Random Forest explained

Random Forest explained

A random forest merges a collection of independent decision trees to get a more accurate and stable prediction

Random forest is a type of ensemble method, ensemble method combine several machine learning models in order to decrease both bias and variance.

Multiple models are fit and each key in  a different aspect of the data, and combining all these models ultimately generates better predictions together than any of the single models on their own.

In Training Data , this algorithm will take N sample from the training data with replace so one record can be present in multiple samples , then next step is to take sample of features to be used for each of the data samples.

So we have N subset of our overall data and those data contains both subset of rows and subset of column(features).

Then for each feature and data subset the algorithm will built a decision tree to try generate the most accurate results, these decision trees are developed independently. So that each decision tree  key in different relationship in the data , this will ultimately result in best prediction.

No alt text provided for this image




To test the model, Now we will pass example A from the test data to all n trees built in training data, based on the features the model will give a prediction 'Survive'. Now the splits and features on the 2nd decision tree will be totally different, so it will be possible to get a different Prediction 'Did Not Survive' . So now you will have n predictions for example A , one from each tree , then the random forest algorithm will basically just aggregate all the predictions together and then  based on voting, with majority rules will make a final predication - So in this case it will say that the person has 'Survived'

* Illustration from Titanic data and prediction is  for each person whether he/she will survive or not survive


No alt text provided for this image

In the next article will write how to tune the Random forest model and interpret the results, Please like , share and comment if any questions.







要查看或添加评论,请登录

Mohammad Arshad的更多文章

社区洞察

其他会员也浏览了