• Journal of Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.113-127
• /
• 2016

Among the techniques for recommendation, collaborative filtering (CF) is commonly recognized to be the most effective for implementing recommender systems. Until now, CF has been popularly studied and adopted in both academic and real-world applications. The basic idea of CF is to create recommendation results by finding correlations between users of a recommendation system. CF system compares users based on how similar they are, and recommend products to users by using other like-minded people's results of evaluation for each product. Thus, it is very important to compute evaluation similarities among users in CF because the recommendation quality depends on it. Typical CF uses user's explicit numeric ratings of items (i.e. quantitative information) when computing the similarities among users in CF. In other words, user's numeric ratings have been a sole source of user preference information in traditional CF. However, user ratings are unable to fully reflect user's actual preferences from time to time. According to several studies, users may more actively accommodate recommendation of reliable others when purchasing goods. Thus, trust relationship can be regarded as the informative source for identifying user's preference with accuracy. Under this background, we propose a new hybrid recommender system that fuses CF and social network analysis (SNA). The proposed system adopts the recommendation algorithm that additionally reflect the result analyzed by SNA. In detail, our proposed system is based on conventional memory-based CF, but it is designed to use both user's numeric ratings and trust relationship information between users when calculating user similarities. For this, our system creates and uses not only user-item rating matrix, but also user-to-user trust network. As the methods for calculating user similarity between users, we proposed two alternatives - one is algorithm calculating the degree of similarity between users by utilizing in-degree and out-degree centrality, which are the indices representing the central location in the social network. We named these approaches as 'Trust CF - All' and 'Trust CF - Conditional'. The other alternative is the algorithm reflecting a neighbor's score higher when a target user trusts the neighbor directly or indirectly. The direct or indirect trust relationship can be identified by searching trust network of users. In this study, we call this approach 'Trust CF - Search'. To validate the applicability of the proposed system, we used experimental data provided by LibRec that crawled from the entire FilmTrust website. It consists of ratings of movies and trust relationship network indicating who to trust between users. The experimental system was implemented using Microsoft Visual Basic for Applications (VBA) and UCINET 6. To examine the effectiveness of the proposed system, we compared the performance of our proposed method with one of conventional CF system. The performances of recommender system were evaluated by using average MAE (mean absolute error). The analysis results confirmed that in case of applying without conditions the in-degree centrality index of trusted network of users(i.e. Trust CF - All), the accuracy (MAE = 0.565134) was lower than conventional CF (MAE = 0.564966). And, in case of applying the in-degree centrality index only to the users with the out-degree centrality above a certain threshold value(i.e. Trust CF - Conditional), the proposed system improved the accuracy a little (MAE = 0.564909) compared to traditional CF. However, the algorithm searching based on the trusted network of users (i.e. Trust CF - Search) was found to show the best performance (MAE = 0.564846). And the result from paired samples t-test presented that Trust CF - Search outperformed conventional CF with 10% statistical significance level. Our study sheds a light on the application of user's trust relationship network information for facilitating electronic commerce by recommending proper items to users.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.19-38
• /
• 2017

Collaborative filtering (CF) algorithm has been popularly used for implementing recommender systems. Until now, there have been many prior studies to improve the accuracy of CF. Among them, some recent studies adopt 'hybrid recommendation approach', which enhances the performance of conventional CF by using additional information. In this research, we propose a new hybrid recommender system which fuses CF and the results from the social network analysis on trust and distrust relationship networks among users to enhance prediction accuracy. The proposed algorithm of our study is based on memory-based CF. But, when calculating the similarity between users in CF, our proposed algorithm considers not only the correlation of the users' numeric rating patterns, but also the users' in-degree centrality values derived from trust and distrust relationship networks. In specific, it is designed to amplify the similarity between a target user and his or her neighbor when the neighbor has higher in-degree centrality in the trust relationship network. Also, it attenuates the similarity between a target user and his or her neighbor when the neighbor has higher in-degree centrality in the distrust relationship network. Our proposed algorithm considers four (4) types of user relationships - direct trust, indirect trust, direct distrust, and indirect distrust - in total. And, it uses four adjusting coefficients, which adjusts the level of amplification / attenuation for in-degree centrality values derived from direct / indirect trust and distrust relationship networks. To determine optimal adjusting coefficients, genetic algorithms (GA) has been adopted. Under this background, we named our proposed algorithm as SNACF-GA (Social Network Analysis - based CF using GA). To validate the performance of the SNACF-GA, we used a real-world data set which is called 'Extended Epinions dataset' provided by 'trustlet.org'. It is the data set contains user responses (rating scores and reviews) after purchasing specific items (e.g. car, movie, music, book) as well as trust / distrust relationship information indicating whom to trust or distrust between users. The experimental system was basically developed using Microsoft Visual Basic for Applications (VBA), but we also used UCINET 6 for calculating the in-degree centrality of trust / distrust relationship networks. In addition, we used Palisade Software's Evolver, which is a commercial software implements genetic algorithm. To examine the effectiveness of our proposed system more precisely, we adopted two comparison models. The first comparison model is conventional CF. It only uses users' explicit numeric ratings when calculating the similarities between users. That is, it does not consider trust / distrust relationship between users at all. The second comparison model is SNACF (Social Network Analysis - based CF). SNACF differs from the proposed algorithm SNACF-GA in that it considers only direct trust / distrust relationships. It also does not use GA optimization. The performances of the proposed algorithm and comparison models were evaluated by using average MAE (mean absolute error). Experimental result showed that the optimal adjusting coefficients for direct trust, indirect trust, direct distrust, indirect distrust were 0, 1.4287, 1.5, 0.4615 each. This implies that distrust relationships between users are more important than trust ones in recommender systems. From the perspective of recommendation accuracy, SNACF-GA (Avg. MAE = 0.111943), the proposed algorithm which reflects both direct and indirect trust / distrust relationships information, was found to greatly outperform a conventional CF (Avg. MAE = 0.112638). Also, the algorithm showed better recommendation accuracy than the SNACF (Avg. MAE = 0.112209). To confirm whether these differences are statistically significant or not, we applied paired samples t-test. The results from the paired samples t-test presented that the difference between SNACF-GA and conventional CF was statistical significant at the 1% significance level, and the difference between SNACF-GA and SNACF was statistical significant at the 5%. Our study found that the trust/distrust relationship can be important information for improving performance of recommendation algorithms. Especially, distrust relationship information was found to have a greater impact on the performance improvement of CF. This implies that we need to have more attention on distrust (negative) relationships rather than trust (positive) ones when tracking and managing social relationships between users.

• Journal of Intelligence and Information Systems
• /
• v.20 no.4
• /
• pp.107-120
• /
• 2014

This study proposes a novel recommender system using the structural hole analysis to reflect qualitative and emotional information in recommendation process. Although collaborative filtering (CF) is known as the most popular recommendation algorithm, it has some limitations including scalability and sparsity problems. The scalability problem arises when the volume of users and items become quite large. It means that CF cannot scale up due to large computation time for finding neighbors from the user-item matrix as the number of users and items increases in real-world e-commerce sites. Sparsity is a common problem of most recommender systems due to the fact that users generally evaluate only a small portion of the whole items. In addition, the cold-start problem is the special case of the sparsity problem when users or items newly added to the system with no ratings at all. When the user's preference evaluation data is sparse, two users or items are unlikely to have common ratings, and finally, CF will predict ratings using a very limited number of similar users. Moreover, it may produces biased recommendations because similarity weights may be estimated using only a small portion of rating data. In this study, we suggest a novel limitation of the conventional CF. The limitation is that CF does not consider qualitative and emotional information about users in the recommendation process because it only utilizes user's preference scores of the user-item matrix. To address this novel limitation, this study proposes cluster-indexing CF model with the structural hole analysis for recommendations. In general, the structural hole means a location which connects two separate actors without any redundant connections in the network. The actor who occupies the structural hole can easily access to non-redundant, various and fresh information. Therefore, the actor who occupies the structural hole may be a important person in the focal network and he or she may be the representative person in the focal subgroup in the network. Thus, his or her characteristics may represent the general characteristics of the users in the focal subgroup. In this sense, we can distinguish friends and strangers of the focal user utilizing the structural hole analysis. This study uses the structural hole analysis to select structural holes in subgroups as an initial seeds for a cluster analysis. First, we gather data about users' preference ratings for items and their social network information. For gathering research data, we develop a data collection system. Then, we perform structural hole analysis and find structural holes of social network. Next, we use these structural holes as cluster centroids for the clustering algorithm. Finally, this study makes recommendations using CF within user's cluster, and compare the recommendation performances of comparative models. For implementing experiments of the proposed model, we composite the experimental results from two experiments. The first experiment is the structural hole analysis. For the first one, this study employs a software package for the analysis of social network data - UCINET version 6. The second one is for performing modified clustering, and CF using the result of the cluster analysis. We develop an experimental system using VBA (Visual Basic for Application) of Microsoft Excel 2007 for the second one. This study designs to analyzing clustering based on a novel similarity measure - Pearson correlation between user preference rating vectors for the modified clustering experiment. In addition, this study uses 'all-but-one' approach for the CF experiment. In order to validate the effectiveness of our proposed model, we apply three comparative types of CF models to the same dataset. The experimental results show that the proposed model outperforms the other comparative models. In especial, the proposed model significantly performs better than two comparative modes with the cluster analysis from the statistical significance test. However, the difference between the proposed model and the naive model does not have statistical significance.

• Journal of Intelligence and Information Systems
• /
• v.21 no.2
• /
• pp.1-18
• /
• 2015

Collaborative filtering(CF) algorithm has been popularly used for recommender systems in both academic and practical applications. A general CF system compares users based on how similar they are, and creates recommendation results with the items favored by other people with similar tastes. Thus, it is very important for CF to measure the similarities between users because the recommendation quality depends on it. In most cases, users' explicit numeric ratings of items(i.e. quantitative information) have only been used to calculate the similarities between users in CF. However, several studies indicated that qualitative information such as user's reviews on the items may contribute to measure these similarities more accurately. Considering that a lot of people are likely to share their honest opinion on the items they purchased recently due to the advent of the Web 2.0, user's reviews can be regarded as the informative source for identifying user's preference with accuracy. Under this background, this study proposes a new hybrid recommender system that combines with users' review mining. Our proposed system is based on conventional memory-based CF, but it is designed to use both user's numeric ratings and his/her text reviews on the items when calculating similarities between users. In specific, our system creates not only user-item rating matrix, but also user-item review term matrix. Then, it calculates rating similarity and review similarity from each matrix, and calculates the final user-to-user similarity based on these two similarities(i.e. rating and review similarities). As the methods for calculating review similarity between users, we proposed two alternatives - one is to use the frequency of the commonly used terms, and the other one is to use the sum of the importance weights of the commonly used terms in users' review. In the case of the importance weights of terms, we proposed the use of average TF-IDF(Term Frequency - Inverse Document Frequency) weights. To validate the applicability of the proposed system, we applied it to the implementation of a recommender system for smartphone applications (hereafter, app). At present, over a million apps are offered in each app stores operated by Google and Apple. Due to this information overload, users have difficulty in selecting proper apps that they really want. Furthermore, app store operators like Google and Apple have cumulated huge amount of users' reviews on apps until now. Thus, we chose smartphone app stores as the application domain of our system. In order to collect the experimental data set, we built and operated a Web-based data collection system for about two weeks. As a result, we could obtain 1,246 valid responses(ratings and reviews) from 78 users. The experimental system was implemented using Microsoft Visual Basic for Applications(VBA) and SAS Text Miner. And, to avoid distortion due to human intervention, we did not adopt any refining works by human during the user's review mining process. To examine the effectiveness of the proposed system, we compared its performance to the performance of conventional CF system. The performances of recommender systems were evaluated by using average MAE(mean absolute error). The experimental results showed that our proposed system(MAE = 0.7867 ~ 0.7881) slightly outperformed a conventional CF system(MAE = 0.7939). Also, they showed that the calculation of review similarity between users based on the TF-IDF weights(MAE = 0.7867) leaded to better recommendation accuracy than the calculation based on the frequency of the commonly used terms in reviews(MAE = 0.7881). The results from paired samples t-test presented that our proposed system with review similarity calculation using the frequency of the commonly used terms outperformed conventional CF system with 10% statistical significance level. Our study sheds a light on the application of users' review information for facilitating electronic commerce by recommending proper items to users.

• Journal of Intelligence and Information Systems
• /
• v.19 no.3
• /
• pp.127-139
• /
• 2013

With an advent of recent knowledge-based society, the interest in intellectual property has increased. Firms have tired to result in productive outcomes through continuous innovative activity. Especially, ICT firms which lead high-tech industry have tried to manage intellectual property more systematically. Firm's interest in the patent has increased in order to manage the innovative activity and Knowledge property. The patent involves not only simple information but also important values as information of technology, management and right. Moreover, as the patent has the detailed contents regarding technology development activity, it is regarded as valuable data. The patent which reflects technology spread and research outcomes and business performances are closely interrelated as the patent is considered as a significant the level of firm's innovation. As the patent information which represents companies' intellectual capital is accumulated continuously, it has become possible to do quantitative analysis. The advantages of patent in the related industry information and it's standardize information can be easily obtained. Through the patent, the flow of knowledge can be determined. The patent information can analyze in various levels from patent to nation. The patent information is used to analyze technical status and the effects on performance. The patent which has a high frequency of citation refers to having high technological values. Analyzing the patent information contains both citation index analysis using the number of citation and network analysis using citation relationship. Network analysis can provide the information on the flows of knowledge and technological changes, and it can show future research direction. Studies using the patent citation analysis vary academically and practically. For the citation index research, studies to analyze influential big patent has been conducted, and for the network analysis research, studies to find out the flows of technology in a certain industry has been conducted. Social network analysis is applied not only in the sociology, but also in a field of management consulting and company's knowledge management. Research of how the company's network position has an impact on business performances has been conducted from various aspects in a field of network analysis. Social network analysis can be based on the visual forms. Network indicators are available through the quantitative analysis. Social network analysis is used when analyzing outcomes in terms of the position of network. Social network analysis focuses largely on centrality and structural holes. Centrality indicates that actors having central positions among other actors have an advantage to exert stronger influence for exchange relationship. Degree centrality, betweenness centrality and closeness centrality are used for centrality analysis. Structural holes refer to an empty place in social structure and are defined as efficiency and constraints. This study stresses and analyzes firms' network in terms of the patent and how network characteristics have an influence on business performances. For the purpose of doing this, seventy-four ICT companies listed in S&P500 are chosen for the sample. UCINET6 is used to analyze the network structural characteristics such as outdegree centrality, betweenness centrality and efficiency. Then, regression analysis test is conducted to find out how these network characteristics are related to business performance. It is found that each network index has significant impacts on net income, i.e. business performance. However, it is found that efficiency is negatively associated with business performance. As the efficiency increases, net income decreases and it has a negative impact on business performances. Furthermore, it is shown that betweenness centrality solely has statistically significance for the multiple regression analysis with three network indexes. The patent citation network analysis shows the flows of knowledge between firms, and it can be expected to contribute to company's management strategies by analyzing company's network structural positions.

• Journal of Intelligence and Information Systems
• /
• v.21 no.1
• /
• pp.99-118
• /
• 2015

Object-oriented programming languages have been widely selected for developing modern information systems. The use of concepts relating to object-oriented (OO, in short) programming has reduced efforts of reusing pre-existing codes, and the OO concepts have been proved to be a useful in interpreting system requirements. In line with this, we have witnessed that a modern conceptual modeling approach supports features of object-oriented programming. Unified Modeling Language or UML becomes one of de-facto standards for information system designers since the language provides a set of visual diagrams, comprehensive frameworks and flexible expressions. In a modeling process, UML users need to consider relationships between classes. Based on an explicit and clear representation of classes, the conceptual model from UML garners necessarily attributes and methods for guiding software engineers. Especially, identifying an association between a class of part and a class of whole is included in the standard grammar of UML. The representation of part-whole relationship is natural in a real world domain since many physical objects are perceived as part-whole relationship. In addition, even abstract concepts such as roles are easily identified by part-whole perception. It seems that a representation of part-whole in UML is reasonable and useful. However, it should be admitted that the use of UML is limited due to the lack of practical guidelines on how to identify a part-whole relationship and how to classify it into an aggregate- or a composite-association. Research efforts on developing the procedure knowledge is meaningful and timely in that misleading perception to part-whole relationship is hard to be filtered out in an initial conceptual modeling thus resulting in deterioration of system usability. The current method on identifying and classifying part-whole relationships is mainly counting on linguistic expression. This simple approach is rooted in the idea that a phrase of representing has-a constructs a par-whole perception between objects. If the relationship is strong, the association is classified as a composite association of part-whole relationship. In other cases, the relationship is an aggregate association. Admittedly, linguistic expressions contain clues for part-whole relationships; therefore, the approach is reasonable and cost-effective in general. Nevertheless, it does not cover concerns on accuracy and theoretical legitimacy. Research efforts on developing guidelines for part-whole identification and classification has not been accumulated sufficient achievements to solve this issue. The purpose of this study is to provide step-by-step guidelines for identifying and classifying part-whole relationships in the context of UML use. Based on the theoretical work on Meta-model Formalization, self-check forms that help conceptual modelers work on part-whole classes are developed. To evaluate the performance of suggested idea, an experiment approach was adopted. The findings show that UML users obtain better results with the guidelines based on Meta-model Formalization compared to a natural language classification scheme conventionally recommended by UML theorists. This study contributed to the stream of research effort about part-whole relationships by extending applicability of Meta-model Formalization. Compared to traditional approaches that target to establish criterion for evaluating a result of conceptual modeling, this study expands the scope to a process of modeling. Traditional theories on evaluation of part-whole relationship in the context of conceptual modeling aim to rule out incomplete or wrong representations. It is posed that qualification is still important; but, the lack of consideration on providing a practical alternative may reduce appropriateness of posterior inspection for modelers who want to reduce errors or misperceptions about part-whole identification and classification. The findings of this study can be further developed by introducing more comprehensive variables and real-world settings. In addition, it is highly recommended to replicate and extend the suggested idea of utilizing Meta-model formalization by creating different alternative forms of guidelines including plugins for integrated development environments.

• Journal of Intelligence and Information Systems
• /
• v.29 no.4
• /
• pp.271-286
• /
• 2023

The global COVID-19 pandemic has transformed face masks from situational accessories to indispensable items in daily life, prompting a shift in public perception and behavior. While the relaxation of mandatory mask-wearing regulations is underway, a significant number of individuals continue to embrace face masks, turning them into a form of personal expression and identity. This phenomenon has given rise to the Fashion Mask industry, characterized by unique designs and colors, experiencing rapid growth in the market. However, existing research on masks is predominantly focused on their efficacy in preventing infection or exploring attitudes during the pandemic, leaving a gap in understanding consumer preferences for mask design. We address this gap by investigating consumer perceptions and preferences for two prevalent mask designs-horizontal fold flat masks and vertical fold flat masks. Through a comprehensive approach involving surveys and eye-tracking experiments, we aim to unravel the subtle differences in how consumers perceive these designs. Our research questions focus on determining which design is more appealing and exploring the reasons behind any observed differences. The study's findings reveal a clear preference for vertical fold flat masks, which are not only preferred but also perceived as unique, sophisticated, three-dimensional, and lively. The eye-tracking analysis provides insights into the visual attention patterns associated with mask designs, highlighting the pivotal role of the fold line in influencing these patterns. This research contributes to the evolving understanding of masks as a fashion statement and provides valuable insights for manufacturers and marketers in the Fashion Mask industry. The results have implications beyond the pandemic, emphasizing the importance of design elements in sustaining consumer interest in face masks.

• Korean Journal of Radiology
• /
• v.24 no.8
• /
• pp.807-820
• /
• 2023

Objective: To assess whether computed tomography (CT) conversion across different scan parameters and manufacturers using a routable generative adversarial network (RouteGAN) can improve the accuracy and variability in quantifying interstitial lung disease (ILD) using a deep learning-based automated software. Materials and Methods: This study included patients with ILD who underwent thin-section CT. Unmatched CT images obtained using scanners from four manufacturers (vendors A-D), standard- or low-radiation doses, and sharp or medium kernels were classified into groups 1-7 according to acquisition conditions. CT images in groups 2-7 were converted into the target CT style (Group 1: vendor A, standard dose, and sharp kernel) using a RouteGAN. ILD was quantified on original and converted CT images using a deep learning-based software (Aview, Coreline Soft). The accuracy of quantification was analyzed using the dice similarity coefficient (DSC) and pixel-wise overlap accuracy metrics against manual quantification by a radiologist. Five radiologists evaluated quantification accuracy using a 10-point visual scoring system. Results: Three hundred and fifty CT slices from 150 patients (mean age: 67.6 ± 10.7 years; 56 females) were included. The overlap accuracies for quantifying total abnormalities in groups 2-7 improved after CT conversion (original vs. converted: 0.63 vs. 0.68 for DSC, 0.66 vs. 0.70 for pixel-wise recall, and 0.68 vs. 0.73 for pixel-wise precision; P < 0.002 for all). The DSCs of fibrosis score, honeycombing, and reticulation significantly increased after CT conversion (0.32 vs. 0.64, 0.19 vs. 0.47, and 0.23 vs. 0.54, P < 0.002 for all), whereas those of ground-glass opacity, consolidation, and emphysema did not change significantly or decreased slightly. The radiologists' scores were significantly higher (P < 0.001) and less variable on converted CT. Conclusion: CT conversion using a RouteGAN can improve the accuracy and variability of CT images obtained using different scan parameters and manufacturers in deep learning-based quantification of ILD.

• Journal of Intelligence and Information Systems
• /
• v.23 no.1
• /
• pp.47-67
• /
• 2017

Steel plate faults is one of important factors to affect the quality and price of the steel plates. So far many steelmakers generally have used visual inspection method that could be based on an inspector's intuition or experience. Specifically, the inspector checks the steel plate faults by looking the surface of the steel plates. However, the accuracy of this method is critically low that it can cause errors above 30% in judgment. Therefore, accurate steel plate faults diagnosis system has been continuously required in the industry. In order to meet the needs, this study proposed a new steel plate faults diagnosis system using Simultaneous MTS (S-MTS), which is an advanced Mahalanobis Taguchi System (MTS) algorithm, to classify various surface defects of the steel plates. MTS has generally been used to solve binary classification problems in various fields, but MTS was not used for multiclass classification due to its low accuracy. The reason is that only one mahalanobis space is established in the MTS. In contrast, S-MTS is suitable for multi-class classification. That is, S-MTS establishes individual mahalanobis space for each class. 'Simultaneous' implies comparing mahalanobis distances at the same time. The proposed steel plate faults diagnosis system was developed in four main stages. In the first stage, after various reference groups and related variables are defined, data of the steel plate faults is collected and used to establish the individual mahalanobis space per the reference groups and construct the full measurement scale. In the second stage, the mahalanobis distances of test groups is calculated based on the established mahalanobis spaces of the reference groups. Then, appropriateness of the spaces is verified by examining the separability of the mahalanobis diatances. In the third stage, orthogonal arrays and Signal-to-Noise (SN) ratio of dynamic type are applied for variable optimization. Also, Overall SN ratio gain is derived from the SN ratio and SN ratio gain. If the derived overall SN ratio gain is negative, it means that the variable should be removed. However, the variable with the positive gain may be considered as worth keeping. Finally, in the fourth stage, the measurement scale that is composed of selected useful variables is reconstructed. Next, an experimental test should be implemented to verify the ability of multi-class classification and thus the accuracy of the classification is acquired. If the accuracy is acceptable, this diagnosis system can be used for future applications. Also, this study compared the accuracy of the proposed steel plate faults diagnosis system with that of other popular classification algorithms including Decision Tree, Multi Perception Neural Network (MLPNN), Logistic Regression (LR), Support Vector Machine (SVM), Tree Bagger Random Forest, Grid Search (GS), Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). The steel plates faults dataset used in the study is taken from the University of California at Irvine (UCI) machine learning repository. As a result, the proposed steel plate faults diagnosis system based on S-MTS shows 90.79% of classification accuracy. The accuracy of the proposed diagnosis system is 6-27% higher than MLPNN, LR, GS, GA and PSO. Based on the fact that the accuracy of commercial systems is only about 75-80%, it means that the proposed system has enough classification performance to be applied in the industry. In addition, the proposed system can reduce the number of measurement sensors that are installed in the fields because of variable optimization process. These results show that the proposed system not only can have a good ability on the steel plate faults diagnosis but also reduce operation and maintenance cost. For our future work, it will be applied in the fields to validate actual effectiveness of the proposed system and plan to improve the accuracy based on the results.

• 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.