Floods have destroyed infrastructure worth millions of dollars in the past, and this is still an issue today. Despite all of the research, there is currently no single, worldwide system that can be used to gather data, store it, analyze it, and predict floods. Researchers all across the world are attempting to create a solution that will allow them to collect, store, and analyze massive volumes of flood data in order to forecast the outcomes of flood-based prediction systems. This study created a water and disaster management system based on the Internet of Things, which used a deep learning model and a hybrid categorization technique. First, the input data is drawn from a vast collection of data called as the flood big data set. The system was built using four Internet of Things sensors: the Water Flow (WF) sensor, the Water Level sensor (WL), the Rain Sensor (RS), and the Humidity sensor (HS). Following that, HDFS map-reduce is utilized to reduce the quantity of redundant data in the IoT sensed data. Following the removal of repetitive data, the data is pre-processed using missing value imputation and a normalizing algorithm. As a consequence, a rule is created that makes use of a mix of attributes and attributes approach. At the last step of the classification process, a hybrid classifier that blends Convolution Deep Neural Network (CDNN) and Artificial Neural Network (ANN) classifiers categorizes the rules as a) probabilities of a flood occurring and b) odds of no flooding occurring. Several criteria, including sensitivity, specificity, accuracy, precision, recall, and F-score, are used to compare the results of the proposed approach. Furthermore, when compared to current algorithms, the suggested technique yields considerably more accurate findings.