There exists a lot of chaotic patterns in the natural kingdom and social society. Databases recording this kind of data also c ontain the chaotic patterns. Generally, the chaotic patterns are characteristics of greatly random fluctuation although they often appear between deterministic and stochastic patterns in KDD. Therefore, the chaotic patterns are always treat ed as random fluctuation distributions and are ignored. Such viewpoint illustra ted by J.Zytkow's Forty-Niner is that a pattern discovery platform based on st atistics. A new network method to discover chaotic pattern in databases is propo sed in this paper. This method together with the Forty-Niner can not only disco ver the chaotic pattern, but also forecast it efficiently. In addition, this met hod is very suitable to deal with large databases and has extensive applicable p rospects in the vivid researching fields of KDD.