Inspired by the healthcare industry, IQVIA commits to providing solutions that enable life sciences companies to innovate with confidence, maximize their opportunities and ultimately drive human health outcomes forward.
In this period of 3-months internship at IQVIA, I developed a semi-supervised rare disease detection framework using a generative adversarial network (GAN). Our model leverage the large amount of unlabeled dataset with a limited labeled dataset, and applies generative model to boost the prediction performance. The result of our best model beats the common supervised-learning methods (logistic regression, neural network, and random forest) by around 5% in terms of precision-recall AUC score. We also build up a web API for user to interact with our model using Flask.