Transactions on Machine Intelligence2821-16933320200901A New Approach in Designing Electrostatic Energy Harvesters with Precise Modeling of Capacitors in 3D Interdigitated Electrode Structures Considering Full Edge Capacitance Effects13614820894910.47176/TMI.2020.136ENM.JahangiriMEng Student, Department of Electronics, School of Electrical Engineering, Iran University of Science and Technology, Tehran, IranJ.Yavand HasaniAssistant Professor, Department of Electronics, School of Electrical Engineering, Iran University of Science and Technology, Tehran, IranJournal Article20200503The analysis and mathematical modeling of energy harvesting structures facilitate the optimal design of these generators. The capacitive effect between two metal plates in an energy harvester plays a crucial role in its performance. Therefore, mathematically modeling this capacitive effect is essential for the design, performance analysis, and optimization of these generators. In fixed-voltage energy harvesters, interdigitated electrode structures are used. This paper first presents novel closed-form mathematical relations to calculate the capacitive effect in energy harvesters with interdigitated structures. Then, using the derived formulas, a new approach to optimizing the structure and dimensions of these harvesters is provided. The obtained formula for modeling the capacitor in interdigitated structures is highly accurate, considers the 3D structure, and includes the effect of the number of electrodes. The aim of optimizing the target harvester structure is to maximize the harvested energy for a given dimension. The mathematical relations presented for calculating capacitive effects are initially validated using COMSOL and MATLAB software, demonstrating their high accuracy. Ultimately, the proposed optimization approach yields a 23% increase in output power compared to reported references.https://www.tmachineintelligence.ir/article_208949_a57be5188068bfcf96519375ed7c7d30.pdfTransactions on Machine Intelligence2821-16933320200901Model for Detecting Fake News on Twitter14915820899410.47176/TMI.2020.149ENM.Narangi FardDepartment of Network Science and Technology, School of Advanced Science and Technology, University of Tehran, IranZ.HeshmatiAssistant Professor, Department of Network Science and Technology, School of Advanced Science and Technology, University of Tehran, IranJournal Article20200504Due to the widespread use of social networks by people of all ages, distinguishing between fake and real news on these platforms has become a significant challenge in today's world. Individuals who disseminate fake news on social networks often aim to achieve various commercial, political, and economic goals. Therefore, identifying and distinguishing real news from fake news is crucial in addressing this issue. The objective of this research is to present an intelligent model for detecting fake news using a news propagation tree in the social network Twitter. The dataset used in this study is sourced from the political news section of the Fake News Net website. Initially, a news propagation tree was constructed for both real and fake news using this dataset, followed by the development of features from structural, temporal, and syntactic perspectives based on the news propagation tree. Finally, machine learning algorithms were employed to build a model for predicting fake and real news. The results indicated that among the algorithms used for modeling, the Random Forest algorithm, with an accuracy of 75.8%, was the best model for distinguishing fake news from real news.https://www.tmachineintelligence.ir/article_208994_404dd771bf265cbd4142d1ec60ba177a.pdfTransactions on Machine Intelligence2821-16933320200901All-Optical Photonic Crystal Decoder Based on Nonlinear Ring Resonator15916420899510.47176/TMI.2020.159ENH.SharifiAssistant Professor, Department of Computer Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, IranJournal Article20200302The evolution of photonic crystal technology has catalyzed advancements in optical signal processing, particularly through the integration of nonlinear ring resonators. This literature review synthesizes recent research findings related to the development of all-optical photonic crystal decoders, emphasizing their potential applications and identifying significant knowledge gaps that warrant further exploration. In this paper, a photonic crystal structure for designing an all-optical decoder using a ring resonator is proposed. The proposed structure employs a nonlinear ring resonator with nano-silicon crystal material for switching operations. The structure is simulated using the finite-difference time-domain (FDTD) method and the plane wave expansion (PWE) method. This logic gate operates at a rate exceeding 660 gigabits per second.https://www.tmachineintelligence.ir/article_208995_1cdef448045c62522c83b8e9b3f532d2.pdfTransactions on Machine Intelligence2821-16933320200901Comparison of Maximum Power Point Tracking Algorithms in the Presence of Boost and SEPIC Converters16517520899610.47176/TMI.2020.165ENA.AzadianPh.D. student in Electrical Engineering, Imam Khomeini International University, Qazvin, IranJournal Article20200604The Maximum Power Point Tracking (MPPT) control system is a method that can bring a photovoltaic (PV) system closer to its maximum achievable efficiency. Given the various algorithms and classifications available, comparing methods in a specific situation can help select the appropriate algorithm. This paper first provides a comprehensive categorization of MPPT methods and then tracks the maximum power point using three algorithms: P&O, IC, and Fuzzy-PI, in the presence of Boost and SEPIC converters. The P&O, IC, and Fuzzy-PI algorithms are executed once with the Boost converter and once with the SEPIC converter. In both cases, the accuracy and speed of tracking the algorithms and the impact of the converters on output power are examined. By comparing oscillations around the desired point and the settling time in each scenario, the use of a PI algorithm, with coefficients calculated by the Fuzzy method and applied to the SEPIC converter, demonstrates better performance than others. Analysis and simulation of the system were conducted using the Simulink unit in MATLAB software.https://www.tmachineintelligence.ir/article_208996_027b124f9aee2730cd1b6f216da32872.pdfTransactions on Machine Intelligence2821-16933320200901A Novel Multifunctional Multi-Port Integrated Converter for Fuel-Cell Hybrid Electric Vehicles17619020899810.47176/TMI.2020.176ENH.Soltani GohariDepartment of Electrical Engineering, K.N.Toosi University of Technology, Tehran, IranK.AbbaszadehProfessor, Department of Electrical Engineering, K.N.Toosi University of Technology, Tehran, Iran0000-0002-5388-8280Journal Article20200503The effective design of the powertrain for Fuel Cell Hybrid Electric Vehicles (FCHEVs) holds paramount importance in achieving high efficiency. Traditional FCHEV powertrains typically employ separate converters, a configuration that adversely impacts overall system efficiency. This paper presents a novel approach by proposing a multifunctional multi-port integrated converter designed for fuel cell-based hybrid electric vehicles. Specifically, the converter is tailored for FCHEVs incorporating an ultra-capacitor and a battery alongside the fuel cell stack to enhance system efficiency and dynamic response during startup. The versatility of the proposed converter is evident in its ability to operate in three distinct modes: battery charging, propulsion, and regenerative braking. Notably, the regenerative braking mode facilitates the return of energy to the battery, optimizing its state of charge (SOC). A standout feature of this converter is its capacity to charge the battery with a pure sinusoidal input current and controllable active and reactive power. This capability arises from its unique topology and the detailed control strategy elucidated in this paper. To complement the converter's performance, an energy management strategy is introduced, contributing to its efficient operation. The topology, coupled with the control strategy, is rigorously simulated using MATLAB/SIMULINK, with results affirming the thoroughness of the system analysis.https://www.tmachineintelligence.ir/article_208998_d5a57ac4e73490bf83a916e85326929f.pdfTransactions on Machine Intelligence2821-16933320200901A Model for Classifying Social Commerce Texts Using Deep Learning19119820899910.47176/TMI.2010.191ENA.NaserasadiAssistant Professor, Department of Computer Science, Zaran Higher Education Complex0000-0003-1357-732XM.EstilaeiLecturer, Department of Engineering, Payame Noor University, Tehran, IranJournal Article20200601With the rapid expansion of online commerce, a significant volume of data related to these activities is generated and shared daily on social media platforms. Analyzing and processing these data can have numerous applications in enhancing and strengthening social commerce. One such processing task is the classification of social commerce texts, which has notable effects in areas such as better customer experience management, online advertisement generation, and increasing customer demand. In this paper, we propose a model for classifying social commerce texts using deep learning and relevant pre-trained language models. This model first utilizes a pre-trained language model to extract text feature vectors and then uses them for accurate text classification. The results obtained from applying the proposed model to benchmark datasets show that the introduced classification algorithm performs well in classifying social commerce texts, with an average precision score of 0.725 and an average recall score of 0.708.https://www.tmachineintelligence.ir/article_208999_dfeebef9f69c397b94e36cc393c4fb4f.pdf