(function(doc, html, url) { var widget = doc.createElement("div"); widget.innerHTML = html; var script = doc.currentScript; // e = a.currentScript; if (!script) { var scripts = doc.scripts; for (var i = 0; i < scripts.length; ++i) { script = scripts[i]; if (script.src && script.src.indexOf(url) != -1) break; } } script.parentElement.replaceChild(widget, script); }(document, '
Related
There have been different efforts to predict epileptic seizures and most of them are based on the analysis of electroencephalography (EEG) signals; however, recent publications have suggested that functional Near-Infrared Spectroscopy (fNIRS), a relatively new technique, could be used to predict seizures. The objectives of this research are to show that the application of fNIRS to epileptic seizure detection yields results that are superior to those based on EEG and to demonstrate that the application of deep learning to this problem is suitable given the nature of fNIRS recordings. A Convolutional Neural Network (CNN) is applied to the prediction of epileptic seizures from fNIRS signals, an optical modality for recording brain waves. The implementation of the proposed method is presented in this work. Application of CNN to fNIRS recordings showed an accuracy ranging between 96.9% and 100%, sensitivity between 95.24% and 100%, specificity between 98.57% and 100%, a positive predictive value between 98.52% and 100%, and a negative predictive value between 95.39% and 100%. The most important aspect of this research is the combination of fNIRS signals with the particular CNN algorithm. The fNIRS modality has not been used in epileptic seizure prediction. A CNN is suitable for this application because fNIRS recordings are high dimensional data and they can be modeled as three-dimensional tensors for classification.
Related
This paper presents the theory, methodology and application of a new predictive model for time series within the financial sector, specifically data from 20 companies listed on the U.S. stock exchange market. The main impact of this article is (1) the proposal of a recommender system for financial investment to increase the cumulative gain; (2) an artificial predictor that beats the market in most cases; and (3) the fact that, to the best of our knowledge, this is the first effort to predict time series by learning redundant dictionaries to sparsely reconstruct these signals. The methodology is conducted by finding the optimal set of predicting model atoms through two directions for dictionaries generation: the first one by extracting atoms from past daily return price values in order to build untrained dictionaries; and the second one, by atom extraction followed by training of dictionaries though K-SVD. Prediction of financial time series is a periodic process where each cycle consists of two stages: (1) training of the model to learn the dictionary that maximizes the probability of occurrence of an observation sequence of return values, (2) prediction of the return value for the next coming trading day. The motivation for such research is the fact that a tool, which might generate confidence of the potential benefits obtained from using formal financial services, would encourage more participation in a formal system such as the stock market. Theory, issues, challenges and results related to the application of sparse representation to the prediction of financial time series, as well as the performance of the method, are presented.
Related
Recent research on Parkinson disease (PD) detection has shown that vocal disorders are linked to symptoms in 90% of the PD patients at early stages. Thus, there is an interest in applying vocal features to the computer-assisted diagnosis and remote monitoring of patients with PD at early stages. The contribution of this research is an increase of accuracy and a reduction of the number of selected vocal features in PD detection while using the newest and largest public dataset available. Whereas the number of features in this public dataset is 754, the number of selected features for classification ranges from 8 to 20 after using Wrappers feature subset selection. Four classifiers (k nearest neighbor, multi-layer perceptron, support vector machine and random forest) are applied to vocal-based PD detection. The proposed approach shows an accuracy of 94.7%, sensitivity of 98.4%, specificity of 92.68% and precision of 97.22%. The best resulting accuracy is obtained by using a support vector machine and it is higher than the one, which was reported on the first work to use the same dataset. In addition, the corresponding computational complexity is further reduced by selecting no more than 20 features.