Open utterances-bot opened 1 year ago
Nice write-up! Another idea could be to fiddle with the threshold, and maybe not use the default 0.5 for class prediction. This doesn't improve the model as such, but results in shifts in the confusion matrix, and lower false negatives.
Thanks! It's definitely something I'd like to have another look at - and changing thresholds is a great idea!
Detecting heart murmurs from time series data in R | Nicola Rennie
Time series analysis can uncover hidden structures in data collected over time. In this blog post, I'll use {tsfeatures} to extract time series features and {tidymodels} to predict which sound recordings of heartbeats contain heart murmurs, using those time series features.
https://nrennie.rbind.io/blog/detecting-heart-murmurs-time-series-r-tidymodels/