House price prediction is of utmost importance in forecasting residential property prices, particularly as the demand for high-quality housing continues to rise. Accurate predictions have implications for real estate investors, financial institutions, urban planners, and policymakers. However, accurately predicting house prices is challenging due to the complex interplay of various influencing factors. Previous studies have primarily focused on basic property information, leaving room for further exploration of more intricate features, such as amenities, traffic, and social sentiments in the surrounding environment. In this paper, we propose a novel approach to house price prediction from a multi-source data fusion perspective. Our methodology involves analyzing house characteristics and incorporating factors from diverse aspects, including amenities, traffic, and emotions. We validate our approach using a dataset of 28550 real-world transactions in Beijing, China, providing a comprehensive analysis of the drivers influencing house prices. By adopting a multi-source data fusion perspective and considering a wide range of influential factors, our approach offers valuable insights into house price prediction. The findings from this study possess the capability to improve the accuracy and effectiveness of house price prediction models, benefiting stakeholders in the real estate market.