franciscorafart / python-ml-course

Udemy machine learning with python course

Python-Data-Science-and-Machine-Learning-Bootcamp

Repo for Python Data Science and Machine Learning Bootcamp

Personal notes

Numpy arrays (Vectores and Matrices)

np.arange(start, finish, step) Example: np.arange(0,10,2) => array([0,2,4,6,8,9])

np.linspace(start, finish, n_of_steps) Example: np.linspace(0, 5, 9) => array([0, 0.625, 1.25 , 1.875,2.5 , 3.125, 3.75 , 4.375, 5])

np.zeros(rows, columns) => matrix of zeros np.ones(rows, columns) => matrix of ones

np.eye(rows, columns) => Identity matrix

• Random numbers np.random.rand(rows) => One-dimensional array of 5 random numbers between 0 and 1 with uniform distribution np.random.randn(rows, columns (optional)) => One or two dimensional arrau of 5 random numbers with standard normal (gaussian) distribution np.random.randint(low, high, amount) => Returns ramndom int between low and high(non-inclusive) range. Thirud optional argument to return array of random ints in that range.

arr.reshape(rows,columns) => Returns data of an array arranged with the specific number of rows and columns. Has to match the amount of elements. For example: arr = np.arrange(0,16) // [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15] arr.reshape(4,4) // [[0,1,2,3], [4,5,6,7], [8,9,10,11], [12,13,14,15]]

• Max and min values arr.max() / arr.min() => Max / Min value of array arr.argmax() / arr.argmin() => Index of max / min value

• Array slices Numpy arrays are different than python arrays because they're able to broadcast, or setting values with index ranges. In numpy arrays, when you slice an array and set it to a variable, you're referencing the same array in memory, not a new one. For example: arr = np.arange(0,100) arr2 = arr[:10]

To copy, you need to specify: arr2 = arr[:10].copy()

• Indexing of matrices (2d arrays) We can do double bracker format: arr[1][2] or single bracket format: arr[1, 2]

We can do matrix slicing, for example: arr2d = arr[:2, :1] // This would grab until row 2, and columns until 1 (not included)

• Fancy indexing: We can select whole rows in any order arr2d = [[9, 4, 6, 8]] // This will select rows 9, 4, 6 and 8 and return them as a matrix in that order.

• Applying operations to arrays arr = array([3,4,5,6,7]) bool_arr = arr > 5 // [False, False, False, True, True]

• Conditional selection arr[bool_arr] // [6,7]

Or more concise: arr[arr > 5] // arr > 5 creates a boolean array map that can be used to index the original array, like a filter.

• numpy operations

Array addition, multiplication, division, substraction (not concat) arr = [1, 2, 3] arr + arr // [2, 4, 6]

Scalar operations arr + 1 // [2, 3, 4]

Universal array functions -> Math operations that broadcast through all elements of an array. For example: np.sqrt(arr) / makes the square root of all elements of array np.sin(arr) / np.log(arr)

PANDAS

• Series

Pandas series are like hashes / dictionaries. We can pass any data structure into them (list, dictionary, numpy array, function) and it will store them in order, with number index (key) by default.

my_data = [10,20,30] pd.Series(data=my_data)

0 10 1 20 2 30 dtype: int64

We can also specify on the second argument index=the keys by which we will reference the data

labels = ['a', 'b', 'c'] pd.Series(data=my_data, index=labels)

a 10 b 20 c 30 dtype: int64

Data Frames

• Data Frame notes in the corresponding Francisco Jupyter notebooks

• MultiIndex notes in the corresponding Francisco Jupyter notebooks

• Missing data If data frame has missing NaN data, we can eliminate those rows or fill in the data with a) df.dropna() # Drops rows with NaN data df.dropna(thresh=3) # Drops rows with 3 or more NaN values df.dropna(axis=1) # Drops columns with NaN values b) df.fillna(value=3) # fills NaN data with value of 3 df['A'].fillna(value=df['A'].mean()) # fills NaN values in column A with the mean of all available values in column A

• Group By method in the corresponding Francisco Jupyter notebook