Effectiveness of document representation for classification

Conventionally, document classification researches focus on improving the learning capabilities of classifiers. Nevertheless, according to our observation, the effectiveness of classification is limited by the suitability of document representation. Intuitively, the more features that are used in representation, the more comprehensive that documents are represented. However, if a representation contains too many irrelevant features, the classifier would suffer from not only the curse of high dimensionality, but also overfitting. To address this problem of suitableness of document representations, we present a classifier-independent approach to measure the effectiveness of document representations. Our approach utilises a labelled document corpus to estimate the distribution of documents in the feature space. By looking through documents in this way, we can clearly identify the contributions made by different features toward the document classification. Some experiments have been performed to show how the effectiveness is evaluated. Our approach can be used as a tool to assist feature selection, dimensionality reduction and document classification.

WOMAC 20, 50, 70 response levels in patients treated with Rofecoxib for osteoarthritis

Aim of study: The goal of this post-hoc analysis was to examine the difference between treatment groups when varying the target response level from at least a 20% improvement from baseline, to at least 50% and 70% improvements in Phase III studies of rofecoxib in patients with osteoarthritis. Methods: The analysis focused on results from two 6-week, placebo-controlled, ibuprofen-comparator, Phase III osteoarthritis studies. These studies employed a flare design requiring a minimum level of symptoms at entry following discontinuation of prior analgesics. Two definitions of ‘‘patient improved’’ from baseline were used: (1) WOMAC-P: a reduction in the WOMAC pain score and (2) WOMAC-PFS: a reduction in the WOMAC pain score and either a reduction in the WOMAC stiffness or function score. The improvement target was increased from 20% to 50% to 70%, relative to baseline, to investigate how the increase affects the ability to detect the differences between treatment groups. Analyses were conducted on the average and last of all measurements collected during a 6-week treatment period. Results: In the ibuprofen-comparator studies, 1545 patients were randomized to placebo, rofecoxib 12.5 mg once daily, rofecoxib 25 mg once daily, and ibuprofen 800 mg three times daily in a 1:3:3:3 ratio. The percentages of patients who met the improvement targets decrease as the target increases from 20% to 50% to 70%. There were meaningful differences between the active treatment and placebo that were inversely related to the improvement target. For example, there was a 31 (P ! 0.001), 21 (P ! 0.001), and 12 (P ! 0.001) percentage-point difference between rofecoxib 25 mg and placebo for the 20%, 50%, and 70% targets for WOMAC-P. For WOMAC-PFS, the differences between rofecoxib 25 mg and placebo were 33 (P ! 0.001), 18 (P ! 0.001), and 9 (P ! 0.01) percentage points for the 20%, 50%, and 70% improvement targets. Conclusions: Meaningful differences between active treatments and placebo were detected at all three response levels associated with the WOMAC-P and WOMAC-PFS endpoints. The differences between groups were more dramatic at the 20% and 50% response levels. The WOMAC (20,50,70)-P and WOMAC (20,50,70)-PFS endpoints further confirm, at an individual patient level, the clinical benefit of rofecoxib in the treatment of osteoarthritis that was previously reported as a difference in means.