The determination of a small set of biomarkers to make a diagnostic call can be formulated as a feature subset selection (FSS) problem to find a small set of genes with high relevance for the underlying classification task and low mutual redundancy. However, repeated application of a heuristic, evolutionary FSS technique usually fails to produce consistent results. Here, we introduce COMB-PSO-LS, a novel hybrid (wrapper-filter) FSS algorithm based on Particle Swarm Optimization (PSO) that features a local search strategy to select the least dependent and most relevant feature subsets. In particular, we employ a Randomized Dependence Coefficient (RDC)-based filter technique to guide the search process of the particle swarm, allowing the selection of highly relevant and consistent features. Classifying cancer samples through patient gene expression profiles, we found that COMB-PSO-LS provides highly stable and non-redundant gene subsets that are relevant for the classification process, outperforming standard PSO methods.

Fluorescence spectroscopic excitation-emission matrices (EEMs) can be used to characterize dissolved organic matter in water with high sensitivity. Our goal is to predict the standard biosensor-based measurement Microtox® utilizing EEMs, pH, turbidity and conductivity, among other variables. EEMs h ave been modeled using novel latent fluorescent Dirichlet allocation (LFDA) based probabilistic graphical model. We found that nonparametric techniques offer a better mapping from, LFDA based EEMs scores and other measurements, to Microtox® measurements. The final decision on Microtox® measurement is given using an evidence fusion mechanism. In general, the novel LFDA based graphical model can be utilized in analyzing two dimensional data.

In this paper, we present a novel latent Dirichlet allocation (LDA) based probabilistic graphical approach for modeling and analyzing fluorescent spectroscopy excitation-emission Matrices (EEMs). By viewing the EEMs as being generated from an underlying hidden pool of flourophore compounds, the pro posed method provides a latent flourophore-space representation of an EEM. We show that this LDA-based model can increase classification performance, especially when paired with parallel factor analysis (PARAFAC) which may be regarded as perhaps the most popular and widely used tool for dealing with EEMs. Our experiments show that the proposed LDA-based algorithm is in some cases more robust than PARAFAC to certain types of noise and data disturbances. We also observe that pairing this LDA-based method with PARAFAC leads to an improvement in classification performance and to added robustness at high peak-signal-to-noise-ratio (PSNR) values.

Over the past decade or so, several research groups have addressed the problem ofmulti-label classificationwhere each example can belong to more than one class at the same time. A common approach, calledBinary Relevance (BR), addresses this problem by inducing a separate classifier for each class. Research has shown that this framework can be improved if mutual class dependence is exploited: an example that belongs to class$X$is likely to belong also to class$Y$; conversely, belonging to$X$can make an example less likely to belong to$Z$. Several works sought to model this information by using the vector of class labels as additional example attributes. To fill the unknown values of these attributes during prediction, existing methods resort to using outputs of other classifiers, and this makes them prone to errors. This is where our paper wants to contribute. We identified two potential ways to prune unnecessary dependencies and to reduce error-propagation in our new classifier-stacking technique, which is namedPruDent. Experimental results indicate that the classification performance ofPruDentcompares favorably with that of other state-of-the-art approaches over a broad range of testbeds. Moreover, its computational costs grow only linearly in the number of classes.

Hierarchical multi-label classification is a relatively new research topic in the field of classifier induction. What distinguishes it from earlier tasks is that it allows each example to belong to two or more classes at the same time, and by assumin

An interesting issue in machine learning is induction in multi-label domains where each example can be labeled with two or more classes at the same time. In a work focusing on text categorization, we followed the most commonly used approach and induced a binary classifier for each class. Analyzing the results, we noticed that performance had been impaired by two factors. First, in text domains, each class is characterized by a different set of attributes; an appropriate attribute-selection technique thus has to be applied separately to each of them. Second, the individual classes often have to be induced from imbalanced training sets, a circumstance we addressed here by majority-class undersampling. The paper provides details of the induction system and reports the results of systematic experimentation.

Automated Collaborative Filtering (ACF) refers to a group of algorithms used in recommender systems, a research topic that has received considerable attention due to its e-commerce applications. However, existing techniques are rarely capable of dealing with imperfections in user-supplied ratings. When such imperfections (e.g., ambiguities) cannot be avoided, designers resort to simplifying assumptions that impair the system's performance and utility. We have developed a novel technique referred to as CoFiDS—Collaborative Filtering based on Dempster-Shafer belief-theoretic framework—that can represent a wide variety of data imperfections, propagate them throughout the decision-making process without the need to make simplifying assumptions, and exploit contextual information. With its DS-theoretic predictions, the domain expert can either obtain a "hard” decision or can narrow the set of possible predictions to a smaller set. With its capability to handle data imperfections, CoFiDS widens the applicability of ACF to such critical and sensitive domains as medical decision support systems and defense-related applications. We describe the theoretical foundation of the system and report experiments with a benchmark movie data set. We explore some essential aspects of CoFiDS' behavior and show that its performance compares favorably with other ACF systems.

The behavior of the genetic algorithm (GA), a popular approach to search and optimization problems, is known to depend, among other factors, on the fitness function formula, the recombination operator, and the mutation operator. What has received less attention is the impact of the mating strategy that selects the chromosomes to be paired for recombination. Existing GA implementations mostly choose them probabilistically, according to their fitness function values, but we show that more sophisticated mating strategies can not only accelerate the search, but perhaps even improve the quality of the GA-generated solution. In our implementation, we took inspiration from the "opposites-attract” principle that is so common in nature. As a testbed, we chose the problem of 1-NN classifier tuning where genetic solutions have been employed before, and are thus well-understood by the research community. We propose three "instinct-based” mating strategies and experimentally investigate their behaviors.

Existing research in association mining has focused mainly on how to expedite the search for frequently co-occurring groups of items in ldquoshopping cartrdquo type of transactions; less attention has been paid to methods that exploit these ldquofrequent itemsetsrdquo for prediction purposes. This paper contributes to the latter task by proposing a technique that uses partial information about the contents of a shopping cart for the prediction of what else the customer is likely to buy. Using the recently proposed data structure of itemset trees (IT-trees), we obtain, in a computationally efficient manner, all rules whose antecedents contain at least one item from the incomplete shopping cart. Then, we combine these rules by uncertainty processing techniques, including the classical Bayesian decision theory and a new algorithm based on the Dempster-Shafer (DS) theory of evidence combination.

Text categorization systems often use machine learning techniques to induce document classifiers from preclassified examples. The fact that each example document belongs to many classes often leads to very high computational costs that sometimes grow exponentially in the number of features. Seeking to reduce these costs, we explored the possibility of running a "baseline induction algorithm" separately for subsets of features, obtaining a set of classifiers to be combined. For the specific case of classifiers that return not only class labels but also confidences in these labels, we investigate here a few alternative fusion techniques, including our own mechanism that was inspired by the Dempster-Shafer Theory. The paper describes the algorithm and, in our specific case study, compares its performance to that of more traditional mechanisms.

For managing the performance of database management systems, we need to be able to estimate the size of queries. Query size estimation (QSE) is difficult if the queries are associated with more than one attribute. Here, we propose, and experimentally evaluate, a novel technique that builds on cluster analysis. Empirical results indicate that, in particular, density-based clustering QSE techniques are beneficial for medium and large sized databases where they compare favourably with partitioning clustering QSE ones such as k-means. This is observed especially in the case of noisy and dense datasets

Association mining techniques search for groups of frequently co-occurring items in a market-basket type of data and turn these groups into business-oriented rules. Previous research has focused predominantly on how to obtain exhaustive lists of such associations. However, users often prefer a quick response to targeted queries. For instance, they may want to learn about the buying habits of customers that frequently purchase cereals and fruits. To expedite the processing of such queries, we propose an approach that converts the market-basket database into an itemset tree. Experiments indicate that the targeted queries are answered in a time that is roughly linear in the number of market baskets, N. Also, the construction of the itemset tree has O(N) space and time requirements. Some useful theoretical properties are proven.

Successful implementations of radial-basis function (RBF) networks for classification tasks must deal with architectural issues, the burden of irrelevant attributes, scaling, and some other problems. This paper addresses these issues by initializing RBF networks with decision trees that define relatively pure regions in the instance space; each of these regions then determines one basis function. The resulting network is compact, easy to induce, and has favorable classification accuracy.

Reasoning with uncertain information is a problem of key importance when dealing with information about the real world. Obtaining the precise numbers required by many uncertainty handling formalisms can be a problem. The theory of rough sets makes it possible to handle uncertainty without the need for precise numbers, and so has some advantages in such situations. This paper presents an introduction to various forms of reasoning under uncertainty that are based on rough sets. In particular, a number of sets of numerical and symbolic truth values which may be used to augment propositional logic are developed, and a semantics for these values is provided based upon the notion of possible worlds. Methods of combining the truth values are developed so that they may be propagated when augmented logic formulae are combined, and their use is demonstrated in theorem proving.