Skip to content

Latest commit

 

History

History
114 lines (69 loc) · 3.22 KB

course-notes-02-python.md

File metadata and controls

114 lines (69 loc) · 3.22 KB
Raw

Introduction to Python for Data Science - Course Notes

January 2019 Chris Cameron

I am auditing Microsoft Professional Program for Artificial Intelligence track.

The second of 10 courses is Introduction to Python for Data Science.

Since I know a lot of Python already I will probably skim the first 3-4 chapters and focus on Matplotlib and Control Flow/Pandas since I haven't used matplotlib in years and I've never used Pandas.

Start Here

nothing noteworthy

Python Basics

nothing noteworthy

Python Lists

nothing noteworthy

Functions and Packages

nothing noteworthy

Numpy

Numpy Arrays

the justification for using numpy is being able to type (arr1 / arr2**2) where arr1 and arr2 are numpy arrays. You could do this with lists too if you wanted to but I understand that the performance and statistical usage of numpy arrays is important.

he does arr1 > 23 to get an array of booleans

then he does arr1[arr1 > 23] to get the specific values that were greater than 23.

neat

2D Numpy Arrays

side note: it's pretty obvious (and a bit hilarious) these transcriptions were done by AI.

you can retrieve elements arr2d[n][m] or arr2d[n,m]. Works with slices too like arr2d[:,1:3]

Numpy Basic Statistics

  • numpy.mean,
  • numpy.median
  • numpy.corrcoef (correlation coefficient, I believe)
  • numpy.std standard deviation
  • numpy.sum
  • numpy.sort
  • numpy.round
  • numpy.random.normal random number
  • numpy.column_stack (creates a columnar 2d numpy array)

Matplotlib

Basic Plots with Matplotlib

  • Shows Hans Rosling's GDP vs Life Expectancy bubble chart.
  • matplotlib.pyplot as plt subpackage
  • plt.plot(...)
  • plt.show() to display it
  • plt.scatter(...)

Histograms

  • plt.hist()

Customization

  • label your axes plt.xlabel('Year'), plt.ylabel('Population')
  • plt.title("blah")
  • plt.yticks()
  • plt.fill_between()

Further Readings

They bring up ggplot as an alternative plotting library.

They mention a gallery of nice plots that matplotlib supports

Control flow and Pandas

Boolean Logic and Control Flow

introduces or and if else elif

Pandas

store data in a DataFrame

typically you import your DataFrame as opposed to declaring it manually (I wonder if this is true generally - this course has been pretty basic)

  • import pandas as pd
  • pd.read_csv()
  • pd.read_csv('filepath', index_col=0) (this makes sure you don't use an index column as a data column by accident)
  • df['column_header'] works with data frames
  • df['column_header'] = [some list]
  • it's based on numpy so you can do like df['density'] = df['population'] / df['area'] * 1000000 to create new columns
  • to access rows you have to use df.loc['rowname'] to find that row
  • you can combine it either way like df.loc['row']['col'] or df['col'].loc('row')

Further Reading

http://pandas.pydata.org/

Final Lab

Unfortunately all labs and exercises in this course were locked by the fee. Fortunately this course is so basic it's not worth the cost to unlock this stuff.