rengine 0.1.5

About this package

Collaborative Filtering is a method that is often used as a recommendation engine. Many industries have used this algorithm to recommend their products, this library can use cosine similarity or even centered cosine similarity.

Depedencies

• Python >= 3
• numpy

How to use

• cos_similarity(var1, var2, centered)

Parameters

var1 : iterable, array, np.array

The single arrray value you want to compare

var2 : iterable, array, np.array

The multi arrray values, you can convert from dataframe or table on your database

centered : bool

By default it will give true, if true you will use centered cosine similarity

Example case

• We'll try to figuring out this single matrix value, on the multi array

[4, 0, 0, 5, 1, 0, 0]

• The example dataset that we had

[[5, 5, 4, 0, 0, 0, 0],
[0, 0, 0, 2, 4, 5, 0],
[0, 3, 0, 0, 0, 0, 3]]

Example of code

from rengine.method import CollaborativeFiltering
import numpy as np
clf =  CollaborativeFiltering()
print(clf.cos_similarity(np.array([5,  4,  0,  0]),  [[1,  0,  3,  2],  [2,  1,  1,  0],  [4,  5,  0,  1]]))

Example of output

[-0.5, 0.866, -0.24] #row 2 in multi array data has more similarity

Another of output

[0.092, -0.559, nan] #nan means there's a number divided by zero