Ensemble Learning - Bagging and Bossting is a free tutorial by Gopal Prasad Malakar from R course
Link to this course(Special Discount):
[ Ссылка ]
This is the best R Course
Course summary:
Get Crystal clear understanding of decision tree
Understand the business scenarios where decision tree is applicable
Become comfortable to develop decision tree using R statistical package
Understand the algorithm behind decision tree i.e. how does decision tree software work
Understand the practical way of validation, auto validation and implementation of decision tree
English
Let me introduce you to ensemble learning also explain you what is bagging and what is boosting. Ensemble learning is the term which is used a lot in the context of machine learning, machine learning usage various algorithms like linear regression, logistic regression, decision tree, K Means, random forest and ensemble learning actually is nothing but application of group learning in these contexts. Ensemble word means group so essentially here multiple models are built and these multiple models come together to bring forth a model that is much more accurate. So, in one-line ensemble models means multiple models coming together to give you the final model. Now let me explain you two popular terms which is used in this context. One it's called bagging. So, begging is all about making models in parallel. So here you are making models in parallel on various samples, these samples may have the same variables may have in fact different variables also and these models then finally vote to give the final model and that becomes the prediction. So, if you look at very much the way you would do in the random forest each of the model is predicting Y...N that is the class of dependent variable and whichever has majority that becomes the final class, here 4 are saying Y and 1 is saying N and that's why this becomes the final prediction. So here multiple models were built in parallel and on various datasets and maybe on various variables and they predict to give you the final model. Now look at boosting, in case of boosting models are built in series. So, if you look at in case of bagging, models were built in parallel, in case of boosting models are built in series and in each consecutive model the weights are adjusted based on the learning of previous models. So, let's take a look what you do you divide the data into five parts let's use 1 3 4 and 5 parts to make the model validate on dataset 2, whatever model comes the new model now you take that model, again those variables you are taking on 1 2 portion 4 and 5 and you will validate on portion third. If it validates it's good enough and that's how you'll keep doing. So, in many ways K-fold validation is nothing but a boosting technique.
Ensemble Learning - Bagging and Bossting Analytics/ Supervised Machine Learning/ Data Science (applied statistics): CHAID / CART / GINI/ ID3/ Random Forest etc.
