A Deep Learning-Cuckoo Search Method for Missing Data Estimation in High-Dimensional Datasets

This study brings together two related areas: deep learning and swarm intelligence for missing data estimation in high-dimensional datasets. The growing number of studies in the deep learning area warrants a closer look at its possible application in the aforementioned domain. Missing data being an unavoidable scenario in present day datasets results in different challenges which are nontrivial for existing techniques which constitute narrow artificial intelligence architectures and computational intelligence methods. This can be attributed to the large number of samples and high number of features. In this paper, we propose a new framework for the imputation procedure that uses a deep learning method with a swarm intelligence algorithm, called Deep Learning-Cuckoo Search (DL-CS). This technique is compared to similar approaches and other existing methods. The time required to obtain accurate estimates for the missing data entries surpasses that of existing methods, but this is considered a worthy bargain when the accuracy of the said estimates in a high dimensional setting are taken into consideration.