Pattern co-occurrence matrix to reduce the low frequency problem and effective pattern discovery

Useful information from a text document can be extracted using text mining techniques. Many text mining methods are generally uses term or keyword based approach. They all have advantages, but get stuck into the issue of synonymy: two terms have the same meaning and polysemy: one term has many meanings. So phrase based approaches are suggested, but they are not performing better than term based approaches. Instead of term based methods pattern mining methods are suggested, but there is still an open research issue of utilizing and modifying these discovered patterns. Pattern mining methods suffer from pattern misinterpretation and low frequency problem. This paper presents an effective technique to extract patterns that applies the pattern taxonomy model to get close sequential patterns. Pattern deploying process is used to reduce the pattern misinterpretation problem. Pattern co-occurrence matrix (PCM) is applied get the important relationship between close sequential patterns and removes patterns with zero co-occurrence weight. PCM and Inner Pattern Evolution process gives the benefit of reducing low frequency problem. This increases effectiveness of extracting interesting and accurate to the topic information.