Key Assumptions :

  • Binary Targets : (e.g., Pass/Fail).
  • Independent Observations : Observations are independent of one another.
  • Low Multicollinearity : The features of the input data should not be highly correlated.
  • Linearity : Linear relationship between features and the log-odds.
  • Sufficient Sample Size : Maximum Likelihood demands a large sample for accurate estimation of the parameters.
    • Rule of thumb : At least 10 cases with the least frequent class for each feature.