More than one billion dollars' worth of work takes place every year in online workplaces like Upwork.com and Elance.com. We analyze the structure of these jobs and build a classifier using logistic regression and gradient tree boosting to identify jobs in trouble. We then report the effectiveness of this classifier in a user experiment. This submission to the ICML crowdsourcing workshop is part of a longer work involving detecting and intervening on bad jobs and preventing bad jobs in the future.
展开▼
