While multi-label classification can be widely applied for problems where multiple classes can be assigned to an object, its effectiveness may be sacrificed due to curse of dimensionality in the feature space and sparseness of dimensionality in the label space. As a solution, this paper presents two alternative methods, namely Dependent Dual Space Reduction and Independent Dual Space Reduction, to reduce dimensions in the dual spaces, i.e. the feature and label spaces, using singular value decomposition (SVD). The first approach constructs the cross-covariance matrix to represent dependency between the features and labels, projects both of them into a single reduced space, and then performs prediction on the reduced space. On the other hand, the second approach handles the feature space and the label space separately by constructing a covariance matrix for each space to represent feature dependency and label dependency before performing SVD on dependency profile of each space to reduce dimension and for noise elimination and then predicting classes using their reduced dimensions. A number of experiments show that prediction on the reduced spaces for both dependent and independent reduction approaches can obtain better classification accuracy as well as faster computation, compared with the prediction using the original spaces. Finally, the results of our methods are also compared with the prediction using other types of dimensionality reduction, with consideration of the factors of covariance and threshold selection.

UNLABELLED: sMOL Explorer is a 2D ligand-based computational tool that provides three major functionalities: data management, information retrieval and extraction and statistical analysis and data mining through Web interface. With sMOL Explorer, users can create personal databases by adding each small molecule via a drawing interface or uploading the data files from internal and external projects into the sMOL database. Then, the database can be browsed and queried with textual and structural similarity search. The molecule can also be submitted to search against external public databases including PubChem, KEGG, DrugBank and eMolecules. Moreover, users can easily access a variety of data mining tools from Weka and R packages to perform analysis including (1) finding the frequent substructure, (2) clustering the molecular fingerprints, (3) identifying and removing irrelevant attributes from the data and (4) building the classification model of biological activity. AVAILABILITY: sMOL Explorer is an Open Source project and is freely available to all interested users at http://www.biotec.or.th/ISL/SMOL/.