Methodology for assessing the impact of data quality on the effectiveness of machine learning models in the task of estimation of the implementation of project checkpoints

Abstract

The paper examines the process of estimation of the implementation of checkpoints of national and federal projects based on machine learning technologies. Federal information systems contain large volumes of data, including data on the progress of national projects, which allows them to be used for machine learning using various models. One of the main tasks is to control the progress of projects and track it at specified checkpoints. Despite organizational and technical measures, there are occasional delays in completing checkpoints. The data on the reasons for delays available in the system made it possible to form a feature space and assess the degree of their influence on the result. However, the classical application of machine learning approaches to solve the classification problem does not allow obtaining a result suitable for practical ap-plication. This is explained by the presence of ambiguities in the original data. Typically, machine learning methods are demonstrated and improved on typical data sets, but in real systems, the quality of the data used for machine learning must be given significant consideration. The study is devoted to the development of a technique for assessing the impact of data quality on the effectiveness of machine learning models in the task of estimation of the implementation of national projects’ checkpoints. The study was conducted on real, anonymized, standardized, coded data from the federal monitoring system. The obtained results demonstrate the effectiveness of the de-veloped technique.