Macroeconomic forecasts enable the policy-makers to foresee the future economic trends and take prompt measures to ensure longer economic growth and quicker economic recovery. Accurate and timely macroeconomic forecasts may also help the enterprises to make better long-term business strategies. Social media has a swift and sensitive response to economic dynamics through online news reports, interviews and individual comments. Economic Indices extracted from social media data are more immediate and comprehensive, but lack of stability and credibility in empirical studies. This research proposed a mix frequency modelling approach to incorporate only the recent high frequency part of social media data in traditional econometrics based macroeconomic forecasting with support of a multisource based macroeconomic forecast system. A mixed data sampling (MIDAS) model is constructed and an empirical evaluation is presented to show how to incorporate Google search queries into Chinese CPI forecasting. The empirical results indicate a satisfactory improvement in forecasting performance. The multisource modelling and forecasting framework offers a practical and implementable solution for involving social media data sources into macroeconomic forecasting systems. This research contributes to future development of decision support systems for governmental policy-making and enterprises’ operational decisions in the Big Data era.