Pub Date : 2023-09-30DOI: 10.9717/kmms.2023.26.9.1115
Tae-Hee Park
{"title":"U-Net Based Single Image Deraining Using the Wavelet Residue Channel Fusion Strategy","authors":"Tae-Hee Park","doi":"10.9717/kmms.2023.26.9.1115","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.9.1115","url":null,"abstract":"","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135040395","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-09-30DOI: 10.9717/kmms.2023.26.9.1149
Minhee Park, Mahnwoo Kwon, Mikyung Hwang, Hyeonseong Kim
Today, with the advent of various communication media, the importance of the visual component of the book cover is increasing as the purchasing behavior pattern changes from paper books to e-books and from offline to online and mobile channels. In particular, social and environmental changes such as COVID-19 caused consumers
{"title":"Analysis of Visual Attention on Korean Essay Book Covers Using Eye-tracking Devices","authors":"Minhee Park, Mahnwoo Kwon, Mikyung Hwang, Hyeonseong Kim","doi":"10.9717/kmms.2023.26.9.1149","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.9.1149","url":null,"abstract":"Today, with the advent of various communication media, the importance of the visual component of the book cover is increasing as the purchasing behavior pattern changes from paper books to e-books and from offline to online and mobile channels. In particular, social and environmental changes such as COVID-19 caused consumers","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135040365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.911
Joon-Min Lee, Kyeong-Tae Kim, Jae-Young Choi
Air pollution, particularly fine dust, poses a significant threat to public health and necessitates accurate prediction models for effective mitigation strategies. In this paper, we propose a so-called attention-based ConvLSTM-DNN networks for fine dust concentration prediction. Our proposed model integrates the feature extraction capabilities of a 2D Convolutional Neural Network (CNN) with the long-term memory retention of an LSTM, capturing spatial and temporal dependencies in the input data. We apply an attention mechanism to enhance the model
{"title":"Attention Based-ConvLSTM-DNN Networks for Fine Dust Concentration Prediction","authors":"Joon-Min Lee, Kyeong-Tae Kim, Jae-Young Choi","doi":"10.9717/kmms.2023.26.8.911","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.911","url":null,"abstract":"Air pollution, particularly fine dust, poses a significant threat to public health and necessitates accurate prediction models for effective mitigation strategies. In this paper, we propose a so-called attention-based ConvLSTM-DNN networks for fine dust concentration prediction. Our proposed model integrates the feature extraction capabilities of a 2D Convolutional Neural Network (CNN) with the long-term memory retention of an LSTM, capturing spatial and temporal dependencies in the input data. We apply an attention mechanism to enhance the model","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.995
Yejin Shin, Min-Jin Lee, Helen Hong, Sung-Il Hwang
In this paper, we propose a prostate cancer aggressiveness prediction model using self-supervised learning based on SimCLR with multi-parametric MR images. Self-supervised learning model is initially trained on the STL10 dataset, and then fine-tuned on the ProstateX dataset, which is similar to the downstream task dataset. To predict prostate cancer aggressiveness, downstream tasks are performed using each sequence of images from the multi-parametric MR dataset. The predicted results are combined using either majority voting or average voting for ensembling. Experimental results demonstrate that the self-supervised learning model fine-turned with similar images improves the performance by an average of 4.56% in accuracy, 20.69% in sensitivity, and 12.02% in negative predictive value. The ensemble method using majority voting with the self-supervised learning model fine-turned on similar images from the multi-parametric MR dataset yields the best performance in terms of accuracy (72.58%), balance accuracy (72.16%), and sensitivity (67.86%).
{"title":"Improvement of Prostate Cancer Aggressiveness Prediction Performance Using a Self-Supervised Learning Model Fine-Turned on Similar Medical Images from Multi-Parametric MR Images","authors":"Yejin Shin, Min-Jin Lee, Helen Hong, Sung-Il Hwang","doi":"10.9717/kmms.2023.26.8.995","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.995","url":null,"abstract":"In this paper, we propose a prostate cancer aggressiveness prediction model using self-supervised learning based on SimCLR with multi-parametric MR images. Self-supervised learning model is initially trained on the STL10 dataset, and then fine-tuned on the ProstateX dataset, which is similar to the downstream task dataset. To predict prostate cancer aggressiveness, downstream tasks are performed using each sequence of images from the multi-parametric MR dataset. The predicted results are combined using either majority voting or average voting for ensembling. Experimental results demonstrate that the self-supervised learning model fine-turned with similar images improves the performance by an average of 4.56% in accuracy, 20.69% in sensitivity, and 12.02% in negative predictive value. The ensemble method using majority voting with the self-supervised learning model fine-turned on similar images from the multi-parametric MR dataset yields the best performance in terms of accuracy (72.58%), balance accuracy (72.16%), and sensitivity (67.86%).","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.956
Dong-Ho Shin, Hyung-Gyu Lim, Myoung-Nam Kim, Ki-Woong Seong
This study introduces a novel piezoelectric transducer for bone conduction implants that combines piezoelectric elements with a rhombus mechanism to enhance compatibility with magnetic resonance environments. To derive the optimal design of the rhombus structure, various parameters were investigated using theoretical analysis and finite element analysis. A theoretical model of the rhombus structure was employed to identify the parameters affecting displacement amplification magnitude. Based on this, a parametric analysis was performed to calculate the displacement amplification ratio according to these parameters. The results showed that as the beam thickness and width increased, the amplification ratio reduced, while with an increased length, the amplification ratio was increased. Therefore, the optimal rhombus structure for the transducer featured beam dimensions of 0.15 mm thickness, 2 mm width, 3.5 mm length, and 5.5° inclination. This configuration amplified piezoelectric element displacement by a factor of 7.02. The amplification ratio remained constant as long as the mass applied to the rhombus frame to control mechanical resonance did not exceed the blocking force of the piezoelectric element. When a mass of 0.3 g was applied to the frame, mechanical resonance occurred at a frequency of 2 kHz, making it suitable as a transducer for bone conduction implants.
{"title":"Design of Bone Conduction Implants Piezoelectric Transducer Based on Rhombus Mechanism for Magnetic Resonance Compatibility Improvement","authors":"Dong-Ho Shin, Hyung-Gyu Lim, Myoung-Nam Kim, Ki-Woong Seong","doi":"10.9717/kmms.2023.26.8.956","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.956","url":null,"abstract":"This study introduces a novel piezoelectric transducer for bone conduction implants that combines piezoelectric elements with a rhombus mechanism to enhance compatibility with magnetic resonance environments. To derive the optimal design of the rhombus structure, various parameters were investigated using theoretical analysis and finite element analysis. A theoretical model of the rhombus structure was employed to identify the parameters affecting displacement amplification magnitude. Based on this, a parametric analysis was performed to calculate the displacement amplification ratio according to these parameters. The results showed that as the beam thickness and width increased, the amplification ratio reduced, while with an increased length, the amplification ratio was increased. Therefore, the optimal rhombus structure for the transducer featured beam dimensions of 0.15 mm thickness, 2 mm width, 3.5 mm length, and 5.5° inclination. This configuration amplified piezoelectric element displacement by a factor of 7.02. The amplification ratio remained constant as long as the mass applied to the rhombus frame to control mechanical resonance did not exceed the blocking force of the piezoelectric element. When a mass of 0.3 g was applied to the frame, mechanical resonance occurred at a frequency of 2 kHz, making it suitable as a transducer for bone conduction implants.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.1003
Joong-Hyun Park
In the past, research related to corporate bankruptcy has primarily conducted empirical analyses through bankruptcy prediction models using financial ratios. However, with the advancement of ICT technology, there has been a growing trend in applying artificial intelligence. In this study, both traditional corporate bankruptcy prediction methodologies and machine learning and deep learning methodologies from the field of deep learning were applied to present the results of corporate bankruptcy prediction models and their predictive power. The dataset used included corporate characteristics, including financial ratios and non-financial information, as well as macroeconomic indicators to account for economic conditions. Five models, SVM, RF, DNN, CNN, and LSTM, were designated, and the model reliability and prediction accuracy for each model were analyzed. The LSTM model demonstrated superior performance and the highest prediction accuracy among the models. When comparing different approaches using only financial ratios (Set 1), using financial ratios and corporate characteristics together (Set 2), and incorporating financial ratios, corporate characteristics, and macroeconomic indicators (Set 3), which included all of these factors, consistently exhibited the highest model reliability and prediction accuracy.
{"title":"Research on Corporate Bankruptcy Prediction Analysis Based on Financial and Non-Financial Information Using Deep Learning","authors":"Joong-Hyun Park","doi":"10.9717/kmms.2023.26.8.1003","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.1003","url":null,"abstract":"In the past, research related to corporate bankruptcy has primarily conducted empirical analyses through bankruptcy prediction models using financial ratios. However, with the advancement of ICT technology, there has been a growing trend in applying artificial intelligence. In this study, both traditional corporate bankruptcy prediction methodologies and machine learning and deep learning methodologies from the field of deep learning were applied to present the results of corporate bankruptcy prediction models and their predictive power. The dataset used included corporate characteristics, including financial ratios and non-financial information, as well as macroeconomic indicators to account for economic conditions. Five models, SVM, RF, DNN, CNN, and LSTM, were designated, and the model reliability and prediction accuracy for each model were analyzed. The LSTM model demonstrated superior performance and the highest prediction accuracy among the models. When comparing different approaches using only financial ratios (Set 1), using financial ratios and corporate characteristics together (Set 2), and incorporating financial ratios, corporate characteristics, and macroeconomic indicators (Set 3), which included all of these factors, consistently exhibited the highest model reliability and prediction accuracy.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.1075
Je-Kyung Lee, Jeoung-Gi Kim, Jeong-In Ahn, Ji-Yeon Lim, Kyung-Ae Cha
In this paper, we propose a method to reduce the amount of manual work in webtoon creation and utilize creative contents derived from AI learning through deep learning-based technology that generates background images of various styles. To achieve this goal, we train CartoonGAN and AnimeGAN models that are specialized in creating images in the style of webtoons and animations, and create background images that can be used for webtoons. Recently, various Generative Adversarial Network (GAN) models have been actively used to create digital content, but cartoon-style images should be created with simplified textures and sharp outlines. In addition, when converting a real image into a cartoon style, it is necessary to create a simple and abstract image while maintaining the content expressed by the image. We build training data suitable for the production of these webtoon-style images, and analyze whether the images generated by the two GAN models can be used for webtoon production, and seek ways to utilize generative AI.
{"title":"Generation and Analysis of Webtoon Background Images Using GAN","authors":"Je-Kyung Lee, Jeoung-Gi Kim, Jeong-In Ahn, Ji-Yeon Lim, Kyung-Ae Cha","doi":"10.9717/kmms.2023.26.8.1075","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.1075","url":null,"abstract":"In this paper, we propose a method to reduce the amount of manual work in webtoon creation and utilize creative contents derived from AI learning through deep learning-based technology that generates background images of various styles. To achieve this goal, we train CartoonGAN and AnimeGAN models that are specialized in creating images in the style of webtoons and animations, and create background images that can be used for webtoons. Recently, various Generative Adversarial Network (GAN) models have been actively used to create digital content, but cartoon-style images should be created with simplified textures and sharp outlines. In addition, when converting a real image into a cartoon style, it is necessary to create a simple and abstract image while maintaining the content expressed by the image. We build training data suitable for the production of these webtoon-style images, and analyze whether the images generated by the two GAN models can be used for webtoon production, and seek ways to utilize generative AI.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991687","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.1064
Nam-Hee Kim, Soon-Bum Lim
Hearing-impaired people should be able to enjoy virtual reality content without discrimination. The development of metaverse has led to the success of virtual environment contents, including virtual concerts, but services for the deaf are still insufficient in this environment. In this paper, we present an auxiliary solution for hearing-impaired people to enjoy metaverse concerts and made video test samples for usability evaluation. To demonstrate that the solutions presented in this study help hearing-impaired people enjoy metaverse concerts, we made video test materials that provide secondary functions such as motion captions and haptic vibration. Test materials that provide basic subtitles were compared with test materials that applied motion captions, haptic vibrations, and both. To this end, an effectiveness evaluation and satisfaction evaluation using the Likert scale were conducted, and then a subjective satisfaction-oriented interview was conducted. Providing motion captions and haptic vibration together resulted in a relatively high usability evaluation effectiveness score, and specific requirements, problems, and improvements were identified through user interviews. If actual programs are developed and distributed based on this study, the quality of cultural life can be improved in the virtual environment of the hearing-impaired people.
{"title":"An Analysis of Usability on Secondary Functions for the Hearing Impaired to Enjoy Metaverse Contents － Focusing on popular music concerts －","authors":"Nam-Hee Kim, Soon-Bum Lim","doi":"10.9717/kmms.2023.26.8.1064","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.1064","url":null,"abstract":"Hearing-impaired people should be able to enjoy virtual reality content without discrimination. The development of metaverse has led to the success of virtual environment contents, including virtual concerts, but services for the deaf are still insufficient in this environment. In this paper, we present an auxiliary solution for hearing-impaired people to enjoy metaverse concerts and made video test samples for usability evaluation. To demonstrate that the solutions presented in this study help hearing-impaired people enjoy metaverse concerts, we made video test materials that provide secondary functions such as motion captions and haptic vibration. Test materials that provide basic subtitles were compared with test materials that applied motion captions, haptic vibrations, and both. To this end, an effectiveness evaluation and satisfaction evaluation using the Likert scale were conducted, and then a subjective satisfaction-oriented interview was conducted. Providing motion captions and haptic vibration together resulted in a relatively high usability evaluation effectiveness score, and specific requirements, problems, and improvements were identified through user interviews. If actual programs are developed and distributed based on this study, the quality of cultural life can be improved in the virtual environment of the hearing-impaired people.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.925
Oh-Seol Kwon
In this paper, a low-resolution object detection algorithm was proposed based on X-transfer learning on GAN model. The proposed method is effective in improving detection of micro objects by optimizing with GAN network for super-resolution and an object recognition network. In addition, the proposed X-transfer learning technique alternately uses transfer learning and curriculum learning to overcome the lack of training data. This method can improve the accuracy, robustness, and localization performance of object recognition based on rich visual information on entire network. The proposed model was evaluated with remote sensing data sets. It was confirmed that the proposed method is more accurate than existing methods in terms of mAP@0.5 and F1 scores.
{"title":"Deep Learning Network Algorithm Based on X-transfer Learning for Micro Object Detection","authors":"Oh-Seol Kwon","doi":"10.9717/kmms.2023.26.8.925","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.925","url":null,"abstract":"In this paper, a low-resolution object detection algorithm was proposed based on X-transfer learning on GAN model. The proposed method is effective in improving detection of micro objects by optimizing with GAN network for super-resolution and an object recognition network. In addition, the proposed X-transfer learning technique alternately uses transfer learning and curriculum learning to overcome the lack of training data. This method can improve the accuracy, robustness, and localization performance of object recognition based on rich visual information on entire network. The proposed model was evaluated with remote sensing data sets. It was confirmed that the proposed method is more accurate than existing methods in terms of mAP@0.5 and F1 scores.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"148 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-31DOI: 10.9717/kmms.2023.26.8.1027
Hamdi Abdurhman Ahmed, Jihwan Lee, Donghyun Kim, ByeongSeok Yu
First, we propose a class of efficient models classed as choice-based recommendation (CBR) for parametric metrics, such as a logit model as a recommendation system using nonparametric approaches. The rest of the papers is organized as follow : we used a simple, streamlined architecture that uses a nonparametric approach such as a feedforward deep neural network (DNN). The study implemented a method to deal with a choice set with a fixed and variable-length option, investigate deep learning methods that consider each choice set as one sample point, the effect of embedding categorical features and accuracy impact, and the efficiency of batch normalization toward a more stable network. To check the performance of our approach, we conducted extensive experiments on multiple datasets and used the top-k accuracy as a metric. We then show the effectiveness of CBR across two industrial applications and use cases, including hotel booking and airline itineraries. The results show that the DNN outperforms the multinomial logit model (MNL) with significant top-k accuracy. The top-k accuracy was further divided into three different DNN models. Among the models, a model that included a layer with batch normalization embedding outperforms with top-k accuracy compared with the model that does not include both batch normalization and embedding layer in the proposed DNN architecture.
{"title":"Deep Learning Ar chitectur e for Choice-based Recommendation System: A Case Study of Flight Sear ch Engine","authors":"Hamdi Abdurhman Ahmed, Jihwan Lee, Donghyun Kim, ByeongSeok Yu","doi":"10.9717/kmms.2023.26.8.1027","DOIUrl":"https://doi.org/10.9717/kmms.2023.26.8.1027","url":null,"abstract":"First, we propose a class of efficient models classed as choice-based recommendation (CBR) for parametric metrics, such as a logit model as a recommendation system using nonparametric approaches. The rest of the papers is organized as follow : we used a simple, streamlined architecture that uses a nonparametric approach such as a feedforward deep neural network (DNN). The study implemented a method to deal with a choice set with a fixed and variable-length option, investigate deep learning methods that consider each choice set as one sample point, the effect of embedding categorical features and accuracy impact, and the efficiency of batch normalization toward a more stable network. To check the performance of our approach, we conducted extensive experiments on multiple datasets and used the top-k accuracy as a metric. We then show the effectiveness of CBR across two industrial applications and use cases, including hotel booking and airline itineraries. The results show that the DNN outperforms the multinomial logit model (MNL) with significant top-k accuracy. The top-k accuracy was further divided into three different DNN models. Among the models, a model that included a layer with batch normalization embedding outperforms with top-k accuracy compared with the model that does not include both batch normalization and embedding layer in the proposed DNN architecture.","PeriodicalId":16316,"journal":{"name":"Journal of Korea Multimedia Society","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135991685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：