使用机器学习算法基于用户的社交媒体使用情况预测用户情绪

# 写前端用的库streamlit import streamlit as st # 导入pandas 和 numpy 用这两个库进行数据处理 import pandas as pd import numpy as np # 将用来画动态图 import time # mpl将plt画的图设置中文 import matplotlib as mpl # 用sns画图，看看最开始的数据长什么样 import seaborn as sns # PCA将数据降维 from sklearn.decomposition import PCA # 混淆矩阵 from sklearn.metrics import confusion_matrix # 决策树 from sklearn.tree import DecisionTreeClassifier # 用于画出整个决策树 from sklearn.tree import export_graphviz import graphviz # k-nn（k-近邻） from sklearn.neighbors import KNeighborsClassifier # 支持向量机 from sklearn.svm import SVC # 集成学习 from sklearn.ensemble import BaggingClassifier # 保存模型 import joblib # 各种评价标准 from sklearn.metrics import accuracy_score from sklearn.metrics import precision_score from sklearn.metrics import recall_score from sklearn.metrics import f1_score # 用来画图 import matplotlib.pyplot as plt # 将plt画出来的图设置为中文简体 mpl.rcParams['font.family'] = 'SimHei' # 读取数据集，数据集可以直接再 test_x = pd.read_csv(r"D:\social-media-usage-and-emotional-well-being\pythonProject\Data preprocessing\test_x.csv") test_y = pd.read_csv(r"D:\social-media-usage-and-emotional-well-being\pythonProject\Data preprocessing\test_y.csv") train_y = pd.read_csv(r"D:\social-media-usage-and-emotional-well-being\pythonProject\Data preprocessing\train_y.csv") train_x = pd.read_csv(r"D:\social-media-usage-and-emotional-well-being\pythonProject\Data preprocessing\train_x.csv") st.header("根据社交媒体使用情况判断用户情绪") choice = st.sidebar.selectbox( label='请选择您想进行的操作', options=('预测情绪', '调试模型'), index=0, format_func=str, ) if choice == '调试模型': st.sidebar.subheader('模型选择与调参') model = st.sidebar.selectbox( label='请选择您想使用的模型', options=('决策树 (Decision_Tree)', 'k-近邻 (knn)', '支持向量机 (SVM)', '集成学习 (Ensemble Learning)'), index=0, format_func=str, help='目前只提供这四种模型' ) if model == '决策树 (Decision_Tree)': st.subheader('决策树') iterations = st.sidebar.slider("max_depth（最大深度）", 1, 30, 20, 1, help='决策树的最大深度，限制树的生长，防止过拟合.') min_samples_split1 = st.sidebar.slider("min_samples_split（最小分裂样本数）", 2, 50, 2, 1, help='内部节点再划分所需的最小样本数。') min_samples_leaf1 = st.sidebar.slider("min_samples_leaf（最小叶子节点样本数）", 1, 50, 1, 1, help='叶节点所需的最小样本数。') progress_bar = st.empty() with st.spinner('加载中...'): time.sleep(0.5) # 实例决策树 Decision_Tree = DecisionTreeClassifier(criterion='entropy', max_depth=iterations, min_samples_split=min_samples_split1, min_samples_leaf=min_samples_leaf1, random_state=2024) # 训练决策树 Decision_Tree.fit(train_x, train_y) joblib.dump(Decision_Tree, 'D:\social-media-usage-and-emotional-well-being\pythonProject\models\Decision_Tree_model.pickle') # 预测 Decision_Tree_pred = Decision_Tree.predict(test_x) # precision成绩 Decision_Tree_precision = precision_score(test_y, Decision_Tree_pred, average='macro') # recall成绩 Decision_Tree_recall = recall_score(test_y, Decision_Tree_pred, average='micro') # f1成绩 Decision_Tree_f1 = f1_score(test_y, Decision_Tree_pred, average='weighted') # 准确率 Decision_Tree_accuracy = accuracy_score(test_y, Decision_Tree_pred) # 混淆矩阵 conf_matrix = confusion_matrix(test_y, Decision_Tree_pred) # 指标名称列表 metrics = ['Accuracy', 'Precision', 'Recall', 'F1'] # 对应的指标值列表 values = [Decision_Tree_accuracy, Decision_Tree_precision, Decision_Tree_recall, Decision_Tree_f1] # 绘制条形图 fig, ax = plt.subplots() colors = ['lightcoral', 'lightgreen', 'lightskyblue', 'lightsalmon'] ax.bar(metrics, values, color=colors) # 添加标题 ax.set_title("Decision Tree Metrics") # 添加x轴标签 ax.set_xlabel("Metrics") # 添加y轴标签 ax.set_ylabel("Values") ax.set_ylim(0, 1) # 显示图形 # st.pyplot(fig) fig1, ax = plt.subplots() sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='Blues') ax.set_xlabel('Predicted') ax.set_ylabel('Actual') # 将两个图放在一行 col1, col2 = st.columns(2) with col1: st.pyplot(fig) with col2: st.pyplot(fig1) if st.button('显示整个决策树'): with st.spinner('加载中...'): time.sleep(0.5) target_names = list('Age, Daily_Usage_Time(minutes), Posts_Per_Day, Likes_Received_Per_Day, Comments_Received_Per_Day, Messages_Sent_Per_Day, Gender_Male, Gender_Non - binary, Platform_Instagram, Platform_LinkedIn, Platform_Snapchat, Platform_Telegram, Platform_Twitter, Platform_Whatsapp'.split(",")) # 将决策树模型导出为graphviz格式的dot文件 dot_data = export_graphviz(Decision_Tree, out_file=None, feature_names=target_names, class_names=list(set(test_y['Dominant_Emotion'])), filled=True, rounded=True, special_characters=True) # 使用graphviz库的Source函数将dot文件转换为可视化图形 graph = graphviz.Source(dot_data) st.graphviz_chart(dot_data) st.code(''' if model == '决策树 (Decision_Tree)': st.subheader('决策树') iterations = st.sidebar.slider("max_depth（最大深度）", 1, 30, 20, 1, help='决策树的最大深度，限制树的生长，防止过拟合.') min_samples_split1 = st.sidebar.slider("min_samples_split（最小分裂样本数）", 2, 50, 2, 1, help='内部节点再划分所需的最小样本数。') min_samples_leaf1 = st.sidebar.slider("min_samples_leaf（最小叶子节点样本数）", 1, 50, 1, 1, help='叶节点所需的最小样本数。') progress_bar = st.empty() with st.spinner('加载中...'): time.sleep(0.5) # 实例决策树 Decision_Tree = DecisionTreeClassifier(criterion='entropy', max_depth=iterations, min_samples_split=min_samples_split1, min_samples_leaf=min_samples_leaf1, random_state=2024) # 训练决策树 Decision_Tree.fit(train_x, train_y) joblib.dump(Decision_Tree, 'D:\social-media-usage-and-emotional-well-being\pythonProject\models\Decision_Tree_model.pickle') # 预测 Decision_Tree_pred = Decision_Tree.predict(test_x) # precision成绩 Decision_Tree_precision = precision_score(test_y, Decision_Tree_pred, average='macro') # recall成绩 Decision_Tree_recall = recall_score(test_y, Decision_Tree_pred, average='micro') # f1成绩 Decision_Tree_f1 = f1_score(test_y, Decision_Tree_pred, average='weighted') # 准确率 Decision_Tree_accurac