The original data consist of 116208 records each including 8 attributes: Account No, DATE, TRANSACTION DETAILS, CHQ.NO., VALUE DATE, WITHDRAWAL AMT, DEPOSIT AMT, BALANCE AMT with lots of Null values.
- Normalizing numeric attributes
- Filling “empty” attributes based on some convention (e.g. most common element)
- Erasing “unreasonable” spaces at the beginning and the end of attribute values;
- Making all the strings lower-case
- Deleting columns of high variability (let us say that column is of high variability if 70% of the rows have unique value).
- Deleting all columns that contains some date format and reformulate the column accordingly.
- Fixing date format for all the rows
- Splitting the columns containing both text and number
- Dropping duplicated transactions, if 70% of their attributes are equal
The Isolation Forest has been used for detection the anomaly in dataset.
BankDataAnomalyDetection- All the pre-processing and analysis steps have been accumulated in.bank.xlsx- Bank Data
- Using other approaches to detect the anomaly like
One Class SVMorNeural Network.