The dataset is a full ranked binary dataset. Therefore, Y = 1 or 0and X's = 1 or 0. Y represents whether a car has an engine failure or not. X represents the features of the car, for example X1 is engine type A, X2 engine type B, X3 gear type etc. I want to know from your experience, which models work well on those type of datasets. I read somewhere, that LDA or KNNare not recommended, as well as SVM.

Which information criterion would you use for forward selection? AIC, BIC..?

The data looks like this

       Engine_A Engine_B Engine_C Color_R Color_B Color_G  Y_Failure
Car1     1         0        0        0       1       0         1
Car2     0         1        0        1       0       0         0 
Car3     0         1        0        0       0       1         1

Would you factorize all variables X and Y?

  • 2
    $\begingroup$ This is not answerable without a lot more detail about your specific problem. $\endgroup$ – mkt - Reinstate Monica Aug 7 '19 at 10:34
  • 1
    $\begingroup$ What is the goal of your analysis? Sample size? How many predictors? ... meanwhile, if you are interested in modeling the probability of $Y=1$, conditional on $x$, look into logistic regression ... $\endgroup$ – kjetil b halvorsen Aug 7 '19 at 11:12
  • $\begingroup$ The goal is to predict Y as good as possible. I already used logistic regression. Let me ask the other way around: which models are not suitable for that particular problem? $\endgroup$ – Textime Aug 7 '19 at 11:59

Browse other questions tagged or ask your own question.