• The Journal of Korean Association of Computer Education
• /
• v.13 no.1
• /
• pp.45-52
• /
• 2010

With rapid increase in the number of web documents, the problem of information overload in Internet search is growing seriously. In order to improve web search results, previous research studies employed user queries/preferred words and the number of links in the web documents. In this study, performance of the search results exploiting these two criteria is examined and other preference criteria for web documents are analyzed. Experimental results show that personalized web search results employing queries and preferred words yield up to 1.7 times better performance over the current search engine and that the search results using the number of links gives up to 1.3 times better performance. Although it is found that the first of the user's preference criteria for web documents is the contents of the document, readability and images in the document are also given a large weight. Therefore, performance of web search personalization algorithms will be greatly improved if they incorporate objective data reflecting each user's characteristics in addition to the number of queries and preferred words.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.7
• /
• pp.441-447
• /
• 2017

In order to provide personalized services on the Web and for mobile services, it is necessary to collect and analyze information processed by users. Typically, information collected by users is managed online. Using information collected online may be sufficient to provide personalized service. However, in terms of O2O services, which are currently mixed with online and offline services, user information from the offline service can also be an important part of personalized service. Therefore, this study suggests an architecture to collect offline user information to provide more precise personalization services. The collection architecture includes Node Analyzer, Distance Checker, Holding Time Checker, and Cross Analyzer as core elements. This study also offers proposals for processing algorithms of key components that make up the proposed architecture. A case study collects user information of interest based on BLE in order to verify the proposed architecture and algorithms.

• Journal of Advanced Navigation Technology
• /
• v.18 no.6
• /
• pp.588-593
• /
• 2014

Recently, the smart devices has been possessed by a large majority of the adult, and offered various personalization services. This paper proposed the lightweight Intelligent Image Security System that notice the existence of any intruder in real time at the place of requiring the security by using smart device. The proposed image security system recognized whether or not intruder exists using the difference frame on the basis of Integral Image and the dynamic background updating algorithms. The intrusion notification is achieved by using the GCM push server that send massages in the application unit of user mobile device, and the SMTP mail server which is use of e-mail standard protocol. In case of the occurrence of intruder, GCM push server send an push-massage by the private mobile device, and SMTP mail server send the intruder's photograph and intrusion time. By the convergence of the various image processing algorithms and the performance of smart device, The proposed image security system can be applied to the various Intelligent Image Security field.

• Journal of Korea Multimedia Society
• /
• v.10 no.9
• /
• pp.1106-1116
• /
• 2007

Recently e-commerce trade on the web has grown rapidly in scale and complexity, just as web site designs and web servers have become more complicated. In view of these complexities, it is obviously difficult to analyse web user's data since they web users employ so many different web paths. The existing association rule investigation algorithms identify all items with a high correlation. However even though users often only want to find items in which they have interest, it is still difficult to find the rules they want out of all of the many association rules found by existing algorithms. In this paper, we propose a system linking each node with the sequence association rule, linking all routes after finding a path corresponding to a user with the association rule-one of the data mining techniques which identify user patterns in web user paths. The suggested system helps us construct individualized or customer-subdivided sites using the sequence association rule in order to harmonize the paths of web users with user characters.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.8
• /
• pp.39-48
• /
• 2023

This study was conducted to provide a solution to the problem of sizing errors occurring in online purchases due to discrepancies and non-standardization in clothing sizes. This paper discusses an implementation approach for a machine learning-based recommender system capable of providing personalized sizes to online consumers. We trained multiple validated collaborative filtering algorithms including Non-Negative Matrix Factorization (NMF), Singular Value Decomposition (SVD), k-Nearest Neighbors (KNN), and Co-Clustering using purchasing data derived from online commerce and compared their performance. As a result of the study, we were able to confirm that the NMF algorithm showed superior performance compared to other algorithms. Despite the characteristic of purchase data that includes multiple buyers using the same account, the proposed model demonstrated sufficient accuracy. The findings of this study are expected to contribute to reducing the return rate due to sizing errors and improving the customer experience on e-commerce platforms.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2008.10a
• /
• pp.410-414
• /
• 2008

Providing recommendations is acknowledged as one of important features of a business-to-consumer online storefront. Though there have been many studies on algorithms and operational procedures of personalized recommender systems, there is still a lack of empirical evidence demonstrating relationships between social presence and two important outcome variables of recommender systems: reuse intention and trust. To test the existence of a causal link between social presence and reuse intention, and mediating role of trust between these two variables, this study performed experiments varying level of social presence while providing personalized recommendations to users based on their explicit preferences. This study also compared these effects in two different product contexts: hedonic and utilitarian product. The results show that the provision of higher social presence increases both the reuse intention and trust of the recommender systems. In addition, the influence of social presence on reuse intention in the setting of recommending utilitarian products is less than that in the setting of recommending hedonic products.

• Journal of the Korean BIBLIA Society for library and Information Science
• /
• v.28 no.4
• /
• pp.501-518
• /
• 2017

Reducing the time it takes for researchers to acquire knowledge and introduce them into research activities can be regarded as an indispensable factor in improving the productivity of research. The purpose of this research is to cluster the information usage patterns of KOSEN users and to suggest optimization method of personalized recommendation service algorithm for grouped users. Based on user research activities and usage information, after identifying appropriate services and contents, we applied a Spark based big data analysis technology to derive a personal recommendation algorithm. Individual recommendation algorithms can save time to search for user information and can help to find appropriate information.

• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.131-141
• /
• 2012

In this paper, we propose a new method for collaborative filtering (CF)-based recommender systems. Traditional CF-based recommendation algorithms have applied constant settings such as a reference group (neighborhood) size and a significance level to all users. In this paper we develop a new method that identifies optimal personalized settings for each user and applies them to generating recommendations for individual users. Personalized parameters are identified through iterative simulations with 'training' and 'verification' datasets. The method is compared with traditional 'constant settings' methods using Netflix data. The results show that the new method outperforms traditional, ordinary CF. Implications and future research directions are also discussed.

• The KIPS Transactions:PartB
• /
• v.19B no.4
• /
• pp.231-236
• /
• 2012

According to consumer behavior theory, human propensity can be divided into two regulatory focus types: promotion and prevention. These two types have much influence on the consumer's decision in many diverse areas. In this research, we apply regulatory focus theory to personalized recommendation to minimize the cold start problem and to improve the performance of recommendation algorithms. To achieve this goal, we extract the consumer behavior variables and information exploration activity index from web shopping logs. We then use them for classifying regulatory focus of the consumer. This research has the contribution to show the possibility of systematization of consumer behavior theory as an interdisciplinary research tool of social science and information technology. Based on this attempt, we will extend the research to IT services adapting theories on other areas.

• Journal of the Korea Convergence Society
• /
• v.8 no.9
• /
• pp.285-292
• /
• 2017

This paper describes a recursive least-squares based convergent algorithm for driving characteristic classification for personalized autonomous driving. Recently, various researches on autonomous driving technology have been conducted for level 4 fully autonomous driving. In order for commercialization of the autonomous vehicle, personalized autonomous driving is required to minimize passenger's insecureness to the autonomous vehicle. To address this problem. this study proposes mathematical model that represents driving characteristics and recursive least-squares based algorithm that can estimate the defined characteristics. The actual data of two drivers has been used to derive driving characteristics and the hypothesis testing method has been used to classify two drivers. It is shown that the proposed algorithms can derive driving characteristics and classify two drivers reasonably.