Tropical climate prediction method combining random forest and feature fusion

Tropical cyclones pose significant threats to coastal populations, causing destruction and loss of life. Precisely forecasting the frequency and arrival dates is still a challenge. This research presents a technique for feature extraction and integration using a random forest (RF) model with a cascaded convolutional neural network. The approach combines different meteorological maps and uses a feature fusion technique to improve prediction accuracy. The RF model is optimized by a grid search algorithm. The results show that the proposed model outperforms conventional models to achieve a mean absolute error of 0.48 and a mean relative error of 14.14%.