The purpose of this paper is to develop new hybrid admission decision prediction models by using Support Vector Machines (SVM) combined with a feature selection algorithm to investigate the effect of the predictor variables on the admission decision of a candidate to the School of of Physical Education and Sports at Cukurova University. Experiments have been conducted on the dataset, which contains data of participants who applied to the School in 2006. The dataset has been randomly split into training and test sets using 10-fold cross validation as well as different percentage ratios. The performance of the prediction models for the datasets has been assessed using classification accuracy, specificity, sensitivity, positive predictive value (PPV) and negative predictive value (NPV). The results show that a decrease in the number of predictor variables in the prediction models usually leads to a parallel decrease in classification accuracy.Keywords: machine learning; prediction; physical ability test; feature selection;
According to some estimates of World Health Organization, in 2014, more than 1.9 billion adults were overweight. About 13% of the world's adult population were obese. 39% of adults were overweight. The worldwide prevalence of obesity more than doubled between 1980 and 2014. Nowadays, mobile applications recording food intake of people become popular. If an improved food classification system is introduced, users take the photo of their meals and system classifies photos into the categories. Hence, we proposed a deep convolutional neural network structure trained from scratch and compared its performance with pre-trained structures Alexnet and Caffenet in INISTA 2017. This study is the extended version of it. Three different deep convolutional neural networks were trained from scratch by using different learning methods: stochastic gradient descent, Nesterov's accelerated gradient and Adaptive Moment Estimation, and compared with Alexnet and Caffenet fine-tuned with the same learning algorithms. Train, validation and test datasets were generated from Food11 and Food101 datasets. All tests were implemented through NVIDIA Digit interface on GeForce GTX1070. According to the test results, although pre-trained models provided better results than proposed structures, their performances were comparable. Moreover, learning optimization methods accelerated and improved the performances of all the compared models.
The purpose of this paper is to develop new hybrid admission decision prediction models by using Support Vector Machines (SVM) combined with a feature selection algorithm to investigate the effect of the predictor variables on the admission decision of a candidate to the School of of Physical Education and Sports at Cukurova University. Experiments have been conducted on the dataset, which contains data of participants who applied to the School in 2006. The dataset has been randomly split into training and test sets using 10-fold cross validation as well as different percentage ratios. The performance of the prediction models for the datasets has been assessed using classification accuracy, specificity, sensitivity, positive predictive value (PPV) and negative predictive value (NPV). The results show that a decrease in the number of predictor variables in the prediction models usually leads to a parallel decrease in classification accuracy.
Drug-target interaction prediction provides important information that could be exploited for drug discovery, drug design, and drug repurposing. Chemogenomic approaches for predicting drug-target interaction assume that similar receptors bind to similar ligands. Capturing this similarity in so-called “fingerprints” and combining the target and ligand fingerprints provide an efficient way to search for protein-ligand pairs that are more likely to interact. In this study, we constructed drug and target fingerprints by employing features extracted from the DrugBank. However, the number of extracted features is quite large, necessitating an effective feature selection mechanism since some features can be redundant or irrelevant to drug-target interaction prediction problems. Although such feature selection methods are readily available in the literature, usually they act as black boxes and do not provide any quantitative information about why a specific feature is preferred over another. To alleviate this lack of human interpretability, we proposed a novel feature selection method in which we used an autoencoder as a symmetric learning method and compared the proposed method to some popular feature selection algorithms, such as Kbest, Variance Threshold, and Decision Tree. The results of a detailed performance study, in which we trained six Multi-Layer Perceptron (MLP) Networks of different sizes and configurations for prediction, demonstrate that the proposed method yields superior results compared to the aforementioned methods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.