基于Python决策树算法的学生学习行为数据分析.zip

import numpy as np import pandas as pd from sklearn.model_selection import train_test_split from sklearn import tree import graphviz import pydotplus import matplotlib.pyplot as plt def data_preprocessing(data): # 数据清洗，采用用平均值填充空缺值 for column in list(data.columns[data.isnull().sum() > 0]): mean_val = data[column].mean() data[column].fillna(mean_val, inplace=True) data = np.array(data.values) feature = data[:,0:3] label = data[:,3] for i in range(np.size(label)): if label[i]<60: label[i] = 3 elif label[i]<80 and label[i]>59: label[i] = 2 else: label[i] = 1 return feature,label def structural_model(feature,label): # 数据集划分，70%训练数据，30%测试数据 feature_train, feature_test, label_train, label_test = train_test_split(feature, label, test_size=0.3) # 选取最合适的深度 max_depths = [] for max_depth in range(10): clf = tree.DecisionTreeClassifier(criterion='entropy',max_depth=max_depth+1) clf.fit(feature_train, label_train) # 拟合 score = clf.score(feature_test,label_test) max_depths.append(score) best_depth = max_depths.index(max(max_depths))+1 # plt.figure(figsize=(20, 8), dpi=80) # plt.plot(range(1, 11), max_depths) # plt.xlabel('max depth') # plt.ylabel('evaluate score') # plt.show(); # 选取最合适的最小叶子树 min_samples = [] for min_sample in range(30): clf = tree.DecisionTreeClassifier(criterion='entropy',max_depth=best_depth,min_samples_leaf=min_sample+5) clf.fit(feature_train, label_train) # 拟合 score = clf.score(feature_test, label_test) min_samples.append(score) best_min_samples_leaf = min_samples.index(max(min_samples))+5 # plt.figure(figsize=(20, 8), dpi=80) # plt.plot(range(4, 34), min_samples) # plt.xlabel('min samples leaf') # plt.ylabel('evaluate score') # plt.show(); # 根据最合适的参数构建模型 mytree = tree.DecisionTreeClassifier(criterion='entropy',max_depth=best_depth,min_samples_leaf=best_min_samples_leaf) mytree.fit(feature_train,label_train) return mytree def tree_visualise(mytree): # 可视化 dot_data = tree.export_graphviz(mytree, out_file=None,\ feature_names=["Attendance","Preview","Job"],\ class_names=["excellent","good","poor"],\ rounded=True) graph = pydotplus.graph_from_dot_data(dot_data) graph.write_pdf("dtree10.pdf") if __name__ == '__main__': # 加载数据 data = pd.read_csv('student_data.csv') # 数据预处理 feature,label = data_preprocessing(data) # 构造模型 mytree = structural_model(feature,label) # 可视化训练好的决策树 tree_visualise(mytree) # 计算预测正确率 feature_train, feature_test, label_train, label_test = train_test_split(feature, label, test_size=0.3) rate = np.sum(mytree.predict(feature_test) == label_test) / mytree.predict(feature_test).size print('训练集数量：',label_train.size) print('测试集数量：',label_test.size) print('正确率：',rate)