Temporal Variations in Monsoon Rainfall: A Comparative Study by Using Machine Learning Techniques

In this study, machine learning techniques-based models used for specifically predicting the “tomorrow” weather (predominantly day rain) as an important measure. Rain in any weather conditions is considered a crucial indicator. Out of the rain-related attributes, the most trusted ones are usually identified through the use of graphical representation. Such visualizations and work out provide insightful analysis about rain and temporal variations. In the modelling phase, the researchers will make the models primary i.e. decision trees, theoretical methods, for this purpose to see their initial prediction capabilities. For seasonal variability in rainfall, it means the changes and shocks in rainfall which follow specific patterns over a year. It represents a range of patterns such as the rain regime which varies with time, space, and time scales, evolving and reflecting the nature of weather systems. Temporal variation knowledge should be of utmost importance for successful water resource management, pluviculture and climate change study. Hence, we will focus on performing an analysis that investigates the impact of the historical weather on prediction through the Seasonal Auto-Regressive Integrated Moving Average (SARIMA) & Long Short-Term Memory (LSTM). The forecasts of the rainfall forecasted by those models are considered valid after the comparisons with the levels of accuracy and efficiency study and the previous models. The ANN (LSTM) model comes out better as compare to ARIMA model. The ANN (LSTM) model with time series showed higher prediction accuracy than ARIMA model for short interval.