To address the issue of short-term forecasting errors for heating-type electric loads in demand response scenarios, this paper proposes a data-knowledge fusion-driven forecasting method. First, an interactive response model for heating-type electric loads under demand response is developed to extract demand response signals. Additionally, a user response rate model with uncertainty quantification is constructed to better align with actual user power consumption behaviors. By incorporating interactive response knowledge as input features, the method integrates three data-driven models to form a data-knowledge fusion-driven forecasting framework. Through three designed forecasting scenarios, experimental results demonstrate that the integration of demand response signals and the consideration of user response rates with uncertainty significantly improve forecasting accuracy, verifying the strong generalization performance of the data-knowledge fusion-driven approach.