A Hybrid Time Series Rainfall Prediction Model Using Neural Prophet and LS TM
G. Lakshmi Vara Prasad, B. Ravi Teja, Talluri Haribabu, G Naga Pavani, Dammu Karunamma, Kolla Vivek
Abstract
India has a reputation for having frequent torrential downpours. Andhra Pradesh (AP) is inextricably linked to the catastrophe when the downpour is very heavy since it is the crucial area in Telugu-speaking states where the center of the government and non-government sectors is situated. Hence to safeguard the populace from the prospect of a flood catastrophe, the AP Ministry needs an evidence-based approach to address probable catastrophes that might occur each year. Annual rainfall forecasts can be used as a guide to estimate the likelihood of significant loss and damage from catastrophe risks. Finding an analytical approach that works in this situation is still difficult at the time. In this study, a hybrid time series system is used to forecast the downpour in AP by fusing LSTM (Long Short Term Memory) with Neural Prophet (NP). The model uses past meteorological data to accomplish the forecast process and generates types of downpours along with wind speed and temperature. This study introduces a fresh line of inquiry for LS TM + NP downpour quantity prediction. Using the datasets that are freely available in internet sources, the LS TM+NP model was compared to existing individual models of traditional frameworks in the investigations presented in this study. The outcomes of the findings demonstrated that the LS TM+NP hybrid system outperformed other time series forecasting measures. RMS E of 0.25 and MAE of 0.16 was attained using the LS TM+NP technique.