International Journal of Intelligent Engineering and Systems

1

Authors: Anand M, Meenakshi Sundaram A

Diabetic retinopathy stands out as one of the highly prevalent causes of vision loss in working people worldwide. In computer vision, deep learning based strategies are seen as a viable solution for efficient diabetic retinopathy detection. We present a UNet-based deep learning architecture for diabetic retinopathy segmentation of blood vessels, exudates, and microaneurysms. Traditional methods often consider the features only from the last convolution unit and discard the remaining features, resulting in low-quality feature maps. However, boundary information plays important role in medical image segmentation. To overcome this, we introduce a skip connection mechanism to concatenate all attributes from each layer. Additionally, we utilize an upsampling layer to aggregate the features at the final sigmoid layer. Finally, we apply channel and spatial attention mechanisms to generate the semantic feature map. Therefore, the proposed approach overcomes the issues of existing methods by incorporating dense skip connection along with channel and spatial attention mechanism which helps to retain the substantial information of image. We tested proposed approach on several publicly available datasets such as IDRiD, DIARETDB1, STARE, ChaseDB1, DRIVE, and HRF datasets. The comparative analysis shows that the proposed approach achieves superior performance, with an average accuracy of 98.10%, average sensitivity of 97.60%, and average specificity of 98.2% for segmentation.

39

Authors: Gouri Bompem, Dhanalakshmi Pandluri

The uncontrolled growth of cells in human brain can lead to the formation of tumors, which can occur in all age people. The tumor in brain can affect nerve cells, soft tissues and blood vessels. The early detection of brain tumor is necessary to aid doctors in treating cancer patients to increase their survival rate. For this various deep learning models are created and discovered for efficient brain tumor detection and classification. In this paper, the Convolutional Neural Network is proposed for efficient brain tumor classification in MRI images using BRATS 2019, 2020 and 2021 dataset. The min-max normalization is used in this research for data preprocessing and fed to the segmentation process. The mask region-based CNN is employed for segmenting brain tumors; Followed by that, Batch normalization is applied to enhance the training process and minimize the overfitting issues. The obtained result shows that the proposed CNN model achieves better accuracy of 99.55% on BRATS 2019, 99.80% on BRATS 2020 and 99.29% on BRATS 2021 dataset which ensures accurate classification compared with other existing methods like 3D U-Net and CapsNet + latent-dynamic condition random field (LDCRF) + post-processing.

61

Authors: Phani Vidhyadhar Rallapalli, Sribindu Sattu, Anil Kumar Kotthuru, Rajanidevi Merigala, Praveen Kumar Devulapalli, Dhanalakshmi Gopal

The multimedia wireless sensor networks (MWSN) need high bit-error rates and high transmission rates in order to send high-quality images through smart devices. The multiple-input multiple-output (MIMO) technology, on which MWSN is based, significantly depends on cooperative communication to properly exploit its benefits. By simultaneously broadcasting multiple symmetrical channels to prevent interference, large-scale wireless networks use multi-radio, multi-channel technology to increase performance. Significant interference can prevent cooperative communication from progressing in large networks. Additionally, since each node in the network is mobile, cooperative multimedia wireless sensor networks have serious issues with routing and transmission latency. The MWSNs become more dynamic as a result of mobility, which is reflected in the overhead control traffic. An adaptive clustering algorithm (ACA) was proposed by concurrently considering (i) relay selection by using channel capacity, mobility and distance metrics and (ii) channel assignment (CA) utilising a dynamic global table (GT). Further to improve the stability of a cluster an adaptive reliable clustering algorithm (ARCA) is present based on a priority neighbour-following technique to choose the ideal neighbour nodes to join the same cluster during the cluster head selection phase to increase cluster stability and reliability passively. Our approach selects a relay-node based on mobility and the maximum available channel capacity in order to reduce end-to-end transmission latency while maintaining aggregate throughput. Simulation results shows that the proposed approach improves the aggregate throughput of the network by 150%, 225%, and 240% when compared to that of the Cluster-based cooperative multi-hop optimal relay selection (CCORS), energy aware cooperative image transmission (EACIT), and efficient cooperative image transmission (ECIT) respectively.

83

Authors: Rania Hamdy Elabd, Ahmed Jamal Abdullah Al-Gburi, Khaled Alhassoon, Jamil Abedalrahim Jamil Alsayaydeh, Tale Saeidi, Naba Jasim Mohammed, Zahriladha Zakaria

This paper presents a high-selectivity, low-insertion-loss open-loop resonator (OLR)-based microstrip diplexer designed for wireless communication systems. We employed a microstrip transmission line with two serial capacitive gaps, incorporating rectangular half-wavelength open-loop resonators, to create a bandpass filter (BPF) with remarkable selectivity. The intended BPFs are interconnected by a T-junction combiner, which is matched to both filters and the antenna port, thereby creating the modelled diplexer. The system is implemented on a Rogers TMM4 substrate featuring a loss tangent of 0.002, a dielectric constant of 4.7, and a thickness of 1.52 mm. The suggested diplexer has dimensions of (90×70) mm². It achieves a modest frequency spacing ratio of R=0.1646 in both transmit and receive modes, with resonance frequencies of ft = 2.191 GHz and fr = 2.584 GHz, respectively. The simulated structure demonstrates favorable insertion losses of approximately 1.2 dB and 1.79 dB for the two channels at fractional bandwidths of 1.24% (2.191 GHz) and 0.636% (2.584 GHz). The simulated isolation values at 2.191 GHz and 2.584 GHz are 53.3 dB and 66.5 dB, respectively. Notably, the simulated and measured results exhibit a high degree of consistency. However, minor variations between the simulated and actual results have been noted, primarily attributable to the inherent fabrication tolerances.

110

Authors: Eko Purwanto Aribowo, Albertus Joko Santoso, Pranowo

Quality inspection of solder connections in electronic circuit manufacturing is commonly performed using automatic optical inspection (AOI) technology. The utilization of deep learning in AOI has demonstrated high accuracy and fast computation, yet it requires expensive graphics processing unit (GPU) equipped computers. To enable broader and more cost-effective utilization, there is a need for an AOI model that can be embedded in simple central processing unit (CPU) based computers. To pursue this objective, several research efforts have been undertaken to develop AOI models based on classical machine learning techniques. However, the accuracy and speed achieved by these models have not yet matched deep learning-based AOI models. This study aims to enhance the computational efficiency of classic machine learning by processing only pixels containing textural information. The effectiveness enhancement is achieved through the application of log-polar transformation in the extraction of texture features using the gray level co-occurrence matrix (GLCM) to detect defects in trough hole technology (THT) solder joint connections. By transforming cartesian coordinates into polar coordinates, the textural areas to be detected assume a square shape, facilitating efficient texture feature extraction. To ensure a significant improvement in performance, a comparative performance evaluation is conducted on classic machine learning-based AOI models with and without the log-polar transformation. The texture features extracted from both models are classified using the support vector machine (SVM) method. Model testing and evaluation reveals that the proposed enhancement effort is capable of increasing accuracy levels to 95% and reach computation time by 120 milliseconds.

134

Authors: Ravichandra Torthi, Ajay Dilip Kumar Marapatla, Soumya Mande, Harish Kumar Varma Gadiraju, Chalapathiraju Kanumuri

Nowadays, heart diseases have become a leading cause of mortality worldwide and it affects a huge number of individuals. The early and accurate prediction of heart disease risk factors plays a crucial role in preventing opposing results. Additionally, it is necessary to recognize heart disease quickly and accurately by analyzing patient’s data. This paper proposed a novel approach for predicting heart disease through machine learning techniques. The proposed Bat Algorithm (BA) and Particle Swarm Optimization (PSO) based Random Forest (RF), named BAPSO-RF is utilized for selecting optimum features that can enhance the heart-disease prediction accuracy. The proposed BAPSO-RF is evaluated on UCI heart disease dataset which contains 14 attributes and 270 records. The proposed BAPSO-RF model attains better results by utilizing metrics like accuracy, precision, recall, and f1-score values of about 98.71%, 98.67%, 98.23%, and 98.45% correspondingly which ensures early and accurate prediction of heart disease compared to existing techniques like hybrid of Genetic Algorithm and Particle Swarm Optimization (PSO) with Random Forest (GAPSO-RF), stacked Genetic Algorithm (GA) and Genetic Algorithm with Radial Basis Function (GA-RBF).

159

Authors: Susmitha Pothireddy, Nikhila Peddisetty, Pachipala Yellamma, Gitanjani Botta, Kailash Nadh Gottipati

Hybrid encryption techniques are gaining popularity in the cloud platform due to their ability to address security concerns and data storage challenges. In order to ensure confidentiality at every stage of processing, this article presents an innovative strategy that merges fully homomorphic encryption (FHE) and secure hash algorithm 3 (SHA-3), offering robust protection against cyber threats. FHE and SHA-3 offer a solution to this problem. We conducted experiments using different datasets to test the performance and security strength of this hybrid approach. Our method takes into account factors such as computational overheads, complexity requirements, key management challenges, and limited support from vendors/providers offering specialized hardware/software capable of supporting complex mathematical functions required by each technique separately before combined into a hybrid form suitable enough scaled-out horizontally across different nodes within clouds themselves today globally among end-users relying upon them heavily day-to-day tasks involving mission-critical workloads running inside virtualized machines hosted on public/private clouds ecosystems worldwide. The results show that implementing FHE together with SHA-3 can significantly improve an organization's overall cyber security posture when operating critical infrastructure components residing within cloud environments. According to the results, it performs better than other methods that are currently in used complexity of encryption and decoding, including homomorphic encryption and RSA, AES & DSA, and AES & RSA. The hybrid algorithm we are proposed using fully homomorphic & SHA3 with 2048 bits 0.789 seconds of encryption time and decryption time with 0.001 seconds and execution time with 4.25 seconds. Therefore, we conclude that our proposed hybrid encryption technique is the best approach to address cyber security risks associated with storing sensitive data in cloud computing environments.

180

Authors: Triyanna Widiyaningtyas, Aji Prasetya Wibawa, Utomo Pujianto, Wahyu Caesarendra

Collaborative filtering has emerged as one of the most prevalent techniques for various commercial recommendations. Utilizing a similarity measure to identify similar neighbors is essential to collaborative filtering. Behavior scores and user ratings have recently become increasingly important factors in determining similarity. However, the added users’ behavior scores generate a more complex computation. This research proposes a new similarity technique incorporating the matrix factorization and users’ behavior score-based similarity to minimize computation time. The matrix factorization technique uses singular value decomposition (SVD), and the users’ behavior score-based similarity employs normalized cumulative genre (NCG). Compared to the previous algorithm (i.e., users’ scores probability-based collaborative filtering), the experimental findings with the MovieLens 1M and 100k datasets demonstrated a faster computing time. In addition, with these datasets, our similarity reduces the root mean square error (RMSE) by 8.14% and 11.99% and the mean absolute error (MAE) by 13.52% and 15.81%.

201

Authors: Narendra Babu Kattepogu, G Saravanan, A Rama Koteswara Rao

In this paper, an enhanced sparrow search algorithm (ESSA) is proposed to solve the optimal power flow (OPF) problem of power systems considering renewable energy systems. Two modifications were introduced to the basic SSA to overcome its weak global search capability. Elite reverse learning strategy for diversifying the population and mutation strategy of firefly algorithm (FA) to achieve a fast convergence rate. The effectiveness of the ESSA is analysed for two types of power systems: conventional power systems with only thermal power plants and modern power systems with thermal and renewable energy power plants. A multi-objective optimisation problem considering the operational cost of power plants and greenhouse gas (GHG) emissions was formulated with multiple power system variables and its operational equal and unequal constraints. The effectiveness of ESSA was verified on standard IEEE 14-bus and IEEE 30-bus test systems without RE plants and compared with the literature. In addition, a real-time Andhra Pradesh, an Indian power system with 14-bus is simulated with RE systems. In the 14-bus test system, the cost is reduced by 89.97 $/h, the losses are reduced by 4.106 MW, the voltage deviation is reduced by 0.1276 p. u., and GHG emissions are reduced by 8407.86 lb/MWh. In the 30-bus test system, the cost was reduced by 73.1 $/h, the losses were reduced by 3.962 MW, the voltage deviation was reduced by 0.00459 p. u., and GHG emissions were reduced by 8113 lb/MWh. On the other hand, in a real-time system, the cost is reduced by 114.41 $/h, the losses are reduced by 1.177 MW, the voltage deviation is reduced by 0.1116 p. u., and GHG emissions are reduced by 75560.13 lb/MWh. The obtained results confirm that the proposed ESSA outperforms the basic SSA and other competitive meta-heuristics, namely the grasshopper optimisation algorithm (GOA), whale optimisation algorithm (WOA), and moth-flame optimisation (MFO), to solve the OPF problem. However, embedding RE plants in the OPF problem has resulted in significant reductions in operating costs and GHG emissions, which are the need of the present world with rising fossil fuel costs and increasing global warming.

223

Authors: Rahmat Hendrawan, Z. K. A. Baizal, Gia Septiana Wulandari

Constructing a multi-day travel itinerary is a notable challenge, particularly for individuals planning extended trips. This study addresses the complexity by aiming to automatically generate an optimal multi-day travel itinerary that satisfies user interests. The problem is framed as a Capacitated Vehicle Routing Problem with Time Window (CVRPTW), and user interests are shaped by attributes such as rating and cost of points of interest (POIs), travel duration, the number of POIs in the itinerary, and penalty attributes (POI penalty and time penalty). The generated solution must adhere to constraints like daily travel duration limits and the operational hours of POIs. To ensure alignment with user interests, the Multi-Utility Attribute Theory (MAUT) is employed as the fitness function. This study proposes a VRP approach utilizing the hybrid Ant Colony System (ACS) and Brainstorm optimization (BSO) algorithm (the hybrid ACS-BSO) for multi-day travel itinerary generation, addressing the Traveling Salesman Problem (TSP) approach limitations. The hybrid ACS-BSO outperforms conventional algorithms, such as Genetic Algorithm (GA), Tabu Search (TS), and Simulated Annealing (SA), across 5 sets of random POIs with an average fitness value of 0.6704. moreover, the hybrid ACS-BSO outperforms the other conventional algorithms in optimizing each attribute. In terms of travel duration attribute, the hybrid ACS-BSO generate an itinerary requiring only 6 days to visit 40 POIs, while the other algorithms need 7 days. In terms of cost and rating attributes, the hybrid ACS-BSO achieves the best fitness values compared to the others. Furthermore, the hybrid ACS-BSO outperforms the standalone algorithms (ACS and BSO) across varying numbers of POIs, but it faces a maximum 311 seconds running time for 87 POIs, indicating a time complexity weakness. Comparatively, ACS, BSO, and the hybrid ACS-BSO in VRP approach surpass their TSP counterparts, affirming the effectiveness of the VRP approach.

251

Authors: Dustakar Surendra Rao, L. Koteswara Rao, Vipparthi Bhagyaraju

The discovery of Microsoft Kinect sensors opened a new research direction to human action recognition (HAR) from videos. However, the depth maps and body postures are of noisy and less reliable in their original form. Hence, they contribute limited contribution toward the actions recognition, especially in real time scenarios. Moreover, the HAR from single data modality have several limitations. Hence, this paper proposes a multi-modal HAR framework based on a simple and effective deep learning model. The proposed model considered depth maps and body postures as input and describes each action through two newly proposed descriptors namely improved motion history image (IMHI) and spatio-temporal posture descriptor (STPD). IMHI removes undefined motion regions and STPD ensure the provision of complete Spatio-temporal motion information to the training system. Alongside, a new temporal segmentation is also proposed to ensure robustness against speed variations. Finally, different fusion rules are adapted to determine the correct action based on different policies. We conduct extensive Simulations on three benchmark standard datasets namely MSRAction3D, MAD and PKU-MMD and the average accuracy obtained is observed as 95.2333%, 91.9945% and 93.3141% respectively. Experimental results prove that the proposed framework is discriminative for the actions with similar movements.

277

Authors: Berlian Muhammad Galin Al Awienoor, Erwin Budi Setiawan

In the current phase of technological development, Netflix has become a popular platform for entertainment. Often people feel overwhelmed when choosing a movie because of the variety of genres. To overcome this problem, a recommendation system is needed that can help people find the best movie according to their preferences. In addition, Twitter was used to collect tweets related to movies, which were then processed into rating values. The crawled dataset consisted of 855 movies and 44 user reviews (including data from IMDB, rotten tomatoes, and metacritic websites), for a total of 23,130 records. This research proposed to use the switching hybrid filtering (SHF) method combined with recurrent neural network (RNN) as classification. In SHF, the emphasis on rating prediction was initiated by the content based filtering method with RoBERTa, followed by switching to item-based collaborative filtering. This situation arose because the dataset had a sparseness of 74.46%. SHF provided accurate rating prediction with an MAE of 0.0617 and an RMSE of 0.1178. Nadam optimization with optimal learning rate in RNN classification gave the best results with an accuracy of 86.11%. The research successfully developed a method that proved effective and provided positive results, contributing to the development of a recommendation system designed to assist users in choosing movies based on their preferences.

311

Authors: Ahmed Ibrahim Turki, Saad Talib Hasson

Intelligent transport systems (ITS) include a broad range of applications that require proactive strategies and predictive data driven by artificial intelligence and big data. The objective of this paper is to improve the accuracy of traffic flow prediction by utilizing a novel approach that combines feedforward neural-networks with the Quasi-Newton (QN) optimization method. The proposed method decreases the error factor based on the lagrange multiplier and Jacobian vector. This enhancement has resulted in a faster convergence during the learning process. The sample was chosen utilizing the dataset provided by the England Highway (HE) traffic monitoring systems in 2023. In order to assess the proposed model, the research findings are compared to other standard prediction techniques. As a regression model, the proposed optimized multi-layer perceptron neural network method achieves an average root-mean-squared-error of 0.143 compared to 0.319, 0.459, and 0.406 achieved by (random forest, Naïve bays, k-nearest neighbour) respectively. That is, the proposed model achieved an average percentage of improvement in prediction of are approximately 55.17%, 68.81%, and 64.78%, respectively, compared to other standard techniques. Finally, the superiority of the proposed model was evaluated by the coefficient of determination (R²) and mean-absolute-error measure, and its performance was better than other forecasting techniques as well.

333

Authors: Nagaraj Dhivya, Palaniappan Sharmila

Lung diseases have been a significant concern throughout history, necessitating early disease prediction using high-level knowledge. Deep Learning models have proven effective in diagnosing lung disorders using clinical imaging modalities like Computerized Tomography (CT) and Chest X-Ray (CXR) images. The Ensemble Deep Lung Disease Predictor (EDEPLDP) framework has been proposed for the rapid detection of various diseases using CT and CXR images. However, the U-Net model used for segmentation tasks lacks sufficient low-level localization abilities. To address this, a Semantic Location enhanced Swin Transformer-based U-Net (SLST-U-Net)+EDEPLDP model is proposed in this article. This model leverages Location Attention (LA) and De Mejora Progresiva (DMP) to enhance feature discrimination at the level of spatial information and semantic position. The Contextual Guidance Attention (CGA) method combines spatial and semantic information. The DMP enhances feature discrimination by increasing edge data inference and providing a richer depiction of the target position. The CGA reduces the semantic gap and effectively fuses spatial texture information and semantic information. The LA mechanism improves computational capacity for semantic features and precision of semantic position data, enabling retrieval of long-range contextual data in channel and geographic contexts. Additionally, Swin Transformer (ST) is added in the encoder and decoder section of U-net to increase the finer details of spatial and semantic information. Finally, the features extraction and classification part of EDEPLDP is employed to detect and classify the lung diseases. Experimental results revealed that the proposed SLST-U-Net+EDEPLDP model outperforms the CNN, E2E-DNN LungNet22, EfficientNet-SE, LDDNet and EDepLDP models with an accuracy of 94.94% and 95.42% on CXR and CT images, respectively.

365

Authors: Haider D. Albonda

The integration of 5G New Radio (5G-NR) and Wi-Fi networks in heterogeneous environments holds significant promise for delivering critical services with relatively high reliability and low latency. In this paper, we tackle the critical challenge of establishing robust multi-connectivity within heterogeneous systems integrating 5G NR and Wi-Fi systems. In this context, our focus is on optimizing a new, distributed multi-Connectivity solution, with a main emphasis on identifying the ideal base stations (BSs) and Wi-Fi access points (APs) with their respective spectrum associations. This solution is essential for ensuring low-latency and reliable communication required, particularly for mission-critical services. The simulation results demonstrate the significant advantages of the proposed algorithm compared to reference schemes. It leads to network performance improvements by enhancing system utility function, reducing service access time, and increasing packet success rates, through high-quality connectivity with relatively low interference probability. In particular, for mission-critical services, the proposed algorithm leads to a 16 % and 41% increase in effective network throughput when the offered load is 190 Mbps, which outperforms the benchmark approaches. Furthermore, it improves the utility function by approximately 0.12, and 0.19 compared to the reference approaches. This reflects a substantial improvements in success delivery rates for mission-critical services by 10 % and 18 %, respectively, when compared to the reference approaches, particularly evident at an offered load of approximately 190Mbps.

390

Authors: Mike Yuliana, Wildan Dharma Walidaniy

The previous global pandemic has shifted many of these proposed needs to remote interactions, including document authentication. The solution for remote document authentication is the use of digital signatures. However, in reality, most implementations are inefficient and even weaken already secure methods. This study presents a novel efficient digital signature method based on a multi-signature method that incorporates the edwards-curve digital signature algorithm (EdDSA) algorithm, a quick-response (QR) code, and the cryptographically secure pseudorandom number generator (CSPRNG) for key generation. This study assessed the performance efficiency of this method by running it on a large number of executions in terms of generation, signature generation, and signature generation time for three signers and a verifier. The results show that this proposed method outperforms all the existing methods that this study compared across all parameters. In key generation performance assessments, this proposed method stands out by employing CSPRNG that already been proven for its efficiency and security for cryptographic use. The usage of the multi-signature method in the EdDSA algorithm has made this proposed method superior in signature generation and signature verification performance assessment. On paper, EdDSA is the latest algorithm that has surpassed its predecessors regarding security and efficiency. By using the multi-signature method, this study further improves the signature generation and signature verification efficiency as signatures only need to be generated and verified once to represent all signers. With the performance assessment done on this proposed device, this proposed method has an average improvement of 76.27% across all parameters against the existing method. Additionally, using QR codes in the method facilitates real-life signature verification by simplifying the scanning process for verifiers. This work provides a secure and efficient solution to document file authentication.

412

Authors: K. Kowsalyadevi, N. V. Balaji

Industrial 4.0 technological breakthroughs highly impact healthcare 4.0 and enable transformative impact on the healthcare system by shifting towards efficient, patient-centric, data-driven, and robust global healthcare services. This paper presents a robust security framework; federated learning (FL) based RPL vulnerability analysis and attack detection (FRVA), for ensuring secure Healthcare 4.0. The FRVA is proposed to defend the RPL-healthcare 4.0 against multiple attacks by applying deep learning-based fuzzing and FL-enabled hybrid learning. RPL vulnerabilities are analyzed using randomly generated inputs by deep learning-based fuzzing. Further, it feeds the RPL vulnerability-rich fuzzed output dataset to the FL-hybrid learning model. The second model improved the customized local learning models using globally shared information according to FL, resulting in high learning accuracy with precise attack detection. The proposed FRVA runs the vulnerability analysis and attack detection at the edges to prolong the network lifetime with high security. Moreover, the performance of the FRVA is validated through Python-based simulations using different metrics. The simulation results demonstrate that the proposed FL-based hybrid CNN-LSTM strategy enhances the accuracy by 5.55% and 12.9%, respectively, compared with the individual CNN and LSTM methods. It also enhances the accuracy by 25.83% and 5.97% than the other conventional FL-based detection strategies.

439

Authors: Mohammed Hussein Ali, Ghanim A. Al-Rubaye

The coding of channels is a core part of any communication network. Future wireless systems will need high performance codes with low complexity encoding and decoding in order to meet a wide variety of requirements, from running in highly dependable situations with short information messages and low code rates to running in high throughput scenarios with lengthy messages and high code rates. The investigation of polar codes in fading channels has important implications in the context of implementing polar codes in wireless communications. A significant development in the fields of information and coding theory, polar code allows for the flexible adjustment of code rates and block lengths without compromising error correction efficiency. Recently, the Third Generation Partnership Project (3GPP) New Radio (NR) group has supported polar codes to provide channel coding in the physical control channels of the 5G, wireless standard. Channel coding, which exists inside the physical layer, is essential in determining the latency and dependability of a communication system. However, error correction performance diminishes with shorter message durations. The polar encoding scheme can achieve channel capacity because it is based on the practical use of channel polarization operation. In this research, a polar code is included in the proposed system for data transmission over various multipath fading channel models (Nakagami-m, Rayleigh, and Lognormal fading channels) in the presence of AWGN, and in this design, the Successive Cancellation List (SCL) decoding algorithms for polar code are used in order for the suggested NR-polar QAM-OFDM system to perform better than the uncoded QAM-OFDM system. According to Monte Carlo simulations, QAM constellation NR-Polar-COFDM using LLRs computed due to the normal distribution over different types of channel PDFs performs better than the uncoded QAM-OFDM system in terms of bit error rat BER. To further enhance the system performance, CRC bits are used to enhance the efficiency of SCL decoding by improving distance characteristics. The simulation results show that BER is approximately 10-6 at SNR of 1dB in the AWGN channel, and in the multipath fading channel BER approaches 10-6 at SNR of 22.5dB in the Rayleigh and Nakagami-m fading channels, while in the lognormal fading channel at SNR of 14dB for short information block lengths at code rates (1/3) in the uplink and downlink control channels. The findings additionally confirm the superiority of Polar coding in the situation of transmitting information with short block lengths in the AWGN compared to the NR-LDPC-OFDM.

464

Authors: Duvva Naresh Kumar, M. Kezia Joseph

Segmentation of Lung Region is an important step in the clinical diagnosis of lung diseases like Pneumonia, Tuberculosis, and COVID-19. However, most the methods employed Deep learning algorithms without providing any additional information at the segmentation phase. Hence this paper proposes a new segmentation Mechanism based on the Fuzzy C-Means (FCM) and different features. A new Variant of FCM called as Dual FCM is introduced which forms an objective function based on two different Sub-Objective Functions (SOFs). The first SOF segments the images through gray pixel intestines while the second SOF used features for segmentation. Two different features are derived namely morphological features and gradient features from every CXR image. Finally, the segmented results at individual SOFs are fused to get a final segmented result. Extensive experiments have been carried out over the proposed approach through two publicly available datasets namely Montgomery County (MC) Dataset and Shenzhen Dataset. The performance is assessed through two evaluation measures namely Jaccard Index (JI) and Dice Coefficient (DC) and the average obtained is noticed as 92.985% and 95.295% respectively. Further the proposed segmentation mechanism is compared with several past lung image segmentation approaches. The approximate improvement in the JI and DC from past methods is noticed as 4.2210% and 2.3150% respectively.

489

Authors: Md. Zahid Hassan, Shahid Md. Asif Iqbal, Asaduzzaman

Reducing control packets, especially in proactive routing protocols, needed to establish routes can lower network overhead in Mobile Ad-hoc Networks (MANETs). In Optimized Link State Routing (OLSR) protocol, each Multipoint Relay (MPR) node propagates Topology Control (TC) messages to advertise neighbor information. However, OLSR controls the TC messages by reducing the number of MPR nodes. In this study, we propose an efficient MPR node selection mechanism to reduce the TC message volume leading to a minimized routing overhead. Each node selects the lowest cost node heuristically from its first hop neighbors as the MPR node for any destination. The same MPR node can be selected for multiple destinations if it costs the lowest for each destination node. The selection technique is realized by modifying only the default OLSR TC and Hello messages. The proof-of-concept implementation in the NS3 simulator reveals that the proposed methodology reduces the routing overhead by selecting around 55%, 28% and 49% (on average) fewer MPR nodes compared to the traditional OLSR, SSTB and M-OLSR protocol respectively, without negotiating packet delivery ratio, throughput and delay.

511

Authors: Nandeesha R, Somashekar K

The inherent issue associated with any digital image is its underlying redundancy and large dimension, which require a huge amount of storage space and higher bandwidth for transmission over a wireless channel. This factor has inspired the researchers to arrive at the optimal resolution that compresses digital images based on their content type and also achieves better visual quality. The proposed work aims at compressing the images according to the contrast variations by hybridizing discrete wavelet transform (DWT) and machine learning (ML) techniques for quality reconstruction. It includes two operation phases. In the initial phase, the ML model is first trained with a training set containing various image data. The second stage is mainly subjected to performing image compression and decompression by employing the joint approach of the transform technique and a trained ML model. Compression involves the application of two-level DWT to the image, then each sub-band is divided into non-overlapping blocks, and a decision is made for each block based on the block variance. Each block is subjected to the extraction of structural features, that are next Huffman encoded, and finally, these features are employed in image reconstruction. Reconstruction involves querying each block feature with a pre-trained K-nearest neighbour (K-NN) model. Experiments have been conducted, and the effectiveness of the proposed system is analyzed in terms of peak signal-to-noise ratio (PSNR), mean squared error (MSE), structural similarity index measure (SSIM), and compression ratio (CR). Comparative assessments against prior methods reveal superior visual quality metrics and computation time, with custom real-time image testing confirming its outperformance over current approaches. Notably, the proposed work achieves a PSNR of 53.2093 dB for the “Saras” image, SSIM values of 0.9997 for “Cameraman,” and a nearly 5% increase in PSNR levels compared to existing techniques, substantiating its efficacy in content-based image compression.

544

Authors: Amiq Fahmi, Diana Purwitasari, Surya Sumpeno, Mauridhi Hery Purnomo

Real-world datasets frequently compromise considerably on noise, resulting in the emergence of outlier data. Detecting and removing outliers in large and imbalanced datasets is a challenging and exciting study in machine learning, especially in healthcare, for accurate prediction. Therefore, it is essential to handle outliers properly, as their presence in classification datasets leads to more difficult, inaccurate, and lower predictive modelling performance. The study proposes methods to enhance prediction accuracy in an imbalanced real-world health dataset of dengue infection cases. First, use a two-layer ensemble method called IsFLOF, which involves an isolation forest (IsF) and a local outlier factor (LOF) to find and accurately eliminate global and local outliers. This approach overcomes the limitations of the IsF algorithm, which is only sensitive to global outliers but vulnerable to local outliers, while LOF excels in local outlier detection but has high complexity. Second, once a dataset with correctly measured value distributions was obtained by eliminating outliers, a resampling process was conducted to prevent prediction bias caused by imbalanced instance data in the multi-class setting. Subsequently, insignificant features were filtered out to further refine the dataset. In the end, eight machine learning algorithms are used to test the robustness and effectiveness of the proposed method. The experimental results showed that the AdaBoost classifier, combined with selected features from the Fast Correlation-Based Filter (FCBF), achieved 93.5% and 95.1% accuracy in training and testing, respectively. In a more distant context, the proposed method is tested and compared with recent methods, including using a public dataset of imbalanced hypothyroid cases. It showed higher and more acceptable prediction accuracy than the original and synthetic data.

572

Authors: Shwetha K S, Chandramouli H

Huge demand on business intelligence applications over large volume of enterprise data has resulted in rapid adoption of High performance data analytics. Hadoop based high performance computing environment are optimized for large files. Data centric execution with localization of computing proximal to data provided higher performance for large files. But for small files, the performance reduced and overhead increased due to the way Hadoop handles files. Resource allocation and scheduling policies of Hadoop has to be improvised to handle small files. This work proposes an integrated data, task, and resource management technique to speed up processing of small files in Hadoop based high performance computing environment. As part of data management, the data placement is made dynamic to access frequency and inherent data semantics. As part of task management, tasks were grouped based on fine grained data semantics correlation they process. As part of resource management, data blocks or storage nodes are replicated to improve the access latency with minimal cost overheads. With this integrated management, the proposed solution is able to increase the speedup of handling small files by at least 8.7% compared to most recent works on individual management of data, task and resources.

595

Authors: Yaddarabullah, Yoga Alviando, Dewi Lestari, Aedah Binti Abd Rahman, Amna Saad

The purpose of this study is to introduce a novel approach to predict induction motor frequency adjustments in Air Handling Units (AHU). This is essential as traditional methods have frequently been unable to effectively address the complex seasonal and stochastic fluctuations that are inherent in these environments. To overcome this challenge, the research focuses on utilizing experimental Chen’s Fuzzy Time Series model, specifically designed to incorporate temporal and seasonal patterns into the predictive analysis. Various predictive models, including Seasonal Autoregressive Integrated Moving Average (SARIMA), Holt-Winters Exponential Smoothing (HWES), neural network (NN) based ensemble model, hybrid of artificial neural network (ANN) and SARIMA, and Seasonal Autoregressive Integrated Moving Average with Exogenous factors (SARIMAX), are compared to determine the most effective model in optimizing AHU induction motor frequency. Results indicate that the modified Chen's Fuzzy Time Series model demonstrated high efficacy with an R-squared value of 0.9945 in a one-hour time interval in the seasonal pattern, showing an almost perfect fit between predicted outcomes and actual data compared to other prediction models. Furthermore, the modified Chen's model achieved a Mean Absolute Percentage Error (MAPE) of 2.41% and a Root Mean Square Error (RMSE) of 0.72, significantly outperforming other models in predictive accuracy and reliability. The modified Chen's model showed an efficiency improvement of 86% in MAPE and 87% in RMSE compared to other prediction models.

621

Authors: Eman A. Radhi, Mohammed Y. Kamil

Breast ultrasound is an important technology for detecting breast lesions, but it faces a challenge in the form of speckle noise that negatively affects the quality of images. Effective methods are needed to eliminate this noise without compromising the image's fine edges and important features. In this work, we present a cutting-edge methodology based on a deep understanding of the dynamics of breast ultrasound images and the challenges of speckle noise. The method uses two separate stages of processing with two nonlinear filters. The first filter, the anisotropic diffusion filter, smooths out edges and boosts image contrast by lowering noise while keeping the tissue's structure. In the second stage, the SRAD filter is applied to remove residual noise and refine the image, increasing its clarity and improving the ability to visualize subtle lesions. This improved approach was evaluated using a comprehensive set of quantitative indices. The results confirmed the significant performance improvement provided, with the lowest MSE value of 0.2 and the highest PSNR of 59.3. Also, it reached the optimal value in the UQI and SSI, which indicates the robustness of the method in maintaining the quality of the structural image. The results confirm the value of the proposed approach and herald its promising potential for improving medical diagnostic results using breast ultrasound.

645

Authors: M. Gokiladevi, Sundar Santhoshkumar

Chronic Kidney Disease (CKD) is a progressive condition that may cause kidney failure, so earlier diagnosis is critical for proper management. The condition has a large fatality, particularly in developing nations. CKD often remains unnoticed because there is no obvious earlier-stage symptom. Meanwhile, earlier diagnosis and on-time clinical intervention are essential for reducing the disease progression. Detecting CKD using deep learning (DL) methods and feature selection (FS) could be a useful application of artificial intelligence (AI) in healthcare. DL algorithms can provide cost-effective and efficient computer-aided diagnoses (CAD) to assist clinicians in accomplishing earlier CKD recognition. Therefore, this study develops an automated CKD detection using Henry Gas Optimization Algorithm with DL (CKDD-HGSODL) approach. The drive of the CKDD-HGSODL approach is to classify and detect the presence of CKD utilizing FS and hyperparameter tuning strategies. In the presented CKDD-HGSODL technique, min-max scaling can be used to normalize the input data. In addition, the CKDD-HGSODL technique utilizes the HGSO model for selecting optimal features. For the CKD detection process, an attention-based gated recurrent unit (AGRU) model can be utilized. At last, the slime mould algorithm (SMA) can be utilized for the optimal hyperparameter selection of the AGRU approach that aids in improving the classifier results. To validate the outcome of the CKDD-HGSODL approach, a comprehensive simulation value is made on the benchmark CKD database. The obtained performances depict the enhanced detection results of the CKDD-HGSODL approach on CKD diagnosis.

670

Authors: Purba Daru Kusuma, Meta Kallista

Many recent swarm-based metaheuristics are trapped in the exploitation of the highest quality as the main or the only reference and the neighbourhood search with the reduction of local search space during the iteration. Regarding to this issue, this paper introduces a novel metaheuristic called swarm space hopping algorithm (SSHA). SSHA consists of three searches. First, a directed search toward the highest quality is performed. Second, the directed search toward the resultant of better agents or away from the other agent is performed. Third, the arithmetic crossover between the agent and a randomized solution in the first half or second half of space is performed. In this work, three evaluations are performed to assess the performance of SSHA. The first evaluation is the benchmark evaluation to compare the performance of SSHA with other recent metaheuristics: northern goshawk optimization (NGO), zebra optimization algorithm (ZOA), clouded leopard optimization (CLO), osprey optimization algorithm (OOA), and total interaction algorithm (TIA). The result exhibits that SSHA is better than NGO, ZOA, CLO, OOA, and TIA in 21, 20, 17, 17, and 21 functions. In the second evaluation, the individual search evaluation to compare the contribution between the first and second searches is performed, with the result that the second search outperforms the first search. The third evaluation is performed to assess the contribution of the third search in the optimization process, and the result shows that the contribution of the third search is significant only in three functions.

700

Authors: Muhammad Rivai, Dava Aulia, Sheva Aulia

Exhaled breath analysis comprises chemical compounds that can be utilized for diagnostic purposes, including asthma detection. An electronic nose can be offered as a means of monitoring patient circumstances. A significant problem often occurs when determining the appropriate number of gas sensors while maintaining high accuracy. The firefly algorithm (FA) is very effective because of its exploratory capabilities, presents theories that are easy to understand and has relatively fewer parameters. This study aims to reduce and determine the appropriate number of gas sensors for an electronic nose in differentiating healthy and asthmatic subjects using the FA and exhaled breath analysis. The experimental results indicate that the FA provides only four gas sensors that still maintain high performance. The convolutional neural network model was favored for its ability to classify the entire asthma dataset, making it the best machine learning model for the electronic nose, with an accuracy of 97.8%.

728

Authors: Yunifa Miftachul Arif, Muhammad Khairul Amin Imron, Ajib Hanani, Hani Nurhayati, Ahmad Fahmi Karami, Mokhamad Amin Hariyadi

The study explores an Ethereum blockchain-based data-sharing system for a computer assembly simulation game, emphasizing the relationship between gas limits and transaction speeds. The research integrates smart contracts for secure data storage of scores and player profiles. A significant challenge identified was the complexity of blockchain's variable transaction speeds for the average user. The research investigated how different gas limits affected transaction times, with experiments conducted across three networks. Results show that a gas limit of 200,000 to 300,000 resulted in transaction speeds of approximately 30 seconds. Increasing the gas limit to 400,000 to 500,000 reduced transaction times to 15-30 seconds, while a limit of 600,000 to 700,000 led to speeds below 15 seconds. These findings suggest a direct correlation between higher gas limits and quicker transaction validations. The research concludes that investing in higher gas can significantly reduce transaction times, presenting a trade-off between cost and speed in blockchain data-sharing for educational simulation media.