基于TabNet的森林覆盖类型预测方法

Forest Cover Type Prediction Method Based on TabNet

森林覆盖类型的分类识别对于研究森林资源变化、合理利用森林资源具有重要意义,构建准确、鲁棒的分类模型是此类研究的关键。采用TabNet构建预训练模型对输入特征进行重构,然后使用重构后的特征表征训练TabNet分类器,最后与LightGBM模型进行集成,提升了模型效果。实验显示该模型的分类性能最高,准确率达到97.95%,提出的模型具有良好的可解释性,对筛选森林覆盖类型影响因素有着指导意义。

The classification and identification of forest cover types is of great significance to the study of forest resource changes and the rational use of forest resources.Building an accurate and robust classification model is the key to this research.In this paper,we use TabNet to construct a pre-training model to reconstruct the input features,then use the reconstructed feature representation to train the TabNet classifier,and finally integrate with the LightGBM model to improve the model effect.Comparative experiments show that the classification performance of this model is the highest,with an accuracy of 97.95%.The proposed model has good interpretability and has guiding significance for screening the influencing factors of forest cover types.