Selecting and cleansing the dataset
We have selected the transaction history from a customer account as our dataset, which we will use to build the model for fraud detection. A large bank can have millions of customers. The historical transaction data of these customers can be used to build models that are unique for each customer, based upon his spending patterns.
We will start with the last 3 years of transaction history. Let's take a look at a single transaction to understand the information captured in it:
Datum;Naam / Omschrijving;Rekening;Tegenrekening;Code;AfBij;Bedrag (EUR);MutatieSoort;Mededelingen 20151022;ZIGGO SERVICES BV;NL54INGB07XXX32XXX;NL98INGB0000845745;IC;Af;52,5;Incasso;Europese Incasso, doorlopend IBAN: NL98INGB0000845745 BIC: INGBNL2A Naam: ZIGGO SERVICES BV ID begunstigde: NL30ZZZ333034790000 SEPA ID machtiging: 000788255400101112009000000941 Kenmerk: 030176416129000 Omschrijving: 270934155
The preceding listing contains information about a single...