Drought Prediction Using Machine Learning Forecasting Model in the Context of Bangladesh During 1981–2018

Drought is a short engagement in water or moisture availability substantially under the regular or expected. The amount for particular length drought occasions is recognized month-to-month. In this study, rainfall records from Bangladesh Meteorological Department Bangladesh with proved standard precipitation index (SPI) between 1981 and 2017 are used. Historical document of drought is received from Bangladesh Bureau of Statistics. The International Disaster Database may be used to validate the SPI result, and SPI is calculated at the District of Dinajpur in Bangladesh. The SPI can observe droughtphenomena on a nearby scale. For this study, we have calculated SPI scale value for the dataset of rainfall like SPI month-1; month-3, month-6, month-9, and month-12, and we have used the short-term scale value for detecting the drought scale. We have calculated in this one sub-area throughout the country to detect the drought using shiny packages which is server based. In our study, we have observed that moderate drought frequency is excessive anywhere in the country. Based on the forecasting method FB Prophet which is also known as time series algorithm for forecasting, we have predicted the future year SPI value for the month which has given most moderate drought according to our study. Our main purpose is to be able to detect the drought of upcoming year so that we can be able to take necessary steps to prevent it. In our study, the forecasting methods are most accurate to predict the drought forecast for future.