Search

At present, the demand for information in distributed database systems is large and growing day by day. While this is happening, new challenges arise to improve the performance of the databases. Data fragmentation and replication methods have a leading role in distributed systems over the cloud, which is why this paper presents the design of a horizontal fragmentation, allocation and replication method in the cloud. This research proposes an algorithm for solving the problem of overlapping horizontal fragments in a data fragmentation and replication method in the cloud. The design of a Web application using the aforementioned method is also presented, this allows the fragmentation, assignment and replication scheme proposed by the method to be directly applied on a distributed database.

Plant identification belongs to a specific application domain of data mining. Images of plant leaves are usually used as the main element to distinguish a plant from another. For proper identification, feature extraction is necessary. In the literature, most plant recognition systems use the features along with a classification method, which has been adapted or modified to face this type of application. In this paper, we propose three new geometric features that describe the vertical and horizontal symmetry of leaves. These features are simple to extract from images. According to the results of experiments, when these features are used in conjunction with other well-known geometric characteristics, the performance of classical classification methods is remarkably improved. To show the effectiveness of the proposal, we test seven classifiers with images of leaves publicly available on the Internet.

Support vector machine (SVM) is not suitable for classification on large data sets due to its training complexity. Convex hull can simplify SVM training, however the classification accuracy becomes lower when there are inseparable points. This paper introduces a novel method for SVM classification, called convex-concave hull. After a grid processing, the convex hull is used to find extreme points. Then we detect a concave (non-convex) hull, the vertices of it are used to train SVM. We applied the proposed method on several problems. Experimental results demonstrate that our approach has good classification accuracy while the training is significantly faster than other state of the art methods.

Support Vector Machine (SVM) is an important classification method used in a many areas. The training of SVM is almost O(n^{2}) in time and space. Some methods to reduce the training complexity have been proposed in last years. Data selection methods for SVM select most important examples from training data sets to improve its training time. This paper introduces a novel data reduction method that works detecting clusters and then selects some examples from them. Different from other state of the art algorithms, the novel method uses a decision tree to form partitions that are treated as clusters, and then executes a guided random selection of examples. The clusters discovered by a decision tree can be linearly separable, taking advantage of the Eidelheit separation theorem, it is possible to reduce the size of training sets by carefully selecting examples from training sets. The novel method was compared with LibSVM using public available data sets, experiments demonstrate an important reduction of the size of training sets whereas showing only a slight decreasing in the accuracy of classifier.

In this paper, we present an algorithm to speed up the training of SVM. The proposed algorithm is based on SV candidates selection strategy, exploiting the observation that typically from a set of elements with the same label, if there exist SV, then most of them are on the boundary of the set. We compute the non convex hull sets that envelop the elements with the same label, this sets have in general a few elements compared with the entire data set. To train the SVM we use only the non convex hulls sets, which improves the training time. According to the results, our algorithm gives good accuracy and the training time is reduced considerably (under certain run conditions), the proposed algorithm is suitable for datasets with small number of features yet.