数据挖掘k-means和PAM算法区别.docxVIP

专 业： 智能科学与技术 学生姓名： 肖万发 学 号： 14020510039 k-means和PAM算法区别：k-means算法找的聚类中心是类别的平均值而PAM算法聚类中心是找已经分好的类别中代价最小的一个样本点K-means程序：importnumpy as npimportmatplotlib.pyplot as pltfrom mpl_toolkits.mplot3d import Axes3Dimport randomimport mathtemp1=input(输入你要生成的数据个数（大于10001的偶数）：)temp2=input(输入你的噪声百分比（直接输入数字）:)fig=plt.figure() ax=Axes3D(fig)pi=math.pinum1=int(temp1)lable=np.zeros(num1)num2=int(num1/(100/int(temp2)))num3=int(int(temp1)/2)R=np.zeros(num1)a=np.zeros(num1)b=np.zeros(num1)x=np.zeros(num1)y=np.zeros(num1)z=np.zeros(num1)fori in range(num1):a[i]=random.uniform(0,2*pi)b[i]=random.uniform(0,pi)if(i=num3): R[i]=np.random.randint(1,50)x[i]=R[i]*math.sin(b[i])*math.cos(a[i])y[i]=R[i]*math.sin(b[i])*math.sin(a[i])z[i]=R[i]*math.cos(b[i])lable[i]=0elif(inum3): R[i]=np.random.randint(50,100)x[i]=R[i]*math.sin(b[i])*math.cos(a[i])y[i]=R[i]*math.sin(b[i])*math.sin(a[i])z[i]=R[i]*math.cos(b[i])lable[i]=1c0=np.random.normal(size=num1)c1=np.random.normal(size=num1)c2=np.random.normal(size=num1)fori in range(num2):if(i=int(num2/2)):x[i]=x[i]+c0[i]y[i]=y[i]+c1[i]z[i]=z[i]+c2[i]elif(iint(num2/2)):x[num3+i-1]=x[num3+i-1]+c0[i]y[num3+i-1]=y[num3+i-1]+c1[i]z[num3+i-1]=z[num3+i-1]+c2[i]data=np.zeros((num1,3))fori in range(num1):data[i,0]=x[i]data[i,1]=y[i]data[i,2]=z[i]x0=np.zeros(num1)y0=np.zeros(num1)z0=np.zeros(num1)x1=np.zeros(num1)y1=np.zeros(num1)z1=np.zeros(num1)K=np.zeros((2,3))K[0,:]=data[100,:]K[1,:]=data[15000,:]for n in range(8): sum0=0 num00=np.zeros((1,3)) sum1=0 num11=np.zeros((1,3))for q in range(num1): e0=(K[0,:]-data[q,:])*(K[0,:]-data[q,:]) f0=e0[0]+e0[1]+e0[2] e1=(K[1,:]-data[q,:])*(K[1,:]-data[q,:]) f1=e1[0]+e1[1]+e1[2]if(f0=f1):lable[q]=0 num00=num00+1 sum0=sum0+data[q,:]elif(f1f0):lable[q]=1 num11=num11+1 sum1=sum1+data[q,:]K[0,:]=sum0/num00K[1,:]=sum1/num11for p in range(num1):if(lable[p]==0):x0[p]=x[p]y0[p]=y[p]z0[p]=z[p]elif(lable[p]==1):x1[p]=x[p