Knowledge discovery by accuracy maximization

Abstract

Here we describe KODAMA (knowledge discovery by accuracy maximization), an unsupervised and semisupervised learning algorithm that performs feature extraction from noisy and high-dimensional data. Unlike other data mining methods, the peculiarity of KODAMA is that it is driven by an integrated procedure of crossvalidation of the results. The discovery of a local manifold's topology is led by a classifier through a Monte Carlo procedure of maximization of cross-validated predictive accuracy. Briefly, our approach differs from previous methods in that it has an integrated procedure of validation of the results. In this way, the method ensures the highest robustness of the obtained solution. This robustness is demonstrated on experimental datasets of gene expression and metabolomics, where KODAMA compares favorably with other existing feature extraction methods. KODAMA is then applied to an astronomical dataset, revealing unexpected features. Interesting and not easily predictable features are also found in the analysis of the State of the Union speeches by American presidents: KODAMA reveals an abrupt linguistic transition sharply separating all post-Reagan from all pre-Reagan speeches. The transition occurs during Reagan's presidency and not from its beginning.

Automatic Band Selection Using Spatial-Structure Information and Classifier-Based Clustering

“…Recently, Cacciatore et al proposed a new method-knowledge discovery by accuracy maximization [34], which is used for feature extraction from noised and high-dimensional data. The peculiarity of this algorithm is that it can find the topology of the local manifold of the data through maximization of predictive accuracy with a classifier.…”

“…The peculiarity of this algorithm is that it can find the topology of the local manifold of the data through maximization of predictive accuracy with a classifier. Inspired by this, we use the maximization of cross-validated accuracy of [34] to cluster similar bands, which can automatically determine the number of clusters according to the data distribution.…”

A Convex Matrix Optimization for the Additive Constant Problem in Multidimensional Scaling with Application to Locally Linear Embedding

“…We denote it by LLE(X). However, in practice, the data matrix such as the genetic data studied in [34,5] often has missing values. A common approach is to impute all the missing values.…”

Hypothesis: The Sound of the Individual Metabolic Phenotype? Acoustic Detection of NMR Experiments

“…To compare the efficiency of the audio recognition process with standard chemometrics analysis, the corresponding eight NMR spectra were analyzed, after Fourier transformation and pre-processing, using the clustering algorithm k-means, which is one of the most widely used and best performing clustering algorithms (Cacciatore et al, 2014;MacQueen, 1967). The clustering obtained by k-means algorithm achieved an ARI score of 0.16, which is larger than the score for non-musicians, but interestingly, much smaller than those obtained by musicians.…”

“…Its individuality is given by the interplay between different concentration levels and by covariance patterns of different molecules. Disentangling and extracting the constituent building blocks of the metabolic fingerprint from high-dimensional data requires the application of multivariate statistical tools and/or machine learning algorithms (Cacciatore et al, 2014;Jansen et al, 2005;Saccenti et al, 2013;Trygg et al, 2007;Weckwerth et al, 2005), the latter often requiring a large number of samples in order to build and train predictive models (Stockwell et al, 2002).…”