In the era of digital intelligence, China"s Internet enterprises" overseas mergers and acquisitions have risen strongly, but the risks are huge. Machine learning and non-financial information can provide new ideas for such risk early warning. This study selects the big data of 56 overseas mergers and acquisitions of 45 Chinese Internet listed companies from 2013 to 2020, and uses the Stacking integrated learning model to mine the big data financial risk early warning factors. The research found that the stacking integrated learning model has better big data early warning effect than other machine learning models; Traditional financial indicators such as operational capacity are still the preferred indicators for big data early warning of financial risks of overseas mergers and acquisitions of Internet enterprises, but innovative non-financial indicators such as stock bar reviews also have important early warning value. This study provides empirical evidence that Stacking machine learning and stakeholder big data information can help to early warning the financial risks of overseas mergers and acquisitions of Internet enterprises, and provides important reference for Internet enterprises, investors, regulators, etc. to make financial risk control decisions of overseas mergers and acquisitions.