QnA - 1

 GBM 
NaiveBayes

- difference between logit and probit models?

The main difference between the logit and probit models is the assumption of the distribution of the errors in a model.

The logit model follows the cumulative distribution function of the logistic distribution, while the probit model uses the cumulative distribution function of the standard normal distribution.

- KNN

- Regression : regression is way to find coefficients to fit into the equation, or how many apples to get to fit into a basket

- Random Forest vs Gradient Boosting - asked in HDW

Random forest is a supervised learning algorithm that utilizes multiple decision trees. Decision trees are made of decision nodes which further lead to either another decision node or a final value (leaf node).

Gradient boosting tree uses boosting algorithm. Boosting algorithm trains a sequence of weaker models to increase the accuracy of the overall model by learning from the previous model.

Random Forest is a bagging based algorithm (merging independent trees that have been created using random subsets of training data), while Gradient boosting is a boosting based algorithm (combining weak models iteratively to produce one strong model).

As seen, both random forest and gradient boosted trees are trying to reduce error but do it in a slightly different way. There is also a difference in how both models reduce error with respect to bias and variance.

Gradient boosting is used to reduce the bias as the size of the decision trees grows larger. Random forest is used to reduce the variance as the number of decision trees increases.