As the volume and complexity of data generated by diverse applications continue to grow, anomaly detection in Big Data environments has become a vital challenge. This paper presents an innovative approach that leverages deep learning techniques to enhance predictive analytics for anomaly detection. We introduce a hybrid model that combines convolutional neural networks (CNNs) with long short-term memory networks (LSTMs) to capture both spatial and temporal dependencies in large-scale datasets. The model is trained on labeled data to learn normal patterns and then employs unsupervised learning to detect anomalies. Experimental evaluations on real-world Big Data sets show that our approach outperforms traditional methods, highlighting the adaptability of deep learning in addressing the challenges posed by the dynamic nature of Big Data. This research advances anomaly detection methodologies in complex, high-dimensional datasets, laying the groundwork for robust systems capable of identifying irregularities across a range of Big Data applications.