def Evolutionize(self):
bestGeneration = []
bestCentroid = None
bestFitness = np.inf
for i in range(self.itter):
tempBestGen = np.inf
for j in range(self.populationsize):
particles = [k for k in range(self.populationsize)
if k != j] #not current
randomparticle = []
for a in range(3):
ran = random.choice(particles)
randomparticle.append(ran)
particles.remove(ran)
a, b, c = randomparticle
mutant = []
for z in range(self.centroidNum):
mutant.append([])
for x in range(self.dimension):
#print(len(self.population[a].centroid))
#print(len(self.population[a].centroid[z]))
tempa = self.population[a].centroid[z][x]
tempb = self.population[b].centroid[z][x]
tempc = self.population[c].centroid[z][x]
temp = tempa - self.mut * (tempb - tempc)
mutant[z].append(temp)
#temp = [self.population[a].centroid[z][itt] +
# self.mut * (self.population[b].centroid[z][itt]
# - self.population[c].centroid[z][itt]) for itt in range(len(self.population[0].centroid[z])-1)]
#print (temp)
#print(mutant)
for a in range(len(mutant)):
for b in range(len(mutant[a])):
mutant[a][b] = np.clip(mutant[a][b], self.bounddown[a],
self.boundup[a])
crossTruth = [
np.random.rand(self.dimension) < self.crossp
for a in range(self.centroidNum)
]
#print(crossTruth)
trial = self.population[j].centroid.copy()
for a in range(len(crossTruth)):
for b in range(len(crossTruth[a])):
if (crossTruth[a][b]):
trial[a][b] = mutant[a][b]
temp = self.population[j].getfitness(trial)
print("=========================")
print("particle number: " + str(j))
print("fitness trial : " + str(temp))
print("self best : " + str(self.population[j].lastfitness))
#print("%s, %s" %(self.population[j].lastfitness,tempBestGen))
#print("%s" %(self.population[j].lastfitness<tempBestGen))
if self.population[j].lastfitness < tempBestGen:
#recording best fitness each generation
tempBestGen = self.population[j].lastfitness
if (self.population[j].lastfitness < bestFitness):
bestFitness = self.population[j].lastfitness
bestCentroid = self.population[j].centroid
if temp < self.population[j].lastfitness:
self.population[j].lastfitness = temp
self.population[j].centroid = trial
bestGeneration.append(tempBestGen)
print("=====================================")
print("Itteration : %s" % (i))
print("=====================================")
print("bestfitness : %s" % (min(self.fitness)))
for i in range(len(bestGeneration)):
print("best fitness at generation -%s : %s" %
(i, bestGeneration[i]))
print("---------------------------------------------------")
print("Centroid")
print("---------------------------------------------------")
for i in range(len(bestCentroid)):
print("centroid-%s :%s" % (i, bestCentroid[i]))
clust = KMeans(self.data.data, bestCentroid, self.ansdict)
clust.run(self.itterkmeans)
print("===================================================")
print("====================K-Means========================")
print("===================================================")
print("SSE : %s" % (clust.getaccsse()))
print("Akurasi : %s" % (clust.getacc()))
print("Assigned data : %s" % (clust.assignedNum))
print("---------------------------------------------------")
print("Centroid")
print("---------------------------------------------------")
for i in range(len(clust.centroid)):
print("centroid-%s :%s" % (i, clust.centroid[i]))
clust.print_Output()
import random
from dataset import Dataset
from KMeans import KMeans
Data = Dataset('nursery.txt')
Data.run()
numCen = 5
centroid = [
Data.data[random.randrange(0, len(Data.data))].copy()
for i in range(numCen)
]
Clust = KMeans(Data.data, centroid, 5)
Clust.run(10)
print(Clust.getaccsse())
print(Clust.getacc())
class Ant():
def __init__(self,data, ansdict,numCen, tabuindex, pheromap, alpha, beta, max_itter, decay):
self.centroid = [data[random.randrange(0,len(data))].copy()for i in range(numCen)]
self.defCentroid = self.centroid.copy()
self.clust = KMeans(data, self.centroid, ansdict)
self.max_itter = max_itter
self.alpha = alpha
self.beta = beta
self.pheromap = pheromap
self.tabu = {}
self.decay = decay
self.fitness = 0
self.numCen = numCen
self.data = data
self.tabuIndex = tabuindex
def init(self):
self.centroid = self.defCentroid.copy()
self.tabu = {}
def spreadPheromon(self):
self.clust.getRandomData(5)
self.possibleMove = self.clust.random
self.distMove = self.clust.randomDist
#print(self.possibleMove)
#print(type(self.pheromap))
for nodeList in range(len(self.possibleMove)):
for node in range(len(self.possibleMove[nodeList])):
a = self.possibleMove[nodeList][node]
try:
self.pheromap[a] = self.pheromap[a] *self.decay
try:
self.pheromap [a]+=1.0/self.distMove[nodeList][node]
except:
self.pheromap [a]+=0
except KeyError:
self.pheromap[a] = self.distMove[nodeList][node]
#print(self.pheromap)
#print(self.distMove)
def probMov(self):
self.poss_mov = [[] for i in range(self.numCen)]
self.prob_mov=[[]for i in range(self.numCen)]
for nodeList in range(len(self.possibleMove)):
for node in range(len(self.possibleMove[nodeList])):
a = self.possibleMove[nodeList][node]
try:
temp = self.tabu[a]
except KeyError:
self.poss_mov[nodeList].append(a)
temp = (self.pheromap[a]**self.alpha)*(self.distMove[nodeList][node]**self.beta)
self.prob_mov[nodeList].append(temp)
temp_mov = len(self.possibleMove)
for a in range(len(self.poss_mov)):
for b in range(len(self.poss_mov[a])):
self.prob_mov[a][b] = self.prob_mov[a][b] / (temp_mov)
#print(self.prob_mov)
def decaytabu(self):
todel = []
for node in self.tabu:
self.tabu[node]-=1
if (self.tabu[node]==0):
todel.append(node)
for node in todel:
del self.tabu[node]
def mov(self):
self.decaytabu()
newcen = []
#print(self.poss_mov)
for i in range(self.numCen):
a = list(range(len(self.poss_mov[i])))
movement = choice(a,1,self.prob_mov[i])
movement = movement[0]
#print(self.poss_mov[i][movement])
newcen.append(self.data[self.poss_mov[i][movement]])
self.tabu[self.poss_mov[i][movement]]=self.tabuIndex
self.centroid = newcen.copy()
self.clust.newcentroid(self.centroid)
#self.clust.clusterize()
#print(self.centroid)
def run(self):
for a in range(self.max_itter):
self.spreadPheromon()
self.probMov()
self.mov()
self.fitness = self.clust.getaccsse()
