In the last years, the banks realized that several opportunities can be exploited out of their data and they started to explore in depth the field of data science. This thesis analyzes in details the bank transactional data: from the single bank transactions, we reconstructed the payment relationships in a country and basically, by considering the whole network, we built the graph that represents the value chain. In particular, we focused on Romania and we extracted insights out of the graph addressing several business cases. Firstly, we identified business opportunities by analyzing the chain of suppliers and customers of the companies. Since these suggestions were to be implemented by the colleagues in Commercial Planning, we had to make them explainable and easy to visualize. For this purpose, we used Gephi, a graph dedicated software, which could be integrated into UniCredit services through its API. Another usage was to study the intrinsic risk of a company for the bank, given the structure of its value chain: several methods were implemented among which a random forest that performed best. Then, we extended the analysis by observing the risk for the whole network, given the default of a company: which would be the impact on direct suppliers and customers and cascading on the remaining companies? Eventually, we observed the trend of the business relationships to understand how they evolve in time and to be able to predict possible liquidity shortages for our clients. A simple interface was created to help the end users to visualize the discovered information. Summing up, the payments network was extremely useful to obtain additional information about the health of a country and of the companies operating there and to discover interesting commercial opportunities for the bank. As soon as the new PSD2 (Payments Service Directive 2) normative will become active, this analysis could be furtherly enhanced, reaching a complete knowledge of the network.
Negli ultimi anni le banche si sono rese conto delle molteplici opportunità che i loro dati riservano e hanno iniziato a esplorare più in profondità il campo della scienza dei dati e della loro analisi. Questo lavoro va ad analizzare in dettaglio i dati relativi alle transazioni bancarie: dai singoli bonifici si ricostruiscono le relazioni di pagamento in una nazione e di fatto, considerando tutto il network, si va a costruire un grafo che rappresenta la catena del valore. In particolare ci siamo focalizzati sulla Romania e dal grafo abbiamo estratto informazioni rivolte a più business cases. In primis, abbiamo identificato opportunità commerciali analizzando la catena di fornitori e clienti delle aziende. Dato che poi questi suggerimenti andavano rivolti ai colleghi della Pianificazione Commerciale, abbiamo dovuto rendere facilmente spiegabili e visualizzabili i metodi utilizzati. A questo proposito, abbiamo utilizzato principalmente Gephi, un software dedicato ai grafi, che, tramite le sue API, é direttamente integrabile in altri servizi di UniCredit. Un altro utilizzo é stato studiare il rischio insito in ogni compagnia cliente per la banca, data la struttura della catena di valore: diversi metodi sono stati implementati fra cui una foresta casuale che ha ottenuto i migliori risultati. Abbiamo poi esteso l’analisi osservando il rischio per il network nella sua interezza in caso di fallimento di una delle compagnie presenti: quale sarebbe l’impatto sui diretti clienti/fornitori ma anche a cascata sulle restanti aziende. Infine abbiamo osservato il trend degli ammontare delle relazioni commerciali, per capire come evolvano nel tempo ed essere in grado di prevedere possibili ammanchi di liquidità per i nostri clienti. Una semplice interfaccia é stata creata per aiutare gli utilizzatori finali a visualizzare le informazioni così scoperte. Riassumendo, il network dei pagamenti è stato utilissimo per ottenere informazioni aggiuntive sulla salute del sistema-paese e delle singole aziende e anche per evidenziare interessanti opportunità commerciali per la banca. Non appena la nuova normativa PSD2 entrerà in vigore, questa analisi potrà essere potenziata ulteriormente, raggiungendo una conoscenza del network totale.
From payment transactions to a structured network : key companies identification, risk assessment and graph mining
COMORETTO, LUCA
2018/2019
Abstract
In the last years, the banks realized that several opportunities can be exploited out of their data and they started to explore in depth the field of data science. This thesis analyzes in details the bank transactional data: from the single bank transactions, we reconstructed the payment relationships in a country and basically, by considering the whole network, we built the graph that represents the value chain. In particular, we focused on Romania and we extracted insights out of the graph addressing several business cases. Firstly, we identified business opportunities by analyzing the chain of suppliers and customers of the companies. Since these suggestions were to be implemented by the colleagues in Commercial Planning, we had to make them explainable and easy to visualize. For this purpose, we used Gephi, a graph dedicated software, which could be integrated into UniCredit services through its API. Another usage was to study the intrinsic risk of a company for the bank, given the structure of its value chain: several methods were implemented among which a random forest that performed best. Then, we extended the analysis by observing the risk for the whole network, given the default of a company: which would be the impact on direct suppliers and customers and cascading on the remaining companies? Eventually, we observed the trend of the business relationships to understand how they evolve in time and to be able to predict possible liquidity shortages for our clients. A simple interface was created to help the end users to visualize the discovered information. Summing up, the payments network was extremely useful to obtain additional information about the health of a country and of the companies operating there and to discover interesting commercial opportunities for the bank. As soon as the new PSD2 (Payments Service Directive 2) normative will become active, this analysis could be furtherly enhanced, reaching a complete knowledge of the network.File | Dimensione | Formato | |
---|---|---|---|
2019_12_Comoretto.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
1.62 MB
Formato
Adobe PDF
|
1.62 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/152204