Near miss is defined as an extraordinary event that could have reasonably resulted in an accident or incident. Near miss must be found and removed for preventing major accident due to the strong relationship between near miss and real accident. This paper suggests a new method of managing accident data, which efficiently handles numerous safety data coming from real chemical processing plants. Actual 2,200 data from petrochemical company A were analyzed and the ''detailed classification sheets'' were produced. The result of the actual data classification was similar to Heinrich''s 1:29:300 theorem. yHAMS(Yonsei Hazard Analysis Management System) was developed to classify and identify potential hazards and suggest the prority of investment of improve the safety of the plant. The data classification method was improved to adapt domestic accident situation and the result was 5 major and 60 detailed classes. The system was applied to another 1,770 accident data during the next six months and the number of safety accidents was sharply decreased. yHAMS was proved to be effective in reducing safety accidents in a chemical plant through efficient management of near misses.