Linear/Logistic...Decision trees?...Random Forests?

Question

1) Im getting anxious because i dont feel confident about Decision trees and Random Forests as much as i do for logistic n linear algorithms. Also, aren't all the stated above algorithms just 'estimators' (if you know) ? And i do know that random forest is just multiple decision trees but still i just quiet cant get it in my head.
2) Sklearn has so many classification n regression algos that i dont understand what to use and if decision trees r supposed to be classification how can they have regression algos? im going to guess to make myself feel better that its some kinda hack made possible due to python syntax, plz do correct me.
3) what is model_selection class exactly? In sololearn's course we start from train_test_split then go to kfold (which is basically for loop for the previous model_selector) then to GridSearch. Why are all these called model_selector anyway, they just split data into train/test and loop it. Isnt it better to call them data_splitter or something?
Plz help.

Answer 1

If you want to learn machine learning ,go to Coursera.
Course is free.
Search machine learning (Andrew ng)