CSSI Elements: Development of Assumption-Free Parallel Data Curing Service for Robust Machine Learning and Statistical Predictions

Develop an assumption-free, general-purpose missing data curing service on parallel computing environment for robust machine learning and statistical inference using large/big incomplete data in broad science and engineering. Parallel Fractional Hot Deck Imputation (P-FHDI) is being developed in conjunction with hybrid parallelisms and sure independence screening technique. Results demonstrate a promising performance of the P-FHDI in improving subsequent machine learning and statistical prediction with large incomplete data with favorable scalability.