The process of urinary bladder cancer recurrence in 33 patients with 113 tumors, analyzed by changes in IHC features of 80 pairs of subsequent tumors.

e16507 Background: The aim was to define IHC changes between the two subsequent urinary bladder cancers (UBC). Methods: IHC data on EGFR, HER2, HER3, Ki-67, MLH1, MSH2, MSH6 and PMS2 in 113 UHC from 24 male and 9 female patients (1 to six recurrences) were used. Except for the Ki-67 value, other markers were stratified: “0” for no positive cells; “1” < = 10% positive cells; “2” 1%-30% positive cells; and “3” 31%-100% positive cells. Data of consecutive tumors were paired in 80 processes of recurrence (PoR). Changes between the latter and the former tumor were calculated: in +/- % for Ki-67 values, and as integer sums of absolute changes in expression for HER markers and for Lynch markers. EM clustering was applied for recognition of relevant IHC changes. Results: Three aspects were tested: the speed of recurrence, was it the first, or later recurrence and whether PoRs depended on the total number of tumors in that patient. Early and late PoRs clustered along the Lynch score, while the intermediate clustered along the delta Ki-67 value. The first PoRs clustered along the HER score and all subsequent PoRs depended on the Lynch score. Conclusions: Even in this limited group of patients Lynch and HER markers showed complex differences between early and late recurrent tumors. The speed of the recurrence and changes of IHC features depended mainly on the rate of change in Lynch markers, suggesting that they should be tested as predictors of UBC recurrence.[Table: see text]

A Learning-Based EM Clustering for Circular Data with Unknown Number of Clusters

Clustering is a method for analyzing grouped data. Circular data were well used in various applications, such as wind directions, departure directions of migrating birds or animals, etc. The expectation & maximization (EM) algorithm on mixtures of von Mises distributions is popularly used for clustering circular data. In general, the EM algorithm is sensitive to initials and not robust to outliers in which it is also necessary to give a number of clusters a priori. In this paper, we consider a learning-based schema for EM, and then propose a learning-based EM algorithm on mixtures of von Mises distributions for clustering grouped circular data. The proposed clustering method is without any initial and robust to outliers with automatically finding the number of clusters. Some numerical and real data sets are used to compare the proposed algorithm with existing methods. Experimental results and comparisons actually demonstrate these good aspects of effectiveness and superiority of the proposed learning-based EM algorithm.

Sistema de Apoio à Avaliação de Atividades de Programação por Reconhecimento Automático de Modelos de Soluções

O desenvolvimento de um programa de computador é um processo de resolução de problema que resulta em várias possibilidades de soluções. Dessa forma, a avaliação de exercícios de programação demanda muito esforço do professor tanto na avaliação manual, quando analisam-se várias possibilidades de soluções, quanto na avaliação automática, quando vários modelos de soluções devem ser fornecidos como entradas. Com o objetivo de auxiliar professores na identificação de modelos de soluções a partir de programas desenvolvidos por alunos, este trabalho propõe um sistema baseado em clustering para reconhecimento de modelos de soluções e para mapeamento dessas soluções em escores atribuídos por professores. Os primeiros experimentos de aplicação desse sistema em duas bases de programas desenvolvidos por estudantes de programação apresentaram resultados promissores.

Supervised Multilevel Clustering for Rainfall forecasting using Meteorological data

Atmospheric science focuses on weather processes and forecasting. Numerical and statistical analysis plays an important role in meteorological research. Meteorological data will be used to predict the changes in climatic patterns by using forecasting models and weather forecasting instruments. Data mining techniques have more scope to discover future weather patterns by analyzing past weather dimensions. In our study two techniques Multiple Linear Regression (MLR) and Expectation Maximization (EM) clustering algorithms are combined for rainfall forecasting. MLR interprets most important parameters of rainfall for clustering algorithm. EM clustering algorithm will find correctly and incorrectly clustered instances when applied on selected partitioned attributes. The model was able to forecast less rainfall, medium rainfall and high rainfall by analyzing past meteorological observations. Standard deviation is used as a measure of error correction to improve the cluster results. Data normalization helps to improve model performance. These findings are useful to determine future climate expectation.