Deep learning multilayer perceptron (MLP) for flood prediction model using wireless sensor network based hydrology time series data mining

Flood disaster is a frequent disaster in Indonesia and causes many victims. To reduce the number of victims, it is necessary to build an accurate flood prediction system. Wireless Sensor Network (WSN) is a component of information retrieval that provides more optimal results in obtaining data time series. Multilayer Perceptron (MLP) as part of the deep learning method is the most vastly used neural network in time series data forecasting. Some contexts which is used in predicting the water elevation level in downstream area are rainfall and water elevation level on weir. The flood prediction system consists of two main parts, namely the remote site and control center. Remote site means equipment in the field / in place of data measurement, while the control center is on a web server that can be opened on any computer via the internet. Multilayer Perceptron can be used as one of the algorithms for predicting flood events based on rainfall time series data, and water levels in a weir. MLP resulted MAPE value of 3.64%, this means the error generated in the system built is 3.64% compared with the real value used as testing. When compared to the multiple regression linier, MLP has better results in predicted water elevation level on downstream canal.