An improved graph based association rules mining (ARM) approach to extract association rules fromtext databases is proposed in this paper. The text document in the proposed technique is read only once to lookfor the terms whose occurrences are greater than some threshold value, these terms are stored in a file with theirfrequencies, then they are represented as nodes in a weighted directed graph where edges represent relationsbetween these terms, the edges will denote the associations between terms while the edges' weights denote thestrength or confidence of these rules. The proposed method is called Dynamic Graph based Rule Mining fromText (DGRMT) because the graph is built level by level according the length of a sentence (number of frequentterms). Weighted subgraph mining is used to ensure the efficiency and throughput of the proposed technique;only the most frequent subgraphs are extracted. The proposed technique is validated and evaluated using realworld textual data sets and compared with one of the best graph based rule mining technique, which is algorithmfor Generating Association Rules based on Weighting scheme(GARW). The results determine that the proposed approach is better than GARW on almost all textual datasets.
Feature Classification Method Of Frequency Cepstrum Coefficient Based On Weighted Extreme Gradient Boosting
