Comparative and Hybrid Analysis of CNN and LSTM Architectures in Deep Learning
Keywords:
Convolutional Neural Networks (CNN); Long Short-Term Memory (LSTM); Deep Learning; HybridAbstract
Background: Recent advancements in deep learning have significantly improved artificial intelligence applications, particularly through Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. However, standalone models often fail to effectively capture both spatial and temporal dependencies in complex data. Objective: This study aims to compare CNN and LSTM architectures and propose a hybrid CNN–LSTM model to improve performance in tasks involving spatiotemporal data. Methodology: A quantitative experimental approach was adopted using benchmark datasets, including CIFAR-10 for image data and the UCI Human Activity Recognition (HAR) dataset for sequential data. Individual CNN and LSTM models were developed and compared with a hybrid CNN–LSTM architecture. Standard preprocessing techniques, feature engineering, and evaluation metrics such as accuracy, precision, recall, and F1-score were applied. Results: The hybrid CNN–LSTM model achieved superior performance, with an accuracy of 88%, outperforming standalone CNN (80%) and LSTM (81%) models. The results demonstrate improved robustness and generalization through combined spatial and temporal learning. Conclusion: The findings confirm that hybrid architectures provide a more effective solution for complex real-world problems involving multimodal and sequential data. This study contributes to the development of efficient and scalable deep learning systems with potential applications in healthcare, activity recognition, and intelligent systems.




