Opioid death projections with AI-based forecasts using social media language - npj Digital Medicine

. Overall we found that each model’s error was mostly stable over both years, but did see a clear trend in 2016 being a harder year to forecast.Mean absolute error and 95% confidence intervals across our 2 test years for each statistical model. TOP appeared to be in having a more robust prediction of 2016, with a reduction of 0.6 MAE when compared to RNN. However, all models followed the same trend of 2016 being a bit harder to predict than 2017.

with a 95% confidence interval. In general, we saw a decrease in error as the number of tweets increases with the effects tailing off at 160,000 tweets.Impact of the number of tweets on error, as measured by deaths per 100k, per county for the test year . The line on the graph is fit with a LOWESS regressionOP shows a trend of decreasing error as counties have increasingly more language to build topic representations from.

We investigated the utility of such an approach by comparing our model’s performance using only a univariate input versus a multivariate one; as shown in Table. Here, we found that all models saw a considerable drop in predictive power when the language features were removed. These results highlight how each model benefited from the inclusion of language-based features from Twitter. Some models, such as the linear autoregressive ridge saw only a small increase in error, 0.

Table 3 Comparison of performance across all of our proposed machine learning models in a univariate context and their ideal history length in the number of years they have access to.To gain insight into the individual language patterns that reliably predicted future opioid deaths, we evaluated the relationship between changes in each of the topics prior to changes in the outcomes for both 2016 and 2017.

physorg_com stonybrooku NaturePortfolio Not only just that... Not at all.

Technology Technology Latest News, Technology Technology Headlines

Similar News:You can also read news stories similar to this one that we have collected from other news sources.

Towards artificial intelligence-based learning health system for population-level mortality prediction using electrocardiograms - npj Digital Medicinenpj Digital Medicine - Towards artificial intelligence-based learning health system for population-level mortality prediction using electrocardiograms
Source: medical_xpress - 🏆 101. / 51 Read more »