Aggregated Analysis of Procurement Linked Data

Analysis Scenarios

The data on public contracts, in combination with external data retrieved from the linked data cloud, can be submitted to aggregated analysis. The beneficiaries of such analysis can be:

• Journalists and NGOs: the data may help them reveal corruption and clientelism in public sector.

• Official government bodies: both specific supervisory bodies that address the issues of transparency and fair competition and statistical offices that collect data as part of aggregated information on the national economy.

• Bidders: analysing the previous successful and unsuccessful tenders may be helpful when preparing a new one; in long term, the companies may also actively plan their bidding strategies based on procurement market trends (revealed by automated analysis).

• Contracting authorities: they want to understand the supply side in order to know how to formulate the contract conditions, in view of successful matchmaking. Good progress of a future contract may be derived from previous experience with certain bidders. An additional goal may be to attract an adequate number of bidders; excessively many bidders bring large overheads to the awarding process, while too low a number may reduce competition (and, under some circumstances, even lead to contract canceling by a supervisory body, due to an anti-monopoly action).

Analytical Methods

A straightforward approach to aggregated analysis is via summary tables and charts expressing the relationship between, e.g., the number of contracting authorities, contractors, contracts, tenders, lots, or geographical localities. The value of contracts can be calculated as a sum or average per authority, contractor, region, kind of delivery, classification of goods etc. Charts can be generated for presentation of these statistics split by various dimensions (e.g. bar charts) or showing the evolution (e.g. line charts, timeline). The geographical dimension is best presented on maps: detailed data can be shown as points on the map, e.g., pointers with shaded tooltips on OpenStreetMap. The data for such analysis are normally provided by SPARQL SELECT queries, which allow to both retrieve the data and perform basic aggregation operations.

More sophisticated analysis can be provided by data mining tools, which automatically interrelate multiple views on data, often based on contingency table. As an example, see a fragment of analysis of U.S. procurement data with respect to the impact various attributes of a contract notice may have on the subsequent number of tenders (Fig. 6).

The association rules listed in the table fragment regard both a CPV code of the contract object (mainObject attribute), originating from one of the core procurement dataset, and the population density attribute, originating from DBpedia. It indicates that contracts for 'Research and Development in the Physical, Engineering, and Life Sciences' in localities with higher population density tend to attract a high number of tenders (as higher interval values for the former mostly coincide with higher values for the latter, in the individual rules).

Fig. 6. Discovered factors correlated with number of tenders

Integration of Analytical Functionality into PCFA

Although the central role in the PCFA scenarios is reserved to matchmaking, there are also reserved slots for invocation of analytical features. Since this part of implementation has been ongoing till the final months of the project, it is not yet functional at the time of completing this chapter of the book. The analytical functionality will be at the disposal of the buyer (contracting authority), and will amount to:

• Interactively exploring, in graphical form, the linked data about

– the current notice

– a (matching) historical notice/contract

– a relevant supplier, including its contracts.

Viewing suggested values for the remaining pieces of contract notice information based on the already provided ones. The values will be provided by an inductively trained recommender.

Getting an estimate of the number of bidders for (as complete as possible) contract notice information. For this, a predictive ordinal classifier will be developed.

When the integration of analytical functionality has been completed, usability testing by several contract authorities' representatives will take place.

< Prev   CONTENTS   Next >