ESLModels

Algorithm from The Elements of Statistical Learning book implement by Python 3 code.

Until now, I finish chapter 3, 4, 7. I am working on chapter 11.

The Algorithm model is placed in esl_model.chx.models, x means the number of chapter, for example, esl_model.ch3.models

To run the code, you must install Python >= 3.5, because I use @ operate instead of numpy.dot. See pep-0465

from esl_model.ch3.models import LeastSquareModel

# import prostate data set
from esl_model.datasets import ProstateDataSet

data = ProstateDataSet()

lsm = LeastSquareModel(train_x=data.train_x, train_y=data.train_y)
lsm.pre_processing()
lsm.train()

# after pre_processing and train, you can get the beta_hat
print(lsm.beta_hat)

# predict
y_hat = lsm.predict(data.test_x)

# get the test result
test_result = lsm.test(data.test_x, data.test_y)

# get the mean of square error
print(test_result.mse)

# get standard error
print(test_result.std_error)

You can find the source in esl_model.ch3.models

I try to make the code clean and simple so that people can understand the algorithm easily.

class LeastSquareModel(LinearModel):
    def _pre_processing_x(self, X):
        X = self.standardize(X)
        X = np.insert(X, 0, 1, axis=1)
        return X

    def train(self):
        x = self.train_x
        y = self.train_y
        self.beta_hat = self.math.inv(x.T @ x) @ x.T @ y

    def predict(self, X):
        X = self._pre_processing_x(X)
        return X @ self.beta_hat

How to

I also write some article describe how to write some algorithm.

How to write Reduced Rank LDA: http://littlezz.github.io/how-to-write-reduced-rank-linear-discriminant-analysis-with-python.html
How to use travis with numpy and pytest: http://littlezz.github.io/travis-ci-with-numpy-and-pytest.html

Install

pip(3) install git+https://github.com/littlezz/ESL-Model

littlezz/ESL-Model

ESLModels

How to

Install

Reference