Past floods have caused millions of dollars in damage to vital infrastructure. While much research has gone into the topic, as of yet, there is no global system that can be utilized to collect, store, assess, and forecast floods. Researchers throughout the world are trying to find a way to gather, store, and analyze massive amounts of data on floods to better anticipate the results of flood-based prediction systems. In this study, a deep learning model and a hybrid classification algorithm were utilized to build an Internet of Things-based water and disaster management system. First, the large data gathering from the flood is used to gather the input data. The system was built using the water flow (WF), water level (WL), rain sensor (RS), and humidity (HS) sensors. The redundant information is then removed from the IoT data by using HDFS map-reduce. Data are pre-processed using missing value imputation and a normalizing approach after being cleared of duplicates. Therefore, an approach that combines attributes and characteristics is used to construct a rule. A Convolutional Deep Neural Network (CDNN) and Artificial Neural Network (ANN) hybrid classifier sorts the rules into two groups: a) those with a high likelihood of a flood happening, and b) those with a low probability. The proposed method’s efficacy is evaluated using standard statistical measures of reliability and validity. In addition, the proposed technique provides a much more precise result than other algorithms.