A cloud-enabled predictive analytics model for assessing health risks under climate variation

The relationship between climatic conditions and disease spread is a crucial aspect of public health. Factors such as temperature, humidity, precipitation, seasonal changes, and air pollution directly and indirectly influence the transmission of infectious diseases, including respiratory illnesses, heat-related disorders, vector-borne diseases like malaria and dengue, and foodborne infections. Large-scale climate variations can further intensify these impacts. To address these challenges, effective prediction models are essential for minimizing health risks and supporting timely decision-making. This paper introduces a novel method called the Weighted Iterative Piecewise Regression–based Improved Deep Multilayer Perceptron Classifier (WIPR-IDMPC) for predicting health impacts under varying climatic conditions using cloud data centers. The primary objective of the proposed model is to enhance prediction accuracy while reducing time consumption and error rates. The framework consists of three phases: data acquisition, feature selection, and classification. In the data acquisition phase, IoT devices collect large volumes of weather and health-related data, which are transmitted to the cloud for processing. The cloud employs an improved deep multilayer perceptron classifier consisting of an input layer, multiple hidden layers, and an output layer. The Weighted Piecewise Regressive Relief method is used to select relevant features by removing redundant and irrelevant attributes, thereby reducing computational time. The selected features are fed into the classifier, which processes training and testing samples to categorize health impact levels such as very low, low, medium, high, and very high. The performance analysis indicates improvements of accuracy by 98.24% and precision by 98.18% and recall by 98.69%. Additionally, the proposed method achieves a significant reduction in RMSE by 1.75, along with a decrease in prediction time by 68.38 ms compared to traditional approaches. • Introduce a predictive analytics model that improves climate-related health impact forecasting accuracy and efficiency. • Utilize large-scale cloud data with sensor inputs for real-time health risk assessment and decision making. • Employ feature selection to isolate influential climate variables and reduce processing effort for learning models. • Enhance predictive performance using a deep multilayer classifier for multi-level health severity categorization. • Demonstrate measurable improvements in accuracy, error reduction, and prediction speed for timely public health decisions.