An Auto-ML Framework Based on GBDT for Lifelong Learning

Automatic Machine Learning (Auto-ML) has attracted more and more attention in recent years, our work is to solve the problem of data drift, which means that the distribution of data will gradually change with the acquisition process, resulting in a worse performance of the auto-ML model. We construct our model based on GBDT, Incremental learning and full learning are used to handle with drift problem. Experiments show that our method performs well on the five data sets. Which shows that our method can effectively solve the problem of data drift and has robust performance.