A Hamiltonian Monte Carlo Model for Imputation and Augmentation of Healthcare Data

• DOI: 10.48550/arxiv.2103.02349
• Publication Date: 2021-01-01
• Authors: Narges Pourshahrokhi; Samaneh Kouchaki; Kord M. Kober; Christine Miaskowski; Payam Barnaghi
• Keywords: Imputation (statistics)

Missing values exist in nearly all clinical studies because data for a variable or question are not collected available. Inadequate handling of missing can lead to biased results and loss statistical power analysis. Existing models usually do consider privacy concerns utilise the inherent correlations across multiple features impute values. In healthcare applications, we confronted with high dimensional sometimes small sample size datasets that need more effective augmentation imputation techniques. Besides, processes traditionally conducted individually. However, imputing augmenting significantly improve generalisation avoid bias machine learning models. A Bayesian approach creating augmented samples is proposed this work. We propose folded Hamiltonian Monte Carlo (F-HMC) inference as practical process cross-dimensional relations by applying random walk dynamics adapt posterior distribution generate large-scale samples. The method applied cancer symptom assessment dataset confirmed enrich quality precision, accuracy, recall, F1 score, propensity metric.