The presence of missing data is a perennial challenge in clinical datasets, compromising the quality of analyses and the performance of predictive models. This paper addresses this problem by proposing a methodology to identify a threshold for excluding variables with missing data. Instead of using a fixed percentage of absence, the approach is based on evaluating the performance of machine learning models trained to predict the values of each missing variable. The analysis of predictive capability, combined with the percentage of missing data and the correlation structure among variables, offers a criterion based on evidence to decide between excluding and imputing a variable, a notorious challenge in clinical prognostic models. While the method is illustrated with clinical data, its flexibility allows it to be applied to a variety of other contexts, such as environmental, social, and financial data, where the interdependence between variables is crucial for imputation..

The number of people diagnosed with autism spectrum disorder (ASD) has been increasing significantly. However, the early treatment has become restricted, reducing the potential benefits from this intervention and increasing family and social expenses as a result of late interventions. Underdeveloped countries have proportionally fewer people diagnosed with ASD. Thus, the accessibility to mechanisms used to identification of ASD is important in theses countries. The aim of this paper is to evaluate the use of convolutional neural networks to assist in the identification of ASD, using static two-dimensional facial images as input. The results show success for the studied approach. MobileNet and DenseNet201 obtained the best results with an average of 90.7% accuracy and standard deviations of 0.68% and 1.64%, respectively. DenseNet201 achieved an accuracy of 93.5% in the best cases.

Self-labeled techniques, a semi-supervised classification paradigm (SSC), are highly effective in alleviating the scarcity of labeled data used in classification tasks through an iterative process of self-training. This problem was addressed by several approaches with different assumptions about the features of the input data, examples of these approaches being self-training, co-training, STRED, among others. This paper presents a framework for data self-labeling based on deep autoencoder combined with a self-labeled technique that takes into consideration cross-entropy. The model uses the Encoder to reduce the dimensionality of the input that is submitted to a labeling layer. The weights of this layer are adjusted through the learning from a clustering performed in the Z space, which is the reduced dimensionality space. Results showed that the proposed method obtained competitive performance in relation to classic methods that are found in the literature.