Digital Library

Vol. 18, No. 5, Oct. 2022  Cover

1. Feature Extraction of Non-proliferative Diabetic Retinopathy Using Faster R-CNN and Automatic Severity Classification System Using Random Forest Method
   Younghoon Jung, Daewon Kim
   Vol. 18, No. 5, pp. 599-613, Oct. 2022  
   10.3745/JIPS.04.0252
   Details Full Text PubReader
   Keywords: Faster R-CNN, Classification, Machine Learning, Non-proliferative Diabetic Retinopathy, Random Forest
   Show / Hide Abstract

   Non-proliferative diabetic retinopathy is a representative complication of diabetic patients and is known to be a major cause of impaired vision and blindness. There has been ongoing research on automatic detection of diabetic retinopathy, however, there is also a growing need for research on an automatic severity classification system. This study proposes an automatic detection system for pathological symptoms of diabetic retinopathy such as microaneurysms, retinal hemorrhage, and hard exudate by applying the Faster R-CNN technique. An automatic severity classification system was devised by training and testing a Random Forest classifier based on the data obtained through preprocessing of detected features. An experiment of classifying 228 test fundus images with the proposed classification system showed 97.8% accuracy.


2. Aspect-Based Sentiment Analysis with Position Embedding Interactive Attention Network
   Yan Xiang, Jiqun Zhang, Zhoubin Zhang, Zhengtao Yu, Yantuan Xian
   Vol. 18, No. 5, pp. 614-627, Oct. 2022  
   10.3745/JIPS.04.0254
   Details Full Text PubReader
   Keywords: Aspect-Based Sentiment Analysis, Attention Mechanism, long short-term memory network, Position Information, Word Embedding
   Show / Hide Abstract

   Aspect-based sentiment analysis is to discover the sentiment polarity towards an aspect from user-generated natural language. So far, most of the methods only use the implicit position information of the aspect in the context, instead of directly utilizing the position relationship between the aspect and the sentiment terms. In fact, neighboring words of the aspect terms should be given more attention than other words in the context. This paper studies the influence of different position embedding methods on the sentimental polarities of given aspects, and proposes a position embedding interactive attention network based on a long short-term memory network. Firstly, it uses the position information of the context simultaneously in the input layer and the attention layer. Secondly, it mines the importance of different context words for the aspect with the interactive attention mechanism. Finally, it generates a valid representation of the aspect and the context for sentiment classification. The model which has been posed was evaluated on the datasets of the Semantic Evaluation 2014. Compared with other baseline models, the accuracy of our model increases by about 2% on the restaurant dataset and 1% on the laptop dataset.


3. Applying Token Tagging to Augment Dataset for Automatic Program Repair
   Huimin Hu, Byungjeong Lee
   Vol. 18, No. 5, pp. 628-636, Oct. 2022  
   10.3745/JIPS.04.0251
   Details Full Text PubReader
   Keywords: Augment Dataset, Automatic Program Repair, Machine Learning, Token Tagging
   Show / Hide Abstract

   Automatic program repair (APR) techniques focus on automatically repairing bugs in programs and providing correct patches for developers, which have been investigated for decades. However, most studies have limitations in repairing complex bugs. To overcome these limitations, we developed an approach that augments datasets by utilizing token tagging and applying machine learning techniques for APR. First, to alleviate the data insufficiency problem, we augmented datasets by extracting all the methods (buggy and non-buggy methods) in the program source code and conducting token tagging on non-buggy methods. Second, we fed the preprocessed code into the model as an input for training. Finally, we evaluated the performance of the proposed approach by comparing it with the baselines. The results show that the proposed approach is efficient for augmenting datasets using token tagging and is promising for APR.


4. Decision Support System for Mongolian Portfolio Selection
   Enkhtuul Bukhsuren, Uyanga Sambuu, Oyun-Erdene Namsrai, Batnasan Namsrai, Keun Ho Ryu
   Vol. 18, No. 5, pp. 637-649, Oct. 2022  
   10.3745/JIPS.04.0255
   Details Full Text PubReader
   Keywords: Data Mining, decision support system, K-Means Clustering
   Show / Hide Abstract

   Investors aim to increase their profitability by investing in the stock market. An adroit strategy for minimizing related risk lies through diversifying portfolio operationalization. In this paper, we propose a six-step stocks portfolio selection model. This model is based on data mining clustering techniques that reflect the ensuing impact of the political, economic, legal, and corporate governance in Mongolia. As a dataset, we have selected stock exchange trading price, financial statements, and operational reports of top-20 highly capitalized stocks that were traded at the Mongolian Stock Exchange from 2013 to 2017. In order to cluster the stock returns and risks, we have used k-means clustering techniques. We have combined both k-means clustering with Markowitz's portfolio theory to create an optimal and efficient portfolio. We constructed an efficient frontier, creating 15 portfolios, and computed the weight of stocks in each portfolio. From these portfolio options, the investor is given a choice to choose any one option.


5. Feature Analysis for Detecting Mobile Application Review Generated by AI-Based Language Model
   Seung-Cheol Lee, Yonghun Jang, Chang-Hyeon Park, Yeong-Seok Seo
   Vol. 18, No. 5, pp. 650-664, Oct. 2022  
   10.3745/JIPS.02.0182
   Details Full Text PubReader
   Keywords: Artificial intelligence, Fake Review, GPT-2, Language Model, Machine Learning, software engineering
   Show / Hide Abstract

   Mobile applications can be easily downloaded and installed via markets. However, malware and malicious applications containing unwanted advertisements exist in these application markets. Therefore, smartphone users install applications with reference to the application review to avoid such malicious applications. An application review typically comprises contents for evaluation; however, a false review with a specific purpose can be included. Such false reviews are known as fake reviews, and they can be generated using artificial intelligence (AI)-based text-generating models. Recently, AI-based text-generating models have been developed rapidly and demonstrate high-quality generated texts. Herein, we analyze the features of fake reviews generated from Generative Pre-Training-2 (GPT-2), an AI-based text-generating model and create a model to detect those fake reviews. First, we collect a real human-written application review from Kaggle. Subsequently, we identify features of the fake review using natural language processing and statistical analysis. Next, we generate fake review detection models using five types of machine-learning models trained using identified features. In terms of the performances of the fake review detection models, we achieved average F1-scores of 0.738, 0.723, and 0.730 for the fake review, real review, and overall classifications, respectively.


6. Optimization Method of Knapsack Problem Based on BPSO-SA in Logistics Distribution
   Yan Zhang, Tengyu Wu, Xiaoyue Ding
   Vol. 18, No. 5, pp. 665-676, Oct. 2022  
   10.3745/JIPS.01.0090
   Details Full Text PubReader
   Keywords: Binary Particle Swarm, Knapsack Problem, Logistics Distribution, Logistics Simulation, simulated annealing
   Show / Hide Abstract

   In modern logistics, the effective use of the vehicle volume and loading capacity will reduce the logistic cost. Many heuristic algorithms can solve this knapsack problem, but lots of these algorithms have a drawback, that is, they often fall into locally optimal solutions. A fusion optimization method based on simulated annealing algorithm (SA) and binary particle swarm optimization algorithm (BPSO) is proposed in the paper. We establish a logistics knapsack model of the fusion optimization algorithm. Then, a new model of express logistics simulation system is used for comparing three algorithms. The experiment verifies the effectiveness of the algorithm proposed in this paper. The experimental results show that the use of BPSO-SA algorithm can improve the utilization rate and the load rate of logistics distribution vehicles. So, the number of vehicles used for distribution and the average driving distance will be reduced. The purposes of the logistics knapsack problem optimization are achieved.


7. TsCNNs-Based Inappropriate Image and Video Detection System for a Social Network
   Youngsoo Kim, Taehong Kim, Seong-eun Yoo
   Vol. 18, No. 5, pp. 677-687, Oct. 2022  
   10.3745/JIPS.04.0253
   Details Full Text PubReader
   Keywords: CNN, Intelligent Image and Video Detection System, Tree-Structured Convolutional Neural Networks (TsCNNs)
   Show / Hide Abstract

   We propose a detection algorithm based on tree-structured convolutional neural networks (TsCNNs) that finds pornography, propaganda, or other inappropriate content on a social media network. The algorithm sequentially applies the typical convolutional neural network (CNN) algorithm in a tree-like structure to minimize classification errors in similar classes, and thus improves accuracy. We implemented the detection system and conducted experiments on a data set comprised of 6 ordinary classes and 11 inappropriate classes collected from the Korean military social network. Each model of the proposed algorithm was trained, and the performance was then evaluated according to the images and videos identified. Experimental results with 20,005 new images showed that the overall accuracy in image identification achieved a high-performance level of 99.51%, and the effectiveness of the algorithm reduced identification errors by the typical CNN algorithm by 64.87 %. By reducing false alarms in video identification from the domain, the TsCNNs achieved optimal performance of 98.11% when using 10 minutes frame-sampling intervals. This indicates that classification through proper sampling contributes to the reduction of computational burden and false alarms.


8. Designing a Magnetically Controlled Soft Gripper with Versatile Grasping Based on Magneto-Active Elastomer
   Rui Li, Xinyan Li, Hao Wang, Xianlun Tang, Penghua Li, Mengjie Shou
   Vol. 18, No. 5, pp. 688-700, Oct. 2022  
   10.3745/JIPS.04.0256
   Details Full Text PubReader
   Keywords: Biomimetics, Magnetic Actuation, Magnetic Particle-Filled Composite, Soft Gripper
   Show / Hide Abstract

   A composite bionic soft gripper integrated with electromagnets and magneto-active elastomers is designed by combining the structure of the human hand and the snake’s behavior of enhancing friction by actively adjusting the scales. A silicon-based polymer containing magnetized hard magnetic particles is proposed as a soft finger, and it can be reversibly bent by adjusting the magnetic field. Experiments show that the length, width, and height of rectangular soft fingers and the volume ratio of neodymium–iron–boron have different effects on bending angle. The flexible fingers with 20 vol% are the most efficient, which can bend to 90° when the magnetic field is 22 mT. The flexible gripper with four fingers can pick up 10.51 g of objects at the magnetic field of 105 mT. In addition, this composite bionic soft gripper has excellent magnetron performance, and it can change surface like snakes and operate like human hands. This research may help develop soft devices for magnetic field control and try to provide new solutions for soft grasping.


9. Semantic Image Segmentation for Efficiently Adding Recognition Objects
   Chengnan Lu, Jinho Park
   Vol. 18, No. 5, pp. 701-710, Oct. 2022  
   10.3745/JIPS.02.0183
   Details Full Text PubReader
   Keywords: Image Segmentation, Machine Learning, Object Detection
   Show / Hide Abstract

   With the development of artificial intelligence technology, various methods have been developed for recognizing objects in images using machine learning. Image segmentation is the most effective among these methods for recognizing objects within an image. Conventionally, image datasets of various classes are trained simultaneously. In situations where several classes require segmentation, all datasets have to be trained thoroughly. Such repeated training results in low training efficiency because most of the classes have already been trained. In addition, the number of classes that appear in the datasets affects training. Some classes appear in datasets in remarkably smaller numbers than others, and hence, the training errors will not be properly reflected when all the classes are trained simultaneously. Therefore, a new method that separates some classes from the dataset is proposed to improve efficiency during training. In addition, the accuracies of the conventional and proposed methods are compared.


10. Accurate Segmentation Algorithm of Video Dynamic Background Image Based on Improved Wavelet Transform
    Ming Ming
    Vol. 18, No. 5, pp. 711-718, Oct. 2022  
    10.3745/JIPS.02.0184
    Details Full Text PubReader
    Keywords: Dynamic Background Image, Image Segmentation, Improved Wavelet Transform, Video Image
    Show / Hide Abstract

    In this paper, an accurate segmentation algorithm of video dynamic background image (VDBI) based on improved wavelet transform is proposed. Based on the smooth processing of VDBI, the traditional wavelet transform process is improved, and the two-layer decomposition of dynamic image is realized by using two- dimensional wavelet transform. On the basis of decomposition results and information enhancement processing, image features are detected, feature points are extracted, and quantum ant colony algorithm is adopted to complete accurate segmentation of the image. The maximum SNR of the output results of the proposed algorithm can reach 73.67 dB, the maximum time of the segmentation process is only 7 seconds, the segmentation accuracy shows a trend of decreasing first and then increasing, and the global maximum value can reach 97%, indicating that the proposed algorithm effectively achieves the design expectation.