Research of Data-driven Iterative Optimization Scheme for Machine Learning Models

Abstract

With the development of big data technology providing massive data information for machine learning, more researchers are focusing on data methods in big data technology, by constructing better data set to improve machine learning model's performance. This research based on the data-driven and human-machine collaboration methods in big data technology, continuously build data set to optimize machine learning models. We take the growth and risk assessment model of technology companies as an example, design and implement a data-driven machine learning model pipeline. The experiments show that the pipeline can effectively and continuously optimize these machine learning models, and ensure the stability of machine learning application.