生物信息學原理作業第五彈：K-means聚類的實現。
轉載請保留出處！
原理參考：K-means聚類（上）
數據是老師給的，二維，2 * 3800的數據。plot一下可以看到有7類。
怎么確定分類個數我正在學習，這個腳本就直接給了初始分類了，等我學會了再發。
下面貼上Python代碼，版本為Python3.6。
1 # -*- coding: utf-8 -*- 2 """ 3 Created on Wed Dec 6 16:01:17 2017 4 5 @author: zxzhu 6 """ 7 import numpy as np 8 import matplotlib.pyplot as plt 9 from numpy import random 10 11 def Distance(x): 12 def Dis(y): 13 return np.sqrt(sum((x-y)**2)) #歐式距離 14 return Dis 15 16 def init_k_means(k): 17 k_means = {} 18 for i in range(k): 19 k_means[i] = [] 20 return k_means 21 22 def cal_seed(k_mean): #重新計算種子點 23 k_mean = np.array(k_mean) 24 new_seed = np.mean(k_mean,axis=0) #各維度均值 25 return new_seed 26 27 def K_means(data,seed_k,k_means): 28 for i in data: 29 f = Distance(i) 30 dis = list(map(f,seed_k)) #某一點距所有種子點的距離 31 index = dis.index(min(dis)) 32 k_means[index].append(i) 33 34 new_seed = [] #存儲新種子 35 for i in range(len(seed_k)): 36 new_seed.append(cal_seed(k_means[i])) 37 new_seed = np.array(new_seed) 38 return k_means,new_seed 39 40 def run_K_means(data,k): 41 seed_k = data[random.randint(len(data),size=k)] #隨機產生種子點 42 k_means = init_k_means(k) #初始化每一類 43 result = K_means(data,seed_k,k_means) 44 count = 0 45 while not (result[1] == seed_k).all(): #種子點改變，繼續聚類 46 count+=1 47 seed_k = result[1] 48 k_means = init_k_means(k=7) 49 result = K_means(data,seed_k,k_means) 50 print('Done') 51 #print(result[1]) 52 print(count) 53 plt.figure(figsize=(8,8)) 54 Color = 'rbgyckm' 55 for i in range(k): 56 mydata = np.array(result[0][i]) 57 plt.scatter(mydata[:,0],mydata[:,1],color = Color[i]) 58 return result[0] 59 60 data = np.loadtxt('K-means_data') 61 run_K_means(data,k=7)
附上結果圖：
這個算法太依賴於初始種子點的選取了，隨機選點很有可能會得到局部最優的結果，所以下一步學習一下怎么設置初始種子點以及分類數目。