Hello everyone,
i am currently in a seminar regarding credit scoring and we are trying to predict the probability, that a company is going to bankrupt, by looking at ratios from their financial statement (recreating the Olson O-Score with classification)
We got a group of 20 Companies, that went bankrupt in the last few years and a peer group of companies that are still solvent as well as all the relevant numbers to calculate our financial ratios.
Our professor told us, that we can use the lda function to determine the weights in our equation, but so far i am struggling to get an output, that makes sense. (i also havent used r in a while, so my question might also be stupid)
We are basically trying to find out the best wheigtings, so that if you sum up all the variables multiplied by said weighting you get a high sum for the defaulted companies and a low sum for the still solvent ones.
Now when i use the lda function in RStudio this is the output i get:
I have a total of 13 rows, 12 of which are ratios for calculation and the first one being the status (default/solvent) for classification.
I am not exactily sure, if the lda function makes it clear, that the sums in the end should have the biggest possible difference, because when i take the coefficients back into excel, the resulting probablities make absolutely no sense, and i can get better results just by targeted guessing the coefficients.
Does someone know how i can optimize my outpout here to get the desired results?
Thanks alot for any help.