per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233147165articleM. Heydarianmohsentabriz900@gmail.com1Sagar Gorbaniheydarian@azaruniv.ac.ir2Dept. of Info. Tech. and Comp. Eng., Azarbijan Shahid Madani Universityو ،شذقهظو ]قشد Dept. of Info. Tech. and Comp. Eng., Azarbijan Shahid Madani Universityو ،شذقهظو ]قشد <p style="direction: ltr;">In Mobile Internet of Things (MIoT) or Wireless Sensor Network (WSN) networks, which can be fog-based, there are challenges such as energy consumption management, Quality of Service (QoS) improvement, and reliability, which have attracted a lot of research. have given. Limited resources and dynamic topology in these networks have made these challenges more complicated. In the Internet of Things network, wireless sensors use a battery with a limited capacity to supply their energy, so operations such as data collection and routing that cause energy consumption need to be optimized. In order to manage these challenges, the principles of green network design and optimization methods can be used to help increase the life of the network, optimize energy consumption, and increase network efficiency. In this article, linear optimization methods and graph theory-based algorithms are used, and a new routing algorithm is presented that can improve optimal energy consumption, QoS, network lifetime, and efficiency. Mathematical modeling and simulation of the new method show that this algorithm, compared to existing methods, can pass more data through the network by taking shorter routes and use the network resources optimally</p>http://ijece.org/Article/43973Optimal transmission minimum spanning tree wireless sensor network network lifetime.per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233209216articleR. Molaee Fardmolaefard@gmail.com1J. Mohammadzadehj.mohamadzadeh@kiau.ac.ir2payam yarahmadiyarahmadi.pay@gmail.com3Dept. of Comp. Eng., Faculty of Comp. and Inf. Tech., Science and Research Branch, Islamic Azad University, Tehran, IranDept. of Comp. Eng., Karaj Branch, Islamic Azad University, Karaj, IranDept. of Comp. Eng., Taras National University of Kyiv, Kyiv, Ukraine<p style="direction: ltr;">This research presents a method to improve mobile application recommendation systems using collaborative filtering and the cuckoo meta-heuristic algorithm. The SW-DBSCAN algorithm was used for data clustering. This algorithm was able to achieve an efficiency of 99% in the clustering section, which was better than other similar algorithms. Also, the cuckoo algorithm was used to optimize the data, which achieved a performance of 98% and achieved better performance than the firefly, gray wolf, and particle optimization algorithms. For the prediction part, a neural network algorithm was used, which was also able to achieve acceptable performance compared to other similar algorithms. Ultimately, this information is provided to the user using a recommender system based on collaborative filtering. One of the problems that recommender systems face is the problem of having the same meaning. The problem of having the same meaning occurs when an item is represented by two or more names, names that have the same meaning. In such cases, the recommender system cannot distinguish whether these names represent different items or all refer to the same item, which is what this method attempts to solve. Also, according to research, using this recommended method can correctly identify user needs up to 94% of the time and offer appropriate suggestions to the user.</p>http://ijece.org/Article/48615Recommender system mobile application collaborative filtering cuckoo algorithm data mining.per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233166178articleF. Kazem zadehm.hooshmand@imamreza.ac.ir1M. Houshmand Kafashianm_houshmand@yahoo.com2M. Houshmandm_houshmand61@yahoo.com3Imam Reza International UniversityTelecommunication Company of IranImam Reza International University<p style="text-align: left;">&nbsp;Time series forecasting in engineering, telecommunications, and finance is of great importance. Financial time series, which are often multivariate, require precise and optimized algorithms. In recent research, deep neural networks have demonstrated successful results in improving the accuracy of financial time series forecasting. This study investigates the use of LSTM and GRU networks in predicting cryptocurrency prices and examines the bidirectional implementation of these networks, with an emphasis on optimal hyperparameter selection to reduce prediction error and enhance accuracy through Grid Search, RandomizedSearchCV, and Bayesian methods. Simulation results indicate that employing bidirectional LSTM and BiGRU networks reduced the prediction error rate for BTC by up to 3.22%, for ETH by 3.94%, and for LTC by 3.99%.</p>http://ijece.org/Article/49818Time series forecasting deep learning bidirectional neural network cryptocurrency price prediction neural network prediction error price prediction error.per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233191199articleA. Hamedooni Asliabhamedoni1653@gmail.com1Sh. Javidanishjavidani1059@gmail.com2َA. Javidanialijavidanii@gmail.com3Jahad-e-Daneshgahi Institute of Higher Education, Hamedan, IranJahad-e-Daneshgahi Institute of Higher Education, Hamedan, IranDepartment of Computer Engineering, Faculty of Engineering, Bu-Ali Sina University, Hamedan, Iran<p style="text-align: left;">This paper presents a novel approach for recognizing ongoing actions from segmented videos, with the main focus on extracting long-term features for effective classification. First, optical-flow images between consecutive frames are computed and described by a pretrained convolutional neural network. To reduce feature-space complexity and simplify training of the temporal model, PCA is applied to the optical-flow descriptors. Next, a lightweight channel-attention module is applied to the low-dimensional PCA features at each time step to enhance informative components and suppress weak ones. The descriptors of each video are then aligned and followed over time, forming a multi-channel 1D time series from which long-term patterns are learned using a two-layer stacked LSTM. After the LSTM, a temporal-attention module performs time-aware aggregation by weighting informative time steps to produce a coherent context vector for classification. Experiments show that combining PCA with channel and temporal attention improves accuracy on the public UCF11 and jHMDB datasets while keeping the model lightweight and outperforming reference methods. The code is available at: https://github.com/alijavidani/Video_Classification_LSTM</p>http://ijece.org/Article/50755Video Classification Human Action Recognition Deep Learning Convolutional Neural Networks Recurrent Neural Networks Long-Short Term Memory (LSTM)per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233200208articleS. Seydisakinehseydi@ut.ac.ir1M. Salehimersali@ut.ac.ir2University of Tehran<p style="direction: ltr;">Artificial neural networks are a class of computational models inspired by the structure and functionality of biological neural networks in the human brain. Convolutional Neural Networks (CNNs), as a prominent type of these models, are widely applied in various domains such as image classification, object detection, natural language processing, and healthcare.</p> <p style="direction: ltr;">As CNN architectures grow in size, the number of parameters and the volume of data movement increase, leading to higher dependence on off-chip memory, which in turn significantly raises energy consumption. A primary strategy for reducing both energy usage and off-chip memory accesses is to maximize data reuse at every level of the memory hierarchy. Data reuse can be exploited at three levels: (1) data flow and processing elements, (2) loop and computation scheduling, and (3) inter-layer and network-level operations.</p> <p style="direction: ltr;">Tiling is one of the key techniques for improving data reuse at the scheduling level. In this work, we present a precise mathematical formulation for modeling the number of data reuses. We then formulate an optimization problem to determine the optimal parameters for maximizing data reuse for each network configuration. Furthermore, we investigate the relationship between network structural parameters, such as kernel size and stride, and the optimal tile size. Our analysis shows that, in 70% of the network layers examined, the optimal tile size is smaller than four times the kernel size.</p>http://ijece.org/Article/48678Convolutional neural networks energy consumption off-chip memory data reuse tiling.per Iranian Research Institute for Electrical Engineeringفصلنامه مهندسی برق و مهندسی کامپيوتر ايران16823745168237452026-01233179190articleZ. Abbasizabasi@gmail.com1<p>Imbalanced datasets are one of the major challenges in the automatic diagnosis of diseases. The imbalance in data classes leads to failures in diagnosis, which can be particularly dangerous for diseases such as breast cancer. In this study, a modified version of the ReliefF algorithm, which is a feature selection algorithm, is proposed. The modifications have been made to select and balance instances effectively. The proposed algorithm balances the number of instances in breast cancer datasets to improve diagnosis. In this algorithm, instances are weighted and ranked. After ranking them, the dataset is balanced using the proposed oversampling method based on the instance weights. This algorithm has been applied to two breast cancer datasets: Wisconsin Breast Cancer Dataset (WBCD) and Wisconsin Diagnostic Breast Cancer Dataset (WDBCD). The balanced dataset was then classified using various classification algorithms. The classification results show that performance evaluation metrics have improved compared to the classification of the original data. The best results obtained in&nbsp; WBCD dataset are Accuracy = 98.04%, G-Mean = 98.00% and in WDBCD dataset are Accuracy = 98.31%, G-Mean = 98.35%. The obtained results indicate the effectiveness of the proposed algorithm in breast cancer diagnosis.</p>http://ijece.org/Article/49704Imbalanced datasets automated disease diagnosis oversampling.