https://eijms.com/index.php/ms <p><strong><span id="cell-5-name" class="gridCellContainer"><span class="label">International Journal of Mathematics and Statistics (ISSN: 2208-2212) </span></span></strong>publishes a wide range of high quality research articles in the field (but not limited to) given below: mathematics, applied mathematics, applied commutative algebra and algebraic geometry, mathematical biology, physics and engineering, theoretical bioinformatics, experimental mathematics, theoretical computer science, numerical computation etc.</p> ARC Publishing of Books and Other Publication Services L.L.C en-US International Journal of Mathematics and Statistics 2208-2212 https://eijms.com/index.php/ms/article/view/80 <p><em>The core function of data science depends on optimization because it provides the base for both algorithmic speed and statistical modeling and decision processes. Research examines optimization methods from a mathematical perspective through evaluations of their theoretical foundations together with their convergence attributes and calculations requirements. The research investigates classical gradient-based approaches Gradient Descent and Newton’s Method with complete convergence analysis and establishes the role of Karush-Kuhn-Tucker (KKT) conditions and Lagrange duality for analyzing convex optimization problems. The discussion focuses on non-convex optimization challenges because traditional methods fall short for these problems yet metaheuristic approaches including Simulated Annealing, Genetic Algorithms, and Particle Swarm Optimization solve complex high-dimensional problems effectively.</em></p> <p><em>The study recognizes three main mathematical optimization difficulties: solving large-dimensional optimization issues and finding efficient methods for deep learning while achieving the proper balance between exploration and exploitation. Research proposals outline a strategy to connect classical and heuristic optimization methods by integrating machine learning-based techniques that create adaptive and reliable optimization models. This study produces findings that will impact data science along with artificial intelligence as well as computational mathematics since they create a foundation for upcoming developments in optimization-driven methodologies.</em></p> Dr. Charudatta Dattatraya Bele Mobin Ahmad Maheshwari Munigala Dr. Dhananjay Raosaheb Vidhate Copyright (c) 2025 International Journal of Mathematics and Statistics 2025-05-01 2025-05-01 11 1 11 15 10.53555/eijms.v11i1.80 https://eijms.com/index.php/ms/article/view/78 <p>A research examines machine learning algorithm applications for financial data prediction in markets by using synthetic data to minimize reliance on traditional datasets. The study develops a hybrid framework that combines deep learning and reinforcement learning models to enhance forecasting precision. GANs and VAEs work together to generate synthetic financial data through advanced network architectures. The prepared data is improved through logarithmic returns and technical indicator techniques. Transformer-based models deliver better performance than LSTM networks because they decrease prediction errors by 12% compared to LSTM networks. At the same time Deep Q-Networks (DQNs) operating under the reinforcement learning framework generate 23% superior cumulative returns compared to typical trading methods. Research outcomes indicate that artificial intelligence-driven models can improve market predictions and company decision-making functions. The research develops a new synthetic data method that boosts financial prediction while delivering essential information for automated traders and financial establishments.</p> Dr.Ashish Kumar Soni Dr Genu Roney Varghese Mrs Nirmala Roney Varghese Ms. Vallerina Roney Varghese Ms. Vinitaa Roney Varghese Mr. Basavaraj Basannagari Copyright (c) 2025 EPH - International Journal of Mathematics and Statistics 2025-03-04 2025-03-04 11 1 1 10 10.53555/eijms.v11i1.78