Leveraging Machine Learning for accurate Precipitation Forecasting

Forecasting precipitation is a vital in alerting people about potential challenges related to rain and enabling them to take active safety measures. Precipitation prediction will enhance the agricultural productivity, ensures a stable food supply, and maintains good water quality. This study emphases on developing the most accurately prediction model for the daily rainfall, considering the major effects of inconsistent rainfall in urban areas. The complex nature of precipitation, driven by a variety of environmental factors, which makes difficult to predict accurately. In this work, real-time environmental features for the model are fetched from Open Weather API. Subsequently, the Pearson correlation coefficient technique is applied to pick the relevant features which induce rainfall are utilized in the model as an input. The dataset was gathered from the resident meteorological division to evaluate the performance of various machine learning techniques such as Random Forest (RF), Light Gradient-Boosting Machine (LightGBM), Extreme Gradient Boost (XGBoost). In order to efficiently explore more parameter space, the swarm intelligence algorithm Fish Swarm Optimization (FSO) was integrated with a better performing model. To validate the robustness and performance of the models, various classification metrics are utilized. With the accuracy of 87%, the XGBoost-FSO Hybrid Model performs superior than other models in predicting daily rainfall.