Catálogo de publicaciones - libros

In a recommender system where users rate items we predict the rating of items users have not rated. We define a rating graph containing users and items as vertices and ratings as weighted edges. We extend the work of [1] that uses the on the bipartite rating graph incorporating negative edge weights into the calculation of the resistance distance. This algorithm is then compared to other rating prediction algorithms using data from two rating corpora.

This research is focused on developing effective visualization tools for query construction and advanced exploration of temporal relational databases. Temporal databases enable the retrieval of each of the states observed in the past and even planned future states. Several query languages for relational databases have been introduced, but only a few of them deal with temporal databases. Moreover, most users are not highly skilled in query formulation and hence are not able to define complex queries. The visual approach introduced here aims at simplifying the query construction process. It gives the user the option to define complex temporal constructs and provides visual tools with which to explore the returned networks intuitively. The exploration process should provide better insight into networks of entities, reveal patterns between the entities, and enable the user to forecast the behavior of entities in the future. A visual query language as an isolated subsystem is not sufficient in itself for a complete data analysis process. A query’s output should be further explored to find patterns that are hidden in the output.

This paper reports on a first implementation of a webservice that supports image-based queries within the domain of tree taxonomy. As such, it serves as an example relevant to many other possible applications within the field of biodiversity and photo-identification. Without any human intervention matching results are produced through a chain of computer vision and image processing techniques, including segmentation and automatic shape matching. A selection of shape features is described and the architecture of the webservice is explained. Classification techniques are presented and preliminary results shown with respect to the success rate. Necessary future enhancements are discussed. Benefits are highlighted that could result from redesigning image-based expert systems as web services, open to the public at large.

A process of Knowledge Discovery in Databases (KDD) involving large amounts of data requires a considerable amount of computational power. The process may be done on a dedicated and expensive machinery or, for some tasks, one can use distributed computing techniques on a network of affordable machines. In either approach it is usual the user to specify the of the sub-tasks composing the whole KDD process before execution starts.

In this paper we propose a technique that we call . The feature of the technique is due to its capability of generating new sub-tasks of the Data Mining analysis process at execution time. The of sub-tasks of the DM is, therefore, dynamic.

To deploy the proposed technique we extended the Distributed Data Mining system HARVARD and adapted an Inductive Logic Programming system (IndLog) used in a Relational Data Ming task.

As a proof-of-concept, the extended system was used to analyse an artificial dataset of a credit scoring problem with eighty million records.

We present , a privacy preserving algorithm for mining time series data. We assume data is split among several sites. The problem is to find all frequent subsequences of time series without revealing local data to any site. Our solution exploit density estimate and secure multiparty computation techniques to provide privacy to a given extent.

A novel non linear signal prediction method is presented using non linear signal analysis and deterministic chaos techniques in combination with neural networks for a diode resonator chaotic circuit. Multisim is used to simulate the circuit and show the presence of chaos. The Time series analysis is performed by the method proposed by Grasberger and Procaccia, involving estimation of the correlation and minimum embedding dimension as well as of the corresponding Kolmogorov entropy. These parameters are used to construct the first stage of a one step / multistep predictor while a back-propagation Artificial Neural Network (ANN) is involved in the second stage to enhance prediction results. The novelty of the proposed two stage predictor lies on that the backpropagation ANN is employed as a second order predictor, that is as an error predictor of the non-linear signal analysis stage application. This novel two stage predictor is evaluated through an extensive experimental study.

The paper focuses on usage of disjunction of items in association rules mining. We used the GUHA method instead of the traditional algorithm and enhanced the former implementations of the method with ability of disjunctions setting between items. Experiments were conducted in our Ferda data mining environment on data from the medical domain. We found strong and meaningful association rules that could not be obtained without the usage of disjunction.