Ranking Algorithm - Banking Use Case

[Joker_Hazard]

Hello all friends.

I currently work at the banking segment. Today, I'm trying to create an algorithm that points out a rank system of majorly 500 clients based of a few premises:

1. Financial Volume Transactioned (payments) through 5 years. The more cashflow it has, the better the rank.

2. Quantity of single payments through 5 years. The more payments the client does, the better the rank.

3. Total financial funds. Check whether the client is AAA or BBB for example. The more money the client stores with us, the better.

Which tool would you guys recommend? How can I create such a thing with Alteryx?

Thanks in advance!

[IraWatt]

Hey @Joker_Hazard,

I imagine without looking at the data you would first use a Summarize Tool grouping on client to calculate each of those metrics eg. the total cashflow, number of single payments and total funds. Then you would use a Join tool to combine the results for each client. Finally you would use a Formula Tool to calculate each rows (clients) rank. 

You may also need to use a filter tool to restrict the data to the past 5 years when summarizing for the first ad second point.

[Joker_Hazard]

Hi @IraWatt  Thanks for the response!!

Yeah, tha'ts majorly done already. I'm searching for a prescriptive/KNN method.

I honestly don't know if I should standarize all metrics to a scale from 0 to 1 or should I use the whole dataset as input for all neural/predictive tools that alteryx provides.

Have you made use of them someday?

Thanks

[IraWatt]

@Joker_Hazard Cool use case for KNN! I believe the rule of thumb is yes standardise the data Why should we Standardize Data in a KNN model? – Vipanchi (wordpress.com). The tool by default comes with an option to do this automatically:

Looks like there's a good Alteryx blog on the topic here: Tool Mastery | K-Centroids Diagnostics - Alteryx Community

[Joker_Hazard]

Ira, thanks!!! Will take a look to create something alike.

One last question, should I use the customer ID as a variable as well? 

There is Customer ID which is my key field and three predictor fiields. With that in mind, when using KNN the key field should be accounted and flaged on those tools as well?

Thanks

[IraWatt]

@Joker_Hazard no definitely don't select the ID field as it doesn't make sense to say two clients are similar just because their ID's are close numerically. Just leave the box unticked in the configurations :