2018 Neukom Research Prize Winners

Neukom Prize for Outstanding Graduate Research

A Sentiment-Based Gradient Boosting Tree Model for Predicting Alternative Cryptocurrency Price Fluctuations

Tianyu Ray Lia, Anup S. Chamrajnagara, Xander R. Fonga, Nicholas R. Rizika

In this paper, we analyze Twitter signals as a medium for user sentiment to predict the price fluctuations of a small cap alternative cryptocurrency called ZClassic. We extracted tweets on an hourly basis for a period of 3.5 weeks, classifying each tweet as positive, neutral, or negative. We then compiled these tweets into an hourly sentiment index, creating an unweighted and weighted index, with the latter giving larger weight to retweets. These two indices, alongside the raw summations of positive, negative, and neutral sentiment were juxtaposed to ⇠ 400 data points of hourly pricing data to train an Extreme Gradient Boosting Regression Tree Model. Price predictions produced from this model were compared to historical price data, with the resulting predictions having a 0.81 correlation with the testing data.

Our model’s predictive data yielded statistical significance at the p < 0.0001 level. Our model is the first academic proof of concept that social media platforms such as Twitter can serve as powerful social signals for predicting price movements in the highly speculative alternative cryptocurrency, or “alt-coin”, market.

Neukom Prize for Outstanding Undergraduate Research

Machine learning based classification of deep brain stimulation outcomes in a rat model of binge eating using ventral striatal oscillations

Lucas Dwiel

Neuromodulation-based interventions continue to be evaluated across an array of appetitive disorders but broader implementation of these approaches remains limited due to variable treatment outcomes. 

We hypothesize that individual variation in treatment outcomes may be linked to differences in the networks underlying these disorders. Here, Spragu-Dawley rats received deep brain stimulation separately within each nucleus accumbens (NAc) sub-region (core and shell) using a within-animal crossover design in a rat model of binge eating. Significant reductions in binge size were observed with stimulation of either target but with significant variation in effectiveness across individuals. When features of local field potentials (LFPs) recorded from the NAc were used as predictors of the pre-defined stimulation outcomes (response or non-response) from each rat using a machine-learning approach (lasso), stimulation outcomes could be predicted with greater accuracy than expected by chance (effect sizes: core = 1.13, shell = 1.05). Further, these LFP features could be used to identify the best stimulation target for each animal (core vs. shell) with an effect size = 0.96.

These data suggest that individual differences in underlying network activity may contribute to the variable outcomes of circuit based interventions, and measures of network activity have the potential to individually guide the selection of an optimal stimulation target and improve overall treatment response rates.