Korruptsiooni tuvastamine riigihangetes läbi andmeanalüüsi
Files
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Magistritöö eesmärgiks oli uurida, kas ainult andmeanalüüsile tuginedes on võimalik ennustada korruptsiooni võimalikkust Eesti riigihangetes ning tulenevalt eelnevast teha riigile soovitusi, kuidas parandada korruptsiooni tuvastamise võimalusi. Seatud eesmärgi saavutamiseks andis autor muuhulgas ülevaate korruptsioonist ja korruptsioonist riigihangetest ning nende vastu võitlemise strateegiast ja olulisematest viimastel aastatel toimunud arengutest maailmas. Kõrgele korruptsiooniriskile on viidanud nii üleeuroopalised uuringud kui ka meedia. Olemasolevatele andmetele tuginedes tegi autor masinõppe algoritmi, mis hindab korruptsiooni võimalikust riigihangetes Eestis.Rakenduse automatiseeritud lähenemise ja andmete analüüsi tulemusena jõudis autor tulemusteni, mis näitavad, et antud andmetele tuginedes on võimalik hinnata korruptsiooni tõenäosust Eesti riigihangetes. Eelneva põhjal saab seega öelda, et andmeanalüüsi kasutades on võimalik muuta korruptsiooni tuvastamine konkreetsemaks, lihtsamaks ja efektiivsemaks. Lähtudes teooriast ja tehtud praktilisest tööst, esitas autor enda poolsed soovitused riigile, milliste andmete kasutamisel ja analüüsil oleks võimalik korruptsiooniriski täpsemini ennustada ja seeläbi korruptsiooniriski maandada.
Corruption is present in all aspects of the society and it hinders the progress of various sectors of the economy. In this context, corruption is defined as the act of dishonesty for personal gain by those in power. One of the biggest sectors it influences is public procurement. Previous research has shown that corruption is present in public procurement and it reduces the transparency of the process. Taking into account the monetary value of the public procurement sector, it is clear that this is a problem that must be addressed. Various studies have used qualitative analysis to root out the core of the issue, but as it still thrives, it essential that more accurate and acute measures are used. In order to tackle this problem, there have also been studies that try to quantify the likelihood of it, rather than only looking at qualitative research and this is where data analytics comes into play – the core of this study. This thesis aims to determine whether using open data resources and data analytics it is possible to classify corruption in the public procurement processes and therefore suggest a suitable set of data to make the detection of corruption easier and quicker. Building on existing work on corruption, it asks: what data could be analysed in classifying corruption and what methods could be used? Based on a review of the literature on corruption and theories of machine learning, data analytics was used to assess possible corruption in public procurement in Estonia. In the data analytical process the author used machine learning approaches that predict the classification of procurement as corrupt or non-corrupt. The analysis of the results demonstrated that based on available data it is possible to predict corruption in public procurement in Estonia. Furthermore, the results also indicate that some features have a bigger impact on corruption in public procurement. Taking into account the background, related work and the current results, the author suggests that data analytics is vital in the fight against corruption and using machine learning can yield in good results in predicting corruption. Further research is needed to identify other factors that could strengthen the effectiveness of these approaches.
Corruption is present in all aspects of the society and it hinders the progress of various sectors of the economy. In this context, corruption is defined as the act of dishonesty for personal gain by those in power. One of the biggest sectors it influences is public procurement. Previous research has shown that corruption is present in public procurement and it reduces the transparency of the process. Taking into account the monetary value of the public procurement sector, it is clear that this is a problem that must be addressed. Various studies have used qualitative analysis to root out the core of the issue, but as it still thrives, it essential that more accurate and acute measures are used. In order to tackle this problem, there have also been studies that try to quantify the likelihood of it, rather than only looking at qualitative research and this is where data analytics comes into play – the core of this study. This thesis aims to determine whether using open data resources and data analytics it is possible to classify corruption in the public procurement processes and therefore suggest a suitable set of data to make the detection of corruption easier and quicker. Building on existing work on corruption, it asks: what data could be analysed in classifying corruption and what methods could be used? Based on a review of the literature on corruption and theories of machine learning, data analytics was used to assess possible corruption in public procurement in Estonia. In the data analytical process the author used machine learning approaches that predict the classification of procurement as corrupt or non-corrupt. The analysis of the results demonstrated that based on available data it is possible to predict corruption in public procurement in Estonia. Furthermore, the results also indicate that some features have a bigger impact on corruption in public procurement. Taking into account the background, related work and the current results, the author suggests that data analytics is vital in the fight against corruption and using machine learning can yield in good results in predicting corruption. Further research is needed to identify other factors that could strengthen the effectiveness of these approaches.