Flood is the most common disaster in Indonesia and certainly harmful to society in the form of material or psychical. Therefore, it's necessary to identify the potential and flood mitigation earlier to reduce the potential losses suffered by the society after the occurrence of disaster. This is difficult to do with conventional methods so that in this research proposed "Neural Network Learning Vector Quantization as Identification Method of Potential and Mitigation of Flood Disaster". With this algorithm specified four nodes input layer, one hiden layer with two neurons and two output layers where four nodes input layer are elevation, drainage, rainfall and flood events are derived from data of BPS Malang, BMKG Karangploso, and data of BPBN. Data processing and testing will generate two outputs, they are identification of flooding potential area and no flooding potential area in every villages in Malang. The test results by using confution matrix showed the accuracy value at 95.34%, sensitivity value at 100%, specification value at 95.29%, and error rate at 4.68% on 1710 dataset that composed of 70% training data and 30% testing data with learning rate at 0.1, decrement learning rate at 0.01, maximum epoch at 10 and minimum epoch at 0.0000001. I.