Assessment and Classification Models of Regional Investment Projects Implemented through Concession Agreements

Abstract

Imposed wide-ranging sanctions require stricter control over the use of budget funds in order to increase the return on investment and minimise the risks of inappropriate spending. Thus, regional development based on the implementation of investment projects with public participation through concession agreements becomes particularly important. The article considers the construction of classification models for the assessment of such projects to identify high-risk concession agreements. State customers can use these models to make informed decisions when choosing a contractor and to improve the efficiency of public property management. For an objective assessment of the integrity of contractors based on financial and other factors, the study used screening models and built-in tools of the SPARK information and analytical system, as well as the methods of descriptive analysis of big data, machine learning and the nearest neighbours approach for clustering regional investment projects according to the risk of improper execution of concession agreements. The presented approach was tested on 1248 regional investment projects implemented through concession agreements. As a result, the research identified two clusters: projects with low risk (83.8 %) and high risk (16.2 %) of improper performance of obligations by the concessionaire. To assess the models’ accuracy and sensitivity to outliers, the confusion matrix and Spearman’s coefficient were utilised, which showed a sufficiently high accuracy of the resulting classification. The constructed models can be used for selecting regional investment projects, as well as for monitoring implemented projects in order to identify potential risks of their non-completion and timely take necessary response measures.