简单adaboot

#学习自机器学习实战
import numpy as np
import math
def loadSimpleData():
    dataMat=np.matrix([[1.,2.1],[2.,1.1],[1.3,1.],[1.,1.],[2.,1.]])#data
    classLabels=[1.0,1.0,-1.0,-1.0,1.0]#label
    return dataMat,classLabels
def stumpClassify(dataMatrix,dimen,threshVal,threshIneq):
    retArray=np.ones((np.shape(dataMatrix)[0],1))#a easy thresh decide clasifier
    if threshIneq=='lt':
        retArray[dataMatrix[:,dimen]<=threshVal]=-1.0
    else:
        retArray[dataMatrix[:,dimen]>threshVal]=-1.0
    return retArray

def buildStump(dataArr,classLabels,D):
    dataMatrix=np.mat(dataArr);labelMat=np.mat(classLabels).T#initilize
    m,n=np.shape(dataMatrix)# number and voclumn
    numSteps=1.0;bestStump={};bestClasEst=np.mat(np.zeros((m,1)))#nuber of weakerclassifier bestclassier name and
    minError=np.inf#initilize
    for i in range(n):
        rangeMin=dataMatrix[:,i].min();rangeMax=dataMatrix[:,i].max()#range of data
        stepSize=(rangeMax-rangeMin)/numSteps#step
        for j in range(-1,int(numSteps)+1):
            for inequal in ['lt' ,'gt']:
                threshVal=(rangeMin+float(j)*stepSize)#decide thresh try all
                predictedVals=stumpClassify(dataMatrix,i,threshVal,inequal)#predict using weakclassifier
                errArr=np.mat(np.ones((m,1)))#record errar
                print("errArr1" ,errArr)
                errArr[predictedVals==labelMat]=0#record error
                print("errArr2" ,errArr)
                weightedError=D.T*errArr
                if(weightedError<minError):
                    minError=weightedError#record the best value then creat best weakclassifier
                    bestClasEst=predictedVals.copy()
                    bestStump['dim']=i#column feature
                    bestStump['thresh']=threshVal#value
                    bestStump['ineq']=inequal#way
    return bestStump,minError,bestClasEst
def adaBoostTrainDS(dataArr,classLabels,numIt=40):
    weakClassArr=[]#initilize
    m=np.shape(dataArr)[0]#num of train test
    D=np.mat(np.ones((m,1))/m)#initilize weights
    aggClassEst=np.mat(np.zeros((m,1)))
    for i in range(numIt):
        bestStump,error,classEst=buildStump(dataArr,classLabels,D)#get bestweakclassifier
        print("D:" ,D.T)
        alpha=float(0.5*np.log((1.0-error)/max(error,1e-6)))#normalize means less error high alpha
        bestStump['alpha']=alpha
        weakClassArr.append(bestStump)#add about 40
        print("classEst" ,classEst.T)#record
        #expon=np.multiply(-1*alpha*np.mat(classLabels).T,classEst)#-1*a*label*predict
        expon=math.e**(-alpha)#the way the book give i dont know why thw author didn't use it to update
        D=np.multiply(D,expon)#update
        D=D/D.sum()
        aggClassEst+=alpha*classEst
        print("aggClassEst:" ,aggClassEst.T)
        aggErrors=np.multiply(np.sign(aggClassEst)!=np.mat(classLabels).T,np.ones((m,1)))
        errorRate=aggErrors.sum()/m#record
        print("total error:" ,errorRate,"\n")
        if errorRate.all()==0.0:break
    return weakClassArr
def adaClassify(datToClass,classifierArr):
    dataMatrix=np.mat(datToClass)
    m=np.shape(dataMatrix)[0]
    aggClassEst=np.mat(np.zeros((m,1)))
    for i in range(len(classifierArr)):
        #print(classifierArr[i]['dim'],classifierArr[i]['thresh'],classifierArr[i]['ineq'])
        classEst=stumpClassify(dataMatrix,classifierArr[i]['dim'],classifierArr[i]['thresh'],classifierArr[i]['ineq'])#test us every classifer
        #classEst = stumpClassify(dataMatrix,classifierArr[i])
        aggClassEst+=classifierArr[i]['alpha']*classEst#pridect *weight
        print("aggClassEst" ,aggClassEst)
    return np.sign(aggClassEst)#result
if __name__=='__main__':
    x,y=loadSimpleData()
    #x,y=loadDataSet('testSet.txt')
    D=np.mat(np.ones((5,1))/5)#initilize weight
   # a,b,c=buildStump(x,y,D)#get the classifier
    w=adaBoostTrainDS(x,y)
    print(np.shape(w))
    print(adaClassify(x,w))