1. Streamlit概述
Streamlit（/ˈstriːm.lɪt/，斯垂姆利特）是一个开源的Python库，专门为机器学习和数据科学团队设计，用于快速构建和共享美观、交互式的数据应用。自2019年发布以来，Streamlit因其简洁性、高效性和强大的功能集，迅速成为数据科学家和工程师创建数据驱动应用的首选工具之一。
1.1 核心设计理念
Streamlit的设计遵循几个关键原则：
- 纯Python开发：无需前端经验，仅用Python即可创建完整的Web应用即时反馈：每次保存代码变更都会实时反映在应用中极简API：提供直观的API，减少样板代码无缝集成：与主流数据科学生态系统(Pandas, NumPy, Matplotlib等)完美兼容
1.2 技术架构
Streamlit采用独特的"脚本即应用"架构：
- 应用代码从上到下执行，每次交互都触发完整脚本重新运行内置智能缓存机制优化性能基于React的前端渲染，提供现代Web体验使用WebSocket实现前后端实时通信
1.3 代码示例
本文中所有代码示例都是完整可运行的Streamlit应用，只需安装Streamlit库即可运行：
pip install streamlit pandas numpy plotly scikit-learn运行示例
streamlit run 示例文件名.py完全可运行的示例
import streamlit as stimport pandas as pdimport numpy as np# 最小化可运行示例st.title("Streamlit 最小示例")st.write("下面是一个简单的DataFrame展示:")# 创建示例数据data = pd.DataFrame({ 'A': np.random.randn(50), 'B': np.random.randn(50)})st.dataframe(data) # 交互式表格st.line_chart(data) # 折线图2. 核心功能与高级特性
2.1 基础组件
import streamlit as stimport pandas as pdimport numpy as np# 文本显示st.title("完整基础组件演示")st.header("这是一个头部")st.subheader("子头部")st.text("固定宽度文本")st.markdown("**Markdown** 支持 *格式化*")st.latex(r"\sqrt{x^2 + y^2}") # LaTeX公式# 数据展示df = pd.DataFrame({ '日期': pd.date_range(start='2023-01-01', periods=10), '值': np.random.randn(10).cumsum()})st.dataframe(df.style.highlight_max(axis=0)) # 交互式DataFramest.table(df.head(3)) # 静态表格# 指标显示st.metric("温度", "25.6 °C", "1.2 °C") # 当前值，变化值# 图表st.line_chart(df.set_index('日期')) # 折线图st.area_chart(df.set_index('日期')) # 面积图st.bar_chart(df.set_index('日期')) # 柱状图2.2 交互组件
import streamlit as stimport pandas as pdst.title("交互组件演示")# 用户输入name = st.text_input("请输入您的姓名", "张三")age = st.slider("选择年龄", 0, 100, 25)options = st.multiselect( "选择兴趣", ["运动", "音乐", "阅读", "旅行", "美食"], ["音乐", "阅读"])# 日期输入appointment = st.date_input("预约日期")time = st.time_input("预约时间")# 颜色选择器color = st.color_picker("选择颜色", "#00f900")# 按钮与操作if st.button("提交信息"): st.success(f""" 已提交信息: 姓名: {name} 年龄: {age} 兴趣: {', '.join(options)} 预约时间: {appointment} {time} 颜色: {color} """)# 文件上传示例uploaded_file = st.file_uploader("上传CSV文件", type=["csv"])if uploaded_file is not None: df = pd.read_csv(uploaded_file) st.write("上传数据预览:") st.dataframe(df.head())2.3 布局与样式
import streamlit as stimport pandas as pdimport numpy as npst.title("布局与样式演示")# 列布局st.header("列布局示例")col1, col2, col3 = st.columns(3)with col1: st.metric("温度", "25°C", "1.2°C")with col2: st.metric("湿度", "65%", "-3%")with col3: st.metric("风速", "12km/h", "0.5km/h")# 标签页st.header("标签页示例")tab1, tab2, tab3 = st.tabs(["原始数据", "可视化", "分析"])with tab1: df = pd.DataFrame(np.random.randn(20, 3), columns=['a', 'b', 'c']) st.dataframe(df)with tab2: st.line_chart(df)with tab3: st.write("描述统计:") st.write(df.describe())# 侧边栏st.sidebar.header("配置选项")n_samples = st.sidebar.number_input("样本数", 100, 10000, 1000)show_raw = st.sidebar.checkbox("显示原始数据", True)color = st.sidebar.selectbox( "图表颜色", ["blue", "red", "green"])# 扩展器with st.expander("点击查看代码"): st.code("""import streamlit as stst.title("Hello World") """)2.4 高级状态管理
2.4.1 复杂会话状态管理
import streamlit as stimport pandas as pdimport numpy as npst.title("高级状态管理示例")# 初始化复杂会话状态if 'app_state' not in st.session_state: st.session_state.app_state = { 'dataframe': pd.DataFrame(np.random.randn(5, 3), columns=['A', 'B', 'C']), 'settings': { 'color_theme': 'dark', 'font_size': 12, 'show_grid': True }, 'history': [] }# 状态修改函数def update_dataframe(): new_data = pd.DataFrame(np.random.randn(5, 3), columns=['A', 'B', 'C']) st.session_state.app_state['history'].append(st.session_state.app_state['dataframe'].copy()) st.session_state.app_state['dataframe'] = new_datadef undo_last_action(): if st.session_state.app_state['history']: st.session_state.app_state['dataframe'] = st.session_state.app_state['history'].pop()# 界面布局col1, col2 = st.columns(2)with col1: st.subheader("当前数据") st.dataframe(st.session_state.app_state['dataframe']) if st.button("生成新数据", on_click=update_dataframe): st.success("数据已更新!") if st.button("撤销操作", on_click=undo_last_action): st.info("已撤销上一步操作")with col2: st.subheader("应用设置") # 动态修改设置 theme = st.selectbox( "颜色主题", ['dark', 'light', 'blue'], index=['dark', 'light', 'blue'].index(st.session_state.app_state['settings']['color_theme']) ) font_size = st.slider( "字体大小", 8, 24, st.session_state.app_state['settings']['font_size'] ) show_grid = st.checkbox( "显示网格线", st.session_state.app_state['settings']['show_grid'] ) if st.button("保存设置"): st.session_state.app_state['settings'].update({ 'color_theme': theme, 'font_size': font_size, 'show_grid': show_grid }) st.success("设置已保存!") st.subheader("当前设置") st.json(st.session_state.app_state['settings'])# 显示操作历史st.subheader("操作历史记录")history_df = pd.DataFrame({ '操作时间': pd.date_range(end=pd.Timestamp.now(), periods=len(st.session_state.app_state['history'])), '数据形状': [f"{df.shape[0]}行×{df.shape[1]}列" for df in st.session_state.app_state['history']]})st.dataframe(history_df)2.4.2 跨页面状态共享
创建两个文件演示跨页面状态：
- 1_page1.py
import streamlit as stst.title("页面1 - 跨页面状态共享")# 初始化共享状态if 'shared_state' not in st.session_state: st.session_state.shared_state = { 'counter': 0, 'messages': [] }# 修改共享状态def increment_counter(): st.session_state.shared_state['counter'] += 1 st.session_state.shared_state['messages'].append( f"页面1增加计数器到{st.session_state.shared_state['counter']}" )st.write("当前计数器:", st.session_state.shared_state['counter'])st.button("增加计数器", on_click=increment_counter)# 添加消息message = st.text_input("添加共享消息")if st.button("提交消息"): if message: st.session_state.shared_state['messages'].append(f"页面1: {message}") st.success("消息已添加!")st.write("共享消息列表:")st.write(st.session_state.shared_state['messages'])- 2_page2.py
import streamlit as stst.title("页面2 - 跨页面状态共享")# 访问共享状态if 'shared_state' not in st.session_state: st.session_state.shared_state = { 'counter': 0, 'messages': [] }# 修改共享状态def decrement_counter(): st.session_state.shared_state['counter'] -= 1 st.session_state.shared_state['messages'].append( f"页面2减少计数器到{st.session_state.shared_state['counter']}" )st.write("当前计数器:", st.session_state.shared_state['counter'])st.button("减少计数器", on_click=decrement_counter)# 添加消息message = st.text_input("添加共享消息")if st.button("提交消息"): if message: st.session_state.shared_state['messages'].append(f"页面2: {message}") st.success("消息已添加!")st.write("共享消息列表:")st.write(st.session_state.shared_state['messages'])2.5 性能优化技术
2.5.1 智能缓存与计算优化
import streamlit as stimport pandas as pdimport numpy as npimport timest.title("性能优化技术示例")# 1. 数据缓存示例@st.cache_datadef load_large_dataset(size): """模拟加载大型数据集""" st.write("正在生成数据... (这只会运行一次)") time.sleep(2) # 模拟耗时操作 data = pd.DataFrame({ 'id': range(size), 'value': np.random.randn(size), 'category': np.random.choice(['A', 'B', 'C', 'D'], size) }) return data# 2. 资源缓存示例@st.cache_resourcedef train_complex_model(data_size): """模拟训练复杂模型""" st.write("正在训练模型... (这只会运行一次)") time.sleep(3) # 模拟耗时操作 return { 'model_type': 'RandomForest', 'data_size': data_size, 'accuracy': np.random.uniform(0.8, 0.95) }# 3. 计算缓存示例@st.cache_datadef expensive_computation(data, param): """模拟耗时计算""" st.write(f"执行计算... (参数: {param})") time.sleep(1) return { 'mean': data['value'].mean(), 'sum': data['value'].sum(), 'param': param }# 用户界面st.header("1. 数据缓存演示")dataset_size = st.slider("选择数据集大小", 1000, 100000, 10000, step=1000)data = load_large_dataset(dataset_size)st.write(f"已加载 {len(data)} 行数据")st.dataframe(data.head())st.header("2. 资源缓存演示")if st.button("训练模型"): model = train_complex_model(dataset_size) st.success(f"模型训练完成! 准确率: {model['accuracy']:.2%}")st.header("3. 计算缓存演示")compute_param = st.slider("计算参数", 1, 10, 5)result = expensive_computation(data, compute_param)st.write("计算结果:", result)# 4. 选择性渲染示例st.header("4. 选择性渲染优化")show_details = st.checkbox("显示详细分析")if show_details: with st.spinner("正在执行详细分析..."): # 模拟耗时分析 time.sleep(2) st.subheader("详细分析结果") # 使用容器避免重复计算 with st.container(): st.write("值分布:") st.bar_chart(data['value'].value_counts()) st.write("按类别分组统计:") st.table(data.groupby('category').agg({ 'value': ['mean', 'std', 'count'] }))else: st.info("勾选上方复选框显示详细分析")# 5. 渐进式加载示例st.header("5. 渐进式加载技术")if st.button("加载大数据集(渐进式)"): progress_bar = st.progress(0) status_text = st.empty() chunks = [] for i in range(5): # 模拟分块加载 status_text.text(f"正在加载第 {i + 1}/5 部分...") time.sleep(0.5) chunk = pd.DataFrame({ 'id': range(i * 2000, (i + 1) * 2000), 'value': np.random.randn(2000) }) chunks.append(chunk) progress_bar.progress((i + 1) / 5) full_data = pd.concat(chunks) st.session_state['big_data'] = full_data status_text.text("加载完成!") st.dataframe(full_data.head())2.6 自定义组件与扩展
2.6.1 创建自定义可视化组件
import streamlit as stimport pandas as pdimport numpy as npfrom streamlit.components.v1 import htmlimport jsonimport randomst.title("自定义组件开发示例")# 1. 使用HTML/JS创建自定义组件st.header("1. 自定义HTML/JS组件")# 简单的D3.js条形图组件def d3_bar_chart(data, width=600, height=400): """创建自定义D3.js条形图组件""" chart_id = f"chart-{random.randint(0, 10000)}" html_code = f""" <div id="{chart_id}"></div> <script src="https://d3js.org/d3.v7.min.js"></script> <script> // 数据 const data = {json.dumps(data)}; // 设置图表尺寸 const width = {width}; const height = {height}; const margin = {{top: 20, right: 30, bottom: 40, left: 40}}; // 创建SVG const svg = d3.select("#{chart_id}") .append("svg") .attr("width", width) .attr("height", height); // 创建比例尺 const x = d3.scaleBand() .domain(data.map(d => d.category)) .range([margin.left, width - margin.right]) .padding(0.1); const y = d3.scaleLinear() .domain([0, d3.max(data, d => d.value)]).nice() .range([height - margin.bottom, margin.top]); // 添加条形 svg.selectAll("rect") .data(data) .join("rect") .attr("x", d => x(d.category)) .attr("y", d => y(d.value)) .attr("width", x.bandwidth()) .attr("height", d => y(0) - y(d.value)) .attr("fill", "steelblue"); // 添加X轴 svg.append("g") .attr("transform", `translate(0,${{height - margin.bottom}})`) .call(d3.axisBottom(x)); // 添加Y轴 svg.append("g") .attr("transform", `translate(${{margin.left}},0)`) .call(d3.axisLeft(y)); </script> """ return html(html_code, width=width, height=height + 50)# 使用自定义组件sample_data = [ {"category": "A", "value": 30}, {"category": "B", "value": 45}, {"category": "C", "value": 25}, {"category": "D", "value": 60}]st.write("自定义D3.js条形图:")d3_bar_chart(sample_data)# 2. 使用第三方组件库st.header("2. 集成第三方组件")# 使用streamlit-aggrid高级表格try: from st_aggrid import AgGrid, GridOptionsBuilder st.write("使用streamlit-aggrid创建高级表格:") df = pd.DataFrame( np.random.randn(20, 5), columns=['A', 'B', 'C', 'D', 'E'] ) gb = GridOptionsBuilder.from_dataframe(df) gb.configure_pagination(enabled=True) gb.configure_side_bar() gb.configure_default_column( editable=True, groupable=True, filterable=True ) grid_options = gb.build() AgGrid(df, gridOptions=grid_options, height=400)except ImportError: st.warning("请先安装st-aggrid: pip install streamlit-aggrid")# 3. 自定义表单组件st.header("3. 自定义表单组件")def custom_slider(label, min_val, max_val, default_val): """创建自定义滑动条组件""" slider_id = f"slider-{random.randint(0, 10000)}" html_code = f""" <div style="margin: 20px 0;"> <label for="{slider_id}" style="display: block; margin-bottom: 8px;">{label}</label> <input type="range" id="{slider_id}" min="{min_val}" max="{max_val}" value="{default_val}" style="width: 100%; height: 10px;" oninput="updateValue(this.value)"> <div id="{slider_id}-value" style="text-align: center; margin-top: 5px;"> {default_val} </div> </div> <script> function updateValue(val) {{ document.getElementById("{slider_id}-value").innerText = val; // 将值发送回Streamlit parent.postMessage({{ isStreamlitMessage: true, type: "slider_change", value: val, id: "{slider_id}" }}, "*"); }} </script> """ # 处理前端消息 component_value = html(html_code, height=80) # 从前端获取值 if st.session_state.get(f"{slider_id}_value"): return st.session_state[f"{slider_id}_value"] return default_val# 使用自定义滑动条st.write("自定义滑动条组件:")custom_value = custom_slider("选择数值", 0, 100, 50)st.write("当前值:", custom_value)# 4. 嵌入外部应用st.header("4. 嵌入外部应用")# 嵌入Plotly编辑器plotly_editor_html = """<div> <iframe src="https://chart-studio.plotly.com/create/?fid=plotly_demo:1" width="100%" height="600" frameborder="0"> </iframe></div>"""st.write("嵌入Plotly图表编辑器:")html(plotly_editor_html, height=650)2.6.2 创建可复用组件
- custom_components.py
from streamlit.components.v1 import htmlclass CustomComponents: @staticmethod def metric_card(title, value, delta=None, delta_color="normal"): """自定义指标卡片组件""" color_map = { "normal": "#1f77b4", "positive": "#2ecc71", "negative": "#e74c3c" } color = color_map.get(delta_color, "#1f77b4") delta_html = f""" <div style="color: {color}; font-size: 14px; margin-top: 4px;"> {delta} </div> """ if delta is not None else "" html_str = f""" <div style=" border: 1px solid #ddd; border-radius: 8px; padding: 20px; margin: 10px 0; background: white; box-shadow: 0 2px 4px rgba(0,0,0,0.1); "> <div style="color: #666; font-size: 14px;">{title}</div> <div style="font-size: 28px; font-weight: bold; margin: 8px 0;">{value}</div> {delta_html} </div> """ return html(html_str) @staticmethod def progress_circle(percentage, size=100, color="#4CAF50"): """圆形进度条组件""" html_str = f""" <div style="position: relative; width: {size}px; height: {size}px; margin: 20px auto;"> <svg width="{size}" height="{size}"> <circle cx="{size // 2}" cy="{size // 2}" r="{size // 2 - 5}" stroke="#eee" stroke-width="10" fill="none" /> <circle cx="{size // 2}" cy="{size // 2}" r="{size // 2 - 5}" stroke="{color}" stroke-width="10" fill="none" stroke-dasharray="{percentage / 100 * 2 * 3.1415 * (size // 2 - 5)} {2 * 3.1415 * (size // 2 - 5)}" stroke-dashoffset="0" transform="rotate(-90 {size // 2} {size // 2})" /> </svg> <div style=" position: absolute; top: 50%; left: 50%; transform: translate(-50%, -50%); font-size: 20px; font-weight: bold; color: {color}; ">{percentage}%</div> </div> """ return html(html_str)- 使用自定义组件
import streamlit as stfrom custom_components import CustomComponentsst.title("可复用自定义组件示例")# 使用指标卡片组件st.header("1. 指标卡片组件")col1, col2, col3 = st.columns(3)with col1: CustomComponents.metric_card("用户数", "1,234", "+12%", "positive")with col2: CustomComponents.metric_card("留存率", "78%", "-3%", "negative")with col3: CustomComponents.metric_card("平均时长", "2.5h")# 使用圆形进度条st.header("2. 圆形进度条组件")progress = st.slider("选择进度", 0, 100, 75)CustomComponents.progress_circle(progress, size=150, color="#3498db")# 动态更新示例if st.button("随机更新"): import random progress = random.randint(0, 100) st.experimental_rerun()3. 专业应用场景
3.1 机器学习模型部署
import streamlit as stimport pandas as pdimport numpy as npfrom sklearn.linear_model import LinearRegression# 模拟训练数据np.random.seed(42)X = np.random.rand(100, 1) * 10y = 2 * X.squeeze() + np.random.randn(100) * 2# 训练简单模型model = LinearRegression()model.fit(X, y)# 创建预测界面st.title("简单线性回归预测器")st.write("""### 房价预测模型使用简单的线性回归模型预测房价""")# 用户输入sqft = st.slider("房屋面积(平方英尺)", 500, 5000, 1500)# 预测prediction = model.predict([[sqft]])st.metric("预测房价", f"${prediction[0]:,.2f}")# 显示模型信息st.subheader("模型信息")st.write(f"斜率(每平方英尺价格): ${model.coef_[0]:.2f}")st.write(f"截距: ${model.intercept_:.2f}")# 显示训练数据st.scatter_chart(pd.DataFrame({'面积': X.squeeze(), '价格': y}))3.2 数据探索与分析
import streamlit as stimport pandas as pdimport numpy as npimport plotly.express as pxst.title("数据探索分析工具")# 生成示例数据@st.cache_datadef load_data(): np.random.seed(42) data = pd.DataFrame({ '日期': pd.date_range('2023-01-01', periods=100), '销售额': np.random.randn(100).cumsum() * 100 + 1000, '客户数': np.random.poisson(50, 100), '产品类别': np.random.choice(['A', 'B', 'C'], 100), '地区': np.random.choice(['东区', '西区', '南区', '北区'], 100) }) return datadf = load_data()# 交互式控件st.sidebar.header("分析选项")show_raw = st.sidebar.checkbox("显示原始数据", False)columns = st.sidebar.multiselect( "选择分析列", df.select_dtypes(include=['number']).columns.tolist(), default=['销售额', '客户数'])group_by = st.sidebar.selectbox("分组依据", ['无', '产品类别', '地区'])# 数据显示if show_raw: st.subheader("原始数据") st.dataframe(df)# 描述统计st.subheader("描述统计")st.write(df[columns].describe())# 可视化st.subheader("数据可视化")if group_by != '无': fig = px.line(df, x='日期', y=columns, color=group_by)else: fig = px.line(df, x='日期', y=columns)st.plotly_chart(fig)# 相关性分析if len(columns) >= 2: st.subheader("相关性分析") st.write(df[columns].corr()) fig = px.scatter_matrix(df, dimensions=columns) st.plotly_chart(fig)3.3 实验参数调优
import streamlit as stimport pandas as pdfrom sklearn.ensemble import RandomForestClassifierfrom sklearn.datasets import make_classificationfrom sklearn.model_selection import train_test_splitst.title("模型参数调优实验")# 生成模拟数据@st.cache_datadef generate_data(): X, y = make_classification( n_samples=1000, n_features=20, n_informative=3, n_classes=2, random_state=42 ) return train_test_split(X, y, test_size=0.2, random_state=42)X_train, X_test, y_train, y_test = generate_data()# 参数配置st.sidebar.header("模型参数")n_estimators = st.sidebar.slider("树的数量", 10, 200, 100)max_depth = st.sidebar.slider("最大深度", 2, 20, 5)min_samples_split = st.sidebar.slider("最小分裂样本数", 2, 20, 2)bootstrap = st.sidebar.checkbox("Bootstrap采样", True)# 训练按钮if st.sidebar.button("训练模型"): with st.spinner("模型训练中..."): model = RandomForestClassifier( n_estimators=n_estimators, max_depth=max_depth, min_samples_split=min_samples_split, bootstrap=bootstrap, random_state=42 ) model.fit(X_train, y_train) train_acc = model.score(X_train, y_train) test_acc = model.score(X_test, y_test) st.success("训练完成!") col1, col2 = st.columns(2) col1.metric("训练准确率", f"{train_acc:.2%}") col2.metric("测试准确率", f"{test_acc:.2%}", f"{(test_acc - train_acc):.2%}")# 特征重要性if 'model' in locals(): st.subheader("特征重要性") importance = pd.DataFrame({ '特征': [f"特征 {i}" for i in range(X_train.shape[1])], '重要性': model.feature_importances_ }).sort_values('重要性', ascending=False) st.bar_chart(importance.set_index('特征'))4. 完整可运行应用示例
import streamlit as stimport pandas as pdimport numpy as npimport plotly.express as pxfrom sklearn.datasets import load_irisfrom sklearn.ensemble import RandomForestClassifierfrom sklearn.model_selection import train_test_split# 应用配置st.set_page_config( page_title="完整数据科学应用", page_icon="🧊", layout="wide")# 加载数据@st.cache_datadef load_data(): iris = load_iris() df = pd.DataFrame( data=np.c_[iris.data, iris.target], columns=iris.feature_names + ['target'] ) df['species'] = df['target'].map({ 0: iris.target_names[0], 1: iris.target_names[1], 2: iris.target_names[2] }) return dfdf = load_data()# 侧边栏st.sidebar.header("配置选项")analysis_type = st.sidebar.radio( "选择分析类型", ["数据探索", "特征工程", "模型训练"])# 主界面st.title("鸢尾花数据分析应用")if analysis_type == "数据探索": st.header("数据探索") # 数据显示选项 show_options = st.expander("数据显示选项") with show_options: rows = st.slider("显示行数", 5, 50, 10) cols = st.multiselect( "选择列", df.columns.tolist(), default=df.columns.tolist()[:4] ) st.dataframe(df[cols].head(rows)) # 可视化 st.header("数据可视化") chart_type = st.selectbox( "选择图表类型", ["散点图", "箱线图", "直方图"] ) x_axis = st.selectbox("X轴", df.columns.tolist()[:4]) y_axis = st.selectbox("Y轴", df.columns.tolist()[:4]) color_by = st.selectbox("颜色依据", ['species', None]) if chart_type == "散点图": fig = px.scatter(df, x=x_axis, y=y_axis, color=color_by) elif chart_type == "箱线图": fig = px.box(df, x=color_by, y=y_axis) else: fig = px.histogram(df, x=x_axis, color=color_by) st.plotly_chart(fig, use_container_width=True)elif analysis_type == "特征工程": st.header("特征工程") st.write("特征相关性矩阵") numeric_cols = df.select_dtypes(include=['float64']).columns corr = df[numeric_cols].corr() st.dataframe(corr.style.background_gradient(cmap='coolwarm')) fig = px.imshow(corr, text_auto=True) st.plotly_chart(fig, use_container_width=True)elif analysis_type == "模型训练": st.header("模型训练") # 模型参数 st.sidebar.subheader("模型参数") n_estimators = st.sidebar.slider("树的数量", 10, 200, 100) max_depth = st.sidebar.selectbox("最大深度", [None, 5, 10, 20], index=1) # 训练测试分割 X = df[df.columns[:4]] y = df['target'] X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42 ) if st.button("训练模型"): with st.spinner("训练中..."): model = RandomForestClassifier( n_estimators=n_estimators, max_depth=max_depth, random_state=42 ) model.fit(X_train, y_train) train_acc = model.score(X_train, y_train) test_acc = model.score(X_test, y_test) st.success("训练完成!") col1, col2 = st.columns(2) col1.metric("训练准确率", f"{train_acc:.2%}") col2.metric("测试准确率", f"{test_acc:.2%}") # 特征重要性 st.subheader("特征重要性") importance = pd.DataFrame({ '特征': X.columns, '重要性': model.feature_importances_ }).sort_values('重要性', ascending=False) st.bar_chart(importance.set_index('特征')) # 预测示例 st.subheader("预测示例") sample = X_test.iloc[:5].copy() sample['预测类别'] = model.predict(X_test.iloc[:5]) sample['真实类别'] = y_test.iloc[:5].values st.dataframe(sample)总结
Streamlit通过其独特的"脚本即应用"哲学，彻底改变了数据科学应用的开发范式。对于专业人士而言，掌握Streamlit意味着能够将分析结果、机器学习模型和数据处理流程快速转化为可交互、可共享的应用，极大提升了工作效率和成果影响力。虽然它在复杂应用场景下存在一定限制，但其简洁性、Python原生支持和活跃的社区使其成为数据科学工作流中不可或缺的工具。
