danielacargnin
IS-IT--Management
Hi, my name is Daniela Cargnin. I am an M.Sc.Candidate in Computer Science at the "Universidade Federal do Pernambuco - UFPE" here in Brazil.
Im needing some help.
Im working on my dissertation and its related to the process of Data Mining.
Im going to talk about a methodology for the creation of specifcs "data marts" to Data Mining problems. It is known that depending on the DM problem (classification, clustering, ...) we have to do some kind of transformation on the data. One example, is the creation of behavior variables, specially used on behaviour scoring. For the creation of a model to behaviour scoring, we have information about the clients and their shoppings (their contracts and theirs respective parcels). The granularity of the base that we want to aply to the model is "contract". So, we have to aggregate all informations about the client (relative to their contract's parcels) in new variables to improve the quality of the data base (and create the behaviour of the client).
One exemple of behaviour variable is "Quantity of previous contracts", "Quantity of parcels payed with overdue", and so on...
Depending on your DM problem, you have to do other kind of data aggregation and transformation to improve de quality of the model.
My question is if there is any methodology related to this. Im searching for this but I cant find, because I dont know if there is a specifc name to this kind of methodology in english.
Does anybody knows where I can find some material related to this or if is there any methodology about it?
I really appreciate any help.
Best reguards...
Daniela Cargnin
Im needing some help.
Im working on my dissertation and its related to the process of Data Mining.
Im going to talk about a methodology for the creation of specifcs "data marts" to Data Mining problems. It is known that depending on the DM problem (classification, clustering, ...) we have to do some kind of transformation on the data. One example, is the creation of behavior variables, specially used on behaviour scoring. For the creation of a model to behaviour scoring, we have information about the clients and their shoppings (their contracts and theirs respective parcels). The granularity of the base that we want to aply to the model is "contract". So, we have to aggregate all informations about the client (relative to their contract's parcels) in new variables to improve the quality of the data base (and create the behaviour of the client).
One exemple of behaviour variable is "Quantity of previous contracts", "Quantity of parcels payed with overdue", and so on...
Depending on your DM problem, you have to do other kind of data aggregation and transformation to improve de quality of the model.
My question is if there is any methodology related to this. Im searching for this but I cant find, because I dont know if there is a specifc name to this kind of methodology in english.
Does anybody knows where I can find some material related to this or if is there any methodology about it?
I really appreciate any help.
Best reguards...
Daniela Cargnin