When rumors prevail in securities market, it is very difficult for investors to identify valid information. In the meantime, investors have much more ways to access information with the evolution of internet. But there is an overwhelming quantity of information on the Internet, the coexistence of facts and rumors, namely, “widely circulated” and “specious”, yet “unconfirmed officially” vague information, makes it more difficult for investors who with limited rationality to distinguish facts from rumors. Existing studies are mainly devoted in the method of event study, namely screening rumors from “official channels” that clarified, which is neither timely efficient in terms of accessing to rumors nor providing the basis for decision-making. Traditional news has evolved into various forms of social media, including forums, blogs, micro-blogs etc., and users can not only gain quick access to more valuable and timely information, but also amplify information that embed the news effectively by participating in commenting on various social media. Dynamic information creation, sharing and coordination among Web users are exerting increasingly prominent impact on the securities market in now days. Thus, it is very necessary to study the effects of social media as online forums on the securities market. In this paper, the method of machine learning is adopted for the first time to identifying the Internet rumors automatically, and successfully in crawling massive forum data by smart computer technology. Unlike the case study and statistical sampling of rumors, this paper conduct automatic identification of Internet rumors by utilize the smart technology, thus paving the way for more in-depth analysis about the effects of Internet media on the securities market in future.