Flood prediction in Balikpapan city using logistic regression

The relocation of Indonesian New Capital City (IKN) to East Kalimantan province brings significant impact on buffer city of IKN, especially the city of Balikpapan. The development of IKN influence to the economic growth, but it also increase the risk of flooding because of rapid grow o population and deforestation. Flood is one of major disaster in Balikpapan, and the anomalous weather condition makes Balikpapan suffer for high risk of flooding in the future. Flood prediction can be one of tools for mitigating flood as it can give an early warning system. Therefore, this study is conducted to develop flood forecasting technique for Balikpapan. In this study, logistic regression methods is applied to the flood occurrence data which is a dichotomous variable. Meanwhile, the flood occurrence is predicted by influence some weather parameters as the explanatory variable. Flood event considers as a rare event, so there is an unbalance proportion in the dependent variable. An oversampling technique is applied to address the unbalance problems and it provides better value of Pseudo R2. Moreover, the result of this study showed that the significant variable influenced to flood event are minimal temperature, relative humidity, rainfall, and global horizontal irradiation (GHI). This model has lowest AIC and greater Pseudo R2. From the odds ratio, it can be seen that minimum temperature draws significant impact on the predicting flood event. By assuming the other variables are constant, decreasing temperature multiply the odds of flooding by 3.82 times. Meanwhile, the increasing level of the rainfall can affect the odds of flooding by1.13 times.