Àngela Nebot

Feature selection algorithms: a survey and experimental evaluation

In view of the substantial number of existing feature selection algorithms, the need arises to count on criteria that enables to adequately decide which algorithm to use in certain situations. This work assesses the performance of several fundamental algorithms found in the literature in a controlled scenario. A scoring measure ranks the algorithms by taking into account the amount of relevance, irrelevance and redundance on sample data sets. This measure computes the degree of matching between the output given by the algorithm and the known optimal solution. Sample size effects are also studied.

Applying Data Mining Techniques to e-Learning Problems

This chapter aims to provide an up-to-date snapshot of the current state of research and applications of Data Mining methods in e-learning. The cross-fertilization of both areas is still in its infancy, and even academic references are scarce on the ground, although some leading education-related publications are already beginning to pay attention to this new field. In order to offer a reasonable organization of the available bibliographic information according to different criteria, firstly, and from the Data Mining practitioner point of view, references are organized according to the type of modeling techniques used, which include: Neural Networks, Genetic Algorithms, Clustering and Visualization Methods, Fuzzy Logic, Intelligent agents, and Inductive Reasoning, amongst others. From the same point of view, the information is organized according to the type of Data Mining problem dealt with: clustering, classification, prediction, etc.

Combined qualitative/quantitative simulation models of continuous-time processes using fuzzy inductive reasoning techniques

Abstract A new mixed quantitative and qualitative simulation methodology based on fuzzy inductive reasoning is presented. The feasibility of this methodology is demonstrated by means of a simple hydraulic control system. The mechanical and electrical parts of the control system are modeled using differential equations, whereas the hydraulic part is modeled using fuzzy inductive reasoning. The technique is described in detail in the first part of this paper. The example is shown in the second part of the paper. The mixed quantitative and qualitative model is simulated in ACSL, and the simulation results are compared with those obtained from a fully quantitative model. The example was chosen as a simple to describe, yet numerically demanding process whose sole purpose is to prove the concept. Several practical applications of this mixed modeling technique are mentioned in the paper. but their realization has not yet been completed

Mixed quantitative:qualitative modeling and simulation of the cardiovascular system

The cardiovascular system is composed of the hemodynamical system and the central nervous system (CNS) control. Whereas the structure and functioning of the hemodynamical system are well known and a number of quantitative models have already been developed that capture the behavior of the hemodynamical system fairly accurately, the CNS control is, at present, still not completely understood and no good deductive models exist that are able to describe the CNS control from physical and physiological principles. The use of qualitative methodologies may offer an interesting alternative to quantitative modeling approaches for inductively capturing the behavior of the CNS control. In this paper, a qualitative model of the CNS control of the cardiovascular system is developed by means of the fuzzy inductive reasoning (FIR) methodology. FIR is a fairly new modeling technique that is based on the general system problem solving (GSPS) methodology developed by G.J. Klir (Architecture of Systems Problem Solving, Plenum Press, New York, 1985). Previous investigations have demonstrated the applicability of this approach to modeling and simulating systems, the structure of which is partially or totally unknown. In this paper, five separate controller models for different control actuations are described that have been identified independently using the FIR methodology. Then the loop between the hemodynamical system, modeled by means of differential equations, and the CNS control, modeled in terms of five FIR models, is closed, in order to study the behavior of the cardiovascular system as a whole. The model described in this paper has been validated for a single patient only.

Synthesis of an anaesthetic agent administration system using fuzzy inductive reasoning

Control of the depth of anaesthesia is a difficult undertaking. Progress has been made during recent years by use of different methodologies and monitoring systems that suggest the safe amount of an anaesthetic drug, considering the condition of an individual patient. Despite these improvements, anaesthetists still rely heavily on personal experience when suggesting the anaesthetic dosage during surgical operations. The purposes of this paper are twofold. One is a description of the design of an anaesthetic agent control system using a qualitative modelling and simulation methodology called Fuzzy Inductive Reasoning (FIR). A comparison with a system developed for the same application using a neural network approach is also presented. The second purpose is a discussion of the problem of separating system-generic from patient-specific behaviour in the context of inductive modeling using the FIR methodology. In order to be useful, the model generated by FIR should reflect upon system-generic behavioural characteristics exclusively, while suppressing patient-specific behavioural patterns. A technique based on combining knowledge obtained from different patients is designed that makes it possible to derive a single model characterizing a specific class of similar patients undergoing similar operations, preserving the common characteristics of all these patients while filtering out the specific behavioural patterns of any one of the individual patients from whom the data were obtained.

Visual-FIR: A tool for model identification and prediction of dynamical complex systems

Abstract A new platform for the fuzzy inductive reasoning (FIR) methodology has been designed and developed under the MATLAB environment. The new tool, named Visual-FIR, allows the identification of dynamic systems models in a user-friendly environment. FIR offers a pattern-based approach to modeling and predicting either univariate or multivariate time series, obtaining very good results when applied to various areas such as control, biology, and medicine. However, the available implementation of FIR was such that new code had to be developed for each new application studied. Visual-FIR resolves this limitation and offers a high-efficiency implementation. Furthermore, the Visual-FIR platform presents a new vision of the methodology based on process blocks and adds new features, increasing the overall capabilities of the FIR methodology. The DAMADICS benchmark problem is addressed in this research using the Visual-FIR approach.

Modeling and Simulation of the Central Nervous System Control with Generic Fuzzy Models

The analysis of the human cardiovascular system by means of modeling and simulation methodologies is of relevance from a medical point of view because it allows doctors to acquire a better understanding of cardiovascular physiology, offer more accurate diagnostics, and select better suited therapies. The cardiovascular system is composed of the hemodynamical system and the central nervous system (CNS). In this work, two generic models of the CNS for patients with coronary diseases are identified by means of the fuzzy inductive reasoning (FIR) methodology. One of the models is generic only in its structure, whereas the other one is a fully generic model. It is very useful for doctors to have available a generic CNS model for a group of patients with common characteristics because this model can be used to predict the future behavior of new patients with the same characteristics.

Fuzzy inductive reasoning: a consolidated approach to data-driven construction of complex dynamical systems

Fuzzy inductive reasoning (FIR) is a modelling and simulation methodology derived from the General Systems Problem Solver. It compares favourably with other soft computing methodologies, such as neural networks, genetic or neuro-fuzzy systems, and with hard computing methodologies, such as AR, ARIMA, or NARMAX, when it is used to predict future behaviour of different kinds of systems. This paper contains an overview of the FIR methodology, its historical background, and its evolution.

Optimization of fuzzy partitions for inductive reasoning using genetic algorithms

Fuzzy Inductive Reasoning (FIR) is a data-driven methodology that uses fuzzy and pattern recognition techniques to infer system models and to predict their future behavior. It is well known that variations on fuzzy partitions have a direct effect on the performance of the fuzzy-rule-based systems. The FIR methodology is not an exception. The performance of the model identification and prediction processes of FIR is highly influenced by the discretization parameters of the system variables, i.e. the number of classes of each variable and the membership functions that define its semantics. In this work, we design two new genetic fuzzy systems (GFSs) that improve this modeling and simulation technique. The main goal of the GFSs is to learn the fuzzification parameters of the FIR methodology. The new approaches are applied to two real modeling problems, the human central nervous system and an electrical distribution problem.

Local maximum ozone concentration prediction using soft computing methodologies

The prediction of ozone levels is an important task because this toxic gas can produce harmful effects to the population health especially of children. This article describes the application of the Fuzzy Inductive Reasoning methodology and a Recurrent Neural Network (RNN) approach, the Long Short Term Memory (LSTM) architecture, to a signal forecasting task in an environmental domain. More specifically, we have applied FIR and LSTM to the prediction of maximum ozone(O3) concentrations in the East Austrian region. In this article the results of FIR and LSTM on this task are compared with those obtained previously using other types of neural networks (Multilayer Perceptrons (MLPs), Elman Networks (ENs) and Modified Elman Networks (MENs)). The performance of the best LSTM networks inferred are equivalent to the best FIR models identified and both are slightly better than the other Neural Networks studied (MENs, ENs and MLPs, in decreasing order of performance). Cross validation tests are included in this research in order to study more deeply the accuracy of the FIR models and to extract as much information as possible from the available data.