| Abstract | Rainfall prediction has been an important application in meteorology and one of the most scientifically and technologically challenging problem around the world. The objective of this research study is to predict the rainfall into four categories such as low rainfall, medium rainfall, high rainfall & no rainfall, on a short term basis which is limited to tomorrow. In
this research study, analysis of the use of data mining techniques in forecasting the rainfall is done. This study proposes an architecture for the weather information systems which forecast the rainfall using data mining techniques for the country, Australia. This is carried
out by Support Vector Machines (SVM), LightGBM, Naive Bayes & XGBoost classification
algorithms and meteorological data consists of 24 attributes, collected in between the years
2007 to 2017, from 49 different weather stations that cover all over Australia. LightGBM
gave the best accuracy among the four algorithms. The data is collected from Kaggle. Finally, a rainfall prediction system is developed using LightGBM algorithm which has three
major functionalities such as to train, predict and visualize the rainfall data. For training
the system, the admin needs to log in to the system, then upload the data set and train the system. For prediction, users need to first enter the location and enter today’s environment
conditions. Then the system would give output as confidence levels for the four categories of rainfall (low rainfall, medium rainfall, high rainfall & no rainfall). To visualize the rainfall data, users need to select the location and then they can select year and month to get different types of visualizations.
|