This is code for KNN from a text file. I have to explain how I would do unit tests on the implementation of the code.

import pandas as pd import sys import random def pickUnseenData(df, num): # ramdom shuffle the original data in order to randomly get num instances as unseen_instances index_li = [x for x in range(df.shape[0])] random.shuffle(index_li) unseen_df = pd.DataFrame(columns=df.columns, dtype=int) drop_li = [] for i in range(num): unseen_df = unseen_df.append(df.iloc[index_li[i]]) drop_li.append(df.index[index_li[i]]) df = df.drop(drop_li) return (df, unseen_df) # caculate distance between instance_1 and instance_2 # euclidean distance metric # return distance and class def getDistance(ins1, ins2): dis = (ins1-ins2).apply(lambda x: x**2)[:-1].sum() return(dis, ins1["Class"]) def knn(seen_df, unseen_df, k): ans = pd.DataFrame(columns=['Class'], index=unseen_df.index) for unseen_i in range(unseen_df.shape[0]): dis_df = pd.DataFrame(columns=['Distance', 'Class']) unseen_instance = unseen_df.iloc[unseen_i] for seen_i in range(seen_df.shape[0]): seen_instance = seen_df.iloc[seen_i] dis, seen_class = getDistance(seen_instance, unseen_instance) dis_df = dis_df.append( {'Distance': dis, 'Class': seen_class}, ignore_index=True) # get k near neibours k_df = dis_df.sort_values(by='Distance')[:k] if (k_df[k_df['Class'] == 0].shape[0] > k_df[k_df['Class'] == 1].shape[0]): ans.at[unseen_instance.name, 'Class'] = 0 else: ans.at[unseen_instance.name, 'Class'] = 1 return ans if __name__ == "__main__": if(len(sys.argv) < 4): print("Useage: {} ".format( sys.argv[0])) exit(1) df = pd.read_csv(sys.argv[1], sep="\t", header=0, index_col="Sequence.id") seen_df, unseen_df = pickUnseenData(df, int(sys.argv[3])) result = knn(seen_df, unseen_df, int(sys.argv[2])) diff = result-unseen_df print('k: {}; proportion of correctly classfied unknown instances: {}%'.format( sys.argv[2], diff[diff['Class'] == 0].shape[0]/diff.shape[0]*100))