The approach to handle multiple label for each gene is to have a learning problem for each label that appears in \texttt{yeast.labelled}. In each learning problem, a gene is a positive example if it contains that label, otherwise it is a negative example. In this way we learn one classifier for each label. To label unseen genes, we run each generated classifier on the gene data and we assign the label to the gene if the classifiers gives a positive answer. As a classifier, we have used Tilde for its speed and good accuracy. In order to finish the experiments before the deadline we had to consider only a subset of the available data, namely the protein secondary structure data.
A Simple Approach to a Multi-Label Classification Problem
RIGUZZI, Fabrizio
2005
Abstract
The approach to handle multiple label for each gene is to have a learning problem for each label that appears in \texttt{yeast.labelled}. In each learning problem, a gene is a positive example if it contains that label, otherwise it is a negative example. In this way we learn one classifier for each label. To label unseen genes, we run each generated classifier on the gene data and we assign the label to the gene if the classifiers gives a positive answer. As a classifier, we have used Tilde for its speed and good accuracy. In order to finish the experiments before the deadline we had to consider only a subset of the available data, namely the protein secondary structure data.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.