Weather forecasting using DBSCAN clustering algorithm

Abstract

The main objective of this study is the clustering of meteorological parameters and forecasting weather in the region of Annaba (Algeria) using clustering techniques. The proposed two-stage clustering approach is based on the first stage, on the proposition of ANN-DBSCAN, a combination of the DBSCAN algorithm and an Artificial Neural Network (ANN) for grouping the clusters. Internal indices of validation were used to compare and verify the correctness and efficiency of the results. Our experiments identified five groups, each of which was associated with the area’s usual weather parameters. Our proposed incremental DBSCAN is employed in the second stage to determine the data pattern that can predict the future atmosphere. The natural molecules of the measured pollutants (nitrogen dioxide (NO2), ozone (O3), carbon dioxide (CO2), and sulfur dioxide (SO2)) are directly dependent on weather forecasting. The focus of this research is on a section of the Samasafia database. The proposed algorithm is used to determine the weather trend in that database. Advanced numerical analysis was applied to a few prediction tasks.