AI-Based Smart Hospital Analytics Using Hadoop and Machine Learning

Modern hospitals generate large volumes of heterogeneous data from electronic health records (EHR),laboratory systems, imaging archives, pharmacy records, and bedside monitoring devices. Conventional relational systemsstruggle to store and process this data at scale, which limits timely clinical and operational decision-making. This paper presents a smart hospital analytics framework that combines the Hadoop ecosystem (HDFS, MapReduce, Hive, and Spark) with machine learning (ML) to support patient readmission prediction, lengthof-stay estimation, and resource-utilization analysis. The framework follows a layered architecture comprising data acquisition, distributed storage, processing and feature engineering, ML analytics, and visualization layers. LogisticRegression, Decision Tree, Random Forest, and Gradient-Boosted Trees are trained with Spark MLlib on a distributed cluster. Evaluation on a public hospital encounter dataset shows that ensemble models outperform linear baselines in AUC and F1- score, and that the distributed pipeline reduces execution time substantially as worker nodes are added. Privacy, security, and deployment challenges are also discussed.