International Journal of Advances in Soft Computing and its Applications

Multi-Modal Deep Learning for Purchase Decision Modeling: A Comprehensive Framework for E-Commerce Recommendation

2026-03-15T15:31:11+00:00

Understanding customers' purchasing decisions is a fundamental challenge in e-commerce. This work presents a multi-modal deep learning architecture using product image, product description, and user behavior history to predict the probability that a user will purchase a product. We use ResNet-50, BERT, and a bidirectional LSTM to encode features from the three modalities and propose a cross-modal attention mechanism to integrate the features. Our experiments are carried out on the Amazon Electronics dataset. We achieve an ROC-AUC of 0.892, which outperforms the best unimodal model by at least 8\%. Ablation experiments reveal that the different modalities complement one another, with user behavior history being the most important modality.

Integrating Artificial Intelligence with Earned Value Management for Enhanced Performance in Smart Renewable Energy Projects

2026-01-05T16:08:45+00:00

Renewable energy projects in Jordan continue to face persistent cost overruns and schedule delays, largely due to the limitations of traditional project control systems that rely on manual progress reporting, subjective assessments, and delayed decision cycles. This study evaluates the integration of Artificial Intelligence (AI) with Earned Value Management (EVM) as a comprehensive approach to improving performance monitoring and forecasting in smart renewable energy initiatives. A mixed-methods framework was adopted, combining an industry survey of 394 practitioners with an instrumental case study of a solar photovoltaic system upgrade modeled in Autodesk Revit and scheduled in Primavera P6. The case study revealed early and significant deviations, with an SPI of 0.75 and a CPI of 0.56 at Week 4, followed by a sharp decline in cost performance to approximately CPI 0.31, ultimately indicating an unavoidable budget overrun and a negative TCPI. Survey findings strongly support the potential of AI–EVM integration: 61.7% of respondents reported substantial improvements in decision-making, 57.1% emphasized enhanced real-time tracking, and over 55% highlighted gains in forecasting accuracy through AI techniques such as machine learning, computer vision, and natural language processing. Despite these positive perceptions, key barriers—including poor data quality, regulatory limitations, high implementation costs, and organizational resistance—remain significant. Overall, the findings demonstrate that AI-enhanced EVM offers a transformative shift from reactive project control to predictive, data-driven management. The study concludes that adopting such an integrated framework is essential for improving cost efficiency, schedule reliability, and overall project outcomes in Jordan’s rapidly expanding renewable energy sector.

An Intelligent Approach to HTTP Flood DDoS Detection Using Bayes-Entropy

2026-03-01T00:03:10+00:00

One of the most serious cyberattacks on network systems or internet services is the distributed denial of service (DDoS) attack. Even though DDoS attacks can be detected in a variety of ways, the issue still exists. The major presumptions around this gap are put out in this work using mathematical techniques that may effectively identify HTTP flooding DDoS attacks. To stop destructive HTTP flooding DDoS packets from reaching the website, this work presented a powerful mathematical approach based on Bayes-entropy. The traffic will be separated into aggregated packets based on (t) time, and each aggregated packet will be broken down into equal smaller time intervals called events, which will subsequently be grouped into groups based on (t) time (equal packets size with the same inter arrival time). This technique will use the Bayes theorem to calculate the chance of HTTP flooding DDoS attacks inside the group and the entropy equation to calculate the unpredictability within the group. If the computed statistics suggest a high frequency of such attacks and a low amount of randomness in the chosen group under examination, it will be categorized as an HTTP flooding DDoS attack; otherwise, it will be labeled as normal. Experiment results on the ISCX dataset show that the proposed technique produces high accuracy rates of 97.14%.

Protein Tertiary Structure Prediction using a β-Hill Climbing Optimizer

2026-03-25T17:54:41+00:00

Predicting the protein tertiary structure from its amino acids is a big challenge nowadays. There are huge number of protein sequences with unknown structures. However, there are many computational methods which tried to solve this problem, in general, these methods achieve good results for small size proteins, but most of them were not powerful when the conformational space is huge. Many studies have tried to solve this problem using either local search algorithms or global search algorithms; whereas some studies have used hybrid algorithms which make use of local search-based algorithms and population-based meta-heuristic algorithms. This study introduces a new algorithm, β- hill climbing Algorithm, to solve the ab initio protein tertiary structure prediction problem. The β- hill climbing Algorithm is used to find the local optimal solution within the search space by its Iterated Local Search (ILS). Furthermore. The proposed algorithm predicts the tertiary structure of a protein without any prior knowledge but based on its sequence alone. The proposed algorithm has been evaluated using two protein sequences; Met-enkephalin (1PLW) and Plant Seed Protein (1CRN). The results show that the proposed algorithm can find more good solutions compare to other previous studies.

Studying Traffic Accidents Utilizing a Machine Learning Approach: Case Study of Jordan

2026-03-19T22:15:19+00:00

Traffic accidents continue to be a significant public safety concern in Jordan, largely due to the many interaction of human, vehicle, road, and environmental factors. The research presented here is based on a data-driven analysis of fatal traffic crashes in Jordan with a comprehensive dataset using a national dataset covering all recorded crashes between 2018 and 2022. Within this dataset, there is detailed information about driver demographics, vehicle characteristics, conditions of the road, environmental factors and accident attributes, providing the necessary information to perform a complete machine learning analysis. Statistical analyses were conducted using the selected features to identify and quantify the relationships most strongly associated with death resulting from traffic accidents. Machine learning models were developed to predict whether an accident would result in a fatal outcome. The most accurate machine learning models were Random Forest and XGBoost, both of which achieved an accuracy of approximately 0.96 overall, while further evaluation using class-sensitive metrics highlighted differences in their ability to identify fatal cases. To improve interpretability, explainable artificial intelligence techniques were integrated into the analysis. SHAP was used to identify the most influential factors helping in fatal incidents outcomes at a global phase, while LIME gave localized explanations for the prediction of individual. The added value of combining machine learning with explainable models to better understand the mechanisms underlying fatal traffic accidents was of great focus on this work. The results support evidence-based policy interventions within Jordan, including targeted enforcement measures, infrastructure improvements, and strategies addressing behaviors that are related to high-risk driving in order to reduce fatalities and serious injuries resulting from traffic accidents.

Convex Geometry-Driven Vehicle Localization in LiDAR for Advanced Driver Assistance Systems

2026-02-12T21:44:00+00:00

Precise localization of vehicle bounding boxes in LiDAR images is an important part of Advanced Driver Assistance Systems (ADAS). This paper offers a explainable solution, which is a combination of point-cloud clustering and a convex hull algorithm. The approach initially uses a clustering procedure to isolate vehicle areas by grouping LiDAR points, and the points of the same vehicle can be combined with each other, despite noise or partial occlusions. Each cluster is then subject to a convex hull which produces a minimal bounding polygon that geometrically approximates the vehicle footprint. Outlier removal, size-adaptive clustering parameters, and occlusion correction are further used to enhance bounding-box accuracy. Evaluation on the KITTI data has been performed experimentally with a mean absolute error of 0.167, root mean squared error of 0.186 and average percentage error of 8.5% variance among vehicle dimensions. The approach has a Pearson correlation coefficient of 0.9995 with ground-truth annotations, which is high. Also, scores of 1-D Intersection over Union are above 93 on average, which is also a good sign of spatial alignment. The convex-hull-based framework is proposed to be straightforward, strong and efficient in terms of calculation. Although no training data is needed, the performance obtained is high and the results are better than some of the recent learning-based detectors, which has an important practical implication of the method. Sample LiDAR frames can be run on an NVIDIA Xavier platform in less than 30 ms, which can meet the real-time latency of ADAS with low computation costs.

Evaluating and Ranking Simulation Tools for Railway Fire Safety with the Analytic Hierarchy Process

2026-03-07T19:05:09+00:00

Ensuring fire safety in railway systems requires reliable simulation tools capable of reproducing complex physical and behavioral phenomena. Traditional large-scale evacuation drills are costly and difficult to replicate, which makes numerical modeling an indispensable alternative. A wide range of software packages exist for simulating fire dynamics and passenger behavior, each offering different levels of fidelity, accessibility, cost, and validation robustness. Selecting the most suitable tool for railway fire safety applications is therefore a critical decision. This study introduces a structured decision-making framework based on the Analytic Hierarchy Process (AHP) to evaluate and rank simulation tools according to multiple criteria. The proposed methodology provides a reproducible approach for comparing alternatives and identifying the most appropriate software for railway fire safety engineering. Results highlight the advantages and limitations of the platform PyroSim/FDS in comparison with competing software, demonstrating that AHP is an effective method for guiding tool selection.

Beyond Accuracy: Protocol-Dependent Robustness and Reliability in Patch-Based Breast Histopathology Classification

2026-03-29T10:18:37+00:00

Breast cancer is one of the diseases that cause a dead worldwide among women. Early diagnosis can save many lives and this will lead to a proper treatment. Histopathological image analysis is an important diagnostic method for detecting breast cancer. Computer-aided diagnosis of breast images helps radiologists do the task more efficiently and appropriately. The reliable evaluation of deep learning models in computational pathology depends strongly on how the data is split during validation. In patch-based histopathology classification, random image-level splitting can obscure inter-patient heterogeneity and produce overly stable performance estimates.This study investigates the effect of the evaluation protocol on Invasive Ductal Carcinoma (IDC) detection using the Breast Histopathology Images dataset. A lightweight Convolutional Neural Network (CNN) was trained using balanced sampling and evaluated under two 5-fold cross-validation schemes: strict patient-level separation and stratified random image-level splitting. Under patient-level evaluation, the model achieved a mean accuracy of 0.7826 ± 0.0295 and a mean AUC of 0.8691 ± 0.0257, with a Brier score of 0.1582 ± 0.0159. Random splitting produced similar mean accuracy (0.7867 ± 0.0096) and AUC (0.8652 ± 0.0053), with slightly improved Brier score (0.1509 ± 0.0047). While paired statistical testing showed no significant differences in mean performance (p > 0.05), patient-level evaluation exhibited substantially higher fold-wise variability. Calibration analysis indicated moderate reliability with mild overconfidence, and error inspection revealed consistent morphological failure patterns. Overall, the evaluation protocol mainly affects perceived robustness rather than average performance, highlighting the importance of patient-level validation and variability reporting.

ArDSS: An Arabic Down Syndrome Speech Corpus

2026-04-04T22:06:27+00:00

Down syndrome (DS) is associated with characteristic speech impairments, yet Arabic resources remain limited. This paper introduces ArDSS, an annotated Arabic speech corpus of DS speech for acoustic analysis and Automatic Speech Recognition (ASR) research. Recordings were collected from 13 Arabic-speaking children with DS (6 males, 7 females; ages 5–17) in Saudi Arabia using a clinically guided set of 100 phonetically diverse isolated words, captured in 3–4 sessions per participant under controlled conditions. Following quality control and segmentation, ArDSS comprises 5,509 utterances (16 kHz, 16-bit WAV) totaling 14.6 hours. Each utterance is linked to XML metadata and attempt-level annotations, and a standardized pipeline extracts fundamental frequency, formants (F1–F3), and voice-quality measures (harmonics-to-noise ratio, jitter, shimmer) after feature-validity screening, yielding 3,279 productions with complete acoustic profiles. Corpus analyses indicate substantial inter-speaker variability and recurrent Arabic-relevant pronunciation patterns, including liquid substitution, emphatic neutralization, cluster reduction, and affricate simplification. ArDSS provides a foundational resource for Arabic clinical phonetics and for developing DS-aware ASR and assistive applications.

Smart Healthcare system for diagnosing Autism Spectrum Disorder Using Deep Neural Networks

2026-05-02T12:50:02+00:00

The number of people who have a disease that falls on the autistic spectrum is quickly increasing. Unfortunately, it may be difficult for them to obtain an early diagnosis and receive intervention in order to avoid problematic behaviors, which may lead to social isolation and financial loss for the entire family. Children who fall somewhere on the autism spectrum often have difficulties with social reciprocity, shared attention, and effective language expression. Consequently, autistic children commonly struggle with feelings of isolation. The novelty of the study is the early identification of autism in children based on their behaviors, such as arm flapping, head banging, and spinning, enabling intervention. This study integrates three transfer models—MobileNet, ResNet50, and InceptionResNetV2—with a long short-term memory (LSTM) model to aid early diagnosis of autism spectrum disorder. The Self-Stimulatory Behaviors Dataset (SSBD), which is available to the general public, was used for training. This dataset contains 125 films of persons with autism as well as autistic and non-autistic controls. We used the usual measures of accuracy, specificity, and sensitivity to see how well the three deep learning models worked. The empirical data from the diagnostic system indicated that integrating an LSTM model with ResNet50 achieved a superior accuracy of 93.10%. Every family must have access to an objective, affordable, and easy-to-use diagnostic or screening solution to enable early intervention for their children with ASD. To reach this goal, we developed an early screening tool for autism spectrum disorder that uses deep learning to accurately assess stimming behaviors and children's movements. This enables early diagnosis and treatment of the disorder.

Cybercrime Prediction via Multi-Modal Learning from Behavioral, Network, and Textual Data

2026-04-29T21:00:52+00:00

The increasing sophistication of cybercriminal activities creates significant challenges for traditional security systems, which rely heavily on unimodal detection methods. In this paper, we propose an integrated multi-modal deep learning methodology for better travel predictions by considering time series ordering od behavioral event transaction patterns with network traffic characteristics and textual threat intelligence. However, the suggested framework has modality-specific encoders which include Bidirectional Long Short-Term Memory; for behavioral sequence modeling, Graph Neural Networks, which focus on network topological processing, and BERT-based transformers with a focus on textual feature extraction. A new attention-based fusion mechanism finds latent representations from each modality, allowing the model to adjust source information dynamically depending on its predictive utility. Using three benchmarked datasets, IEEE-CIS Fraud Detection, CICIDS2017, and curated threat intelligence corpora for experimental evaluation of the proposed approach yield an F1-score of 0.912 and ROC-AUC of 0.947 and surpasses the respective single-modalities baselines by margins of 8.3% and 11.7%.

Variable-order fractional calculus, chaotic systems, numerical methods, fractional derivatives, nonlinear dynamics

2026-05-04T16:04:54+00:00

Recently, the V-O-F fractional differential operators have become quite popular for modelling nonlinear dynamical systems with hereditary effects and memory features. However, the V-O-F operator is more suitable for nonlinear dynamical systems with chaotic properties, even though the integer-order derivative operator is a great method for comprehending the behavior of some ordinary differential equations. Here, we present a generalized numerical method to simulate the solution of a broad class of V-O-F differential equations, based on the theorem of fractional calculus and Lagrange interpolation polynomials. The technique may effectively take into account various kernels, including the Atangana-Baleanu-Caputo kernel, the Caputo kernel, and the Caputo-Fabrizio kernel. We use our suggested numerical method to simulate the dynamics of nonlinear chaotic models, specifically a circuit model and a time-varying financial system model.

3D Vision-Based Inspection Using Multi-View Reconstruction and Depth Estimation for Industrial Quality Control

2026-04-26T20:58:14+00:00

Industrial inspection remains difficult when defects are subtle, weakly textured,
or partially occluded, since 2D vision provides limited geometric evidence.
This paper presents a compact 3D inspection framework that combines multi-view reconstruction, depth estimation using the pretrained MiDaS v3.1 DPT-Hybrid model, and Open3D-based geometric analysis for OK/NOK classification and conformity assessment. The pipeline integrates feature matching, pose estimation, reconstruction, dense depth inference, and point-to-reference deviation analysis in one decision process. On a multi-view dataset of mechanical and metallic parts, the method achieves 95.1% accuracy, 94.2% precision, 93.5% recall, 0.46 mm mean geometric error, and 54 ms average processing time per part. Relative to RGB-only and depth-only baselines, the fused framework is more robust to weak-texture and geometry-driven defects while remaining compatible with practical deployment.

Self-Supervised Temporal Transformer Learning for Behavioral Segmentation in Digital Banking

2026-04-19T19:43:27+00:00

Digital banking systems continuously record detailed traces of how customers interact with financial services, offering rich opportunities for behavioral analysis. However, using these data for customer segmentation remains difficult due to the heterogeneity of banking events, strong temporal dependencies, and the lack of labeled segment definitions. This paper introduces a sequence-based segmentation approach that models customer behavior as a stream of multi-modal banking events and learns representations directly from raw interaction logs. The proposed framework relies on a temporal Transformer encoder trained with self-supervised objectives, allowing categorical attributes, numerical values, and inter-event time oinformation to be jointly encoded. Masked event modeling and next-event prediction are used to guide representation learning, while a deep embedded clustering objective integrates segmentation directly into the training process. Experiments on a large-scale real-world digital banking dataset show that the proposed method produces more compact and better separated behavioral segments than traditional aggregation-based techniques and neural baselines. In addition, the resulting segments remain stable across monthly evaluation windows and reflect persistent behavioral characteristics related to activity regularity, spending variability, and engagement patterns. In the future, we also plan to extend this work in several directions, such as incorporating the segmentation objectives with downstream outcomes, performing online segmentation, and conducting a deeper causal analysis of segment jumps. Plus, it would be interesting to inject domain knowledge or expert intervention into the segmentation process.

A Stacking Ensemble Framework for improving Lung Cancer Prediction

2026-04-15T22:27:23+00:00

Lung cancer is the leading cause of cancer-related deaths worldwide and underscores the need to develop effective early detection tools, particularly in health environments with limited resources. This work offers a machine learning model for predicting the risk of lung cancer using non-imaging, formal clinical, and behavioral records from the Kaggle Sanjoli02 dataset. Its methodology uses a systematic, multi-step pipeline consisting of data cleaning, categorical encoding, feature engineering based on correlation, standardization, class-imbalance management with the Adaptive Synthetic (ADASYN), and dimensionality reduction with Principal Component Analysis (PCA) to keep the most discriminative data. A variety of classifiers are then tested, and the most effective heterogeneous learners are stacked in a two-layer ensemble to take advantage of complementary error behavior and increase the final predictive ability. The experimental outcomes prove the usefulness of the suggested framework, where a final classification accuracy of 99.58 is achieved, outperforming the state-of-the-art benchmarks and providing an inexpensive clinical decision-support system that does not require imaging.

Poisoning-Resilient Homomorphic Secure Aggregation for High-Dimensional Cybersecurity Gradients

2026-04-10T23:40:11+00:00

Federated Learning (FL) allows for collaborative model training across several geographically dispersed clients while protecting client data from other users. Although FL is susceptible to poisoning attacks, collusion, and Byzantine adversaries, which can result in decreased performance of the overall model, we present a lightweight FL framework that utilizes homomorphic encryption to protect against poisoning attacks while also providing client data protection and low computational overhead by integrating Additive Homomorphic Encryption (AHE), Trust-Weighted Aggregation (TWA), Anomaly Scoring, and Gradient Clipping. The proposed framework has been extensively simulated with the MNIST and CIFAR-10 datasets. These simulations have demonstrated that the proposed framework can maintain greater than 95% of the overall model's accuracy when the number of malicious clients is as high as 40%. Local epochs are processed at the client side in 15-20 milliseconds; server-side secure aggregation was found to be two times slower than plaintext aggregation; and each round of communication averages approximately 4 MB/client of data transferred. The simulation results provide evidence that the proposed framework will provide robust convergence, scalability, and resilience to a variety of types of adversary behaviors. Existing approaches address either privacy or robustness, but this work proposes a unified framework that simultaneously ensures privacy (via homomorphic encryption) and robustness (via poisoning-resilient aggregation) for applications such as cybersecurity monitoring, IoT, and other privacy sensitive applications.

Case Study: Assessing and Improving the Maturity Level of IT Processes in a Telecom Company using COBIT framework

2026-04-13T09:03:32+00:00

Information technology (IT) is a critical component of most businesses today and will surely expand in importance in the future. Organizations’ boards and senior management are increasingly concerned with successfully managing and administering their IT demands, investments, and resources. This involves addressing the obstacles, opportunities, and challenges of IT management. IT governance is a continual improvement process that strives to preserve value and trust throughout the enterprise, rather than a single-end objective. Many businesses have begun using a risk assessment and mitigation strategy, focusing only on IT governance’s compliance elements, such as COBIT, BASEL III, and SOX. However, they fall short of developing a well-rounded strategy that comprises a thorough framework, a roadmap, and a bottom-up execution plan. Thus, a strategy is required to successfully address the diverse spectrum of IT governance issues and opportunities in a rigorous, coordinated, prioritized, and cost-effective way. This article describes improving eleven IT processes in a telecom firm utilizing the COBIT processes as an enabler, the Process Assessment model, and a full control self-assessment for eleven COBIT processes. The selected case study of a prominent organization was analyzed based on the application of control self-assessment on a telecom company for two consecutive years to measure the achieved maturity improvement in the processes. The proposed implementation is critical for businesses to improve the assessment and enhancement of their IT systems, resulting in a greater level of IT process maturity. The framework may guide each organization to select and modify the appropriate practice based on its situation, strategies, priorities, skills, and resources. The results and implications are obtained from extensive primary and secondary research, as well as evaluating the best business and government practices that exist.

Automated Auditing Framework for Detecting Overfitting and Data Leakage in Machine Learning Models

2026-04-15T09:12:08+00:00

Machine learning now informs decisions in hospitals, banks, and hiring pipelines, where a model's integrity matters as much as its accuracy. Overfitting, data leakage, and feature bias still slip past routine evaluation and surface only after deployment. In this paper, we present an automated auditing framework of six modules that catches these problems beforehand. Overfitting is read from the train-test accuracy gap, stabilized over repeated stratified splits rather than a single partition. Leakage is screened by feature-target correlation and an index-overlap check; feature bias by Mean Decrease Impurity, permutation importance, a model-agnostic SHAP step, and a Gini index. Statistical validation uses stratified cross-validation and paired t-tests corrected for multiple comparisons, and a fairness module reports demographic parity and equalized-odds gaps when a protected attribute is available. In the ablation study the full framework scores 0.677, ahead of every reduced version. We tested it on four datasets with six classifiers. On Breast Cancer Wisconsin it flagged Gradient Boosting and Decision Tree as overfitting; a 30-split analysis confirmed both and placed Random Forest below the 0.05 gap (mean 0.037), where the conventional single-split test would have mislabeled it in about a quarter of splits. Injected leakage was caught at correlation 0.971.

SecMTL-BERT: A Multi-Task Transformer Framework for Unified Information Systems Security Text Intelligence

2026-06-04T15:13:58+00:00

Security operations teams in modern organizations must process volumes of unstructured text incident tickets, threat-intelligence advisories, vulnerability bulletins, and audit logs that far exceed human capacity for manual review. Existing natural language processing (NLP) approaches address these tasks in isolation, requiring separate models and multiple inference passes that are computationally costly and fail to exploit the mutual information shared across related security-text tasks. We propose SecMTL-BERT, a unified multi-task transformer framework that simultaneously performs security incident classification, cyber-entity recognition (CyNER), and threat urgency scoring in a single forward pass. The central architectural innovation is a Cross-Task Attention (CTA) module that enables bidirectional feature sharing between the classification and entity-recognition heads, allowing each task to exploit complementary representations learned by the other. The shared encoder is initialized from BERT-base and subjected to domain-adaptive fine-tuning on a 2.3-million-document security corpus drawn from the National Vulnerability Database, CERT advisories, threat-intelligence reports, and anonymized enterprise SOC tickets. Experiments on three benchmark datasets SecIncident-9K, DNRTI, and ThreatUrgency-8K demonstrate that SecMTL-BERT achieves macro F1 scores of 92.6%, 89.5%, and 90.5% on the three tasks respectively, outperforming the strongest single-task baseline (SecBERT) by 5.7, 6.4, and 6.5 percentage points. Ablation experiments confirm that every architectural component contributes meaningfully to performance. SecMTL-BERT also reduces total inference time by 58.7% relative to running three separate SecBERT models, making real-time deployment in operational security settings feasible. These results have direct implications for IS security managers seeking to automate alert triage, threat intelligence extraction, and advisory prioritization at scale.

IoT-Enabled Fuzzy Logic-based Temperature and Humidity Controller with Irrigation Interval Policy in Outdoor Oyster Mushroom Growing Unit

2026-06-22T22:55:26+00:00

Small-scale farmers and cultivators struggle to maintain optimal growth conditions and reliably produce high yields, as outdoor mushroom farming is highly sensitive to environmental fluctuations. This paper presents an Internet of Things (IoT)-enabled fuzzy logic-based temperature and humidity control system along with an irrigation interval strategy for an outdoor oyster mushroom growing unit. The proposed system employs temperature and humidity sensors for real-time environmental monitoring and a Mamdani-type Fuzzy Inference System (FIS) to dynamically manage a water pump and cooling fan. To address the excessive water consumption observed in conventional FIS-based control, an irrigation interval-based optimization technique is presented. The findings show that even in the face of changing environmental circumstances, the controller successfully keeps the temperature between 24 and 27 °C and the relative humidity over 70%. Stable performance with a low standard deviation and little tracking error from setpoints is demonstrated by statistical analysis. Additionally, while preserving the intended microclimatic conditions, the suggested strategy is affordable, scalable, and greatly increases resource efficiency by reducing water use by almost 80% as compared to the baseline FIS system.

Detecting Plagiarized Text in Images using OCR and NLP-Based Deep Learning Approaches

2026-05-05T21:55:02+00:00

This paper evaluates plagiarism detection using deep learning and natural language processing (NLP) techniques. A novel model, Detecting Embedded Plagiarized Text in Images (DEPTI), is introduced to identify plagiarized text embedded within images, demonstrating high accuracy and robust performance. DEPTI effectively recognizes paraphrased, translated, and artificial intelligence generated content, achieving strong detection capabilities across diverse scenarios. The model integrates PAN-PC-11, TF-IDF, Tesseract OCR, DistilBERT, and LSTM to extract and analyze text from images, enabling advanced plagiarism detection beyond conventional approaches. Experimental results confirm DEPTI’s effectiveness, highlighting its potential as a reliable tool for safeguarding academic integrity in the digital era.