grammarians did not assign it the status of accent, as it was a stereotype that matches accent with intonation.

• Information Systems Review
• /
• v.22 no.1
• /
• pp.59-72
• /
• 2020

Artificial Intelligence is establishing itself as a familiar tool from an intractable concept. In this trend, financial sector is also looking to improve the problem of existing system which includes Fraud Detection System (FDS). It is being difficult to detect sophisticated cyber financial fraud using original rule-based FDS. This is because diversification of payment environment and increasing number of electronic financial transactions has been emerged. In order to overcome present FDS, this paper suggests 3 types of artificial intelligence models, Generative Adversarial Network (GAN), Deep Neural Network (DNN), and Convolutional Neural Network (CNN). GAN proves how data imbalance problem can be developed while DNN and CNN show how abnormal financial trading patterns can be precisely detected. In conclusion, among the experiments on this paper, WGAN has the highest improvement effects on data imbalance problem. DNN model reflects more effects on fraud classification comparatively.

• Journal of Cadastre & Land InformatiX
• /
• v.47 no.2
• /
• pp.233-254
• /
• 2017

The purpose of this study is to analyze the main factors influencing the severity of traffic accidents and to visualize spatiotemporal characteristics of traffic accidents in Seoul. To do this, we collected the traffic accident data that occurred in Seoul for four years from 2012 to 2015, and classified as slight, serious, and death traffic accidents according to the severity of traffic accidents. The analysis of spatiotemporal characteristics of traffic accidents was performed by kernel density analysis, hotspot analysis, space time cube analysis, and Emerging HotSpot Analysis. The factors affecting the severity of traffic accidents were analyzed using decision tree model. The results show that traffic accidents in Seoul are more frequent in suburbs than in central areas. Especially, traffic accidents concentrated in some commercial and entertainment areas in Seocho and Gangnam, and the traffic accidents were more and more intense over time. In the case of death traffic accidents, there were statistically significant hotspot areas in Yeongdeungpo-gu, Guro-gu, Jongno-gu, Jung-gu and Seongbuk. However, hotspots of death traffic accidents by time zone resulted in different patterns. In terms of traffic accident severity, the type of accident is the most important factor. The type of the road, the type of the vehicle, the time of the traffic accident, and the type of the violation of the regulations were ranked in order of importance. Regarding decision rules that cause serious traffic accidents, in case of van or truck, there is a high probability that a serious traffic accident will occur at a place where the width of the road is wide and the vehicle speed is high. In case of bicycle, car, motorcycle or the others there is a high probability that a serious traffic accident will occur under the same circumstances in the dawn time.