Forthcoming and Online First Articles

International Journal of Intelligent Engineering Informatics

International Journal of Intelligent Engineering Informatics (IJIEI)

Forthcoming articles have been peer-reviewed and accepted for publication but are pending final changes, are not yet published and may not appear here in their final order of publication until they are assigned to issues. Therefore, the content conforms to our standards but the presentation (e.g. typesetting and proof-reading) is not necessarily up to the Inderscience standard. Additionally, titles, authors, abstracts and keywords may change before publication. Articles will not be published until the final proofs are validated by their authors.

Forthcoming articles must be purchased for the purposes of research, teaching and private study only. These articles can be cited using the expression "in press". For example: Smith, J. (in press). Article Title. Journal Title.

Articles marked with this shopping trolley icon are available for purchase - click on the icon to send an email request to purchase.

Online First articles are published online here, before they appear in a journal issue. Online First articles are fully citeable, complete with a DOI. They can be cited, read, and downloaded. Online First articles are published as Open Access (OA) articles to make the latest research available as early as possible.

Open AccessArticles marked with this Open Access icon are Online First articles. They are freely available and openly accessible to all without any restriction except the ones stated in their respective CC licenses.

Register for our alerting service, which notifies you by email when new issues are published online.

International Journal of Intelligent Engineering Informatics (30 papers in press)

Regular Issues

  • Fall Detection in Elderly People: Impact of Sensor Position and Feature Selection   Order a copy of this article
    by Sabri Altunkaya 
    Abstract: The most predictive sensors, sensor positions, and features for detecting falls early in elderly people were investigated using feature dataset from three-axis accelerometers placed on the head, pelvis, right and left shank, and pressure-sensing insoles. A feature database containing records of 100 older people (76 non-fallers and 24 fallers) was used. The three different feature selection algorithm was used, and most predictive feature vector obtained. An SVM classification model was developed for each feature vector. As a result, the best classification accuracy was observed for features determined by the feature selection algorithm using the chi-square test. In the classification using ten features of the acceleration signal recorded from the head, 80.17% (
    Keywords: elderly; fall; feature selection; gait analysis.
    DOI: 10.1504/IJIEI.2025.10065988
     
  • Energy Storage Device Degradation and Smart Home Optimal Management   Order a copy of this article
    by Veerpratap Meena 
    Abstract: This article investigates the efficient scheduling of smart home appliances to reduce energy costs and peak-to-average ratios while maintaining homeowner comfort. It addresses the variability in non-essential load operating times based on day-ahead electricity prices in a smart home. The study introduces the concept of energy discharge cost (EDC) to account for energy storage system degradation costs. It presents two scenarios: the first optimally schedules loads, renewable sources, and energy storage without considering EDC, while the second scenario includes EDC in the scheduling process. The research utilises the GAMS environment and the CONOPT solver for modelling and simulation. Results show that scheduling with EDC consideration leads to a significant 20% reduction in overall energy costs and an improved calendar life for energy storage devices. This work fills a gap in optimal energy management literature and demonstrates the benefits of EDC-aware scheduling in smart homes.
    Keywords: Average charging cost; Energy discharging cost; Degradation cost; Mixed integer linear programming.
    DOI: 10.1504/IJIEI.2025.10066166
     
  • Using Advection-Diffusion Model and Auto-eEncoders to Reduce Adversarial Sensor Data Predictions in Water Leak Management   Order a copy of this article
    by Pandian C, P.J.A. Alphonse 
    Abstract: Accurate water leak localisation remains a significant challenge in pipeline maintenance, water distribution networks, and building water systems. Traditional methods using pressure or acoustic monitoring often lack precision, leading to wasted water and infrastructure damage. This paper proposes a novel methodology to enhance leak localisation accuracy by combining machine learning and physical modelling. The strengths of two techniques are leveraged: advection-diffusion models predict the spread of a leak over time based on the movement of water in a pipeline or a building's water system, and autoencoders, a type of neural network, learn encoded representations of the current sensor data under normal operating conditions along with common adversarial patterns. In the proposed work, autoencoders are combined with advection-diffusion models to achieve significantly improved performance in water leak localisation. Experiments underscore that this combined approach can help reduce false alarms in locating water leaks, preventing further damage and reducing water waste. The proposed model shows improved performance in terms of precision, recall, and accuracy over existing systems. The improved performance is due to the effective hybridisation of physical and neural network models for the reduction of false alarm rates in water leak localisation.
    Keywords: water distribution networks; WDNs; water leak localisation; auto-encoders; advection diffusion models; overflow estimation.
    DOI: 10.1504/IJIEI.2025.10066429
     
  • A Novel Approach for Watermarking Medical Images using Electronic Patient Record Data and a Multibit-Quantisation Modulation Method   Order a copy of this article
    by P. J. R. Shalem Raju, Pokkuluri Kiran Sree, Venkata Durga Kiran Kasula, Raja Rao PBV, Prasad Maddula, P. T. Satyanarayana Murty 
    Abstract: The increasing number of medical images and advancements in communication and computer technology has raised concerns about the authenticity of these images in E-health applications. The proposed method is a watermarking scheme for medical images using Electronic Patient Record data. It aims to protect the security and privacy of medical data, as well as to verify the authenticity, integrity, and ownership of the medical image. It involves four stages: pre-processing, feature extraction, generation of ownership share, and copyright authentication. The scheme uses various techniques, such as active contour segmentation, Nearest Neighbor Grayscale Residuals, Logistic-Logistic System, Slantlet Transform, Singular Value Decomposition, and Quantization Index Modulation. The experiment of the proposed method outperforms state-of-the-art methods in terms of Peak Signal-to-Noise Ratio evaluation. Numerous experiments show that the system is resilient to several attacks, as seen by its high Normalised Cross-Correlation values and low Bit Error Rate.
    Keywords: Watermarking; Electronic Patient Record; medical image; Quantization Index Modulation; Nearest Neighbor Grayscale Residuals.
    DOI: 10.1504/IJIEI.2025.10066547
     
  • Feature Selection based Prediction Model using Binary Dung Beetle Optimisation Algorithm and Improved CNN with Regularisation Model   Order a copy of this article
    by R.S. Preyanka Lakshme, K.S. Ganesh 
    Abstract: Feature selection is a crucial process in data analysis, aiming to identify and eliminate irrelevant features to enhance model performance and efficiency. This research proposes a novel approach to address limitations in existing methodologies, such as overlooking feature interactions and struggling with large-scale datasets. Firstly, comprehensive data pre-processing steps including cleaning, transformation, and organisation are implemented to enhance data quality. Next, a binary version of the dung beetle optimisation algorithm is introduced to tackle feature selection issues, leveraging a U-shaped transfer function. Additionally, the dung beetle optimisation algorithm is augmented with L
    Keywords: Feature Selection; Levy Flight; Local Search Algorithm; Improved CNN; Dynamic Chunk-based Max Pooling.
    DOI: 10.1504/IJIEI.2025.10066566
     
  • Dental Image Segmentation for Carrier Detection Using Improved MLP-UNet Model in Dental X-ray Images   Order a copy of this article
    by S. Srividhya Santhi, R. Shoba Rani 
    Abstract: Dental X-ray picture segmentation assists tooth diagnosis. Too much chocolate and unhealthy diets have increased tooth diseases in recent decades. Dental radiography supports clinical diagnosis, treatment, and quality assessment. Clinical quality has been improved by digitalising dental X-ray image analysis systems. This inspires an early-detection dental disease prediction model. Dental X-ray image segmentation for disease diagnosis is gaining attention. Deep learning has grown in various image processing domains. Image segmentation is a key field of computer vision research. The U-Net, a popular picture segmentation method, has been widely applied to medicine. Here, a unique deep learning system, the UNet model improved with ANN, notably MLP, segments dental pictures better. An ANN-enhanced U-Net split the X-ray pictures, and a GLCM selected the features. CNN-mobile net classifier. Four metrics global accuracy, mean accuracy, mean IoU, weighted IoU, and MeanBFScore and classification with precision, recall, Fscore, specificity, and accuracy evaluate segmentation methodology. Our method is accurate and outperforms others in all segmentation and classification tasks.
    Keywords: dental image dataset; convolutional neural network; CNN; grey-level co-occurrence matrix; GLCM; U-Net; artificial neural network; ANN.
    DOI: 10.1504/IJIEI.2025.10066893
     
  • A Unified Framework for Image Denoising using Enhanced Super-Resolution Generative Adversarial Networks   Order a copy of this article
    by Arti Jain, Anand Singh Jalal, Chetan Negi 
    Abstract: Image denoising is a task to remove corrupted pixels from an image. The corrupted pixel betides to the image while capturing and transmitting from various sources. For image denoising, many filtering approaches are used in the literature. Most of the methods are successfully capable to suppress the noisy pixels. However, the resolution of the image deteriorated during the denoising process. Therefore, in the proposed work denoising is unified with the super-resolution technique to get a noise-free informative and magnificent image. In the proposed approach first denoising is performed with the weighted median filter. The weighted median filter has been extensively used and preferred as it minimises loss of information by filtering out background noise without generating unrealistic pixels. The denoised image is then passed to a super-resolution technique to boost the perceiving quality of the denoised image. In light of PSNR and SSIM, it has been evident that the proposed method effectively outcompetes the existing state-of-the-art denoising methods.
    Keywords: image denoising; super-resolution; deep neural networks; generative adversarial network; GAN.
    DOI: 10.1504/IJIEI.2025.10066902
     
  • Enhanced Coyote Optimization Algorithm for Task Scheduling in Computer Systems Using Modified DNN   Order a copy of this article
    by S. Vaaheedha Kfatheen  
    Abstract: Cloud computing has evolved into numerous fields and applications in recent years. However, job and resource scheduling can be improved. Task scheduling that maps incoming tasks to resources is needed for high-performance data mapping in heterogeneous computing systems. Cloud computing lets consumers access computational resources online without infrastructure. A service level agreement (SLA) formalises the connection between cloud service customers (CSC) and CSPs. The SP must provide the best features, response time, and resource utilisation to achieve this SLA. Task scheduling is tough in cloud computing since many operations must be done with limited resources. To overcome this difficulty, employ a smart scheduling strategy with strong algorithms that analyse demands and priorities. Thus, this proposal uses the modified deep neural network (MDNN) and improved coyote optimisation algorithm to schedule tasks efficiently. This research aims to reduce energy use and migration costs. The ICOA will set a multi-objective target to efficiently schedule cloud tasks.
    Keywords: cloud computing; task scheduling; scheduling algorithm; service level agreement; SLA; cloud service customers; CSC; cloud service providers; modified deep neural network; MDNN.
    DOI: 10.1504/IJIEI.2025.10066975
     
  • Spearman Chimp Optimisation Algorithm (SCOA) Feature Selection and Fuzzy Weight Long Short-Term Memory (FWLSTM) Classifier for Cyberbullying Twitter Data   Order a copy of this article
    by Menaka M, Sujatha P 
    Abstract: Social connections developed within narrow cultural limits, such as physical locations, prior to the invention of information and communication technology (ICT). Social technologies have revolutionised online social networks, user-generated content, and rich human behaviour data. Online social networks (OSN) promote social interaction but also trolling, hate speech, and cyberbullying. NLP-based automatic detection is essential to ending cyberbullying. A deep learning algorithm is suggested to detect cyberbullying aggression in this work automatically. Pre-processing, feature extraction, feature selection, and classification are among the processes included in the suggested workflow. The initial pre-processing steps for the Twitter database include noise removal, tokenization, and stemming. The features from the pre-processed database have been extracted using the SAE, TF-IDF, and other techniques. To choose the subset of characteristics, the SCOA is next applied. FWLSTM classifier is then given features. The K Nearest Neighbour (KNN), ANN, Random Forest (RF), and EK-SVM classifiers are contrasted with the FWLSTM classifier. Results are evaluated using precision, recall (sensitivity), specificity, false positive, false discovery, miss, and accuracy.
    Keywords: Stacked Auto-Encoder (SAE); Spearman Chimp Optimization Algorithm (SCOA); Fuzzy Weight Long Short-Term Memory (FWLSTM); Artificial Neural Network (ANN); Enhanced Kernel with Support Vector Machine (EK.
    DOI: 10.1504/IJIEI.2025.10066977
     
  • Intelligence Ensemble Feature Selection (IEFS) and Ensemble Classifier for Cervical Cancer Diagnosis   Order a copy of this article
    by Anjali Kuruvilla, B. Jayanthi 
    Abstract: Cervical cancer is one of the most common diseases affecting women, which may be prevented with an early diagnosis. Numerous recent studies using DL and ML techniques have focused on early-stage prediction of this disease. DL methods, it's critical to understand the importance of factors for appropriately classifying patients with cervical cancer. In other words, FS selects the most important features from the dataset when building a DL model and introduces IEFS and DLE for cervical cancer. IEFS has been introduced to choose the important features and remove redundant features to train a model using the DLE classifier. The IEFS model combines the results of methods like EBFO, EEHO, and KLBWO, creating an optimal best subset. An aggregation function has combined the results of individual methods. DLE is a DL approach that combines several methods (GAN, BGRU, and DWCNN) in cancer diagnosis. The LogitBoost model combines the output from different results of the proposed classifier against current classifiers. MATLABR2020a has been used to simulate and measure the classifiers using precision, recall/sensitivity, F-measure, specificity, and accuracy metrics.
    Keywords: Deep Learning Ensemble (DLE); University of California Irvine (UCI); Deep Learning (DL); Machine Learning (ML); Dynamic Weight Convolutional Neural Network (DWCNN); Generative Adversarial Network.
    DOI: 10.1504/IJIEI.2025.10066978
     
  • A Fuzzy-Enhanced Recursive Feature Elimination for Intrusion Detection In WSN   Order a copy of this article
    by Ashwini Abhale, Jayaram Reddy Avulapalli 
    Abstract: This study addresses the critical need for enhanced security in wireless sensor networks (WSNs), which are integral to modern infrastructure yet highly vulnerable to security breaches. We introduce a ground-breaking intrusion detection system (IDS) that employs recursive feature elimination (RFE) alongside advanced fuzzy classifiers, providing a robust solution against these security threats. The proposed IDS uniquely integrates RFE with three distinct fuzzy classifiers: the adaptive neuro-fuzzy classifier (ANFC) with a 97% accuracy rate, the fuzzy nearest neighbour classifier (FNNC) with 92% accuracy, and the fuzzy decision tree classifier (FDTC) achieving an exceptional 98% accuracy. This innovative approach leverages the strength of RFE in feature selection and the sophisticated pattern recognition capabilities of fuzzy logic. It significantly enhances the system’s ability to accurately differentiate between normal operations and potential security threats in WSNs. The effectiveness of this IDS is highlighted by its remarkable accuracy rates, which are a direct result of the focused analysis of critical attributes facilitated by the RFE process. This research contributes significantly to bolstering the integrity and security of WSNs, presenting a notable methodological advancement in intrusion detection technologies.
    Keywords: recursive feature elimination; RFE; fuzzy classifier; adaptive neuro-fuzzy classifier; ANFC; fuzzy nearest neighbour classifier; FNNC; fuzzy decision tree classifier; FDTC.
    DOI: 10.1504/IJIEI.2025.10067199
     
  • Transfer Learning Based Breast Cancer Detection and Classification using AG2Convnet and Adap-BI Magnification   Order a copy of this article
    by M.D. Bakthavachalam, S.Albert Antony Raj 
    Abstract: Breast cancer (BC) has recently been a major issue. Although numerous BC detection methods are used in medical image processing, the correct detection and classification of benign and malignant micro-calcifications is problematic. This research provides an effective BC detection and classification method using AG2ConvNet and Adap-BI magnification. The input image is first obtained from BUSI and MIAS datasets. Image type conversion improves categorisation after capture. Next, data augmentation uses rotation and flipping. Data is pre-processed using sigmoid scaling and Gaussian filter (GF). Adap-BI magnifies pre-processed data, and the minimisation principle-adapted Canny edge detector detects edges. The AG2ConvNet classifier uses the edge-detected output to diagnose normal, benign, and malignant tumours from the digital mammography picture. Finally, the outbreak prediction model's results are compared to existing methods to verify its efficacy. Results showed that the proposed method outperformed baseline methods. The suggested model outperforms EL-CNN, TL-DRS, and AM-SFF approaches with 95.41% accuracy. This shows the model's superiority and importance in improving breast cancer detection and classification, making it a potential medical image processing tool.
    Keywords: adaptive bicubic interpolation; Adap-BI; minimisation principle adapted canny edge detector; MCED; adaptive glorot GoogLeNet convolutional neural network; AG2ConvNet.
    DOI: 10.1504/IJIEI.2025.10067247
     
  • A Hybrid Deep Learning Architecture for Prediction of Renal Diseases using Residual, Squeeze-Expand and Depth-Wise Convolutional Blocks   Order a copy of this article
    by Niteesh K. R, Vijayakumar Kadappa, Rajeshwari Devi D. V, Divijendranatha Reddy Sirigiri 
    Abstract: In predictive and preventive healthcare, Machine Learning and Deep Learning demonstrate a pivotal role in the early diagnosis of diseases. Kidneys are sophisticated biological filters and play a vital role in maintaining body homeostasis. However, kidney failure leads to many critical and chronic diseases. Early diagnosis of these diseases is essential for better treatment options. The study proposes a hybrid deep learning architecture based on the ideas of residual, squeeze-expand, and depth-wise separable convolution blocks. The proposed model exhibits superior performance over other models in terms of accuracy, precision, recall, F1-score, and AUROC. The hybrid architecture achieves average accuracy of 0.9992 and 0.9975 with 5-fold and 10-fold crossvalidation respectively. The proposed model shows a perfect AUROC of 1.00 on an independent data of 800 images and demonstrates best performance followed by InceptionV3 and MobileNet for renal disease prediction. The proposed model aids clinicians in the early diagnosis of renal diseases.
    Keywords: Artificial Intelligence; Deep Learning; Machine learning; Healthcare; Kidney (Renal) disease prediction.
    DOI: 10.1504/IJIEI.2025.10067400
     
  • Dental Caries Segmentation and Detection using SIU-Net with YOLO v8 on CBCT Images   Order a copy of this article
    by Pradeep T. S, Arul Linsely 
    Abstract: Dental caries is one of the most prevalent and persistent illnesses globally. For radiologists, dental CBCT is an essential diagnostic tool. CBCT is widely used, but diagnosing dental caries remains challenging due to limited datasets and time-consuming annotation processes. To tackle these problems, a novel approach utilising a single input U-Net integrated with YOLO v8 for dental caries segmentation and detection is introduced. The proposed SIU-Net incorporates a multi-scale spatial attention module (MSAM), which enhances the model's ability to focus on critical tooth structures while suppressing irrelevant information, thereby improving the accuracy of caries segmentation. The efficient multiscale channel attention (EMCA) module effectively connects the encoder and decoder of the SIU-Net, facilitating better feature fusion. Furthermore, YOLO v8 is employed in this framework for precise tooth disease detection. This model is rigorously evaluated on CBCT images, and experimental results demonstrate that the proposed approach achieved a dice coefficient of 92%, IoU of 90%, precision of 92%, and accuracy of 94% to show its superior performance in accurately detecting dental caries compared to other state-of-the-art approaches. This proposed model makes significant advancements in dental caries detection and robustness across several dental imaging modalities.
    Keywords: dental caries; single input U-Net; You Only Look Once Version 8; multi-scale spatial attention module; MSAM; efficient multiscale channel attention; EMCA; cone beam computed tomography; CBCT.
    DOI: 10.1504/IJIEI.2026.10067437
     
  • Optimizing of Dynamic aspect Attention for Sequence Modelling via Reinforcement Learning   Order a copy of this article
    by P.M. Diaz, M. Julie Emerald Jiju 
    Abstract: Aspect-based sentiment classification represents a sophisticated approach within sentiment analysis, focusing on identifying sentiment polarity towards specific aspects or entities mentioned in text. Despite the advancements in ABSC techniques, dynamically selecting and re-weighting crucial words in a sentence remains a challenging task. To solve this issue, this research proposes a novel methodology named DRA-SAC, which combines the Dynamic Reweighting Adapter with Soft Actor-Critic, to address these challenges. The DRA component dynamically selects and re-weights important words within a sentence, leveraging an attention mechanism to focus attention on critical aspects of the context. By adapting attention weights, DRA enhances the model's ability to capture aspect-specific sentiment information. Subsequently, SAC guides the learning process, facilitating the adjustment of attention weights through reinforcement learning techniques. The performance of the proposed model is evaluated on four benchmark datasets. The experimental results demonstrate the efficacy and superiority of the proposed approach compared to existing methods.
    Keywords: Aspect based sentiment classification; Dynamic Reweighting Adapter; Soft Actor-critic; Reinforcement learning; BERT.
    DOI: 10.1504/IJIEI.2026.10067440
     
  • CREWLINKS: A Convergence, Robustness Efficient Model towards Water Leak Isolation using Network of Sensors   Order a copy of this article
    by Pandian C, P.J.A. Alphonse 
    Abstract: In water distribution networks (WDN), rerouting is the process of rerouting water flow along different pathways in response to variations in demand, to manage emergencies such as leaks, and to enhance overall performance. This is accomplished by regulating the flow through pumps and valves. Finding the most effective route through pipes and valves to improve resilience, lower water loss, and guarantee a consistent supply is made possible in large part by topology optimisation. This work optimises a genetic algorithm (GA) and uses a convergence-efficient architecture to increase the robustness of IoT-based smart city WDNs. According to simulations, the method decreases resilience loss to only 7.5% as network density increases, and it beats regular GA by 21%.
    Keywords: convergence and robustness efficient WDN architecture; water leak isolation; genetic algorithm; Schneider R; robust re-routing technique.
    DOI: 10.1504/IJIEI.2026.10067591
     
  • Multi-Step Time Series Forecasting of Fish Health in Aquaponics with Combined Attention Enriched CNN-LSTM to Strengthening Sustainability Initiatives for Improved Aquaponics Management   Order a copy of this article
    by G. Safiya Begam, W. Aisha Banu 
    Abstract: Fish health depends on numerous factors, therefore precise estimates are crucial. Historical data on water temperature, pH, dissolved oxygen, and fish health is used. CNNs extracted features from multidimensional input data to predict fish health. The memory neural networks relayed and mined biological and environmental variable feature correlations. The CNN-LSTM model extracted temporal and spatial trends from the time-series aquaponics dataset. AM was added to the time-series prediction model to improve LSTM accuracy. The integrated prediction model uses AM, LSTM, and CNN to focus. To verify accuracy, the CNN-LSTM-AM novel model was tested weekly and monthly. The novel model has RMSE 0.0997, MAE 0.097, MAPE 0.0154, and short-term and long-term R2 values of 92.75 and 90.25. This model allows real-time aquaponics fish health monitoring. CNN-LSTM model, recurrent neural network, long short-term memory neural network, and autoregressive integrated moving average performance were computed and visualised. Increasing aquaponics automation is proposed. Aquaponics multi-step time series fish health forecasting is improved by the provided strategy. This firms sustainability has improved thanks to the unique models accuracy.
    Keywords: Multi-Step Time Series; Fish health Prediction; Attention mechanism; CNN-LSTM; sustainability in Aquaculture; Predictive analytics; Temporal and Spatial analysis; Anomaly Detection (AM).
    DOI: 10.1504/IJIEI.2026.10067592
     
  • Biomedical NLP-based Data Classification for Sensitivity Evaluation of Healthcare Data Using Statistical Features Analysis   Order a copy of this article
    by Manoj Dhawan, Lalit Purohit 
    Abstract: This work aims to improve the performance of multiclass categorisation of biological texts for sensitivity evaluation by combining two distinct feature representation techniques. Bio ALBERT (a domain-specific adaptation of a lite bidirectional encoder representations from transformers) was utilised in this investigation for designing a multiclass classification model for sensitivity. This studys primary contributions include a weighted feature representation technique for biomedical text categorisation. The research is primarily concerned with combining two different feature representation techniques, namely WE and BoW, to improve the performance of a biomedical multiclass text classification system. The experimental results validate the suggested systems theoretical analysis. This work evaluates the effectiveness and efficiency of the proposed task using the MIMIC-III database. Further, MIMIC III and the PubMed dataset are employed to construct the language model. The performance of the proposed weighted feature representation approach for multiclass classification is found to be superior to the conventional techniques.
    Keywords: Natural language processing; transfer learning; Attention Mechanisms; Biomedical NLP; ALBERT; BioALBERT; Electronic Health Records (EHR).
    DOI: 10.1504/IJIEI.2026.10067677
     
  • Study of Meta-Learning Attempts to Select Algorithm for Hard Combinatorial Optimisation Problems   Order a copy of this article
    by Neha Sehta, Urjita Thakar 
    Abstract: Given a large number of techniques to solve a problem, choosing or recommending the most appropriate one for a specific case is difficult and defined as an algorithm selection problem. The machine-learning community identified this as a learning task and named meta-learning. It is applied to algorithm performance data collected from hitherto applications of target algorithms on the problem under study. It aims to identify a mapping from instance characteristics to algorithm performance. Initially applied to classification problems, afterward generalized to other learning and non-learning tasks. In this paper, a review of meta-learning attempts exercised for algorithm selection on different combinatorial optimization problems is presented. Each tuple of metadata contains structural and other features of a problem instance along with performance metrics of target algorithms. Learning techniques have been applied to metadata to realize meta-learning In contrast to more straightforward selection methods, meta-learning models lead to acquiring high-quality solutions.
    Keywords: Machine learning; meta-learning; algorithm selection problem; optimization problems.
    DOI: 10.1504/IJIEI.2025.10067775
     
  • CASD on Enhancing Sentiment Analysis Using Context-Aware Sarcasm Detection on Social Media   Order a copy of this article
    by Paul Davidson, D. Ravindran, R. Anne Pratheeba 
    Abstract: Effective sentiment analysis is crucial for understanding human language, especially when dealing with sarcasm. This study enhances sentiment analysis by integrating BERT contextual embeddings with an ensemble learning classifier to recognise sarcasm. A custom-labelled dataset rich in sarcastic elements was created to train and refine the model, utilising a random forest classifier for its robustness with complex datasets. Comparative analysis against standard models, including BOW-LR, VADER, SVM-TF-IDF, and LSTM, showed that the CASD model significantly improved performance metrics, achieving an accuracy of 0.85 and consistently outperforming baseline models across all sentiment classes. Notably, CASD achieved a precision score of 0.86 for neutral sentiment detection, illustrating its sensitivity to linguistic subtleties. This research introduces a novel framework that effectively accounts for sarcasm, leveraging BERT's contextual understanding and random forest’s ensemble classification to advance sentiment analysis accuracy. This improvement is vital for applications requiring fine-grained sentiment detection, showcasing the potential for sophisticated natural language processing technologies to reflect the complexities of human communication more accurately.
    Keywords: support vector machine with term frequency-inverse document frequency; SVM-TF-IDF; valence aware dictionary and sentiment reasoner; VADER; context-aware sarcasm detection; CASD.
    DOI: 10.1504/IJIEI.2025.10068012
     
  • Optimising Facial Expression Recognition via a Hybrid Ensemble Classifier and Improved Feature Extraction Techniques   Order a copy of this article
    by Sreenivasu Bhukya, L. Nirmala Devi, A. Nageswar Rao 
    Abstract: Facial expression recognition (FER) is crucial for applications in human-computer interaction, security, and healthcare. Traditional methods often struggle with issues such as limited generalisation, high computational demands and handling complex emotions. For addressing these challenges, the research offers a novel methodology which combines advanced image pre-processing, feature extraction, and classification techniques. This initiate with image pre-processing that standardises facial images by reducing noise. Then the feature extraction employs shallow Convolutional Neural Network (CNN) for texture-based features. Then Softmax classifier for emotion prediction and incorporates hyperparameter tuning using the Equilibrium Optimizer (EO) algorithm. Additionally, feature selection is optimized using the Improved Cat Swarm Optimization (ICSO) algorithm to refine the feature set. The integration of an ensemble classifier combining Multi-Layer Perceptrons (MLPs) and ordinal logistic regression improves the robustness of the model, utilising soft labels. Experimental results demonstrate significant improvements over existing methods and achieves the accuracy of 99.12%.
    Keywords: Feature Extraction; Convolutional Neural Networks (CNN); Improved Cat Swarm Optimization (ICSO); Equilibrium Optimizer (EO); Multi-Layer Perceptrons (MLPs).
    DOI: 10.1504/IJIEI.2026.10068320
     
  • Deep Hybridisation of Cross-Space Adaptive Filter with Matrix Completion of Dual Adaptive Jumping Graph Neural Networks for Sentiment classification Systems   Order a copy of this article
    by K.R. Srinath, B. Indira 
    Abstract: In the era of big data and complex social media interactions, sentiment analysis and recommender systems (RSs) face significant challenges due to information overload and the dynamic nature of user preferences. To address these issues, a combination of Clustered Spectral Filtering (CSF) with Matrix Completion and Dual Adaptive Jumping Graph Neural Networks (DualAJGNN) is introduced. The proposed model enhances sentiment classification and provides accurate recommendations by considering users' sentimental attributes and social influence, while overcoming limitations of existing methods such as over-smoothing in Graph Convolutional Networks (GCNs). This research optimises neighbourhood aggregation in GCNs using CSF. Additionally, a Graph Neural Network (GNN) model combines multiple preferences derived from extended user behaviours to create clusters, forming an interest graph for sequential recommendation. The effectiveness of the proposed model is validated using three datasets. The results demonstrate the superiority of the proposed approach compared to existing techniques for RS and Sentiment classification.
    Keywords: Sentiment analysis; Recommender Systems; Graph Convolutional Networks; Clustered Spectral Filtering; Dual Adaptive Jumping Graph Neural Networks; Information overload.
    DOI: 10.1504/IJIEI.2026.10068327
     
  • Deep Learning-Based Multi- Modal Curved Text Detector using Scenic Images: A Multi-Feature Deep Attention Residual Network (DAR-NURD)   Order a copy of this article
    by Prachi Chhabra, Ali Imam Abidi 
    Abstract: Detecting curved texts in natural scenes presents challenges due to variabilities in orientation and distortions during image acquisition. Traditional models often struggle with low accuracy on curved surfaces. To address this, a novel deep learning-based framework, DAR-NURD (deep attention residual network for multi-feature utilisation in realistic detection), is proposed. DAR-NURD enhances text detection accuracy by leveraging advanced feature extraction techniques and integrating attention mechanisms with residual networks. This enables the model to focus on significant image regions, effectively handling curved and distorted text in complex scenarios. Its robust performance is validated through extensive experiments, achieving superior precision (99.31%), recall (99.35%), and F1 score (99.34%) compared to methods like GL-CAM, DBNET++ (RESNET-50), and BSNET. DAR-NURDs applicability extends to real-world scenarios such as autonomous driving, augmented reality, and surveillance, demonstrating its potential to set new benchmarks in curved text detection.
    Keywords: Deep Learning; Image Segmentation; Natural scene images; multi-oriented text; Text Detection.
    DOI: 10.1504/IJIEI.2026.10068366
     
  • Real-time Speech Enhancement Using Temporal Envelope Modulation and Hybrid Neural Network Algorithms for Improved Speech-to-Text Conversion   Order a copy of this article
    by M.R. ANITHA, B. Vijayalakshmi 
    Abstract: The algorithm converts improved speech into text for speech synthesis applications. This approach combines hybrid neural networks and offers superior quality and intelligibility while minimizing word error rates by incorporating temporal envelope modulation analysis at the speech prosody level. The algorithm works with enhanced speech with adaptive gain control having 16-channel BPFs, envelope detectors, and LPFs and combining CNNs and LSTMNs to reduce background noise and enhance accuracy. Then, the output will be text using HMM on the enhanced speech. Various types of noise, such as babble SSN, were tested through experimental evaluations on real-time recorded samples using PRAAT software and TIMIT and Noizeus noise databases at various SNR levels. The algorithm showed a drastic performance improvement, evaluated through PESQ, STOI, and cosine similarity scores, thus showing a gain of 39% in PESQ and 33.3% in STOI. In addition, compared with noisy speech, up to 3% reduction in word error
    Keywords: Speech Enhancement; Convolutional Neural Network; Speech to Text conversion; long short-term memory; Speech intelligibility and Speech quality; Temporal envelopes.
    DOI: 10.1504/IJIEI.2026.10068454
     
  • A Multivariate Multi-Headed Deep LSTM Model: a New Approach to Enhance the Prediction of Methane Production in the Biogas Plant   Order a copy of this article
    by Meena Krishnamoorthy, Srinivasulu Reddy Uyyala, Ilangovan P 
    Abstract: Biogas is a sustainable method for converting diverse waste sources into energy. It is of utmost importance to develop a new approach for enhancing the prediction of methane production in the biogas plant. This study introduces a standalone multi-headed neural network architecture called multivariate multi-headed long short-term memory (MM-LSTM) to improve the prediction accuracy of methane production from biogas digesters. The proposed model treats each input variable independently, employing them as distinct LSTM network models (heads) to improve the prediction accuracy. In the MM-LSTM approach, the outputs of these independent LSTM models are concatenated to forecast high methane production. The present study evaluated the proposed models predictive performance using sensor data from a biogas digester and compared it to state-of-the-art models and single-headed long short-term memory (LSTM) neural network architectures. Based on the experimental results, the MM-LSTM model achieves a prediction accuracy of 99.52%, which is best among all the existing models. The experimental results suggest that the proposed prediction model can be effectively used in biogas plants for biogas production processes and for advancing sustainable energy production.
    Keywords: Multivariate time series; Multi-headed LSTM; Prediction; Neural network model; Methane production; Biogas plants.
    DOI: 10.1504/IJIEI.2026.10068455
     
  • Sarcasm Detection on News Headlines, Twitter and SARC Dataset: a Detailed Evaluation of Shallow and Deep Models   Order a copy of this article
    by Anisha Gupta, Dr. Apeksha Mittal, Rachna Jain 
    Abstract: The presence of sarcasm in online communication can often distort the intended meaning, thereby making sarcasm detection vital An in-depth examination of various shallow and deep sarcasm detection techniques applied to textual data is presented in the paper To facilitate comparison between these techniques, experiments are conducted executing them on two distinct datasets sourced from different platforms: News Headlines and Twitter Ten shallow and seven deep models are run on the datasets recording the performance metrics (accuracy, precision, recall and F1-score) and execution times This work encompassing both performance and execution times lays groundwork for selecting appropriate sarcasm detection model according to dataset characteristics and resource availability SVM leads amongst the shallow models and RoBERTa excels amongst the deep models attaining maximum accuracy Deep models outperform shallow models, but, except for CNN, they require GPU for execution Despite using GPU, their execution times are notably high limiting their practical application.
    Keywords: Sarcasm detection; Natural Language Processing; News Headlines Dataset; Twitter Dataset; Shallow models and Deep Models; Accuracy; Precision; Recall; Performance,.
    DOI: 10.1504/IJIEI.2026.10068512
     
  • A Novel Task Clustering Approach for Workflow Orchestration in a Dynamic Cloud Environment   Order a copy of this article
    by Monika Yadav, Atul Mishra 
    Abstract: In the continually evolving landscape of cloud computing, the effective orchestration of scientific workflows presents a significant challenge. This paper introduces an innovative task clustering approach designed to streamline workflow execution within dynamic cloud environments. The primary objective is to enhance task execution efficiency while consolidating resources. The proposed algorithm performs clustering by considering parent-child relationships and employing ordinal optimisation techniques. Subsequently, it refines clusters based on runtime and the number of shared children, considering the desired resource set size. This optimisation ensures that the clusters are optimally configured for deployment in dynamic cloud environments, thereby maximizing resource efficiency and overall performance. By effectively identifying and grouping related tasks into clusters, the proposed approach minimizes scheduling overhead for fine-grained scientific tasks, thereby enabling improved parallel processing within dynamic cloud environments.
    Keywords: Cloud computing; Task clustering; Directed Acyclic Graph (DAG); WorkflowSim; Workflow orchestration.
    DOI: 10.1504/IJIEI.2026.10068644
     
  • Advanced Neural Network Methods for Soil Type Classification and Crop Prediction using Aerial Images   Order a copy of this article
    by K. Vaishnavi, J. Dhalia Sweetlin, Amar Pratap Singh J, A. Jyotsna 
    Abstract: Agriculture is a great sector for a country. The type of crops to be grown in a region depends on the type of soil, the availability of nutrients, and the prevailing environmental conditions, such as temperature, humidity, and rainfall. Lack of knowledge about nutrient availability lowers productivity. This addresses the problem through recommendations of the best crop, depending on specific parameters. This approach aims to design a model that assists the farmer in identifying an ideal crop for every soil type. This work uses a convolutional neural network to classify soil types from aerial land images. The pH value of the soil is determined by processing these images with the help of the feed forward neural network (FNN). An ensemble classifier predicts the most suitable crop for cultivation in a particular region based on soil type, pH value, and weather data. So far, the proposed CNN model in classifying soil type reaches an accuracy of 92% by using aerial images of this region. The same results were achieved by predicting a crop with an ensemble classifier, using a majority voting of 97%.
    Keywords: Soil Classification; Neural Network; Agriculture; Image Processing; Crop Prediction; Feed Forward Neural Network (FNN); Convolutional Neural Network.
    DOI: 10.1504/IJIEI.2026.10068654
     
  • An Architectural Framework for Intent Identification in Mixed Script Queries based on Roman Transliteration   Order a copy of this article
    by A.N.U. Chaudhary, Rahul Pradhan, Shashi Shekhar 
    Abstract: Mixed script inquiries' various linguistic phrases make intention identification difficult. This study proposes a new method that uses knowledge-based approaches and RNNs to reliably identify mixed script query intentions. The method uses knowledge-based databases of negative and positive terms to describe people and words that change intention. Language nuances are trained using these datasets. Mixed script queries are pre-processed and translated into numerical vector representations using knowledge-based intent areas. These vector representations are processed using RNN architecture to learn complex input-intention mappings. An attention method can emphasise key inquiries to help identify intentions. A diversified dataset of mixed script queries is used to test the suggested approach using accuracy, precision, recall, and F1-score. The results show the model can reliably discern intentions across languages. By solving mixed script query problems and offering a solid framework for multilingual intention recognition, this research advances natural language processing. The method could be used in information retrieval, sentiment analysis, and cross-language communication.
    Keywords: Natural Language Processing (NLP); Machine Learning; Sentiment Analysis; Script Queries; Intention Identification; Cross-Language Communication; Knowledge-Based Regions.
    DOI: 10.1504/IJIEI.2026.10068896
     
  • Optimising Brain Tumour Segmentation and Classification with an Enhanced CNN Model on the BraTS-GOAT 2024 Dataset   Order a copy of this article
    by Vaidehi Satushe, Vibha Vyas, Shilpa Metkar, Davinder Paul Singh 
    Abstract: The BraTS Generalisability Across Tumours (BraTS-GoAT) initiative aims to address the critical need for robust and generalizable models in brain tumour segmentation. An important goal of this study is to come up with a way to make CNN-based brain tumour segmentation models more consistent and useful in more situations. The objective of this study is to enhance the capacity to apply optimized CNN models to large-scale datasets in order to improve generalisability. Using the BraTS 2024 challenge dataset which consists of 3D Magnetic Resonance Images, the suggested model has been 98.47% accurate, had a mean Intersection over Union of 0.8185, an avg dice coefficient of 0.7,an avg Hausdorff 95 score 1.66, 0.9855% precise, sensitivity of 0.9840, a specificity of 0.9952. These findings will contribute to establishing benchmarks for generalisability in medical imaging, promoting the adoption of these technologies in diverse clinical environments.
    Keywords: Brain tumor; generalizability; CNN; Magnetic Resonance Imaging; Intersection over Union.
    DOI: 10.1504/IJIEI.2026.10069000