rss_2.0Sciendo RSS Feed for International Journal on Smart Sensing and Intelligent Systemshttps://sciendo.com/journal/IJSSIShttps://www.sciendo.comhttps://sciendo-parsed.s3.eu-central-1.amazonaws.com/6471fa69215d2f6c89db759f/cover-image.jpghttps://sciendo.com/journal/IJSSIS140216https://sciendo.com/article/10.2478/ijssis-2024-0031<abstract> <title style='display:none'>Abstract</title> <p>Automation is now heavily used in the industry. Human effort is much reduced by implementation of an automation system that monitors and signals system faults that may arise. The number of devices that can be connected to the Internet of Things (IoT) is growing every day. With the aid of the IoT model, a hardware prototype will be developed that will automatically monitor industrial parameters such as humidity, temperature, fire, smoke, pressure, and sense motion, as well as generate alerts and alarms. Each parameter is continuously monitored using an application called Blynk. This proposed work will use IoT to provide information remotely about the current weather and gas concentrations in the plant via email. The collected information can also be used for predictive analysis. The idea behind this proposed work is to accumulate and visualize data about the various parameters utilized in industries by enabling IoT.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00312024-09-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0021<abstract> <title style='display:none'>Abstract</title> <p>Large language models (LLMs) have transformed open-domain abstractive summarization, delivering coherent and precise summaries. However, their adaptability to user knowledge levels is largely unexplored. This study investigates LLMs’ efficacy in tailoring summaries to user familiarity. We assess various LLM architectures across different familiarity settings using metrics like linguistic complexity and reading grade levels. Findings expose current capabilities and constraints in knowledge-aware summarization, paving the way for personalized systems. We analyze LLM performance across three familiarity levels: none, basic awareness, and complete familiarity. Utilizing established readability metrics, we gauge summary complexity. Results indicate LLMs can adjust summaries to some extent based on user familiarity. Yet, challenges persist in accurately assessing user knowledge and crafting informative, comprehensible summaries. We highlight areas for enhancement, including improved user knowledge modeling and domain-specific integration. This research informs the advancement of adaptive summarization systems, offering insights for future development.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00212024-08-16T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0024<abstract> <title style='display:none'>Abstract</title> <p>This study explores the fusion of artificial intelligence (AI) and machine learning (ML) methods within anti–money laundering (AML) frameworks using data from the US Treasury’s Financial Crimes Enforcement Network (FinCEN). ML and deep learning (DL) algorithms—such as random forest classifier, elastic net regressor, least absolute shrinkage and selection operator (LASSO) regression, gradient boosting regressor, linear regression, multilayer perceptron (MLP) classifier, convolutional neural network (CNN), random forest regressor, and K-nearest neighbor (KNN)—were used to forecast variables such as state, year, and transaction types (credit card and debit card). Hyperparameter tuning through grid search and randomized search was used to optimize model performance. The results demonstrated the efficacy of AI/ML algorithms in predicting temporal, spatial, and industry-specific money-laundering patterns. The random forest classifier achieved 99.99% average accuracy in state prediction, while the gradient boosting regressor and random forest classifier excelled in predicting year and state simultaneously, and credit card transactions, respectively. MLP and CNN showed promise in the context of debit card transactions. The gradient boosting regressor performed competitively with low mean squared error (MSE) (2.9) and the highest <italic>R</italic>-squared (<italic>R</italic>²) value of 0.24, showcasing its pattern-capturing proficiency. Logistic regression and random forest classifier performed well in predicting credit card transactions, with area under the receiver operating characteristic curve (ROC_AUC) scores of 0.55 and 0.53, respectively. For debit card prediction, MLP achieved a precision of 0.55 and recall of 0.42, while CNN showed a precision of 0.6 and recall of 0.54, highlighting their effectiveness. The study recommends interpretability, hyperparameter optimization, specialized models, ensemble methods, data augmentation, and real-time monitoring for improved adaptability to evolving financial crime patterns. Future improvements could include exploring the integration of blockchain technology in AML.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00242024-08-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0023<abstract> <title style='display:none'>Abstract</title> <p>This study aims to compare deep learning explainability (DLE) with explainable artificial intelligence and causal artificial intelligence (Causal AI) for fraud detection, emphasizing their distinct methodologies and potential to address critical challenges, particularly in finance. An empirical evaluation was conducted using the Bank Account Fraud datasets from NeurIPS 2022. DLE models, including deep learning architectures enhanced with interpretability techniques, were compared against Causal AI models that elucidate causal relationships in the data. DLE models demonstrated high accuracy (95% for Model A and 96% for Model B) and precision (97% for Model A and 95% for Model B) but exhibited reduced recall (98% for Model A and 97% for Model B) due to opaque decision-making processes. By contrast, Causal AI models showed balanced but lower performance with accuracy, precision, and recall, all at 60%. These findings underscore the need for transparent and reliable fraud detection systems, highlighting the trade-offs between model performance and interpretability. This study addresses a significant research gap by providing a comparative analysis of DLE and Causal AI in the context of fraud detection. The insights gained offer practical recommendations for enhancing model interpretability and reliability, contributing to advancements in AI-driven fraud detection systems in the financial sector.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00232024-08-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0022<abstract> <title style='display:none'>Abstract</title> <p>Vehicular traffic significantly contributes to economic growth but generates frictional noise that impacts urban environments negatively. Road traffic is a primary noise source, causing annoyance and interference. Traditional regression models predict two-dimensional (2D) noise maps, but this study explores the impact and visualization of noise using 2D and three-dimensional (3D) GIS (Geospatial Information Systems) functionalities. Two models were assessed: (i) a 2D noise model for roads and (ii) a 3D noise model for buildings, utilizing limited noise samples. Combining these models produced a comprehensive 3D noise map. Machine learning (ML) models—artificial neural network (ANN), random forest (RF), and support vector machine (SVM)—were evaluated using performance measures: correlation (R), correlation coefficient (R²), and root mean square error (RMSE). ANN outperformed others, with RF showing better results than SVM. GIS was applied to enhance the visualization of noise maps, reflecting average traffic noise levels during weekday mornings and afternoons in the study area.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00222024-08-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0025<abstract> <title style='display:none'>Abstract</title> <p>The study focuses on intelligent driving, emphasizing the importance of recognizing nearby vehicles and estimating their positions using visual input from a single image. It employs transfer learning techniques, integrating deep convolutional networks’ features into a modified CenterNet model for six-degrees-of-freedom (6DoF) vehicle position estimation. To address the vanishing gradient problem, the model incorporates simultaneous double convolutional blocks with skip connections. Utilizing the ApolloCar3D dataset, which surpasses KITTI in comprehensiveness, the study evaluates pretrained models’ performance using mean average precision (mAP). The recommended model, Center-DenseNet201, achieves a mAP of 11.82% for relative translation thresholds (A3DP-Rel) and 39.92% for absolute translation thresholds (A3DP-Abs). These findings highlight the effectiveness of pretrained models in the modified architecture, enhancing vehicle posture prediction accuracy from single images. The research contributes to autonomous vehicle development, fostering safer and more efficient navigation systems in real-world traffic scenarios.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00252024-08-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0027<abstract> <title style='display:none'>Abstract</title> <p>Money laundering has been a global issue for decades. The ever-changing technology landscape, digital channels, and regulations make it increasingly difficult. Financial institutions use rule-based systems to detect suspicious money laundering transactions. However, it suffers from large false positives (FPs) that lead to operational efforts or misses on true positives (TPs) that increase the compliance risk. This paper presents a study of convolutional neural network (CNN) to predict money laundering and employs SHapley Additive exPlanations (SHAP) explainable artificial intelligence (AI) method to explain the CNN predictions. The results highlight the role of CNN in detecting suspicious transactions with high accuracy and SHAP’s role in bringing out the rationale of deep learning predictions.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00272024-07-24T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0026<abstract> <title style='display:none'>Abstract</title> <p>Smart home applications are ubiquitous and have become popular because of the overwhelming use of the Internet of Things (IoT) and artificial intelligence (AI). Living smart with automation and integrated AI-IoT has become more affordable as home automation technologies have matured. In addition, the Internet of Everything (IoE), which involves the interconnection of humans, businesses, and intelligent objects, has the potential to reshape various industries. However, the rising energy cost and demand have led numerous organizations to determine smart ways to monitor, control, and save energy. Hence, this study suggests AI-Enabled Internet of Everything Services (AI-IoES) for efficient smart home energy management. The data have been taken from the Open Smart Home IoT//Energy Dataset for analyzing the energy consumption of home appliances. This paper presents an IoT sensor for energy management to track and control specific loads in smart homes. The deep neural network (DNN) is built for secure demand-side management (DSM) in an IoT-assisted smart grid and trained on the extracted feature from electricity consumption information gathered using an IoT sensor. The system is established with real-time monitoring and a user interface for remote control and access. The experimental outcome demonstrates that the suggested AI-IoES system increases the user experience by 98.9%, energy efficiency ratio (EER) by 97.8%, and accuracy ratio by 97.2%, and reduces energy consumption by 19.2% compared with other existing methods.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00262024-07-23T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0020<abstract> <title style='display:none'>Abstract</title> <p>Stepper motors are extensively used in various automated systems, notably hybrid stepper motors (HSMs), renowned for their high torque density. Optimizing and improving their performance and torque density is highly beneficial. The first step in motor optimization is to perform a sensitivity analysis to identify the parameters most significantly affecting motor performance and torque density. Using Ansys Maxwell, a finite element method (FEM) software, we modeled and analyzed a commercially available HSM. After validating the computer model’s accuracy, we investigated the motor’s sensitivity to various parameters and discussed their effects on performance. Our findings highlight the significance of parameters such as air gap, magnet diameter, wire-turn numbers in the coils, and the shapes of the rotor and stator teeth in influencing HSM sensitivity—conversely, factors like magnet height and stator pole thickness exhibit negligible impact on motor performance.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00202024-07-20T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0019<abstract> <title style='display:none'>Abstract</title> <p>The current assessment of tea quality is considered subjective. This study aims to develop a portable electronic nose to assess the aroma of tea dregs objectively by relying on the aromatic capture process through sensors and using multilayer perceptron (MLP). A MLP with some hyperparameter variations is used and compared with five machine-learning classifiers. The classification using MLP model with ReLU activation function and 3 hidden layers with 100 hidden nodes resulted in the highest accuracy of 0.8750 ± 0.0241. The MLP model using ReLU activation function is better than Sigmoid while increasing the number of hidden layers and hidden nodes does not necessarily enhance its performance. In the future, this research can be improved by adding sensors to the portable electronic nose, increasing the number of datasets used, and using ensemble learning or deep learning models.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00192024-07-20T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0018<abstract> <title style='display:none'>Abstract</title> <p>The rapid growth of natural language processing (NLP) applications, such as text summarization, speech recognition, information extraction, and machine translation, has led to the development of structured query language (SQL) for extracting information from structured data. However, due to limited resources, converting Natural Language (NL) queries to SQL in Bengali is challenging. This article proposes an unsupervised machine learning model to find semantically Bengali closed words that can generate SQL from NL queries in Bengali. The main objective of the proposed system is to provide support in the creation of patient-oriented explanations and educational resources by simplifying intricate medical terminology. The major findings of the proposed system are as follows: The use of machine translation in the field of medicine facilitates the dissemination of healthcare information to a diverse international audience and improves the performance of entity recognition tasks, including the identification of medical conditions, drugs, or procedures within clinical notes or electronic health data. This system allows a naive user to extract health-related information from a healthcare-structured database without any knowledge of SQL. The system accepts a query and generates a response according to the query in Bengali language. Query tokenization and stop word removal are carried out in the preprocessing stage, and unsupervised machine learning techniques are implemented to process the input query sentence. Tokenized words are converted into vectors using the skip-gram model, with noise-contrastive estimation (NCE) applied to discriminate between actual and irrelevant words. Stochastic gradient descent (SGD) optimizes the model by randomly choosing a small amount of data from the dataset and using cosine similarity to measure closer words. The semantically closer words are found using an unsupervised learning method to generate the SQL.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00182024-06-12T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0014<abstract> <title style='display:none'>Abstract</title> <p>Deep learning has changed the approach of urban environmental risk assessment and management. These methods enable solid models for large data sets, enabling early identification, prediction, and description of environmental risks. The current work analyses the advances in deep learning for urban environmental hazard assessments and disaster studies to provide monitoring, management, and mitigation measures. It reports the improvement in self-supervised learning, transformer architectures, persistent learning, attention mechanisms, adversarial robustness, associated learning, meta-learning, and multimodal learning within the domain of urban environmental hazard analysis. These approaches allow the creation of robust models for handling vast data volumes, facilitating early detection, prediction, and characterisation of diverse environmental threats. This trends analysis for urban applications will bring insights for connecting deep-learning models for effective and proactive approaches to tackle urban environmental hazards and disasters.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00142024-05-23T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0016<abstract> <title style='display:none'>Abstract</title> <p>The evolving cyber-attack landscape demands flexible and precise protection for information and networks. Network anomaly detection (NAD) systems play a crucial role in preventing and detecting abnormal activities on the network that may lead to catastrophic outcomes when undetected. This paper aims to provide a comprehensive analysis of NAD using unsupervised learning (UL) methods to evaluate the effectiveness of such systems. The paper presents a detailed overview of several UL techniques, lists the current developments and innovations in UL techniques for network anomaly and intrusion detection, and evaluates 13 unsupervised anomaly detection algorithms empirically on benchmark datasets such as NSL-KDD, UNSW-NB15, and CIC-IDS 2017 to analyze the performance of different classes of UL approaches for NAD systems. This study demonstrates the effectiveness of NAD algorithms, discusses UL approaches' research challenges, and unearths the potential drawbacks in the current network security environment.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00162024-05-19T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0015<abstract> <title style='display:none'>Abstract</title> <p>This paper proposes and evaluates a novel control approach for trajectory tracking, stability enhancement, and vibration reduction of a flexible joint robot (FJR). The FJR is a 2-degree-of-freedom underactuated nonlinear system that is challenging to control due to vibration, underactuation, uncertainties, and external disturbances. The control objectives are high trajectory tracking performance together with suppressing vibration. The proposed control approach uses integral sliding mode control (ISMC) combined with a barrier function based on back-stepping. This ensures robust and smooth performance by eliminating the reaching phase where sliding mode control (SMC) is typically not robust. Robustness is thus guaranteed from the start. The FJR is modeled as a 4th-order system using Lagrangian mechanics and decomposed into two 2nd-order subsystems for control design. Sliding variables are defined for each subsystem. Using these variables, the proposed control achieves robust trajectory tracking, stability, and vibration reduction. Numerical simulations in MATLAB validate the superior performance of the proposed ISMC-barrier function control compared to conventional ISMC for the FJR. This novel control approach addresses the challenges of controlling this FJR.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00152024-05-04T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0008<abstract> <title style='display:none'>Abstract</title> <p>Parkinson's disease (PsD) is a prevalent neurodegenerative malady, which keeps intensifying with age. It is acquired by the progressive demise of the dopaminergic neurons existing in the substantia nigra pars compacta region of the human brain. In the absence of a single accurate test, and due to the dependency on the doctors, intensive research is being carried out to automate the early disease detection and predict disease severity also. In this study, a detailed review of various artificial intelligence (AI) models applied to different datasets across different modalities has been presented. The emotional intelligence (EI) modality, which can be used for the early detection and can help in maintaining a comfortable lifestyle, has been identified. EI is a predominant, emerging technology that can be used to detect PsD at the initial stages and to enhance the socialization of the PsD patients and their attendants. Challenges and possibilities that can assist in bridging the differences between the fast-growing technologies meant to detect PsD and the actual implementation of the automated PsD detection model are presented in this research. This review highlights the prominence of using the support vector machine (SVM) classifier in achieving an accuracy of about 99% in many modalities such as magnetic resonance imaging (MRI), speech, and electroencephalogram (EEG). A 100% accuracy is achieved in the EEG and handwriting modality using convolutional neural network (CNN) and optimized crow search algorithm (OCSA), respectively. Also, an accuracy of 95% is achieved in PsD progression detection using Bagged Tree, artificial neural network (ANN), and SVM. The maximum accuracy of 99% is attained using K-nearest Neighbors (KNN) and Naïve Bayes classifiers on EEG signals using EI. The most widely used dataset is identified as the Parkinson's Progression Markers Initiative (PPMI) database.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00082024-03-12T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0012<abstract> <title style='display:none'>Abstract</title> <p>The paper reviews the milestones and various modern-day approaches in developing phonocardiogram (PCG) signal analysis. It also explains the different phases and methods of the Heart Sound signal analysis. Many physicians depend heavily on ECG experts, inviting healthcare costs and ignorance of stethoscope skills. Hence, auscultation is not a simple solution for the detection of valvular heart disease; therefore, doctors prefer clinical evaluation using Doppler Echo-cardiogram and another pathological test. However, the benefits of auscultation and other clinical evaluation can be associated with computer-aided diagnosis methods that can help considerably in measuring and analyzing various Heart Sounds. This review covers the most recent research for segmenting valvular Heart Sound during preprocessing stages, like adaptive fuzzy system, Shannon energy, time-frequency representation, and discrete wavelet distribution for analyzing and diagnosing various heart-related diseases. Different Convolutional Neural Network (CNN) based deep-learning models are discussed for valvular Heart Sound analysis, like LeNet-5, AlexNet, VGG16, VGG19, DenseNet121, Inception Net, Residual Net, Google Net, Mobile Net, Squeeze Net, and Xception Net. Among all deep-learning methods, the Xception Net claimed the highest accuracy of 99.43 + 0.03% and sensitivity of 98.58 + 0.06%. The review also provides the recent advances in the feature extraction and classification techniques of Cardiac Sound, which helps researchers and readers to a great extent.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00122024-03-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0013<abstract> <title style='display:none'>Abstract</title> <p>In leukemia diagnosis, automating the process of decision-making can reduce the impact of individual pathologists' expertise. While deep learning models have demonstrated promise in disease diagnosis, combining them can yield superior results. This research introduces an ensemble model that merges two pre-trained deep learning models, namely, VGG-16 and Inception, using transfer learning. It aims to accurately classify leukemia subtypes using real and standard dataset images, focusing on interpretability. Therefore, the use of Local Interpretable Model-Agnostic Explanations (LIME) is employed to achieve interpretability. The ensemble model achieves an accuracy of 83.33% in binary classification, outperforming individual models. In multi-class classification, VGG-16 and Inception reach accuracies of 83.335% and 93.33%, respectively, while the ensemble model reaches an accuracy of 100%.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00132024-03-06T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0010<abstract> <title style='display:none'>Abstract</title> <p>The non-orthogonal multiple access (NOMA) multiple access technique, due to its non-orthogonality and providing access to users together, which have the same frequency and time resource, made it a front runner to meet the need of high traffic requirements networks. In this paper, a downlink, NOMA, and cooperative NOMA (CNOMA) are compared with varying different parameters: source transmit power, user transmit power, and power allocation for achievable sum rates. Simulation results show that the CNOMA achieves a higher sum rate as compared to NOMA for all the parameters.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00102024-02-27T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0009<abstract> <title style='display:none'>Abstract</title> <p>Temperature measurement and transmission of a signal safely to the control room for further processing is important for the process industry. In this paper, a modified head-mounted temperature measurement system using a thermocouple with opto-isolation has been developed. Here, the thermocouple is connected to the terminals, mounted on the ceramic base in the head of the thermo-well. It consists of two signal conditioners for thermocouple and AD590, both signal conditioning outputs applied to a summer circuit. The output of the summer circuit which is in the range of 1.73–3.43V, adjusted by a signal conditioning circuit, is applied to the middle electrode of Mach-Zehnder interferometer (MZI). MZI produces normalized optical signals according to the variations in temperature. These optical signals are then transmitted to the control room safely in the inflammable process industry. The transmitted signals are demodulated in the control room and then sent to the PC through an Opto-isolator circuit and DAS card. The necessary theory as well as mathematical equation has been derived. The experimental and simulation results are reported here.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00092024-02-27T00:00:00.000+00:00https://sciendo.com/article/10.2478/ijssis-2024-0007<abstract> <title style='display:none'>Abstract</title> <p>Condition monitoring of oil-immersed in-service transformers to facilitate preventive maintenance is still a challenge. Monitoring of 2-Furfuryldehyde (2-FAL), released in the transformer oil as a result of paper insulation degradation, and moisture ingress can provide insight into the health of the insulation of transformers. Since 2-FAL and moisture are high dielectric constant contamination, capacitive sensor-based detection is a potential solution. A novel Inter digital Capacitive (IDC) sensor is reported in this paper to measure the concentration of 2-FAL and moisture uses Zinc Oxide (ZnO) as a sensing film. The sensor shows good sensitivity, approximately linear characteristics, and low characteristic drift.</p> </abstract>ARTICLEtruehttps://sciendo.com/article/10.2478/ijssis-2024-00072024-02-27T00:00:00.000+00:00en-us-1