International Journal of Electrical and Computer Engineering Systems

Prof. Goran Martinović, PhD

Irena Galić — 2026-01-05

(August 7th, 1969, Orahovica – October 16th, 2025, Osijek)

Assessment of Battery Degradation Using Rainflow Cycle-Counting Algorithm: A Recent Advancement

Mohamad Faizal Yusman Mohd Hanappi — 2026-01-05

Battery based energy storage systems are increasingly popular in power systems as renewable energy continues to grow while ensuring the reliability of power supply. However, battery degradation is a significant issue that can impact power system operations and optimal scheduling strategies. Therefore, estimating the remaining life cycle or assessing the health of batteries due to the degradation process has become a new challenge and research focus in various engineering fields. This topic is relevant in the context of electric vehicles (EVs), where battery degradation caused by continuous and non-continuous operations (i.e., charging and discharging cycles). Degradation can limit the performance of batteries and occur throughout their lifespan whether they are in use or not. The degradation process is complex and influenced by usage and external conditions that are normally measured by state of health (SOH). Therefore, predicting the SOH of batteries is crucial in ensuring the safety, stability, and long-term viability of energy storage and EVs systems. This prediction requires a battery mechanism model that can be established from a complex electrochemical process. Alternatively, a rainflow cycle-counting algorithm (RCCA) has become popular among researchers for battery degradation estimation because of its simplicity. This paper presents a comprehensive review of the battery degradation estimation using RCCA to count the equivalent cycles of charging and discharging profiles.

Leveraging Word2Vec-Enhanced CNN-LSTM Hybrid Architecture for Sentiment Analysis in E-Commerce Product Reviews

Kosala Natarajan — 2026-01-05

The amalgamation of machine learning (ML) techniques and natural language processing (NLP) is leveraged to evaluate the sentiment of textual input. With the increasing popularity of e-commerce platforms like Amazon, product reviews have emerged as an essential source of information for potential purchasers, providing insights into product quality and performance from the consumers' viewpoints. This study aims to systematically organize and analyze customer opinions to effectively capture consumer sentiment based on product reviews. In this study, we propose a deep learning framework that combines a stacked 1D convolutional layer (CNN) with a Long Short-Term Memory (LSTM) network, using pre-trained Word2Vec embedding as fixed input representations. Evaluated on a large Amazon product review dataset, our model — StackedCNN-LSTM-W2V — achieves a classification accuracy of 99%, outperforming traditional CNN, LSTM, and logistic regression baselines.

Sentivolve: Utilizing FastText, CRF, HAN, and Random Forests for Enhanced Sentiment Analysis

T. Anilsagar — 2026-01-05

The objective of this study is to enhance sentiment analysis through an integrative approach termed Sentivolve, which combines FastText embeddings, Conditional Random Fields (CRF), Hierarchical Attention Networks (HAN), and Random Forests (RF). The system aims to improve sentiment classification by leveraging advanced feature extraction, sequence modeling, attention mechanisms, and ensemble learning. FastText captures subword information for better text representation; CRF models sequential dependencies; HAN highlights key textual elements using a hierarchical attention structure; and Random Forests aggregate predictions to ensure consistent sentiment classification. Experimental results demonstrate that Sentivolve outperforms traditional models in both accuracy and generalizability. This integrated approach provides an effective solution for sentiment analysis, especially in handling diverse and complex text data.

A Video Summarization Technique using Multi- Feature DWHT and GMM for CBVR System

Dappu Asha — 2026-01-05

The increasing utilization of multimedia data and digital information in present times presents a vast scope for research in content-based retrieval systems. An improved CBVR System is proposed to extract video streams effectively using DWHT Multi- features and GMM. Our CVBR method performs VSBD for identifying Video shots by computing DWHT on video frames for multi- feature extraction, and then key frames are identified. A summarized frame is developed using the VS algorithm based on GMM on the UCF Dataset. Later, a procedure is applied for the input query video stream, and correlation coefficients are calculated between the query and the database multi-feature vectors, giving us similarity measures. Lastly, our experimental results validate the efficiency of our proposed CBVR System, achieving an average precision of 0.821 and a loss of 0.179, outperforming existing CBVR systems using DCT and optimized perceptual VS, which have precision values of 0.6475 and 0.71, respectively, along with losses of 0.3525 and 0.29.

Privacy Integrity-Aware Blockchain Communication in Federated Edge Learning Platform

Chitresha Jain — 2026-01-05

The blockchain architecture offers transparent security mechanisms in a decentralized manner; due to this, it has attained increasing growth in a federated edge-server learning environment. In federated learning, the data model is executed in multiple edge servers in a collaborative manner, increasing users’ privacy and data-integrity breach because of single point failure attack in the main computational server. Blockchain employing a rewarding mechanism in federated edge-learning platform aids the model to overcome single-point aggregation failure. However, the current method failed to identify selfish and baized workers; further, reaching global consensus model to assure privacy-integrity in blockchain-enabled federated edge-server is difficult. This paper presents privacy-integrity-aware blockchain communication (PIABC) in federated edge-server learning platform. The PIABC model is very effective in comparison with existing blockchain-privacy preserving schemes for identifying the correctly aggregated packets and eliminating malicious packets within the federated edge-server learning platform.

Parallel and Distributed Multi-level Entropy- Based Approach for Adaptive Global Frequent Pattern Mining in Large Datasets

Houda Essalmi — 2026-01-05

Frequent pattern mining in distributed settings remains a significant challenge due to predominantly high computational expenses and high communication overhead. This paper presents AGFPM (Adaptive Global Frequent Pattern Mining), a novel solution that integrates an extensible Master-Slave architecture with an advanced pruning technique that relies on binary entropy and statistical quartiles. AGFPM proposes two primary data structures: the LP-Tree (Local Prefix Tree) and the GP-Tree (Global Prefix Tree). A single pass of each local Slave site is used to build one LP-Tree, and low information value branches are pruned early on by entropy and quartile thresholds. Rather than transferring complete trees, only succinct metadata is sent to the Master site, where the GP-Tree is built from globally sorted items in order of their entropy rankings. A significant aspect of AGFPM is the flexible pruning approach: either the GP-Tree is pruned or not pruned, based on user criteria. This provides a dynamic adjustment between the performance and generality of results, thereby allowing control over the level of compression applied when generating global patterns. Global frequent patterns are then recursively mined from the GP-Tree based on conditional sub-GP-Trees. Frequent patterns are extended at each level of the hierarchy by intersecting the common prefix paths, guided by a Global Header Table. AGFM demonstrates improved performance in execution time, scalability, and robustness against low support thresholds relative to existing methods.