#K-NN分类 import os import sys import time import operator import cx_Oracle import numpy as np import pandas as pd import tensorflow as tf conn=cx_Oracle.connect('doctor/admin@localhost:1521/tszr') cursor = conn.cursor() #获取数据集 def getdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable \ from menzhenZ where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() dataset = [] lables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) dataset.append(temp) lables.append(row[5]) return np.array(dataset),np.array(lables) def gettestdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from \ testZ where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() testdataset = [] testlables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) testdataset.append(temp) testlables.append(row[5]) return np.array(testdataset),np.array(testlables) #K-NN分类 def classify0(inX, dataSet, labels, k): dataSetSize = dataSet.shape[0] diffMat = np.tile(inX, (dataSetSize,1)) - dataSet sqDiffMat = diffMat**2 sqDistances = sqDiffMat.sum(axis=1) distances = sqDistances**0.5 sortedDistIndicies = distances.argsort() classCount={} for i in range(k): voteIlabel = labels[sortedDistIndicies[i]] classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1 sortedClassCount = sorted(classCount.items(), key=operator.itemgetter(1), reverse=True) return sortedClassCount[0][0] #归一化 def autoNorm(dataSet): minVals = dataSet.min(0) maxVals = dataSet.max(0) ranges = maxVals - minVals normDataSet = np.zeros(np.shape(dataSet)) m = dataSet.shape[0] normDataSet = dataSet - np.tile(minVals, (m,1)) normDataSet = normDataSet/np.tile(ranges, (m,1)) return normDataSet, ranges, minVals erace = [] accuc = [] t = [] #启动和检测模型 def datingClassTest(): datingDataMat,datingLabels = getdata("外科","胸外科") normMat, ranges, minVals = autoNorm(datingDataMat) testdataset,testlables = gettestdata("外科","胸外科") testnormMat, testranges, testminVals = autoNorm(testdataset) errorCount = 0.0 start = time.time() for j in [3,5,7,9,11,13]: for i in range(np.shape(testnormMat)[0]): classifierResult = classify0(testnormMat[i,:],normMat,datingLabels,j) print("the classifier came back with: %s, the real answer is: %s" % (classifierResult, testlables[i])) if (classifierResult != testlables[i]): errorCount += 1.0 end = time.time() t.append(end) erace.append(errorCount/float(np.shape(testnormMat)[0])*100) accuc.append((1.0-errorCount/float(np.shape(testnormMat)[0]))*100) print("错误率: %.2f%%" % (errorCount/float(np.shape(testnormMat)[0])*100)) print("准确率: %.2f%%" % ((1.0-errorCount/float(np.shape(testnormMat)[0]))*100)) print("训练和预测一共耗时: %.2f 秒" % (end-start)) datingClassTest() print(accuc) print(erace) print(t)
#探索不同的K值对算法的影响 import matplotlib.pyplot as plt x = [3,5,7,9,11,13] plt.plot(x,erace,c='r') plt.plot(x,accuc,c='g') plt.legend(['error race','accuce race'],loc=9) plt.show() print(accuc) print(erace)
#决策树 import os import sys import time import operator import cx_Oracle import numpy as np import pandas as pd from math import log import tensorflow as tf conn=cx_Oracle.connect('doctor/admin@localhost:1521/tszr') cursor = conn.cursor() #获取数据集 def getdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from menzhenZ where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() dataset = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) temp.append(row[5]) dataset.append(temp) lables = [] lables.append("呼吸急促") lables.append("持续性脉搏加快") lables.append("畏寒") lables.append("血压降低") lables.append("咳血") return dataset,lables def gettestdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from testZ where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() testdataset = [] testlables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) testdataset.append(temp) testlables.append(row[5]) return testdataset,testlables #计算熵值 def calcShannonEnt(dataSet): numEntries = len(dataSet) labelCounts = {} for featVec in dataSet: currentLabel = featVec[-1] if currentLabel not in labelCounts.keys(): labelCounts[currentLabel] = 0 labelCounts[currentLabel] += 1 shannonEnt = 0.0 for key in labelCounts: prob = float(labelCounts[key])/numEntries shannonEnt -= prob * log(prob,2) return shannonEnt #按照给定特征划分数据集 def splitDataSet(dataSet, axis, value): retDataSet = [] for featVec in dataSet: if featVec[axis] == value: reducedFeatVec = featVec[:axis] reducedFeatVec.extend(featVec[axis+1:]) retDataSet.append(reducedFeatVec) return retDataSet #选择最好的属性 def chooseBestFeatureToSplit(dataSet): numFeatures = len(dataSet[0]) - 1 baseEntropy = calcShannonEnt(dataSet) bestInfoGain = 0.0 bestFeature = -1 for i in range(numFeatures): featList = [example[i] for example in dataSet] uniqueVals = set(featList) newEntropy = 0.0 for value in uniqueVals: subDataSet = splitDataSet(dataSet, i, value) prob = len(subDataSet)/float(len(dataSet)) newEntropy += prob * calcShannonEnt(subDataSet) infoGain = baseEntropy - newEntropy if (infoGain > bestInfoGain): bestInfoGain = infoGain bestFeature = i return bestFeature #统计机制 def majorityCnt(classList): classCount={} for vote in classList: if vote not in classCount.keys(): classCount[vote] = 0 classCount[vote] += 1 sortedClassCount = sorted(classCount.items(), key=operator.itemgetter(1), reverse=True) return sortedClassCount[0][0] #创建决策树 def createTree(dataSet,labels): classList = [example[-1] for example in dataSet] if classList.count(classList[0]) == len(classList): return classList[0] if len(dataSet[0]) == 1: return majorityCnt(classList) bestFeat = chooseBestFeatureToSplit(dataSet) bestFeatLabel = labels[bestFeat] myTree = {bestFeatLabel:{}} temp = [] for i in labels: if i != labels[bestFeat]: temp.append(i) labels = temp featValues = [example[bestFeat] for example in dataSet] uniqueVals = set(featValues) for value in uniqueVals: subLabels = labels[:] myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet, bestFeat, value),subLabels) return myTree #使用决策树模型分类 def classify(inputTree,featLabels,testVec): for i in inputTree.keys(): firstStr = i break secondDict = inputTree[firstStr] featIndex = featLabels.index(firstStr) key = testVec[featIndex] valueOfFeat = secondDict[key] if isinstance(valueOfFeat, dict): classLabel = classify(valueOfFeat, featLabels, testVec) else: classLabel = valueOfFeat return classLabel #启动和检测模型 def datingClassTest(): dataSet,labels = getdata("外科","胸外科") myTree = createTree(dataSet,labels) testdataset,testlables = gettestdata("外科","胸外科") errorCount = 0.0 start = time.time() for i in range(np.shape(testdataset)[0]): classifierResult = classify(myTree,labels,testdataset[i]) print("the classifier came back with: %s, the real answer is: %s" % (classifierResult, testlables[i])) if (classifierResult != testlables[i]): errorCount += 1.0 end = time.time() print("错误率: %.2f%%" % (errorCount/float(np.shape(testdataset)[0])*100)) print("准确率: %.2f%%" % ((1.0-errorCount/float(np.shape(testdataset)[0]))*100)) print("训练和预测一共耗时: %.2f 秒" % (end-start)) datingClassTest()
#选取前600条记录生成并打印决策树 dataSet,labels = getdata("外科","胸外科") dataSet = dataSet[0:600] labels = labels[0:600] myTree = createTree(dataSet,labels) print(myTree)
#比较K-NN算法与决策树算法的优劣 import numpy as np import pandas as pd import matplotlib.pyplot as plt x = np.array([10,12]) y = [85.6,87.3] plt.bar(x,y,edgecolor='yellow') for i,j in zip(x,y): plt.text(i-0.2,j-0.2,'%.2f%%' % j) plt.text(9.7,40,'K-NN right race') plt.text(11.7,40,'Tree right race') plt.show()
#使用神经网络探索数据集 import sys import os import time import operator import cx_Oracle import numpy as np import pandas as pd import tensorflow as tf conn=cx_Oracle.connect('doctor/admin@localhost:1521/tszr') cursor = conn.cursor() #one-hot编码 def onehot(labels): n_sample = len(labels) n_class = max(labels) + 1 onehot_labels = np.zeros((n_sample, n_class)) onehot_labels[np.arange(n_sample), labels] = 1 return onehot_labels #获取数据集 def getdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from menzhen where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() dataset = [] lables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) dataset.append(temp) if(row[5]==3): lables.append(0) elif(row[5]==6): lables.append(1) else: lables.append(2) dataset = np.array(dataset) lables = np.array(lables) dataset = dataset.astype(np.float32) labless = onehot(lables) return dataset,labless #获取测试数据集 def gettestdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from test where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() testdataset = [] testlables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) testdataset.append(temp) if(row[5]==3): testlables.append(0) elif(row[5]==6): testlables.append(1) else: testlables.append(2) testdataset = np.array(testdataset) testlables = np.array(testlables) testdataset = testdataset.astype(np.float32) testlabless = onehot(testlables) return testdataset,testlabless dataset,labless = getdata("外科","胸外科") testdataset,testlables = gettestdata("外科","胸外科") dataset = dataset[0:100] labless = labless[0:100] x_data = tf.placeholder("float32", [None, 5]) y_data = tf.placeholder("float32", [None, 3]) weight = tf.Variable(tf.ones([5, 3])) bias = tf.Variable(tf.ones([3])) #使用softmax激活函数 y_model = tf.nn.softmax(tf.matmul(x_data, weight) + bias) #y_model = tf.nn.relu(tf.matmul(x_data, weight) + bias) # loss = tf.reduce_sum(tf.pow((y_model - y_data), 2)) #使用交叉熵作为损失函数 loss = -tf.reduce_sum(y_data*tf.log(y_model)) # train_step = tf.train.GradientDescentOptimizer(1e-4).minimize(loss) #使用AdamOptimizer优化器 train_step = tf.train.AdamOptimizer(1e-4).minimize(loss) #train_step = tf.train.MomentumOptimizer(1e-4,0.9).minimize(loss) #评估模型 correct_prediction = tf.equal(tf.argmax(y_model, 1), tf.argmax(y_data, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float")) init = tf.initialize_all_variables() sess = tf.Session() sess.run(init) start = time.time() for _ in range(10): for i in range(int(len(dataset)/100)): sess.run(train_step, feed_dict={x_data:dataset[i:i+100,:], y_data:labless[i:i+100,:]}) print("模型准确率",sess.run(accuracy, feed_dict={x_data:testdataset , y_data:testlables})) end = time.time() print("模型训练和测试公耗时：%.2f 秒" % (end-start))
#加深一层神经网络 import sys import os import time import operator import cx_Oracle import numpy as np import pandas as pd import tensorflow as tf conn=cx_Oracle.connect('doctor/admin@localhost:1521/tszr') cursor = conn.cursor() #one-hot编码 def onehot(labels): n_sample = len(labels) n_class = max(labels) + 1 onehot_labels = np.zeros((n_sample, n_class)) onehot_labels[np.arange(n_sample), labels] = 1 return onehot_labels #获取数据集 def getdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from menzhen where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() dataset = [] lables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) dataset.append(temp) if(row[5]==3): lables.append(0) elif(row[5]==6): lables.append(1) else: lables.append(2) dataset = np.array(dataset) lables = np.array(lables) dataset = dataset.astype(np.float32) labless = onehot(lables) return dataset,labless def gettestdata(surgery,surgeryChest): sql = "select feature1,feature2,feature3,feature4,feature5,trainLable from test where surgery='%s' and surgeryChest='%s'" % (surgery,surgeryChest) cursor.execute(sql) rows = cursor.fetchall() testdataset = [] testlables = [] for row in rows: temp = [] temp.append(row[0]) temp.append(row[1]) temp.append(row[2]) temp.append(row[3]) temp.append(row[4]) testdataset.append(temp) if(row[5]==3): testlables.append(0) elif(row[5]==6): testlables.append(1) else: testlables.append(2) testdataset = np.array(testdataset) testlables = np.array(testlables) testdataset = testdataset.astype(np.float32) testlabless = onehot(testlables) return testdataset,testlabless dataset,labless = getdata("外科","胸外科") testdataset,testlables = gettestdata("外科","胸外科") dataset = dataset[0:100] labless = labless[0:100] x_data = tf.placeholder("float32", [None, 5]) y_data = tf.placeholder("float32", [None, 3]) weight1 = tf.Variable(tf.ones([5, 20])) bias1 = tf.Variable(tf.ones([20])) y_model1 = tf.matmul(x_data, weight1) + bias1 #加深一层神经网络 weight2 = tf.Variable(tf.ones([20, 3])) bias2 = tf.Variable(tf.ones([3])) y_model = tf.nn.softmax(tf.matmul(y_model1, weight2) + bias2) loss = tf.reduce_sum(tf.pow((y_model - y_data), 2)) # loss = -tf.reduce_sum(y_data*tf.log(y_model)) #train_step = tf.train.GradientDescentOptimizer(1e-4).minimize(loss) train_step = tf.train.AdamOptimizer(1e-4).minimize(loss) # train_step = tf.train.MomentumOptimizer(1e-4,0.9).minimize(loss) correct_prediction = tf.equal(tf.argmax(y_model, 1), tf.argmax(y_data, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float")) init = tf.initialize_all_variables() sess = tf.Session() sess.run(init) start = time.time() for _ in range(10): for i in range(int(len(dataset)/100)): sess.run(train_step, feed_dict={x_data:dataset[i:i+100,:], y_data:labless[i:i+100,:]}) print("模型准确率",sess.run(accuracy, feed_dict={x_data:testdataset , y_data:testlables})) end = time.time() print("模型训练和测试公耗时：%.2f 秒" % (end-start))
#比较决策树与神经网络的优劣 import numpy as np import pandas as pd import matplotlib.pyplot as plt x = np.array([10,12]) y = [87.1,87.4] plt.bar(x,y,edgecolor="yellow") for i,j in zip(x,y): plt.text(i-0.2,j-0.2,"%.2f%%" % j) plt.text(9.7,40,"Tree right race") plt.text(11.7,40,"Net right race") plt.scatter([9.7,11.7],[0.05,0.36],c="r") plt.plot([9.7,11.7],[0.05,0.36],c="g") plt.show()
#统计各种算法处理模型数据 K-NN算法： 当K取[3,5,7,9,11,13]时，对应的： 准确率：[85.6, 72.6, 60.0, 47.4, 34.8, 22.299999999999996] 总耗时：[1554119134.435363, 1554119136.6192698, 1554119138.846019, 1554119141.2507513, 1554119143.4782736, 1554119145.5415804] 决策树： 准确率: 87.10% 训练和预测一共耗时: 0.05 秒 神经网络设计： 1 最小二乘法 softmax GradientDescentOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.16 秒 2 最小二乘法 softmax AdamOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.19 秒 3 最小二乘法 softmax MomentumOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.18 秒 4 最小二乘法 relu GradientDescentOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.17 秒 5 最小二乘法 relu AdamOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.15 秒 6 最小二乘法 relu MomentumOptimizer 模型 模型准确率 0.006 模型训练和测试公耗时：0.19 秒 7 交叉熵 softmax GradientDescentOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.09 秒 8 交叉熵 softmax AdamOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.08 秒 9 交叉熵 softmax MomentumOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.06 秒 10 交叉熵 relu GradientDescentOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.08 秒 11 交叉熵 relu AdamOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.08 秒 12 交叉熵 relu MomentumOptimizer 模型 模型准确率 0.874 模型训练和测试公耗时：0.09 秒 从上面的12种神经网络设计模型中可以看出：最小二乘法 relu MomentumOptimizer 模型 的准确率只有0.006，所以这种模型的设计是失败的。 a = [0.874]*10 print(a)
#计算成功的各种神经网络模型的准确率与耗时的比值： a = [0.874]*11 b = [0.16,0.19,0.18,0.17,0.15,0.09,0.08,0.06,0.08,0.09,0.09] c = [] for i in range(len(a)): c.append(a[i]/b[i]) for i in range(len(c)): print("准确率与耗时的比值:%.4f" % (c[i]))
#K-NN算法 #当K取3、5、7、9、11、13时的准确率饼图分布显示 import numpy as np import pandas as pd import matplotlib.pyplot as plt acc = [85.6, 72.6, 60.0, 47.4, 34.8, 22.2] labels = ['K-3','K-5','K-7','K-9','K-11','K-13'] plt.pie(acc,labels=labels,shadow=True,startangle=90,autopct='%1.4f%%') plt.axis('equal') plt.title('K-NN',fontsize=25) plt.show()
#K-NN算法耗时散点图 import numpy as np import pandas as pd import matplotlib.pyplot as plt from mpl_toolkits.mplot3d import Axes3D x = np.array([1,2,3,4,5,6]) z = np.array([1554119134.435363, 1554119136.6192698,1554119138.846019, 1554119141.2507513, 1554119143.4782736, 1554119145.5415804]) plt.scatter(x,z,c='g') plt.xticks(x+0.4,['KNN-1','KNN-2','KNN-3','KNN-4','KNN-5','KNN-6']) plt.show()
#神经网络算法对应各种有用的模型设计耗时曲线图 import numpy as np import pandas as pd import matplotlib.pyplot as plt from mpl_toolkits.mplot3d import Axes3D x = np.array([1,2,3,4,5,6,7,8,9,10,11]) z = np.array([0.16,0.19,0.18,0.17,0.15,0.09,0.08,0.06,0.08,0.09,0.09]) plt.scatter(x,z,c='r') plt.xticks(x+0.4,['NET-1','NET-2','NET-3','NET-4','NET-5', 'NET-6','NET-7','NET-8','NET-9','NET-10','NET-11']) plt.show()
#K-NN、决策树以及神经网络算法对比 import numpy as np import pandas as pd import matplotlib.pyplot as plt acc = [85.6, 72.6, 60.0, 47.4, 34.8, 22.2,87.10,0.874, 87.4,87.4,87.4,87.4,87.4,87.4,87.4,87.4,87.4,87.4] labels = ['K-3','K-5','K-7','K-9','K-11','K-13','TREE', 'NET-1','NET-2','NET-3','NET-4','NET-5','NET-6','NET-7', 'NET-8','NET-9','NET-10','NET-11'] explode = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0.2,0,0,0] plt.pie(acc,labels=labels,explode=explode,shadow=True,startangle=90,autopct='%1.4f%%') plt.axis('equal') plt.title('K-NN AND TREE AND NET',fontsize=25) plt.show()
