Flood prediction and early warning systems are critical for protecting lives and property during flood disasters. However, traditional forecasting methods often suffer from limited accuracy, data quality issues, and delayed dissemination. This study presents a Flood Prediction and Advisory System (FPAS) that integrates machine learning, blockchain, and persuasive modeling to enhance flood forecasting accuracy and risk communication. The system was developed using a hybrid OOADM–CRISP-DM framework, combining structured software design with data-driven modeling. A 35-year dataset from the Nigerian Meteorological Agency (NiMet) was curated, preprocessed, and analyzed to train and evaluate Logistic Regression, Random Forest (RF), and XGBoost models. Results showed that RF and XGBoost achieved superior predictive performance (AUC ≈ 0.98) and strong probability reliability, as confirmed by calibration and Brier score analysis. The blockchain layer, implemented through a hybrid on-chain/off-chain architecture, ensures transparency, tamper-resistance, and privacy of flood records. A field survey involving 386 participants across Cross River and Kogi States assessed perceptions of persuasive design. Findings indicated broad community support for FPAS adoption, highlighting the potential of behaviorally informed technologies in disaster management. By merging predictive analytics, ethical blockchain data management, and persuasive communication, FPAS demonstrates a replicable model for climate resilience and disaster preparedness. Future enhancements will focus on real-time data integration and gamified persuasion to strengthen proactive community responses in flood-prone regions.