Testing phase finds differences in positive/negative documents by the centroid obtained in training phase by ranking each of them. The simple way to estimate similarity between documents and centroid by summing weights of patterns which are in the documents. VII. Experimental Results To determine accurate measures of similarity or difference between documents you depict results by graph pattern and table pattern. The experimental setup consists of relevant documents that you termed as positive and negative documents .i.e Technology related (Positive) and Science related (Negative). You take into account support factor also consisting of support = 50, support = 75 and support= 90. You also take training set of 6 documents in which you take 3 positive documents and 3 negative documents. Basing on the support factors you have calculated documents weights and also uniquely calculated document weights by PTM method which you discussed earlier. Doc/Sup 50 75 90 Doc 1 1.31556 1.80357 1.80357 Doc 2 1.75182 1.87012 1.87012 Doc 3 2.13338 2.27178 2.27178 Doc 4 1.56941 1.0743 1.0867 Doc 5 1.51786 1.9243 1.8243 Table I. Support Factors for Different Documents In the above Table I you have 3 documents with their relevant weights basing on the support factor. You now …show more content…
The consequences also show that the term classification can be effectively approximated by the proposed clustering method. The proposed methodology is reasonable and robust. This paper demonstrates the new models totally tested and prove the results statistically significant. The paper also proves that the use of unrelated opinion is considerable for improving the performance of relevance feature discovery models. A promising methodology for developing effective text mining models for RFD discovery based on both positive and negative