Non-communicable diseases (NCDs) such as cardiovascular diseases, cancers, chronic respiratory diseases, and diabetes account for a significant portion of global deaths. Early prediction of these diseases is crucial for effective management and cost reduction in healthcare. However, challenges such as temporal dependency in health data pose significant obstacles to accurate early prediction. This study proposes a dynamic prediction model using a sliding window approach that focuses on capturing temporal patterns in sequential health data. The sliding window method preserves the sequence structure of data, allowing for more accurate and timely predictions. Results show that the GRU model achieved the highest performance, with an accuracy of 95%, precision of 94%, recall of 97%, and an F1-score of 95%, outperforming other models such as LSTM, Random Forest, XGBoost, and KNN. The major contribution of this work lies in leveraging the sliding window method to enhance sequential model performance, thereby improving the timeliness and accuracy of early disease predictions and ultimately aiding proactive healthcare management. The performance of various machine learning models trained using this method will be evaluated and compared to traditional approaches to demonstrate its effectiveness in early NCD prediction.