In today's data-driven environments, detecting anomalies in multivariate time series data is crucial across various domains, including healthcare, manufacturing, and finance. Traditional statistical methods often fail to capture complex temporal and inter-variable relationships inherent in such datasets. This research proposes a hybrid deep learning framework that combines autoencoders and Long Short-Term Memory (LSTM) networks to perform unsupervised anomaly detection. The Autoencoder is used for dimensionality reduction and feature extraction, while the LSTM captures temporal dependencies to predict and reconstruct the time series. Anomalies are identified based on deviations in reconstruction error. The model is evaluated on publicly available multivariate datasets and demonstrates superior accuracy and robustness compared to standalone models and conventional anomaly detection techniques. The proposed approach offers a scalable and domain-adaptive solution suitable for real-time monitoring and decision support systems.
Multivariate Time Series, Anomaly Detection, LSTM, Autoencoder, Deep Learning, Reconstruction Error, Predictive Maintenance, Unsupervised Learning, Temporal Modeling, Hybrid Neural Networks