Abstract: In this work, clustering analysis of two partial nitritation-anammox (PN-A) moving bed biofilm reactors (MBBR) containing different types of carrier material was explored for the identification of patterns and operational conditions that may benefit process performance. The systems ran for two years under fluctuations of temperature and organic matter. Ex situ batch activity tests were performed every other week during the operation of these reactors. These datasets and the parameters, which were monitored online and in the laboratory, were combined and analyzed applying clustering analysis to identify non-obvious information regarding the performance of the systems. The initial results were consistent with the literature and from an operational perspective, which allowed the parameters to be explored further. The new information revealed that the oxidation reduction potential (ORP) and the anaerobic ammonium oxidizing bacteria (AnAOB) activity correlated well. ORP also dropped when the reactors were exposed to real wastewater (presence of organic matter). Moreover, operating conditions during nitrite accumulation were identified through clustering, and also revealed inhibition of anammox bacteria already at low nitrite concentrations.

Abstract: The aim was to find craniofacial morphology patterns in a multivariate cephalometric database using a clustering technique. Cephalometric analysis was performed in a sample of 100 teleradiographs collected from Chilean orthodontic patients. Thirty cephalometric measurements were taken from commonly used analysis. The computed variables were used to perform a clustering analysis with the k-means algorithm to identify patterns of craniofacial morphology. The J48 decision tree was used to analyze each cluster, and the ANOVA test to determine the statistical differences between the clusters. Four clusters were found that had significant differences (P<0.001) in 24 of the 30 variables studied, suggesting that they represent different patterns of craniofacial form. Using the decision tree, 8 of the 30 variables appeared to be relevant for describing the clusters. The clustering analysis is effective in identifying different craniofacial patterns based on a multivariate database. The distinct clusters appear to be caused by differences in the compensation process of the facial structure responding to a genetically determined cranial and mandible form. The proposed method can be applied to several databases, creating specific classifications for each one of them.

Abstract: To study concepts that are coded in language, researchers often collect lists of conceptual properties produced by human subjects. From these data, different measures can be computed. In particular, inter-concept similarity is an important variable used in experimental studies. Among possible similarity measures, the cosine of conceptual property frequency vectors seems to be a de facto standard. However, there is a lack of comparative studies that test the merit of different similarity measures when computed from property frequency data. The current work compares four different similarity measures (cosine, correlation, Euclidean and Chebyshev) and five different types of data structures. To that end, we compared the informational content (i.e., entropy) delivered by each of those 4 x 5 = 20 combinations, and used a clustering procedure as a concrete example of how informational content affects statistical analyses. Our results lead us to conclude that similarity measures computed from lower-dimensional data fare better than those calculated from higher-dimensional data, and suggest that researchers should be more aware of data sparseness and dimensionality, and their consequences for statistical analyses.

Abstract: Assembling spatial units into meaningful clusters is a challenging task, as it must cope with a consequential computational complexity while controlling for the modifiable areal unit problem (MAUP), spatial autocorrelation and attribute multicolinearity. Nevertheless, these effects can reveal significant interactions among diverse spatial phenomena, such as segregation and economic specialization. Various regionalization methods have been developed in order to address these questions, but key fundamental properties of the aggregation of spatial entities are still poorly understood. In particular, due to the lack of an objective stopping rule, the question of determining an optimal number of clusters is yet unresolved. Therefore, we develop a clustering algorithm which is sensitive to scalar variations of multivariate spatial correlations, recalculating PCA scores at several aggregation steps in order to account for differences in the span of autocorrelation effects for diverse variables. With these settings, the scalar evolution of correlation, compactness and isolation measures is compared between empirical and 120 random datasets, using two dissimilarity measures. Remarkably, adjusting several indicators with real and simulated data allows for a clear definition of a stopping rule for spatial hierarchical clustering. Indeed, increasing correlations with scale in random datasets are spurious MAUP effects, so they can be discounted from real data results in order to identify an optimal clustering level, as defined by the maximum of authentic spatial self-organization. This allows singling out the most socially distressed areas in Greater Santiago, thus providing relevant socio-spatial insights from their cartographic and statistical analysis. In sum, we develop a useful methodology to improve the fundamental comprehension of spatial interdependence and multiscalar self-organizing phenomena, while linking these questions to relevant real world issues. (c) 2015 Elsevier Ltd. All rights reserved.

Abstract: The objective of territorial design for a distribution company is the definition of geographic areas that group customers. These geographic areas, usually called districts or territories, should comply with operational rules while maximizing potential sales and minimizing incurred costs. Consequently, territorial design can be seen as a clustering problem in which clients are geographically grouped according to certain criteria which usually vary according to specific objectives and requirements (e.g. costs, delivery times, workload, number of clients, etc.). In this work, we provide a novel hybrid approach for territorial design by means of combining a K-means-based approach for clustering construction with an optimization framework. The K-means approach incorporates the novelty of using tour length approximation techniques to satisfy the conditions of a pork and poultry distributor based in the region of Valparaiso in Chile. The resulting method proves to be robust in the experiments performed, and the Valparaiso case study shows significant savings when compared to the original solution used by the company.

Abstract: Although the development of new supervised learning algorithms for machine learning techniques are mostly oriented to improve the predictive power or classification accuracy, the capacity to understand how the classification process is carried out is of great interest for many applications in business and industry. Inductive learning algorithms, like the Rules family, induce semantically interpretable classification rules in the form of if-then rules. Although the effectiveness of the Rules family has been studied thoroughly and new and improved versions are constantly been developed, one important drawback is the effect of the presentation order of the training patterns which has not been studied in depth previously. In this paper this issue is addressed, first by studying empirically the effect of random presentation orders in the number of rules and the generalization power of the resulting classifier. Then a presentation order method for the training examples is proposed which combines a clustering stage with a new density measure developed specifically for this problem. The results using benchmark datasets and a real application of wood defect classification show the effectiveness of the proposed method. Also, since the presentation order method is employed as a preprocessing stage, the simplicity of the Rules family is not affected but instead it enables the generation of fewer and more accurate rules, which can have a direct impact in the performance and usefulness of the Rules family in an expert system context. (C) 2016 Elsevier Ltd. All rights reserved.

Abstract: For developing countries, such as Chile, we study the influential factors for adoption and usage of broadband services. In particular, subsidies on the broadband price are analyzed to see if this initiative has a significant effect in the broadband penetration. To carry out this study, machine learning techniques are used to identify different household profiles using the data obtained from a survey on access, use, and users of broadband Internet from Chile. Different policies are proposed for each group found, which were then evaluated empirically through Bayesian networks. Results show that an unconditional subsidy for the Internet price does not seem to be very appropriate for everyone since it is only significant for some households groups. The evaluation using Bayesian networks showed that other polices should be considered as well such as the incorporation of computers, Internet applications development, and digital literacy training. (C) 2013 Elsevier Ltd. All rights reserved.