A data processing method for a care-giving robot and an apparatus are disclosed, to resolve a prior-art problem that a care-giving robot can perform selection from specified interaction modes only by calculating an emotion status of an interaction object, but cannot provide a more appropriate interaction mode for the interaction object. The method includes: receiving, by a model engine, data of a target object, and generating a growing model capability parameter matrix of the target object (S401)； adjusting, by the model engine, an capability parameter adjustment value in the growing model capability parameter matrix based on an adjustment formula coefficient or based on a standard growing model capability parameter matrix, to determine an adjusted capability parameter adjustment value (S402)； determining, by the model engine, whether the adjusted capability parameter adjustment value exceeds a preset threshold (S403)； and sending, by the model engine, the adjusted capability parameter adjustment value to a machine learning engine if the adjusted capability parameter adjustment value is within a range of the preset threshold, where the machine learning engine provides, for an artificial intelligence module based on the capability parameter adjustment value, an capability parameter required for interacting with the target object (S404).
