[Proposal-ML] Relating Physical Activity to Problematic Internet Use

Killian Conyngham, Jackson M Luckey, Fabian Pawelczyk

Keywords: Deep Learning, Time Series, Actigraphy, Semi-Supervised Learning

Abstract:

The focus of our project is to predict the risk of problematic internet usage (PIU) by minors, based on data of their physical activities. In particular, our goal is to build a model for entrance in the Kaggle competition run by the Child Mind Institute on this topic. Here, one key challenge lies in the high level of missing data—particularly concerning the target variable—and the actigraphy time-series component of the dataset. Our primary objective will be to design and train a deep neural network to effectively address these issues.

Submission Number: 51