Open OmerBelsky opened 5 years ago
When using the explain_prediction function on any observation that has a <0.5 probability of y=1 (as predicted by the model) the \<BIAS> turns negative and the sum of the weights sums up to that same probability... but negative. Help? The ol' notebook: https://github.com/ScifiDeath/super-duper-giggle/blob/master/Negative%20Random%20Forest%20Weight%20Sum.ipynb
When using the explain_prediction function on any observation that has a <0.5 probability of y=1 (as predicted by the model) the \<BIAS> turns negative and the sum of the weights sums up to that same probability... but negative. Help? The ol' notebook: https://github.com/ScifiDeath/super-duper-giggle/blob/master/Negative%20Random%20Forest%20Weight%20Sum.ipynb