GRADIENT BOOSTING DECISION TREE-BASED METHOD AND DEVICE FOR MODEL TRAINING

摘要

Disclosed are a gradient boosting decision tree (GBDT)-based method and device for training a model, the method comprising: dividing a GBDT algorithm process into two phases; in the former phase, acquiring a labelled sample from a data region of a service scenario which is similar to a target service scenario, sequentially training a plurality of decision trees, and determining a training residual generated after undergoing the former phase training; in the latter phase, acquiring a labelled sample from the data region of the target service scenario, and on the basis of the training residual, continuing to train the plurality of decision trees. Finally, the model applied to the target service scenario is actually obtained by integrating the decision trees trained in the former phase and the decision trees trained in the latter phase.