Multi-label text classification with an ensemble feature space

Multi-label text classification aims at assigning more than one class to a given text document, which makes the task more ambiguous and challenging at the same time. The ambiguities come from the fact that often several labels in the prescribed label set are semantically close to each other, making clear demarcation between them difficult. As a consequence, any Machine Learning based approach for developing multi-label classification scheme needs to define its feature space by choosing features beyond linguistic or semi-linguistic features, so that the semantic closeness between the labels is also taken into account. The present work describes a scheme of feature extraction where the training document set and the prescribed label set are intertwined in a novel way to capture the ambiguity in a meaningful way. In particular, experiments were conducted using Topic Modeling and Fuzzy C-means clustering which aim at measuring the underlying uncertainty using probability and membership based measures, respectively. Several Nonparametric hypothesis tests establish the effectiveness of the features obtained through Fuzzy C-Means clustering in multi-label classification. A new algorithm has been proposed for training the system for multi-label classification using the above set of features.

The Methodology of Distinguish Between Random and Chaotic Machine Tool Oscillations

Machine tool oscillations are irregular or aperiodic. Most often, these oscillations are chaotic but, in some cases, they can be quasi-periodic or random. The methodology for characterizing oscillations in the first of two steps uses the nonparametric hypothesis tests which the observed oscillations confirmed as irregular. The methodology for the final characterization of oscillations is based on chaos quantifiers. A time series defined as the measured values of oscillations in the time domain is the basis for calculating the quantifiers of chaos. There are four quantifiers of chaos: the Lyapunov exponent, Kolmogorov entropy, fractal dimension and correlation dimension. The correlation dimension and Kolmogorov entropy are important for distinguishing between random and chaotic oscillations. Other quantifiers of chaos are not used for this purpose. The methodology requires a multidisciplinary approach based on combining Nonlinear Dynamics and Probability Theory and Statistics. The methodology can be applied to many oscillating phenomena. Therefore, the paper mainly used the term oscillations, not vibrations, chatter, etc.

Development of Time Series Disorder Detection Algorithms Based on Nonparametric Hypothesis Testing Criteria

The solution of the problem of detecting, in the online mode, a spontaneous change in the probabilistic characteristics (“disorder” or “breakdown”) of a time series is given. It is pointed out that there is a growing interest in the development of so-called nonparametric disorder detection methods, i.e., methods the application of which does not require the knowledge of the probability distribution function of the controlled process values. It is stated that the majority of the known versions of such methods are based on using a number of standard nonparametric criteria transformed for solving disorder detection problems. It is proposed to use the signs criterion, the series criterion, and the Ramachandran–Ranganathan criterion as a basis for construction of disorder detection algorithms. The methodical aspects of studying the statistical properties and efficiency of the disorder detection algorithms built on their basis are considered. The simulation method was used as a study tool. The plan of carrying out simulation experiments was developed separately for each of the proposed algorithms, taking into account their individual characteristics, but based on the general requirement of fully reproducing the monitoring algorithm performance dynamics under real conditions, when a disorder can appear at any time and there is a transient in the values of the decisive function. By using a simulation experiment for each of the algorithms under consideration, data on their statistical characteristics were obtained and systematized in a scope sufficient for synthesizing a monitoring procedure with the specified properties.