Catálogo de publicaciones - libros

Document clustering methods can be used to structure large sets of text or hypertext documents. Suffix Tree Clustering has been proved to be a good approach for documents clustering. However, the cluster merging algorithm of Suffix Tree Clustering is based on the overlap of their document sets, which totally ignore the similarity between the non-overlap parts of different clusters. In this paper, we introduce a novel cluster merging approach which will combines the cosine similarity and overlap percentage. Using this method, we can get a better clustering result and a comparative small number of clusters.

This paper attempts to present the new approach to design sufficient number of systematic fuzzy linguistics in matrix form and map the Fuzzy Linguistic Variable Matrix, which contains linguistic terms, into numeric domain using Fuzzy Normal Distribution based on the Parabola-based Membership Function. Existing fuzzy set theory is difficult to design the systematic and sufficient fuzzy linguistics. Due to this reason, in most practice, giving insufficient fuzzy linguistics induces inaccurate calculation whilst giving excessive fuzzy linguistics induces the parameter design problems and calculation performance. This paper presents Fuzzy Linguistic Variable Matrix and Parabola-based Fuzzy Normal Distribution (FND) as preferred framework to address the problem.

Principal curve pass through the middle of a multidimensional data set, to express the distributing shape of the points in the data set, we model principal curve for it. The new method of modeling the complex principal curve, based on B-spline network, is proposed. This method combines the polygonal line algorithm of learning principal curve with B-spline network. At one time, the algorithm finding a bifurcate point of the complex principal curve is presented. Our experimental results on simulate data demonstrate that it is feasible and effective.

Clustering ensembles have emerged as a powerful method for improving both the robustness and the stability of unsupervised classification solutions. However, finding a consensus clustering from multiple partitions is a difficult problem that can be approached from graph-based, combinatorial or statistical perspectives. We offer a probabilistic model of consensus using a finite mixture of multinomial distributions in a space of clustering. A combined partition is found as a solution to the corresponding maximum likelihood problem using the GA algorithm. The excellent scalability of this algorithm and comprehensible underlying model are particularly important for clustering of large datasets. This study includes two sections, at the first , calculate correlation matrix. this matrix show correlation between samples and we found the best samples that can be in the center of clusters. In the other section a genetic algorithm is employed to produce the most stable partitions from an evolving ensemble (population) of clustering algorithms along with a special objective function. The objective function evaluates multiple partitions according to changes caused by data perturbations and prefers those clustering that are least susceptible to those perturbations.

Coal and gas outburst is a very complex phenomenon of dynamic disaster in coal mine. There exists a complex non-linear mapping relationship which could not be described with functions between outburst risk and its influential factors. Due to the originality and superiority of artificial neural network (ANN) for modeling and imitating non-linear problems, an ANN model for prediction of outburst risk is set up. Then through practical application, the performance of commonly applied Back Propagation (BP) network for outburst risk prediction is analyzed. Aimed at the weakness of BP algorithm and based on the overall searching characteristic of Genetic Algorithm (GA), an improved compound GA-BP algorithm is used to optimize the model, then both the performance of the network and the predicting reliability of the model are improved.

Currently, companies active in the development of high-tech products has become more and more complex in the age of mass customization. Not only do they have to focus on improving product quality, but rather on gaining experience to modify the current processes in order to streamline the integrated workflow. A real-time process mining system (R-PMS) is developed to analyze the proposed XML based process data for discovering the hid-den relationship between processes. The new feature of this system is the in-corporation of the process mining engine, which is characterized by the combined capabilities of the Online Analytical Processing (OLAP) and fuzzy logic (FL), to form a robust framework for highlighting the undesirable process set-ting and parameters for further improvement in a real-time manner. The simulation results indicate that the OLAP based fuzzy approach is generally superior to those of conventional methods which offer higher flexibility on production process management with decision support ability. In this paper, the de-tailed architecture and a case study are included to demonstrate the feasibility of the proposed system.

Innovation has become recognized as a key factor in the success and even sustainability of an organization but solutions to acquiring knowledge related to innovation are lacking. Strategies such as multidisciplinary teams, suggestion boxes and incentive schemes, flat organizational structures allowing the mail clerk access to the CEO are some of the techniques employed in industry. In this paper, we suggest a psychology-based technique using scenarios to measure innovation expertise. To date we have used our inventory on a novice population, but will soon administer it to an expert population. We present the findings to one of the scenarios and note that the results are contrary to what was actually done or suggested by the innovator.

Kernel functions are used in support vector machines (SVMs) to compute dot product in a higher dimensional space. The performance of classification depends on the chosen kernel. Each kernel function is suitable for some tasks. In order to obtain a more flexible kernel function, a family of RBF kernels is proposed. Multi-scale RBF kernels are combined by including weights. These kernels allow better discrimination in the feature space, and are proved to be the Mercer’s kernels. Then, the evolutionary strategies are applied for adjusting the hyperparameters of SVM. Subsets cross validation is used to be the objective function in evolutionary process, The experimental results show that the accuracy of the proposed method is better than the ordinary approach.

Shipment consolidation is a laborious and, sometimes, tedious task for airfreight forwarders since there is enormous information to be considered and literally quite a number of practical constraints to be fulfilled. In Hong Kong, the unique forwarding operation and rapid cargo flow has further complicated the consolidating process in such a way that local forwarders are almost impossible to achieve the best selection of logistics workflow through the functions of human brain solely. However, none of the currently available intelligent logistics system is able to aid forwarders in making decisions on this crucial operation through the entire supply chain. This paper presents an Iterative Heuristics Expert System (IHES) for solving shipment consolidation problem, adopting rule-based reasoning to provide expert advice for cargo allocation and subsequently applying container loading specific heuristics to support the cargo loading process. Afterwards, the iterative improvement mechanism of IHES undertakes all outcomes until the most optimal solution is found. A presentation of the concept of IHES and its development are included in this paper with a case study conducted in Oriented Delivery Limited (a Hong Kong-based company) to validate its feasibility.

In this paper, we present an interesting filtering algorithm to perform accurate estimation in jump Markov nonlinear systems, in case of multi-target tracking. With this paper, we aim to contribute in solving the problem of model-based body motion estimation by using data coming from visual sensors. The Interacting Multiple Model (IMM) algorithm is specially designed to track accurately targets whose state and/or measurement (assumed to be linear) models changes during motion transition. However, when these models are nonlinear, the IMM algorithm must be modified in order to guarantee an accurate track. In order to deal with this problem, the IMM algorithm was combined with the Unscented Kalman Filter (UKF) [ 6 ]. Even if the later algorithm proved its efficacy in nonlinear model case; it presents a serious drawback in case of non Gaussian noise. To deal with this problem we propose to substitute the UKF with the Particle Filter (PF). To overcome the problem of data association, we propose the use of an accelerated JPDA approach based on the depth first search (DFS) technique [ 12 ]. The derived algorithm from the combination of the IMM-PF algorithm and the DFS-JPDA approach is noted DFS-JPDA-IMM-PF.