DataOps Lifecycle with a Case Study in Healthcare

Abstract

The DataOps methodology has become a solution to many of the difficulties faced by data science and analytics projects. This research introduces a novel DataOps lifecycle along with a detailed description of each phase. The proposed cycle enhances the implementation of data science and analytics projects for achieving business value. As a proof of concept, the new cycle phases are applied in a healthcare case study using the UCI Heart Disease dataset. Two goals are achieved. First, a dataset reduction by features analytic in which the four most effective features are selected. Second, different machine learning algorithms are applied to the dataset. The recorded results show that using the four most effective features is comparable with using the full features (thirteen features), and both approaches show high accuracy and sensitivity. The average accuracy of the highest four features is 82.32%, and the thirteen features is 84.28%. That means that the selected four features affect the applications with 97.67% accuracy. Besides, the average sensitivity of the highest four features is 87.94%, while the thirteen features are 87.12%. The study shows an interesting and significant result that data modeling needn't be done for all data science projects which reduced the dataset.