Catálogo de publicaciones - libros

This study investigates the effectiveness of support vector machines (SVM) approach in detecting the underlying data pattern for the credit card customer churn analysis. This article introduces a relatively new machine learning technique, SVM, to the customer churning problem in attempt to provide a model with better prediction accuracy. To compare the performance of the proposed model, we used a widely adopted and applied Artificial Intelligence (AI) method, back-propagation neural networks (BPN) as a benchmark. The results demonstrate that SVM outperforms BPN. We also examine the effect of the variability in performance with respect to various values of parameters in SVM.

Each stage of the sewage treatment process emits odor causing compounds and these compounds may vary from one location in a sewage treatment works to another. In order to determine the boundaries of legal standards, reliable and efficient odor measurement methods need to be defined. An electronic NOSE equipped with 12 different polypyrrole sensors is used for the purpose of characterizing sewage odors. Samples collected at different locations of a WWTP were classified using a fuzzy clustering technique and a neural network trained with a back-propagation algorithm.

Traffic flow forecasting is an important issue for the application of Intelligent Transportation Systems. Due to practical limitations, traffic flow records may be partially missing or substantially contaminated by noise. In this paper, a robust traffic flow predictor, termed random subspace predictor, is developed integrating the entire spatial and temporal information in a transportation network to cope with this case. Experimental results demonstrate the effectiveness and robustness of the random subspace predictor.

Input/output hidden Markov model (IOHMM) has turned out to be effective in sequential data processing via supervised learning. However, there are several difficulties, e.g. model selection, unexpected local optima and high computational complexity, which hinder an IOHMM from yielding the satisfactory performance in sequence classification. Unlike previous efforts, this paper presents an ensemble learning approach to tackle the aforementioned problems of the IOHMM. As a result, simple IOHMMs of different topological structures are used as base learners in our boosting algorithm and thus an ensemble of simple IOHMMs tend to tackle a complicated sequence classification problem without the need of explicit model selection. Simulation results in text-dependent speaker identification demonstrate the effectiveness of boosted IOHMMs for sequence classification.

We investigate possibilities of inducing temporal structures without fading memory in recurrent networks of spiking neurons strictly operating in the pulse-coding regime. We extend the existing gradient-based algorithm for training feed-forward spiking neuron networks ( [1]) to recurrent network topologies, so that temporal dependencies in the input stream are taken into account. It is shown that temporal structures with unbounded input memory specified by simple Moore machines (MM) can be induced by recurrent spiking neuron networks (RSNN). The networks are able to discover representations of abstract information processing states coding potentially unbounded histories of processed inputs.

Recently, there has been an outburst of interest in extending topographic maps of vectorial data to more general data structures, such as sequences or trees. The representational capabilities and internal representations of the models are not well understood. We concentrate on a generalization of the Self-Organizing Map (SOM) for processing sequential data – the Recursive SOM (RecSOM [1]). We argue that contractive fixed-input dynamics of RecSOM is likely to lead to Markovian organizations of receptive fields on the map. We show that Markovian topographic maps of sequential data can be produced using a simple fixed (non-adaptable) dynamic module externally feeding a standard topographic model designed to process static vectorial data of fixed dimensionality (e.g. SOM). We elaborate upon the importance of non-Markovian organizations in topographic maps of sequential data.

A novel quantum reinforcement learning is proposed through combining quantum theory and reinforcement learning. Inspired by state superposition principle, a framework of state value update algorithm is introduced. The state/action value is represented with quantum state and the probability of action eigenvalue is denoted by probability amplitude, which is updated according to rewards. This approach makes a good tradeoff between exploration and exploitation using probability and can speed up learning. The results of simulated experiment verified its effectiveness and superiority.

Topical crawlers are becoming important tools to support applications such as specialized Web portals, online searching, and competitive intelligence. A topic driven crawler chooses the best URLs to pursue during web crawling. It is difficult to evaluate what URLs downloaded are the best. This paper presents some important metrics and an evaluation function for ranking URLs about pages relevance. We also discuss an approach to evaluate the function based on GA. GA evolving process can discover the best combination of the metrics’ weights. Avoiding misleading the result by a single topic, this paper presents a method which characterization of the metrics’ combination be extracted by mining frequent patterns. Extracting features adopts a novel FP-tree structure and FP-growth mining method based on FP-tree without candidate generation. The experiment shows that the performance is exciting, especially about a popular topic.

A novel quantum swarm evolutionary algorithm is presented based on quantum-inspired evolutionary algorithm in this article. The proposed algorithm adopts quantum angle to express Q-bit and improved particle swarm optimization to update automatically. The simulated effectiveness is examined in solving 0-1 knapsack problem.

In this paper, an algorithm built on the notion of evolutionary system is proposed. During the evolution of distributed tribes, the individuals making up the tribes cooperatively effect pixel communication from one agent to the other in order to improve the homogeneity of the ensemble of the image regions that they represent. The proposed Artificial Coevolving Tribes, containing evolution, mature, propagate, and die, is explored in the context of image. Meanwhile, the given image is recognized and segmented. Stability and scale control of the algorithm is maintained. Results obtained on natural scenes are presented with the evolutionary process.