scikit-learn-1.md

Scikit-Learn Library

(back to Index)

Scikit-Learn is the go-to library for machine learning in Python.

Scikit Intro

• An Introduction to Scikit Learn
• Introduction to Scikit-Learn from Python Data Science Handbook
• Scikit-Learn online documentation is very extensive and thorough

Scikit Algorithms Overview and Cheatsheet

• Scikit cheat sheet
• Another good cheat sheet from datacamp

Setting up Scikit

• Setting up scikit-learn

Explore Built in Datasets

• Dataset documentation
• How to use Scikit-Learn Datasets for Machine Learning

Creating synthetic data

• How to Generate Test Datasets in Python with scikit-learn

Creating train/test splits

• Train-Test Split for Evaluating Machine Learning Algorithms ](https://www.bitdegree.org/learn/train-test-split)

More Reading

• A good ebook on scikit learn

Data Leakage

• Data normalization before or after train-test split?
• Data Leakage in Machine Learning

---

Checklist

You should know the following

• Have Scikit development environment setup
• Explore built-in dataset
• Generate synthetic dataset

---

Exercises

A - Basics

A1 - Make sure you have scikit installed

import sklearn
sklearn.__version__
# you should see version output

B - Builtin Datasets

B1 - Print out all datasets that are included in scikit

B2 - Regression dataset: Load boston dataset

• Find out all the columns
• describe the dataframe
• print out sample data
• Use describe to understand the data

B3 - Classification dataset: Load IRIS dataset

• Find out all the columns
• describe the dataframe
• print out sample data
• Use describe to understand the data

C - Generate Data

C1 - make classification dataset

• Use make_blobs function to make classification dataset
• visualize the dataset

C2 - Use make_regression to make a regression dataset

• Use make_regression to make a regression dataset
• visualize the dataset

D - train/test split

D1 - Simple train/test split

Create data like this

import pandas as pd
from sklearn.model_selection import train_test_split

tip_data = pd.DataFrame({'bill' : [50.00, 30.00, 60.00, 40.00, 65.00, 20.00, 10.00, 15.00, 25.00, 35.00],
                        'tip' : [12.00, 7.00, 13.00, 8.00, 15.00, 5.00, 2.00, 2.00, 3.00, 4.00]})

x = tip_data[['bill']]
y = tip_data[['tip']]

# Use train_test_split function (20% test split)
x_train,x_test,y_train, y_test = train_test_split (x,y,test_size=0.2)

• Print out x_train and x_test
• Visually inspect the data, are there any common elements between train and test data?

D2 - test/train split on housing data

Read the house-sales-simplified.csv.

import pandas as pd

house_sales = pd.read_csv(...)

X = extract all columns except `SalePrice`
y = extract `SalePrice` column

Now split the X,y data into training and testing.

Print out the length of train and test datasets.

Since the data is too large to visually inspect, how can we programmatically ensure there are no common elements between train and test

More exercises

• Scikit exercises

---

Index