We attempted to apply a novel approach to stock market predictions. The Logistic Regression machine learning algorithm (Joseph Berkson) was applied to analyze news article headlines as represented by a bag-of-words (tri-gram and single-gram) representation in an attempt to predict the trends of stock prices based on the Dow Jones Industrial Average. The results showed that a tri-gram bag led to a 49% trend accuracy, a 1% increase when compared to the single-gram representation’s accuracy of 48%.

Suicide is a significant public health problem, with incidence rates and lethality continuing to increase yearly. Given the large human and financial cost of suicide worldwide alongside the lack of progress in suicide prediction, more research is needed to inform suicide prevention and intervention efforts. This study approaches suicide from the lens of suicide note-leaving behavior, which can provide important information on predictors of suicide. Specifically, this study adds to the existing literature on note-leaving by examining history of suicidality, mental health problems, and their interaction in predicting suicide note-leaving, in addition to demographic predictors of note-leaving examined in previous research using data from the National Violent Death Reporting System (NVDRS, n = 98,515). We fit a logistic regression model predicting leaving a suicide note or not, the results of which indicated that those with mental health problems or a history of suicidality were more likely to leave a suicide note than those without such histories, and those with both mental health problems and a history of suicidality were most likely to leave a suicide note. These findings reinforce the need to tailor suicide prevention efforts toward identifying and targeting higher risk populations.

In the following paper, I aim to form relationships between different patient factors and no-show rates. The culmination of these relationships will then be used in a logistic regression model. Data collected from a survey at 26 HonorHealth clinics were analyzed using odds ratios and relative risk methods. Of 310,307 visits collected, 22,280 of them were no shows (7.2%), an 11% decrease from national averages (18.8%). This fueled the study, along with a grant filed by HonorHealth looking at the impact of telehealth on the working poor. A binary logistic regression method was run over the data, and less than 1% of patients' no-shows were predicted correctly. By adding factors, and improving the diversity in the data collected, model accuracy can be improved.