[Go through 10] sklearn usage record

This is this year

Forget it nownet, :)

Chapter 1

The information entropy is more refined (there is a logarithm after the calculation. The calculation will be slower).The Gini index will be higher.Therefore, the decision tree generated by information entropy is relatively fine, and it often leads to overfitting on high-noise data.
The closer R2 is to 1, the better
The smaller the MSE, the better, scoring ="neg_mean_squared_error" is a negative MSE

Chapter Two

In python, if you don't write a loop, you can't write it.
However, you can only use traversal to view the attributes in the model. Because it is imported into pandas, it becomes a string, and it cannot be traversed one by one.
Compile and run the code first, then move the cursor to the method, and press shift+tab to view the parameter description and return value of the method.

Chapter Three

axis=0 1 If you don't understand, just back up the entire inplace=False first, change it right, and then cover it with True
Compared with pd.range, numpy.linspace is used when you don't know the step size(start, end, number)
Embedding method is slow

Chapter Four

Lazy object [*]
First reduce the dimension, and then increase the dimension, you can denoise.

---

Go directly to the case and practice it
I have gone through the Titanic kaggle case code in Chapter 1.Familiar but unfamiliar, like the code for those algorithmic problems.