Computer Vision Application for Effective Flood Forecasting in Public Health Systems

Among the natural disasters, floods are considered as one of the most devastating since they can damage infrastructures, force people leave their homes, and negatively impact the public health. The training of the model seems to be computationally intensive and could prove an interruption where computational resources are a limited commodity. In this paper, we present the Greylag goose driven redefined Long Short-Term Memory (GG-RLSTM) model to be implemented by public health systems in effective flood forecasting. Thus, adapting behavioural patterns in Greylag geese, GG-RLSTM enhances the theoretical structure of LSTM and increases its capability to capture complex relations in the flood processes. From the Kaggle dataset, we gathered meteorological evidence, and geographical features. The results from the experiments indicate that effectiveness of the proposed GG-RLSTM model is higher than the other conventional methods in terms of accuracy (89%), precision (88%), recall (87%), and f1-score (90%). Due to the effectiveness of the model and its applicability in various situations, public health systems can likely adopt it and commence preventive flood measures.