Forecasting Potential Diabetes Complications

Authors: Yang Yang, Walter Luyten, Lu Liu, Marie-Francine Moens, Jie Tang, Juanzi Li

AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the proposed model on a large collection of real medical records. Sparse FGM outperforms (+20% by F1) baselines significantly and gives detailed associations between diabetes complications and lab tests. In this section, we present experimental results to demonstrate the effectiveness of the proposed approach.
Researcher Affiliation Academia Yang Yang Tsinghua University yangyang@keg.cs.tsinghua.edu.cn Walter Luyten Katholieke Universiteit Leuven Walter.Luyten@med.kuleuven.be Lu Liu Northwestern University liulu26@gmail.com Marie-Francine Moens Katholieke Universiteit Leuven sien.moens@cs.kuleuven.be Jie Tang Tsinghua University jietang@tsinghua.edu.cn Juanzi Li Tsinghua University ljz@keg.cs.tsinghua.edu.cn
Pseudocode Yes Algorithm 1: Learning algorithm of Sparse FGM.
Open Source Code Yes All codes used in the paper are publicly available 3http://arnetminer.org/diabetes
Open Datasets No We use a collection of real medical records from a famous geriatric hospital. The data set spans one year, containing 181,933 medical records corresponding to 35,525 unique patients and 1, 945 kinds of lab tests in total.
Dataset Splits No In the experiments, we randomly picked 60% of the medical records as training data and the rest for testing.
Hardware Specification Yes All algorithms were implemented in C++, and all experiments were performed on a Mac running Mac OS X with Intel Core i7 2.66 GHz and 4 GB of memory.
Software Dependencies No All algorithms were implemented in C++, and LIBSVM (Chang and Lin 2011) is employed as the classification model for complication forecasting. However, no specific version numbers for software components like C++ libraries or LIBSVM are provided.
Experiment Setup Yes In all experiments, we empirically set the number of latent variables in Sparse FGM to 100, and set η = 0.1. We employ a gradient descent algorithm to learn the parameters in FGM (Tang, Zhuang, and Tang 2011), and set the learning rate parameter as 0.1.