Incidence of floods is a known risk to several localities across the world. Although floods are a result of a natural hazard due to sudden excessive rainfall, reasons caused by people have increased the risk of floods even due to slight rainfalls. Those reasons are mostly connected with various misuses of the land resources that are unfriendly or obstructing natural means of water flows to lows lands and sea.

This strategy requires the techniques for prediction of floods and flood disaster management system to issue alerts, evacuate people urgently from flooding areas. Study area was selected as Kalu River basin area in Sri Lanka, which is prone for high number of flooding. Historical flood and rainfall data are used as the inputs for training data set. IoT devices monitor the changing weather conditions continuously and the readings from the IoT devices are added to training the model for forecasting predictions.

Crowdsourcing offers a technique for gathering information from public crowd and validation. The gathered data sets are analyzed and validated using statistical analysis techniques to organize and classify the final data sets. In this paper machine learning models are developed together with artificial neural networks for predicting flood occurrences and rainfall based on historical data sets. Furthermore, utilization of IoT device and validated crowdsourcing data sets provides supports early warning system to forecast weather information.