This repository contains two K-Nearest Neighbors (KNN) examples written in Python.
A basic KNN implementation. An unknown movie genre will be calculated with the use of six sample data sets.
([Movie] : title='He's not reall into dudes'; #kisses=2; #kicks=100, genre='Romance', 18.867962264113206)
([Movie] : title='Beautiful Woman'; #kisses=1; #kicks=81, genre='Romance', 19.235384061671343)
([Movie] : title='California Man'; #kisses=3; #kicks=104, genre='Romance', 20.518284528683193)
([Movie] : title='Kevin Longblade'; #kisses=101; #kicks=10, genre='Action', 115.27792503337315)
([Movie] : title='Robo Slayer 3000'; #kisses=99; #kicks=5, genre='Action', 117.41379816699569)
([Movie] : title='Amped'; #kisses=98; #kicks=2, genre='?', 118.92854997854805)
Genre 'Romance' has been added to '?'! Check next line:
[Movie] : title='?'; #kisses=18; #kicks=90, genre='Romance'
This is an extendend version of the previous movie classification.
A .csv-file will be used to support sample data. This file contains various information about several variants of the Portuguese "Vinho Verde" wine (source: https://archive.ics.uci.edu/ml/datasets/Wine+Quality) including a "score" that describes the quality of the wine. The last row of of the .csv-file has been taken as "wine to be classified" - the corresponding score shall be calculated using KNN.
In contrast to the movie classifier, this application is able to normalize the values of the sample data set. Furthermore it's possible to find the smallest valid K-value (see the execution options below).
There are five constants to define the application's behavior:
INPUT_FILE - the csv.file to be read
EXPECTED_SCORE - the original score of the regarded wine (taken from the original .csv-file)
NORMALIZE - enable or disable sample data normalization (boolean)
DEBUG - enable or disable the debug/verbose mode (boolean)
FIND_SMALLEST_K - enable or disable the FIND_SMALLEST_K mode (if disabled, we will just calculate the most likely score)
# ------------------------------------------------------
# File = winequality-red.csv
# K = 500
# Expected result = 6
# FIND_SMALLEST_K = False
# NORMALIZE = False
# DEBUG = True
# ------------------------------------------------------
[DEBUG] The wine to be classified: (6, 0.31, 0.47, 3.6, 0.067, 18, 42, 0.99549, 3.39, 0.66, 11, '?')
[DEBUG] CSV file read! Size of the sample data set: 1598 rows
[DEBUG] These are the top 10 candidates (this is just to show that sorting is working!):
[DEBUG] 0.36 -> 1.893854480180565
[DEBUG] 0.36 -> 1.893854480180565
[DEBUG] 0.6 -> 2.91786671909462
[DEBUG] 0.55 -> 2.957259097897917
[DEBUG] 0.58 -> 3.1041112454936277
[DEBUG] 0.73 -> 3.152948449324853
[DEBUG] 0.83 -> 3.203012871173015
[DEBUG] 0.55 -> 3.3127065991572513
[DEBUG] 0.31 -> 3.3173818143982166
[DEBUG] 0.31 -> 3.3173818143982166
[DEBUG] This is the score count:
[DEBUG] 3 -> 3
[DEBUG] 4 -> 13
[DEBUG] 5 -> 189
[DEBUG] 6 -> 231
[DEBUG] 7 -> 61
[DEBUG] 8 -> 3
--------------------------------------------------------
Nearest Score = 6
Nearest score and expected score are matching!
The nearest score will be evalutated to 5 instead of 6 if NORMALIZATION is enabled.
Hendrik Thorun
[email protected]
Date: 14 June 2016