No abstract

The Kinect sensor is an attachment for the Xbox gaming console which allows players to interact with games through body movement. This paper explores the possibility of using the Kinect in a clinical setting for detection of Parkinson's, postural, flapping, titubation and voice tremors. All of the physical tremors were simulated and the ability of the Kinect to capture them was evaluated. The Kinect was also used to record voice data from real voice tremors patients. Physical and voice data were gathered from healthy persons for comparison. The results showed that the Kinect could reliably detect voice, postural and Parkinson's tremors. A very consistent graph could be obtained repeatedly for both Parkinson's and postural tremors. For voice tremors there was also a consistent pattern that differentiated a normal voice from one with a tremor. We have therefore proven that the Kinect can consistently record Parkinson's, postural and voice tremors.

Health datasets typically comprise of data that are heavily skewed towards the healthy class, thus resulting in classifiers being biased towards this majority class. Due to this imbalance of data, traditional performance metrics, such as accuracy, are not appropriate for evaluating the performance of classifiers with the minority class (disease-affected/unhealthy individuals). In addition, classifiers are trained under the assumption that the costs or benefits associated with different decision outcomes are equal. However, this is usually not the case with health data since it is more important to identify disease affected/unhealthy persons rather than healthy individuals. In this paper we address these problems by examining benefits/costs when evaluating the performance of classifiers. Furthermore, we focus on multiclass classification where the outcome can be one of three or more options. We propose modifications to the Naive Bayes and Logistic Regression algorithms to incorporate costs and benefits for the multiclass scenario as well as compare these to an existing algorithm, hierarchical cost-sensitive kernel logistic regression, and also an adapted hierarchical approach with our cost-benefit based logistic regression model. We demonstrate the effectiveness of all approaches for fetal health classification but the proposed approaches can be applied to any imbalance dataset where benefits and costs are important.

Online credit card fraud is an ongoing problem and with the recent COVID-19 pandemic, there has been a surge of merchants moving their businesses online. It is therefore crucial to identify fraudulent activities before it causes loss to both the bank and its customers. Due to the dynamic nature of fraudsters as well as customer spending behavior, machine learning algorithms are appropriate for this task. However, credit card fraud data is typically imbalanced, favoring the positive class (legitimate transactions), causing traditional machine learning algorithms to err on the side of this majority class; since they consider equal costs and benefits for different decision outcomes when training. Nevertheless, it is more beneficial to correctly identify fraudulent transactions. Therefore, in this paper, we propose a technique for identifying credit card fraud that first accounts for customer spending patterns by aggregating transactions to creative new features based on periodic data. Then, we consider benefits and costs when training an XGBoost classifier in order to achieve maximum benefits. We also evaluate the performance of the classifier using benefits and costs. We demonstrate the effectiveness of our approach using data provided by a bank.