February 16th

Today we learned about the transition matrix. With it you can make predictions about how data will change over time. For example:

You are working for a small bank and need to make a prediction on how many people will be in each credit rating in one year's time. You were given the current number of people in each credit rating, and found a recent survey online that tracked 1000 people in each credit rating over the course of one year.

Current # of people in each credit rating

A	B	C	Default
132	534	638	430

Credit rating change over one year

final credit
Initial credit		A	B	C	Default
	A	900	60	30	10
	B	80	800	80	40
	C	20	100	700	180
	Default	0	0	0	1000

First you'd convert the survey data to a percentage (with each row's percentage being equal to 100%)

final credit
Initial credit		A	B	C	Default
	A	90%	6%	3%	1%
	B	8%	80%	8%	4%
	C	2%	10%	70%	18%
	Default	0%	0%	0%	100%

Then you'd use the survey data to create a transition matrix (with each row being equal to 1).

The current customer data is then put into a matrix.

Finally you'd multiply the two together to get your prediction (rounded to the nearest person in this case).

That's all there is to it!