|
Abstract:
|
With the advent of Microarray data, numerous methods for classification of biological samples have been proposed. However, reproducibility, or consistency of rules used in classifying observations in different samples is seldom achieved. In this paper, we propose a bootstrap resampling method to achieve a set of reproducible rules. More specifically, as an example, we first utilize the K-Top Score Pairs algorithm (K-TSP) to construct classification rules. Then, with the help of a proper rank aggregation method, we obtain rules that are deemed to be more producible. Further, because of comparability of ranks, our method can be applicable to data from different platforms. We illustrate our approach using three lung cancer data from two different platforms. Finally, it is worth noting that although our illustration is based on K-TSP, other classification methods can be utilized as well.
|
