how to get list all the neighbors from KNN for each instance in python

Question

i don't know much about python but for my class i must make a KNN classification where i need to get all the neighbors (not just the predicted one by majority vote) and get percentage of each of the neighbors (percentage of chance of occuring). for example my train data looks like:

zip          lat    long
77339   73730.689   -990323.6834
77339   73731.699   -990323.6834
77345   71679.54137 -998244.7071
77346   71679.54137 -998245.7071

and my test data is:

77388   55410.72694 -994507.7348    
77389   60816.86756 -990211.3075    
77396   68641.10762 -997071.3902    
77380   48762.26134 -978612.6912    

where zip is the class needed to predict. i made two file for train and test set. my goal is to get the neighbor list for each of the test instances. so, if the majority vote says the predict zip for zip 77388 in testset is 77380 (k=3),then i wanna know what is the percentage for occuring 77380 as well as of other neighbors for that zip (other 2 neighbors might be 77388, 77389).

i tried the following code with help of this link http://alexhwoods.com/k-nearest-neighbors-in-scikit-learn/ .

import pandas as pd
import numpy as np
train =pd.read_csv('C:\\train.csv')
test =pd.read_csv('C:\\test.csv')

from sklearn.neighbors import KNeighborsClassifier
cols=['lat','long']
cols2=['tweetzip']

trainArr=train.as_matrix(cols) #trainArr= has training numeric feature (location)
trainRes=train.as_matrix(cols2) # trainRes= contains the correct zip for training data
testArr=test.as_matrix(cols) ##testArr= has location of test data
testRes=test.as_matrix(cols2) # ##testRes= has correct tweetzip for test tada

knn =KNeighborsClassifier(n_neighbors=3,weights='uniform')
knn.fit(trainArr,trainRes.ravel())
output=knn.predict(testArr)
knn.kneighbors(testArr, return_distance=False) ### is it the same as previous line?

tweezip=[]
predictzip=[]
correct=0.0
for i in range(len(output)):
    if testRes[i][0]==output[i]:
        correct+=1

tweezip.append(testRes[i][0])
predictzip.append(output[i])

print tweezip
print predictzip
print 'correct',correct/len(output)

i tried to find something but cannot find the right one that will be helpful for me. Thank you.

Answer 1

graphlab can be easy to implement, free for academic learning. Checkout coursera course on ML with clustering and intro to graphlab here

I find this link on machine learning mastery also easier to understand and implement for a simpler application, written just in python.