• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.157-173
• /
• 2021

With the development of information technology, the amount of available information increases daily. However, having access to so much information makes it difficult for users to easily find the information they seek. Users want a visualized system that reduces information retrieval and learning time, saving them from personally reading and judging all available information. As a result, recommendation systems are an increasingly important technologies that are essential to the business. Collaborative filtering is used in various fields with excellent performance because recommendations are made based on similar user interests and preferences. However, limitations do exist. Sparsity occurs when user-item preference information is insufficient, and is the main limitation of collaborative filtering. The evaluation value of the user item matrix may be distorted by the data depending on the popularity of the product, or there may be new users who have not yet evaluated the value. The lack of historical data to identify consumer preferences is referred to as data sparsity, and various methods have been studied to address these problems. However, most attempts to solve the sparsity problem are not optimal because they can only be applied when additional data such as users' personal information, social networks, or characteristics of items are included. Another problem is that real-world score data are mostly biased to high scores, resulting in severe imbalances. One cause of this imbalance distribution is the purchasing bias, in which only users with high product ratings purchase products, so those with low ratings are less likely to purchase products and thus do not leave negative product reviews. Due to these characteristics, unlike most users' actual preferences, reviews by users who purchase products are more likely to be positive. Therefore, the actual rating data is over-learned in many classes with high incidence due to its biased characteristics, distorting the market. Applying collaborative filtering to these imbalanced data leads to poor recommendation performance due to excessive learning of biased classes. Traditional oversampling techniques to address this problem are likely to cause overfitting because they repeat the same data, which acts as noise in learning, reducing recommendation performance. In addition, pre-processing methods for most existing data imbalance problems are designed and used for binary classes. Binary class imbalance techniques are difficult to apply to multi-class problems because they cannot model multi-class problems, such as objects at cross-class boundaries or objects overlapping multiple classes. To solve this problem, research has been conducted to convert and apply multi-class problems to binary class problems. However, simplification of multi-class problems can cause potential classification errors when combined with the results of classifiers learned from other sub-problems, resulting in loss of important information about relationships beyond the selected items. Therefore, it is necessary to develop more effective methods to address multi-class imbalance problems. We propose a collaborative filtering model using CGAN to generate realistic virtual data to populate the empty user-item matrix. Conditional vector y identify distributions for minority classes and generate data reflecting their characteristics. Collaborative filtering then maximizes the performance of the recommendation system via hyperparameter tuning. This process should improve the accuracy of the model by addressing the sparsity problem of collaborative filtering implementations while mitigating data imbalances arising from real data. Our model has superior recommendation performance over existing oversampling techniques and existing real-world data with data sparsity. SMOTE, Borderline SMOTE, SVM-SMOTE, ADASYN, and GAN were used as comparative models and we demonstrate the highest prediction accuracy on the RMSE and MAE evaluation scales. Through this study, oversampling based on deep learning will be able to further refine the performance of recommendation systems using actual data and be used to build business recommendation systems.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.4
• /
• pp.77-84
• /
• 2012

Face classification has wide applications in intelligent video surveillance, content retrieval, robot vision, and human-machine interface. Pose and expression changes, and arbitrary illumination are typical problems for face recognition. When the face is captured at a distance, the image quality is often degraded by blurring and noise corruption. This paper investigates the efficacy of multi-classifier decision level fusion for face classification based on the photon-counting linear discriminant analysis with two different cost functions: Euclidean distance and negative normalized correlation. Decision level fusion comprises three stages: cost normalization, cost validation, and fusion rules. First, the costs are normalized into the uniform range and then, candidate costs are selected during validation. Three fusion rules are employed: minimum, average, and majority-voting rules. In the experiments, unfocusing and motion blurs are rendered to simulate the effects of the long distance environments. It will be shown that the decision-level fusion scheme provides better results than the single classifier.

• Journal of Intelligence and Information Systems
• /
• v.21 no.1
• /
• pp.143-159
• /
• 2015

Predicting IT trends has been a long and important subject for information systems research. IT trend prediction makes it possible to acknowledge emerging eras of innovation and allocate budgets to prepare against rapidly changing technological trends. Towards the end of each year, various domestic and global organizations predict and announce IT trends for the following year. For example, Gartner Predicts 10 top IT trend during the next year, and these predictions affect IT and industry leaders and organization's basic assumptions about technology and the future of IT, but the accuracy of these reports are difficult to verify. Social media data can be useful tool to verify the accuracy. As social media services have gained in popularity, it is used in a variety of ways, from posting about personal daily life to keeping up to date with news and trends. In the recent years, rates of social media activity in Korea have reached unprecedented levels. Hundreds of millions of users now participate in online social networks and communicate with colleague and friends their opinions and thoughts. In particular, Twitter is currently the major micro blog service, it has an important function named 'tweets' which is to report their current thoughts and actions, comments on news and engage in discussions. For an analysis on IT trends, we chose Tweet data because not only it produces massive unstructured textual data in real time but also it serves as an influential channel for opinion leading on technology. Previous studies found that the tweet data provides useful information and detects the trend of society effectively, these studies also identifies that Twitter can track the issue faster than the other media, newspapers. Therefore, this study investigates how frequently the predicted IT trends for the following year announced by public organizations are mentioned on social network services like Twitter. IT trend predictions for 2013, announced near the end of 2012 from two domestic organizations, the National IT Industry Promotion Agency (NIPA) and the National Information Society Agency (NIA), were used as a basis for this research. The present study analyzes the Twitter data generated from Seoul (Korea) compared with the predictions of the two organizations to analyze the differences. Thus, Twitter data analysis requires various natural language processing techniques, including the removal of stop words, and noun extraction for processing various unrefined forms of unstructured data. To overcome these challenges, we used SAS IRS (Information Retrieval Studio) developed by SAS to capture the trend in real-time processing big stream datasets of Twitter. The system offers a framework for crawling, normalizing, analyzing, indexing and searching tweet data. As a result, we have crawled the entire Twitter sphere in Seoul area and obtained 21,589 tweets in 2013 to review how frequently the IT trend topics announced by the two organizations were mentioned by the people in Seoul. The results shows that most IT trend predicted by NIPA and NIA were all frequently mentioned in Twitter except some topics such as 'new types of security threat', 'green IT', 'next generation semiconductor' since these topics non generalized compound words so they can be mentioned in Twitter with other words. To answer whether the IT trend tweets from Korea is related to the following year's IT trends in real world, we compared Twitter's trending topics with those in Nara Market, Korea's online e-Procurement system which is a nationwide web-based procurement system, dealing with whole procurement process of all public organizations in Korea. The correlation analysis show that Tweet frequencies on IT trending topics predicted by NIPA and NIA are significantly correlated with frequencies on IT topics mentioned in project announcements by Nara market in 2012 and 2013. The main contribution of our research can be found in the following aspects: i) the IT topic predictions announced by NIPA and NIA can provide an effective guideline to IT professionals and researchers in Korea who are looking for verified IT topic trends in the following topic, ii) researchers can use Twitter to get some useful ideas to detect and predict dynamic trends of technological and social issues.

• Journal of Intelligence and Information Systems
• /
• v.21 no.1
• /
• pp.65-82
• /
• 2015

Recently, there has been a surge of interest in finding core issues and analyzing emerging trends for the future. This represents efforts to devise national strategies and policies based on the selection of promising areas that can create economic and social added value. The existing studies, including those dedicated to the discovery of future promising fields, have mostly been dependent on qualitative research methods such as literature review and expert judgement. Deriving results from large amounts of information under this approach is both costly and time consuming. Efforts have been made to make up for the weaknesses of the conventional qualitative analysis approach designed to select key promising areas through discovery of future core issues and emerging trend analysis in various areas of academic research. There needs to be a paradigm shift in toward implementing qualitative research methods along with quantitative research methods like text mining in a mutually complementary manner. The change is to ensure objective and practical emerging trend analysis results based on large amounts of data. However, even such studies have had shortcoming related to their dependence on simple keywords for analysis, which makes it difficult to derive meaning from data. Besides, no study has been carried out so far to develop core issues and analyze emerging trends in special domains like the aviation industry. The change used to implement recent studies is being witnessed in various areas such as the steel industry, the information and communications technology industry, the construction industry in architectural engineering and so on. This study focused on retrieving aviation-related core issues and emerging trends from overall research papers pertaining to aviation through text mining, which is one of the big data analysis techniques. In this manner, the promising future areas for the air transport industry are selected based on objective data from aviation-related research papers. In order to compensate for the difficulties in grasping the meaning of single words in emerging trend analysis at keyword levels, this study will adopt topic analysis, which is a technique used to find out general themes latent in text document sets. The analysis will lead to the extraction of topics, which represent keyword sets, thereby discovering core issues and conducting emerging trend analysis. Based on the issues, it identified aviation-related research trends and selected the promising areas for the future. Research on core issue retrieval and emerging trend analysis for the aviation industry based on big data analysis is still in its incipient stages. So, the analysis targets for this study are restricted to data from aviation-related research papers. However, it has significance in that it prepared a quantitative analysis model for continuously monitoring the derived core issues and presenting directions regarding the areas with good prospects for the future. In the future, the scope is slated to expand to cover relevant domestic or international news articles and bidding information as well, thus increasing the reliability of analysis results. On the basis of the topic analysis results, core issues for the aviation industry will be determined. Then, emerging trend analysis for the issues will be implemented by year in order to identify the changes they undergo in time series. Through these procedures, this study aims to prepare a system for developing key promising areas for the future aviation industry as well as for ensuring rapid response. Additionally, the promising areas selected based on the aforementioned results and the analysis of pertinent policy research reports will be compared with the areas in which the actual government investments are made. The results from this comparative analysis are expected to make useful reference materials for future policy development and budget establishment.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.139-161
• /
• 2019

The utilization of the e-commerce market has become a common life style in today. It has become important part to know where and how to make reasonable purchases of good quality products for customers. This change in purchase psychology tends to make it difficult for customers to make purchasing decisions in vast amounts of information. In this case, the recommendation system has the effect of reducing the cost of information retrieval and improving the satisfaction by analyzing the purchasing behavior of the customer. Amazon and Netflix are considered to be the well-known examples of sales marketing using the recommendation system. In the case of Amazon, 60% of the recommendation is made by purchasing goods, and 35% of the sales increase was achieved. Netflix, on the other hand, found that 75% of movie recommendations were made using services. This personalization technique is considered to be one of the key strategies for one-to-one marketing that can be useful in online markets where salespeople do not exist. Recommendation techniques that are mainly used in recommendation systems today include collaborative filtering and content-based filtering. Furthermore, hybrid techniques and association rules that use these techniques in combination are also being used in various fields. Of these, collaborative filtering recommendation techniques are the most popular today. Collaborative filtering is a method of recommending products preferred by neighbors who have similar preferences or purchasing behavior, based on the assumption that users who have exhibited similar tendencies in purchasing or evaluating products in the past will have a similar tendency to other products. However, most of the existed systems are recommended only within the same category of products such as books and movies. This is because the recommendation system estimates the purchase satisfaction about new item which have never been bought yet using customer's purchase rating points of a similar commodity based on the transaction data. In addition, there is a problem about the reliability of purchase ratings used in the recommendation system. Reliability of customer purchase ratings is causing serious problems. In particular, 'Compensatory Review' refers to the intentional manipulation of a customer purchase rating by a company intervention. In fact, Amazon has been hard-pressed for these "compassionate reviews" since 2016 and has worked hard to reduce false information and increase credibility. The survey showed that the average rating for products with 'Compensated Review' was higher than those without 'Compensation Review'. And it turns out that 'Compensatory Review' is about 12 times less likely to give the lowest rating, and about 4 times less likely to leave a critical opinion. As such, customer purchase ratings are full of various noises. This problem is directly related to the performance of recommendation systems aimed at maximizing profits by attracting highly satisfied customers in most e-commerce transactions. In this study, we propose the possibility of using new indicators that can objectively substitute existing customer 's purchase ratings by using RFM multi-dimensional analysis technique to solve a series of problems. RFM multi-dimensional analysis technique is the most widely used analytical method in customer relationship management marketing(CRM), and is a data analysis method for selecting customers who are likely to purchase goods. As a result of verifying the actual purchase history data using the relevant index, the accuracy was as high as about 55%. This is a result of recommending a total of 4,386 different types of products that have never been bought before, thus the verification result means relatively high accuracy and utilization value. And this study suggests the possibility of general recommendation system that can be applied to various offline product data. If additional data is acquired in the future, the accuracy of the proposed recommendation system can be improved.

• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.47-60
• /
• 2012

Video data comes in the form of the unstructured and the complex structure. As the importance of efficient management and retrieval for video data increases, studies on the video parsing based on the visual features contained in the video contents are researched to reconstruct video data as the meaningful structure. The early studies on video parsing are focused on splitting video data into shots, but detecting the shot boundary defined with the physical boundary does not cosider the semantic association of video data. Recently, studies on structuralizing video shots having the semantic association to the video scene defined with the semantic boundary by utilizing clustering methods are actively progressed. Previous studies on detecting the video scene try to detect video scenes by utilizing clustering algorithms based on the similarity measure between video shots mainly depended on color features. However, the correct identification of a video shot or scene and the detection of the gradual transitions such as dissolve, fade and wipe are difficult because color features of video data contain a noise and are abruptly changed due to the intervention of an unexpected object. In this paper, to solve these problems, we propose the Scene Detector by using Color histogram, corner Edge and Object color histogram (SDCEO) that clusters similar shots organizing same event based on visual features including the color histogram, the corner edge and the object color histogram to detect video scenes. The SDCEO is worthy of notice in a sense that it uses the edge feature with the color feature, and as a result, it effectively detects the gradual transitions as well as the abrupt transitions. The SDCEO consists of the Shot Bound Identifier and the Video Scene Detector. The Shot Bound Identifier is comprised of the Color Histogram Analysis step and the Corner Edge Analysis step. In the Color Histogram Analysis step, SDCEO uses the color histogram feature to organizing shot boundaries. The color histogram, recording the percentage of each quantized color among all pixels in a frame, are chosen for their good performance, as also reported in other work of content-based image and video analysis. To organize shot boundaries, SDCEO joins associated sequential frames into shot boundaries by measuring the similarity of the color histogram between frames. In the Corner Edge Analysis step, SDCEO identifies the final shot boundaries by using the corner edge feature. SDCEO detect associated shot boundaries comparing the corner edge feature between the last frame of previous shot boundary and the first frame of next shot boundary. In the Key-frame Extraction step, SDCEO compares each frame with all frames and measures the similarity by using histogram euclidean distance, and then select the frame the most similar with all frames contained in same shot boundary as the key-frame. Video Scene Detector clusters associated shots organizing same event by utilizing the hierarchical agglomerative clustering method based on the visual features including the color histogram and the object color histogram. After detecting video scenes, SDCEO organizes final video scene by repetitive clustering until the simiarity distance between shot boundaries less than the threshold h. In this paper, we construct the prototype of SDCEO and experiments are carried out with the baseline data that are manually constructed, and the experimental results that the precision of shot boundary detection is 93.3% and the precision of video scene detection is 83.3% are satisfactory.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.