This project was submitted to AI for Global Summit.
Implementing Deep Learning Model for Desertification Prediction
Desertification due to drastic climate change and natural disasters have taken a toll on African countries along with several equatorial nations. However, desertification in South Korea is also a considerable risk, according to Statistics Korea. Thus, in order to assess the severity and predict future outcomes, this study developed a method to predict the effects of desertification on the Korean peninsula. In this study, DRI (Desertification Risk Index) is implemented to measure the degree of the situation and predicted via Recurrent Neural Network. Since the variables of DRI consist of four different statistics (T, P, J’(P), NDVI), each dataset was collected from different sources to calculate the index in a given time epoch. The model trained one year of data, from July 2017 to July 2018, and was made to predict the DRI of the next timestep. The predicted DRI was then evaluated via RMSE loss.
View Project Paper