Miroslav Snorek

Temporal analysis of clusters of supermarket customers: conventional versus interval set approach

Temporal data mining is the application of data mining techniques to data that takes the time dimension into account. This paper studies changes in cluster characteristics of supermarket customers over a 24 week period. Such an analysis can be useful for formulating marketing strategies. Marketing managers may want to focus on specific groups of customers. Therefore they may need to understand the migrations of the customers from one group to another group. The marketing strategies may depend on the desirability of these cluster migrations. The temporal analysis presented here is based on conventional and modified Kohonen self organizing maps (SOM). The modified Kohonen SOM creates interval set representations of clusters using properties of rough sets. A description of an experimental design for temporal cluster migration studies including, data cleaning, data abstraction, data segmentation, and data sorting, is provided. The paper compares conventional and non-conventional (interval set) clustering techniques, as well as temporal and non-temporal analysis of customer loyalty. The interval set clustering is shown to provide an interesting dimension to such a temporal analysis.

2010 Special Issue: Meta-learning approach to neural network optimization

Optimization of neural network topology, weights and neuron transfer functions for given data set and problem is not an easy task. In this article, we focus primarily on building optimal feed-forward neural network classifier for i.i.d. data sets. We apply meta-learning principles to the neural network structure and function optimization. We show that diversity promotion, ensembling, self-organization and induction are beneficial for the problem. We combine several different neuron types trained by various optimization algorithms to build a supervised feed-forward neural network called Group of Adaptive Models Evolution (GAME). The approach was tested on a large number of benchmark data sets. The experiments show that the combination of different optimization algorithms in the network is the best choice when the performance is averaged over several real-world problems.

HyperNEAT controlled robots learn how to drive on roads in simulated environment

In this paper we describe simulation of autonomous robots controlled by recurrent neural networks, which are evolved through indirect encoding using HyperNEAT algorithm. The robots utilize 180 degree wide sensor array. Thanks to the scalability of the neural network generated by HyperNEAT, the sensor array can have various resolution. This would allow to use camera as an input for neural network controller used in real robot. The robots were simulated using software simulation environment. In the experiments the robots were trained to drive with imaximum average speed. Such fitness forces them to learn how to drive on roads and avoid collisions. Evolved neural networks show excellent scalability. Scaling of the sensory input breaks performance of the robots, which should be gained back with re-training of the robot with a different sensory input resolution.

Temporal Web usage mining

Temporal Web usage mining involves application of data mining techniques on temporal Web usage data to discover temporal patterns, which describe the temporal behavior of Web users. Clusters and associations in Web usage mining do not necessarily have crisp boundaries. We introduce the temporal Web usage mining of Web users on one educational Web site, using the adapted Kohonen SOM based on rough set properties [L. J. Pawan et al. (2002)].

NEAT in HyperNEAT substituted with genetic programming

In this paper we present application of genetic programming (GP) [1] to evolution of indirect encoding of neural network weights. We compare usage of original HyperNEAT algorithm with our implementation, in which we replaced the underlying NEAT with genetic programming. The algorithm was named HyperGP. The evolved neural networks were used as controllers of autonomous mobile agents (robots) in simulation. The agents were trained to drive with maximum average speed. This forces them to learn how to drive on roads and avoid collisions. The genetic programming lacking the NEAT complexification property shows better exploration ability and tends to generate more complex solutions in fewer generations. On the other hand, the basic genetic programming generates quite complex functions for weights generation. Both approaches generate neural controllers with similar abilities.

Clustering supermarket customers using rough set based Kohonen networks

This paper describes the creation of intervals of clusters of supermarket customers based on the modified Kohonen self-organizing maps. The supermarket customers from three different markets serviced by a Canadian national supermarket chain were clustered based on their spending and visit patterns. The resulting rough set based clustering captured the similarities and differences between the characteristics of the three regions.

Temporal Cluster Migration Matrices for Web Usage Mining

Time can play a crucial role in the analysis of web usage. Temporal data mining has been an active area of research. However, there is little work on the analysis of cluster memberships over time. Typical clustering operations in web mining involve finding natural groupings of web resources or web users. Changes in clusters can provide important clues about the changing nature of the usage of a web site, as well as changing loyalties of web users. This paper addresses two different types of temporal changes in cluster analysis. The changes in cluster compositions over time and changes in cluster memberships of individual web users. The paper also proposes the concept of temporal cluster migration matrices (TCMM). The proposed matrices are shown to be useful for analyzing the changing nature of a web site as well as changing patronages of individual web users. TCMM can be used as a visualization technique to study results obtained from temporal data mining, which can be more complex because of the additional time dimension.

Modeling of student's quality by means of GMDH algorithms

This article presents a method of modeling based on a data set by means of the GMDH algorithms. The data set contains information about fresh students of the Faculty of Electrical Engineering, Czech Technical University of Prague, Czech Republic (CTU, FEE). The motivation for our investigation was to discover, whether and how a particular student will be successful in his/her university studies. The data set was mined from official study application forms. They reflect students personality, type and results (grades) from high school and entrance examination. These results serve as input vectors for prediction of his/her study results after the 1st semester. The results produced by the model were then compared with the real ones. The second motivation for our investigation was to find the significance of a particular input vector component, because it enables us to identify possible weak points of a student. As appropriate tools the GMDH neural networks (both linear and nonlinear) have been used.

ECG Signal Classification Using GAME Neural Network and Its Comparison to Other Classifiers

Long term Holter monitoring is widely applied to patients with heart diseases. Many of those diseases are not constantly present in the ECG signal but occurs from time to time. To detect these infrequent problems the Holter long time ECG recording is recorded and analysed. There are many methods for automatic detection of irregularities in the ECG signal. In this paper we will comapare the Support Vector Machine (SVM), J48 decision tree (J48), RBF artificial neural network (RBF), Simple logistic function and our novel GAME neural network for detection of the Premature Ventricular Contractions. We will compare and discuss classification performance of mentioned methods. There are also very many features which describes the ECG signal therefore we will try to identify features important for correct classification and examine how the accuracy is affected with only selected features in training set.

Feature Ranking Derived from Data Mining Process

Most common feature ranking methods are based on the statistical approach. This paper compare several statistical methods with new method for feature ranking derived from data mining process. This method ranks features depending on percentage of child units that survived the selection process. A child unit is a processing element transforming the parent input features to the output. After training, units are interconnected in the feedforward hybrid neural network called GAME. The selection process is realized by means of niching genetic algorithm, where units connected to least significant features starve and fade from population. Parameters of new FR algorithm are investigated and comparison among different methods is presented on well known real world and artificial data sets.