def main():
splitRatio = 0.67
pred_acc = []
for attrnum in range(2, 13):
model = Bayesian(attrnum, 3)
# dataset = loadData.loadWine()
# trainingSet, testSet = g.splitDataset(dataset, splitRatio)
# np.savez('Wine.npz', train=trainingSet, test=testSet)
trainingSet = np.load('Wine.npz')['train']
testSet = np.load('Wine.npz')['test']
trainingSet_2, trainingSet_ori = callPCA(trainingSet, 13, attrnum)
testSet_2, testSet_ori = callPCA(testSet, 13, attrnum)
trainingSet = np.array(trainingSet_2)
testSet = testSet_2
print(trainingSet.shape)
summaries = model.summarizeByClass(trainingSet)
predictions, result_prob = model.getPredictions(summaries, testSet)
x, y = g.splitXandY(np.array(testSet), model.attrNum, len(testSet))
# print(x)
# print(x.shape)
confusion_matrix = np.zeros((len(summaries), len(summaries)))
accuracy, confusion_matrix = g.getAccuracy(testSet, predictions,
confusion_matrix)
print(accuracy)
pred_acc.append(accuracy)
plot.ROC(y, result_prob[:, 1])
plt.plot(range(2, 13), pred_acc)
plt.show()
return accuracy
def callFLD(dataset, attrNum):
fld = FLD.FLD(len(dataset[0]) - 1, 1)
X, y = g.splitXandY(dataset, attrNum, len(dataset))
mean = X.mean()
std = X.std()
X_norm = (X - mean) / std
fld.X = {"label": y, "data": X_norm}
fld.initClassData()
fld.reduce()
print(fld.Jw())
m = np.shape(fld.X_f)[1]
for i in range(m):
color = ''
if y[i] == 1: color = 'r'
if y[i] == 2: color = 'g'
if y[i] == 3: color = 'b'
plt.scatter(fld.X_f[0, i], y[i], s=50, c=color)
plt.show()
plt.show()
return np.hstack((fld.X_f.T, y))
def callPCA(dataset, attrNum, k):
X, y = g.splitXandY(dataset, attrNum, len(dataset))
print(k)
finalData, reconMat = PCA.pca(X, k)
# PCA.plotBestFit(finalData, reconMat, y)
return np.hstack((finalData, y)), np.hstack((reconMat, y))
def summarizeByClass(self, dataset):
separated = self.separateByClass(dataset)
summaries = {}
for classValue, instances in separated.items():
x, y = g.splitXandY(np.array(instances), self.attrNum,
len(instances))
summaries[classValue] = self.meanVector(x), self.stdMat(x)
return summaries
def main():
pred_acc = []
for attrnum in range(1, 32):
model = Linear(attrnum, 2, 30)
'''
splitRatio = 0.67
dataset = loadData.loadIono()
trainingSet, testSet = g.splitDataset(dataset, 0.67)
# np.savez('Gender_FLD.npz', train=trainingSet, test=testSet)
'''
trainingSet = np.load('Iono.npz')['train']
testSet = np.load('Iono.npz')['test']
"""
trainingSet = callFLD(np.array(trainingSet), 32)
testSet = callFLD(testSet, 32)
"""
trainingSet_2, trainingSet_ori = callPCA(trainingSet, 32, attrnum)
testSet_2, testSet_ori = callPCA(testSet, 32, attrnum)
trainingSet = trainingSet_2
testSet = testSet_2
for i in range(5000):
if i % 100 == 0:
model.lr = model.lr / 5
batchData = batch(trainingSet, model.batchNum)
x, y = g.splitXandY(batchData, model.attrNum, len(batchData))
model.train(x, y)
x, y = g.splitXandY(np.array(testSet), model.attrNum, len(testSet))
final_output, accuracy = model.predict_test(x, y)
pred_acc.append(accuracy)
# plot.ROC(y, final_output)
plt.plot(range(1, 32), pred_acc)
plt.show()
return
def plot(self):
clusted, label = g.splitXandY(self.data, 13, len(self.data))
separated = separateByClass(clusted, label.flatten())
import matplotlib.pyplot as plt
import itertools
for key, data in separated.items():
print(key, " : ", len(data))
allKey = list(separated.keys())
colors = itertools.cycle(["red", "blue", "green","yellow", "orange"])
for i in range(len(separated.keys())):
color_this = next(colors)
for j in range(len(separated[allKey[i]])):
plt.scatter(separated[allKey[i]][j][0], separated[allKey[i]][j][1], color=color_this, alpha=0.6)
plt.show()
def __init__(self, dataset, type='single', cluster_num=2):
self.dataset = dataset
self.X, self.Y = g.splitXandY(dataset, 13, len(dataset))
self.cluster_num = cluster_num
self.dis_mat = np.zeros((dataset.shape[0], dataset.shape[0]))
self.label = []
self.draw_label = []
self.allCluster = []
self.type = type
# initialize label and distance matrix
for i in range(dataset.shape[0]):
self.label.append(i)
self.draw_label.append(i)
self.allCluster.append(i)
for j in range(dataset.shape[0]):
self.dis_mat[i][j] = np.linalg.norm(self.X[i] - self.X[j],
ord=1)
self.link = []
def main():
orig_attr = 4
splitRatio = 0.67
pred_acc = []
for attrnum in range(2, orig_attr):
dataset = loadData.loadIris()
model = Naive_Bayes(attrnum, 3)
# trainingSet, testSet = g.splitDataset(dataset, splitRatio)
# np.savez('Iris.npz', train=trainingSet, test=testSet)
trainingSet = np.load('Iris.npz')['train']
testSet = np.load('Iris.npz')['test']
# print(trainingSet.shape)
trainingSet_2, trainingSet_ori = callPCA(trainingSet, orig_attr,
attrnum)
testSet_2, testSet_ori = callPCA(testSet, orig_attr, attrnum)
trainingSet = np.array(trainingSet_2)
testSet = testSet_2
summaries = model.summarizeByClass(trainingSet)
# print(summaries)
predictions, result_prob = model.getPredictions(summaries, testSet)
x, y = g.splitXandY(testSet, model.attrNum, len(testSet))
confusion_dim = len(summaries)
confusion_matrix = np.zeros((confusion_dim, confusion_dim))
accuracy, confusion_matrix = g.getAccuracy(testSet, predictions,
confusion_matrix)
print(accuracy)
pred_acc.append(accuracy)
plot.ROC(y, result_prob)
plt.plot(range(2, orig_attr), pred_acc)
plt.show()
def plot(self):
clusted, label = g.splitXandY(self.data, 13, len(self.data))
separated = separateByClass(clusted, label.flatten())
import matplotlib.pyplot as plt
import itertools
for key, data in separated.items():
print(key, " : ", len(data))
allKey = list(separated.keys())
colors = itertools.cycle(["red", "blue", "green","yellow", "orange"])
for i in range(len(separated.keys())):
color_this = next(colors)
for j in range(len(separated[allKey[i]])):
plt.scatter(separated[allKey[i]][j][0], separated[allKey[i]][j][1], color=color_this, alpha=0.6)
plt.show()
dataset = loadData.loadWine()
X,y = g.splitXandY(dataset,13,len(dataset))
dbscan = DBSCAN(dataset,eps=50, minPts=10)
dbscan.find_core_point()
print("eps=1, minPts=20")
dbscan.plot()
print("Adjusted Rand :", metrics.adjusted_rand_score(y.flatten(), dbscan.data[:,-1]))
print("Normalized Mutual Info:", normalized_mutual_info_score(y.flatten(), dbscan.data[:,-1]))