Polar-Zhang_Naive-Bayes_14504_1752736200205.zip

import math import numpy as np import pandas as pd data = pd.read_excel("./data.xlsx") # 构建数据集 def Dataset(Data, rate): # 第一种切行方法 # Train = data.loc[data['diagnosis'] == 'M', :] # print(Train) # 第二种切行方法 M = Data[Data['diagnosis'] == 'M'] Train_M = M.sample(frac=rate, random_state=888, replace=False, axis=0) # 删除指定行列 Test_M = M.drop(Train_M.index, axis=0) B = Data[Data['diagnosis'] == 'B'] Train_B = B.sample(frac=rate, random_state=888, replace=False, axis=0) Test_B = B.drop(Train_B.index, axis=0) # 拼接训练集测试集 Train = pd.concat([Train_B, Train_M]) Test = pd.concat([Test_B, Test_M]) return Train, Test # 计算先验概率 def Priori(Train): S = len(Train) M = len(Train[Train['diagnosis'] == 'M']) B = len(Train[Train['diagnosis'] == 'B']) PM = M / S PB = B / S return PM, PB # 计算均值和方差 def Mean_Var(Train): Mean = Train.iloc[:, 2:].mean() Var = Train.iloc[:, 2:].var() return Mean, Var # 计算恶性高斯概率密度函数 def Gauss_M(Train, Test): gauss_m = [] Train = Train[Train['diagnosis'] == 'M'] mean, var = Mean_Var(Train) for i in range(len(Test)): for j in range(len(mean)): gauss_m.append((1 / math.sqrt(2 * math.pi * var[j])) * math.exp(-(Test.iloc[i, j + 2] - mean[j]) * (Test.iloc[i, j + 2] - mean[j]) / (2 * var[j]))) return gauss_m # 计算良性高斯概率密度函数 def Gauss_B(Train, Test): gauss_b = [] Train = Train[Train['diagnosis'] == 'B'] mean, var = Mean_Var(Train) for i in range(len(Test)): for j in range(len(mean)): gauss_b.append((1 / (math.sqrt(2 * math.pi * var[j]))) * math.exp(-(Test.iloc[i, j + 2] - mean[j]) * (Test.iloc[i, j + 2] - mean[j]) / (2 * var[j]))) return gauss_b # 基于训练集得到的均值、方差以及先验概率对测试集进行分类 def Class(Train, Test): p_m = [] p_b = [] p = [] m = 0 pm, pb = Priori(Train) # 得到先验概率 pm_c, pb_c = pm, pb # 固定先验概率，便于后续使用 gauss_m = Gauss_M(Train, Test) # 计算测试集各独立特征的恶性似然概率 gauss_b = Gauss_B(Train, Test) # 计算测试集各独立特征的良性似然概率 # 计算测试集各独立特征的恶性后验概率 for i in range(0, len(gauss_m), 30): for j in range(i, i + 30, 1): pm = pm * gauss_m[j] p_m.append(pm) pm = pm_c # 计算测试集各独立特征的良性后验概率 for i in range(0, len(gauss_b), 30): for j in range(i, i + 30, 1): pb = pb * gauss_b[j] p_b.append(pb) pb = pb_c # 根据恶性、良性后验概率分类 for i in range(len(p_m)): if p_m[i] > p_b[i]: p.append("M") elif p_m[i] < p_b[i]: p.append("B") else: p.append("NO") # 获得测试集标签 label = np.array(Test['diagnosis']).tolist() # 对比测试集标签与后验概率分类，相同则计数 for i in range(len(p)): if label[i] == p[i]: m = m + 1 # 计算测试集分类准确率 print("测试集准确率：{} %".format(m / len(p) * 100)) if __name__ == '__main__': train, test = Dataset(data, 0.7) Class(train, test)