Aljoscha Klose

Improving naive Bayes classifiers using neuro-fuzzy learning

Naive Bayes classifiers are a well-known and powerful type of classifier that can easily be induced from a dataset of sample cases. However, the strong conditional independence and distribution assumptions underlying them can sometimes lead to poor classification performance. Another prominent type of classifier are neuro-fuzzy classification systems which derive (fuzzy) classifiers from data using neural network inspired learning methods. Since there are certain structural similarities between a neuro-fuzzy classifier and a naive Bayes classifier, the idea suggests itself to mapping the latter to the former in order to improve its capabilities.

Discussing cluster shapes of fuzzy classifiers

Fuzzy classification rules are widely considered a well-suited representation of classification knowledge, as they allow readable and interpretable rule bases. The goal of the paper is to discuss the shapes of the resulting classification borders and thus which class distributions can be represented by such classification systems. 2D and 3D visualizations are used to illustrate the cluster shapes and the borders between distinct classes. Furthermore, general hints concerning the shape of higher dimensional clusters are given.

Soft computing for automated surface quality analysis of exterior car body panels

Today the method for surface quality analysis of exterior car body panels is still characterized by manual detection of local form deviations and evaluation by experts. The new approach presented in this paper is based on 3D image processing. A major step in this process is the classification of the different kinds of surface form deviations. For this purpose, we compared the performance of different soft-computing techniques. Although the dataset was rather small, high dimensional and unbalanced, we achieved promising results with regard to classification accuracies and interpretability of rule bases.

Interactive text retrieval based on document similarities

In this article we present a prototypical implementation of a software tool for document retrieval which groups/arranges (pre-processed) documents based on a similarity measure. The prototype was developed based on self-organising maps to realise interactive associative search and visual exploration of document databases. This helps a user to navigate through similar documents. The navigation, especially the search for the first appropriate document, is supported by conventional keyword search methods. The usability of the presented approach is shown by a sample search.

Data mining with neuro-fuzzy models

Data mining is the central step in a process called knowledge discovery in databases, namely the step in which modeling techniques are applied. Several research areas such as statistics, artificial intelligence, machine learning, and soft computing have contributed to the arsenal of methods for data mining. In this paper, however, we focus on neuro-fuzzy methods for rule learning. In our opinion, fuzzy approaches can play an important role in data mining, because they provide comprehensible results. This goal often seems to be neglected — possibly because comprehensibility is sometimes hard to achieve with other methods.

Semi-supervised learning in knowledge discovery

Recently, semi-supervised learning has received quite a lot of attention. The idea of semi-supervised learning is to learn not only from the labeled training data, but to exploit also the structural information in additionally available unlabeled data. In this paper we review existing semi-supervised approaches, and propose an evolutionary algorithm suited to learn interpretable fuzzy if-then classification rules from partially labeled data. Feasibility of our approach is shown on artificial datasets, as well as on a real-world image analysis application.

Analyzing borders between partially contradicting fuzzy classification rules

Fuzzy classification rules allow the definition of readable and interpretable rule bases. Nevertheless, the shape of the resulting class borders of fuzzy classification rules depends to a great part on the used t-norm and t-conorm and can sometimes even be counter-intuitive. In this paper we discuss the shape of class borders between overlapping rules under consideration of different t-norms and t-conorms and the effect of rule aggregation, i.e. more than one rule defining the same class are overlapping. Furthermore, we discuss the influence of rule weights and point out some aspects of the classification behavior of naive Bayes classifiers, which can be seen as a subset of fuzzy systems. Our main goal is to give the potential user an insight into the classification behavior of fuzzy classifiers. For this, mainly 2D and 3D visualizations are used to illustrate the cluster shapes and the borders between distinct classes.

On the Properties of Prototype-Based Fuzzy Classifiers

The use of natural language rules that are able to handle vague and, possibly, even contradicting knowledge in order to model formal dependences is an intriguing idea. Fuzzy if-then rules have been proposed as classification methods that can easily be defined and interpreted by humans or built automatically by learning algorithms. This paper gives an intuitive insight into the properties and the behavior of prototype-based fuzzy classifiers, using formal descriptions and visualization methods. This can help to avoid some common peculiarities and pitfalls in the manual or automated design of fuzzy classifiers.

Graphical Models for Industrial Planning on Complex Domains

In real world applications planners are frequently faced with complex variable dependencies in high dimensional domains. In addition to that, they typically have to start from a very incomplete picture that is expanded only gradually as new information becomes available. In this contribution we deal with probabilistic graphical models, which have successfully been used for handling complex dependency structures and reasoning tasks in the presence of uncertainty. The paper discusses revision and updating operations in order to extend existing approaches in this field, where in most cases a restriction to conditioning and simple propagation algorithms can be observed. Furthermore, it is shown how all these operations can be applied to item planning and the prediction of parts demand in the automotive industry. The new theoretical results, modelling aspects, and their implementation within a software library were delivered by ISC Gebhardt and then involved in an innovative software system for world-wide planning realized by Corporate IT of Volkswagen Group.

Extracting fuzzy classification rules from partially labeled data

The interpretability and flexibility of fuzzy if-then rules make them a popular basis for classifiers. It is common to extract them from a database of examples. However, the data available in many practical applications are often unlabeled, and must be labeled manually by the user or by expensive analyses. The idea of semi-supervised learning is to use as much labeled data as available and try to additionally exploit the information in the unlabeled data. In this paper we describe an approach to learn fuzzy classification rules from partially labeled datasets.