RSFD: A rough set-based feature discretization method for meteorological data

Meteorological data mining aims to discover hidden patterns in a large number of available meteorological data. As one the most relevant big preprocessing technologies, feature discretization can transform continuous features into discrete ones improve efficiency algorithms. Aiming at problems high interaction multiple attributes, noise interference, and difficulty obtaining prior knowledge data, we propose rough set-based method for (RSFD). First, calculate information gain each candidate breakpoint attribute split intervals. Then, use chi-square test merge these Finally, take variation indiscernibility relation set as evaluation criterion scheme. We scan turn by using strategy splitting first then merging, thus optimal set. compare RSFD with state-of-the-art methods on Experiments show that our achieves better results classification accuracy obtains smaller intervals while ensuring consistency.