The advancement of healthcare databases presents numerous opportunities for the integration of machine learning and artificial intelligence technology. In the medical industry, there are numerous medical devices available; however, clinical errors remain a severe challenge. The process of anomaly detection is to identify unexpected behavior or sequence in the health care report, which can help medical professionals reduce or avoid the patient’s health issue. Anomaly detection algorithms can identify unusual access patterns, find data corruption, malicious activities, and monitor real time data. Machine learning algorithms excel in recognizing complex patterns and finding illnesses faster due to various learning factors. This paper provides a system for locating abnormalities in healthcare data and offering thorough justifications for such anomalies. This paper starts with the discussion of various anomaly detection techniques, a literature survey, and finally proposes hybrid models to enhance the performance of healthcare anomalies using unsupervised learning methods. To achieve this, we employed and evaluated three machine learning algorithms and three hybrid algorithms, which are Autoencoder, Isolation forest, and clustering, in their capability to identify anomalies in diabetics and cancer data points. The experimental results show the Autoencoder-based anomaly detection metrics have the best performance on this task across two vast datasets. Anomaly detection can be a highly valuable method for uncovering epidemiological significant patterns within extensive datasets that span both dimension of time and space. Our analytical approach can be applied to other illness and locations to enhance tracking, enable quick responses, and support efforts to control endemic diseases.