M. Carmen Garrido

Probabilistic inference from arbitrary uncertainty using mixtures of factorized generalized gaussians

This paper presents a general and efficient framework for probabilistic inference and learning from arbitrary uncertain information. It exploits the calculation properties of finite mixture models, conjugate families and factorization. Both the joint probability density of the variables and the likelihood function of the (objective or subjective) observation are approximated by a special mixture model, in such a way that any desired conditional distribution can be directly obtained without numerical integration. We have developed an extended version of the expectation maximization (EM) algorithm to estimate the parameters of mixture models from uncertain training examples (indirect observations). As a consequence, any piece of exact or uncertain information about both input and output values is consistently handled in the inference and learning stages. This ability, extremely useful in certain situations, is not found in most alternative methods. The proposed framework is formally justified from standard probabilistic principles and illustrative examples are provided in the fields of nonparametric pattern classification, nonlinear regression and pattern completion. Finally, experiments on a real application and comparative results over standard databases provide empirical evidence of the utility of the method in a wide range of applications.

Towards the learning from low quality data in a Fuzzy Random Forest ensemble

Imperfect information inevitably appears in real situations for a variety of reasons. Although efforts have been made to incorporate imperfect data into classification techniques, there are still many limitations as to the type of data, uncertainty and imprecision that can be handled. In this paper, we will present a Fuzzy Random Forest ensemble for classification and show its ability to handle imperfect data into the learning and the classification phases. Then, we will describe the types of imperfect data it supports. We will devise an augmented ensemble that can operate with others type of imperfect data: crisp, missing, probabilistic uncertainty and imprecise (fuzzy and crisp) values. Additionally, we will perform experiments with datasets used in other papers to show the advantage of being able to express the true nature of imperfect information.

Gene Priorization for Tumor Classification Using an Embedded Method

The application of microarray technology to the diagnosis of cancer has been a challenge for computational techniques because the datasets obtained have high dimension and a few examples. In this paper two computational techniques are applied to tumor datasets in order to carry out the task of diagnosis of cancer (classification task) and identifying the most promising candidates among large list of genes (gene prioritization). Both techniques obtain good classification results but only one provides a ranking of genes as additional information and thus, more interpretable models, being more suitable for jointly addressing both tasks.

Imputing missing values from low quality data by NIP tool

An important aspect to consider in applications which work with great volumes of data is that frequently these data are of low quality and also cannot be use other types of data. The field of Soft Computing has dealt, among other things, with developing techniques that will be able to work with these types of low quality data in a suitable way, respecting the true origin of these data. In this paper we present a method to carry out the imputation of missing values from information that may be of low quality when another possibility is not available. The method is based on a predictable model. The imputation method developed is incorporated into the software tool NIP increasing its functionality of imputation/replacement of low quality values.

Learning in a Fuzzy Random Forest ensemble from imperfect data

Instrument errors or noise interference during experiments may lead to incomplete data when measuring a specific attribute. Obtaining models from imperfect data is a topic currently being treated with more interest. In this paper, we present the learning phase of a Fuzzy Random Forest ensemble for classification from imperfect data. We perform experiments with imperfect datasets created for this purpose and datasets used in other papers to show the express the true nature of imperfect information.

Towards an App Based on FIWARE Architecture and Data Mining with Imperfect Data

In this work, the structure for the prototype construction of an application that can be framed within ubiquitous sensing is proposed. The objective of application is to allow that a user knows through his mobile device which other users of his environment are doing the same activity as him. Therefore, the knowledge is obtained from data acquired by pervasive sensors. The FIWARE infrastructure is used to allow to homogenize the data flows.

Intelligent Data Analysis, Soft Computing and Imperfect Data

In different real problems the available information is not as precise or as accurate as we would like. Due to possible imperfection in the data (understanding that these contain data where not all the attributes are precisely known, such as missing, imprecise, uncertain, ambiguous, etc. values), tools provided by Soft Computing are quite adequate, and the hybridization of these tools with the Intelligent Data Analysis is a field that is gaining more importance. In this paper, first we present a brief overview of the different stages of Intelligent Data Analysis, focusing on two core stages: data preprocessing and data mining. Second, we perform an analysis of different hybridization approaches of the Intelligent Data Analysis with the Soft Computing for these two stages. The analysis is performed from two levels: If elements of Soft Computing are incorporated in the design of the method/model, or if they are also incorporated to be able to deal with imperfect information. Finally, in a third section, we present in more detail several methods which allow the use of imperfect data both for their learning phase and for the prediction.

A More Realistic K-Nearest Neighbors Method and Its Possible Applications to Everyday Problems

Currently, many of the elements that surround us in daily life need software systems that work from the information available in the domain (data-driven application domains) by performing a process of data mining from it. Between the data mining techniques used in everyday problems we find the k-Nearest Neighbors technique. However, in domains and real situations it is very common to find vague, ambiguous and noisy data, that is, imperfect information.Although this imperfect information is inevitable, most applications have traditionally ignored the need for developing appropriate approaches for representing and reasoning with such data imperfections. The soft computing field has dealt with the development of techniques that can work with this kind of information as discipline whose main characteristic is tolerance to inaccuracy and uncertainty.In this work, we extend the k-Nearest Neighbors technique using concepts and methods provided by Soft Computing. The aim is to carry out the processes of instance selection and classification in everyday problems from imperfect information making the technique more realistic.

Fuzzy Discretization Process from Small Datasets

A classification problem involves selecting a training dataset with class labels, developing an accurate description or a model for each class using the attributes available in the data, and then evaluating the prediction quality of the induced model. In this paper, we focus on supervised classification and models which have been obtained from datasets with few examples in relation with the number of attributes. Specifically, we propose a fuzzy discretization method of numerical attributes from datasets with few examples. The discretization of numerical attributes can be a crucial step since there are classifiers that cannot deal with numerical attributes, and there are other classifiers that exhibit better performance when these attributes are discretized. Also we show the benefits of the fuzzy discretization method from dataset with few examples by means of several experiments. The experiments have been validated by means of statistical tests.

Measuring Data Imperfection in a Neighborhood Based Method

In this paper, we present an extension of k nearest neighbors method so it can perform imputation/classification from datasets with low quality data. The method performs a weighting of neighbors based on their imperfection and distance of classes. Thus the method allows us explicitly to indicate the average degree of imperfection of the neighbors that it is accepted to carry out the imputation/classification and the average distance of classes to the class of example to impute/classify that it is allowed. We carry out several experiments with both real-world and synthetic datasets with low quality data to test the proposed method.