<![CDATA[Floods are natural disasters caused by high rainfall intensity and poor absorption capacity in an area. The impact of floods results in material and human casualties, necessitating a flood disaster mitigation process. Based on this, the purpose of this study is to predict flood disasters with the Deep Learning (DL) concept using the Deep Neural Investigation Network (DNIN) method, which is a CNN–BiLSTM hybrid. The research method used includes Deep Neural Investigation Network (DNIN) combined with feature selection to predict floods. Feature selection is carried out using the SelectKBest method with the ANOVA F-test (f_classif) evaluation function to select features that have the most significant influence on the target flood variable. The DNIN method extracts features from input data and processes the sequence of these features to capture two-way temporal dependencies before being used for prediction. This research dataset consists of 3000 rows of data sourced from Kaggle (https://www.kaggle.com/datasets/yusufginanjar7/banjir-jabodetabek) with fields name_2, name_3, avg_rainfall, max_rainfall, avg_temperature, elevation, landcover_class, ndvi, slope, soil_moisture, year, month banjir, lat long. The results of this study have proven the application of the Deep Neural Investigation Network (DNIN) method with feature selection is able to predict floods. The results show that the application of the DNIN method with feature selection is able to predict flood disasters with an accuracy level of 93%. Based on the results of this study, the application of the Deep Neural Investigation Network (DNIN) method with feature selection is able to provide a significant contribution in predicting flood disasters accurately and can be used as a decision support system in flood disaster risk mitigation and reduction efforts]]>
