International Journal of Advances in Intelligent Informatics
Not a member yet
235 research outputs found
Sort by
Abnormal behavior recognition using SRU with attention mechanism
Full text linkIn response to the critical need for enhanced public safety measures, this study introduces an advanced intelligent surveillance system designed to autonomously detect abnormal behaviors within public spaces. Leveraging the computational efficiency and accuracy of a Simple Recurrent Unit (SRU) integrated with an attention mechanism, this research delivers a novel approach towards understanding and interpreting human interactions in real-time video footage. Distinctively, the model specializes in identifying two primary categories of abnormal behavior: aggressive two-person interactions such as physical confrontations and collective crowd dynamics, characterized by sudden dispersal patterns indicative of distress or danger. The incorporation of Attention mechanism precisely targets critical elements of behavior, thereby enhancing the model's focus and interpretative clarity. Empirical validation across five benchmark datasets reveals that our model not only outperforms traditional Long Short-Term Memory (LSTM) frameworks in terms of speed by a factor of 1.5 but also demonstrates superior accuracy in abnormal behavior recognition. These findings not only underscore the model's potential in preempting potential safety threats but also mark a significant advancement in the application of deep learning technologies for public security infrastructures. This research contributes to the broader discourse on public safety, offering actionable insights and robust technological solutions to enhance surveillance efficacy and response mechanisms in critical public domains
Fastener and rail surface defects detection with deep learning techniques
Full text linkThe railways, which are frequently used by countries for both passenger and freight transportation, should be checked periodically. Controls made with classical methods are slow and there are often overlooked faults.  This work suggests a novel deep learning-based technique for identifying fastener and railway track surface defects. Within the scope of the proposed method, firstly,  The railroad track was observed using an autonomous drone. Shaky images in the recorded video were removed with a video stabilization algorithm. Frames were created and labeled from the video and rail and fastener regions were detected using the Faster R-CNN deep neural network. Fault detection was performed through ResNet101v2-based classification using different datasets for  identifying surface detects in rails and different datasets for detection of fasteners. The proposed method was experimentally shown to have a 98% accuracy rate for detecting rail surface flaws and a 95% accuracy rate for detecting fastener flaws. An user interface was developed to display the identified faulty images on computers, tablets and mobile phones via a mobile application. The system, which was previously proposed in a different study, was retrained by going through the video stabilization step, thus improving the fault detection rate, and the method was also included in the user interface module. This study contributes to the processing of ever-increasing video data with deep learning-based methods. It is also anticipated that it will benefit researchers working in the field of railway non-contact fault detection
Comparative study of predictive models for hoax and disinformation detection in indonesian news
Full text linkAlong with the times, false information easily spreads, including in Indonesia. In Press Release No.485/HM/KOMINFO/12/2021 the Ministry of Communication and Information has cut off access to 565,449 negative content and published 1,773 clarifications on hoax and disinformation content. Research has been carried out regarding this matter, but it is necessary to classify fake news into disinformation and hoaxes. This study presents a comparison between our proposed model, which is an ensemble of shallow learning predictive models, namely Random Forest, Passive Aggressive Classifier, and Cosine Similarity, and the deep learning model that uses BERT-Indo for classification. Both models are trained using equivalent datasets, which contain 8757 news, consisting of 3000 valid news, 3000 hoax news, and 2757 disinformation news. These news were obtained from websites such as CNN, Kompas, Detik, Kominfo, Temanggung Mediacenter, Hoaxdb Aceh, Turnback Hoax, and Antara, which were then cleaned from all unnecessary substances, such as punctuation marks, numbers, Unicode, stopwords, and suffixes using the Sastrawi library. At the benchmarking stage, the shallow learning model is evaluated to increase accuracy by applying ensemble learning combined using hard voting. This results in higher values, with an accuracy of 98.125%, precision of 98.2%, F-1 score of 98.1%, and recall of 98.1%, compared to the BERT-Indo model which only achieved 96.918% accuracy, 96.069% precision, 96.937% F-1 score, and 96.882% recall. Based on the accuracy value, shallow learning model is superior to deep learning model. This machine learning model is expected to be used to combat the spread of hoaxes and disinformation in Indonesian news. Additionally, with this research, false news can be classified in more detail, both as hoaxes and disinformatio
Enhanced personalized learning exercise question recommendation model based on knowledge tracing
Full text linkPersonalized exercise question recommendation is a crucial aspect of smart education used to customize educational exercises and questions to individual students' distinct abilities and learning progress. Integrating cognitive diagnosis with deep learning has shown promising results in personalized exercise recommendations. However, the black-box nature of the deep learning model hinders their interpretability. This makes it challenging for educators and students to understand the reasons behind the model's predictions for the next problem, and this limits their opportunity to take an active role in improving the learning process. To address this limitation, this article presents a novel personalized exercise question recommendation model based on knowledge tracing. The approach incorporates graph convolutional neural networks to model the student's abilities, thus enhancing the interpretability of the model. By employing Bidirectional gate recurrent unit (Bi-GRU), the model effectively traces fluctuations in students' abilities over time and predicts their responses to exercise questions. Experimental results demonstrate the effectiveness of this model, achieving an accuracy of 90.8% and 92.6% on ASSISTment 2009 and ASSISTment 2017 datasets, containing 4218 and 1709 student records, respectively. Moreover, the experiment was also conducted to validate the model's exercise difficulty setting. Results indicate an acceptable level of effectiveness in generating appropriate difficulty-level recommendations for individual students. The proposed model contributes to advancing personalized exercise recommendations by offering valuable insights that can lead to more efficient and effective student learning experiences
Self-supervised few-shot learning for real-time traffic sign classification
Full text linkAlthough supervised approaches for traffic sign classification have demonstrated excellent performance, they are limited to classifying several traffic signs defined in the training dataset. This prevents them from being applied to different domains, i.e., different countries. Herein, we propose a self-supervised approach for few-shot learning-based traffic sign classification. A center-awareness similarity network is designed for the traffic sign problem and trained using an optical flow dataset. Unlike existing supervised traffic sign classification methods, the proposed method does not depend on traffic sign categories defined by the training dataset. It applies to any traffic signs from different countries. We construct a Korean traffic sign classification (KTSC) dataset, including 6000 traffic sign samples and 59 categories. We evaluate the proposed method with baseline methods using the KTSC, German traffic sign, and Belgian traffic sign classification datasets. Experimental results show that the proposed method extends the ability of existing supervised methods and can classify any traffic sign, regardless of region/country dependence. Furthermore, the proposed approach significantly outperforms baseline methods for patch similarity. This approach provides a flexible and robust solution for classifying traffic signs, allowing for accurate categorization of every traffic sign, regardless of regional or national differences
A novel convolutional feature-based method for predicting limited mobility eye gaze direction
Full text linkEye gaze direction is a critical issue since several applications in computer vision technology rely on determining gaze direction, where individuals move their eyes to limited mobility locations for sensory information. Deep neural networks are considered one of the most essential and accurate image classification methods. Several methods of classification to determine the direction of the gaze employ convolutional neural network models, which are VGG, ResNet, Alex Net, etc. This research presents a new method of identifying human eye images and classifying eye gaze directions (left, right, up, down, straight) in addition to eye-closing discrimination. The proposed method (Di-eyeNET) stands out from the developed method (Split-HSV) for enhancing image lighting. It also reduces implementation time by utilizing only two blocks and employing dropout layers after each block to achieve fast response times and high accuracy. It focused on the characteristics of the human eye images, as it is small, so it cannot be greatly enlarged, and the eye's iris is in the middle of the image, so the edges are not important. The proposed method achieves excellent results compared to previous methods, classifying the five directions of eye gaze instead of the four directions. Both the global dataset and the built local dataset were utilized. Compared to previous methods, the suggested method's results demonstrate high accuracy (99%), minimal loss, and the lowest training time. The research benefits include an efficient method for classifying eye gaze directions, with faster implementation and improved image lighting
Enhancement of images compression using channel attention and post-filtering based on deep autoencoder
Full text linkImage compression is a crucial research topic in today's information age, especially to meet the demand for balanced data compression efficiency with the quality of the resulting image reconstruction. Common methods used for image compression nowadays are based on autoencoders with deep learning foundations. However, these methods have limitations as they only consider residual values in processed images to achieve existing compression efficiency with less satisfying reconstruction results. To address this issue, we introduce the Attention Block mechanism to improve coding efficiency even further. Additionally, we introduce post-filtering methods to enhance the final reconstruction results of images. Experimental results using two datasets, CLIC for training and KODAK for testing, demonstrate that this method outperforms several previous research methods. With an efficiency coding improvement of -28.16%, an average PSNR improvement of 34%, and an MS-SSIM improvement of 8%, the model in this study significantly enhances the rate-distortion (RD) performance compared to previous approaches
Link stability - based optimal routing path for efficient data communication in MANET
Full text linkThe paper delves into the complexities of Mobile Ad hoc Networks (MANETs), which consist of a diverse array of wireless nodes. In such networks, routing packets poses a significant challenge due to their dynamic nature. Despite the variety of techniques available for optimizing routing in MANETs, persistent issues like packet loss, routing overhead, and End-to-End Delay (EED) remain prevalent. In response to these challenges, the paper proposes a novel approach for efficient Data Communication (DC) by introducing a Link Stability (LS)-based optimal routing path. This approach leverages several advanced techniques, including Pearson Correlation Coefficient SWIFFT (PCC-SWIFFT), Galois-based Digital Signature Algorithm (G-DSA), and Entropy-based Gannet Optimization Algorithm (E-GOA). The proposed methodology involves a systematic process. Initially, the nodes in the MANET are initialized to establish the network infrastructure. Subsequently, the Canberra-based K Means (C-K Means) algorithm is employed to identify Neighboring Nodes (NNs), which are pivotal for creating communication links within the network. To ensure secure communication, secret keys (SK) are generated for both the Sender Node (SN) and the Receiver Node (RN) using Galois Theory. Following this, PCC-SWIFFT methodologies are utilized to generate hash codes, serving as unique identifiers for data packets or routing information. Signatures are created and verified at the SN and RN using the G-DSA. Verified nodes are subsequently added to the routing entry table, facilitating the establishment of multiple paths within the network. The Optimal Path (OP) is selected using the E-GOA, considering factors such as link stability and network congestion. Finally, Data Communication (DC) is initiated, continuously monitoring LS to ensure optimal routing performance. Comparative analysis with existing methodologies demonstrates the superior performance of the proposed model. In summary, the proposed approach offers a comprehensive solution to enhance routing efficiency in MANETs by addressing critical issues and leveraging advanced algorithms for key generation, signature verification, and path optimizatio
Type-2 Fuzzy ANP and TOPSIS methods based on trapezoid Fuzzy number with a new metric
Full text linkModeling and linguistic representation in the form Interval Type-2 Fuzzy have better accuracy than Type-1 Fuzzy. The type-2 fuzzy set involves more uncertainty than the type-1 fuzzy set. The degree of fuzzy membership is used to explain uncertainty and ambiguity in the real world. This study presents the type-2 Fuzzy Analytic Network Process (ANP) method to determine the weight of each attribute based on the level of interest and the extension method of type-2 Fuzzy TOPSIS to handle problems based on the value of the fuzzy type-2 attribute. Decision-making is based on the assessment of several experts called Multi-Criteria Group Decision Making (MCGDM), using type-2 Fuzzy geometric mean aggregation function. The membership function in this research is type-2 fuzzy based on the trapezoid. The contribution is a hybrid method Type-2 Fuzzy TOPSIS with Fuzzy Type-2 ANP group-based with new metric intervals on fuzzy type-2 for decision making. The results are a hybrid type-2 FANP and FTOPSIS decision-making model to support the best decision-making. Based on a comparison of the accuracy of trapezoid model 1, model 2, and model 3, the best accuracy result is model 3, which is 84%. The research benefits by presenting a hybrid Type-2 Fuzzy TOPSIS and ANP method that improves decision-making accuracy and better handling uncertainty and ambiguity than Type-1 Fuzzy systems
Optimization of use case point through the use of metaheuristic algorithm in estimating software effort
Full text linkUse Case Points estimation framework relies on the complexity weight parameters to estimate software development projects. However, due to the discontinue parameters, it lead to abrupt weight classification and results in inaccurate estimation. Several research studies have addressed these weaknesses by employing various approaches, including fuzzy logic, regression analysis, and optimization techniques. Nevertheless, the utilization of optimization techniques to determine use case weight parameter values has yet to be extensively explored, with the potential to enhance accuracy further. Motivated by this, the current research delves into various metaheuristic search-based algorithms, such as genetic algorithms, Firefly algorithms, Reptile search algorithms, Particle swarm optimization, and Grey Wolf optimizers. The experimental investigation was carried out using a Silhavy UCP estimation dataset, which contains 71 project data from three software houses and is publicly available. Furthermore, we compared the performance between models based on metaheuristic algorithms. The findings indicate that the performance of the Firefly algorithm outperforms the others based on five accuracy metrics: mean absolute error, mean balance relative error, mean inverted relative error, standardized accuracy, and effect size. This research could be useful for software project managers to leverage the practical implications of this study by utilizing the UCP estimation method, which is optimized using the Firefly algorithm