prediction model - prematurely paid mortgages
TRANSCRIPT
Data miningCreating prediction model predicting prematurely paid mortgages
Dávid Kakaš
AssignmentAssignment
“Make production model for predicting prematurely paid mortgages,
so bank will be able to contact clients in advance and persuade them to
stay in the bank.”
Preparation: Financial point Preparation: Financial point of viewof view
I sat down with colleges from product, retail and segmentation department and we put together relevant:1. customer features2. deposit product’s features3. account’s movement statistics4. mortgage statistics5. mortgage refix information6. market & banks interest ratios
Preparation: DWH point of Preparation: DWH point of viewview
During first few weeks I browse bank’s DWH for needed attributesAfter I found every relevant attribute I got back to start and design new attributesNew attributes were computed from raw data in DWH and should help increase information value of data in analysis
Preparation: Technical point Preparation: Technical point of viewof view
I designed automated process of creating suitable train, test and score datasetI drew independent and target variables from 2 years periodI choose „one month“ difference between targets and independent variables so the model will be able to predict desired fact one month in advance
Programming the analysisProgramming the analysis SAS Enterprise Guide:
I programmed thousands of lines of code in SAS macro language, SAS procedures and SQL
SAS Enterprise Miner:
I designed data analysis using SEMMA methodology
OutcomesOutcomes
Automated program which is able to produce train, test and score data sets according to user preferences
Data model which is able to predict prematurely paid mortgages
... so that was one of many interesting projects and task I was working on.
Interested?Interested?
Please do not hesitate to contact me:
Email address: [email protected] Phone: +421 907 110 703