import streamlit as st import pandas as pd import numpy as np from agent import ChurnPredictionAgent, CustomerData # 设置页面标题和布局 st.set_page_config( page_title="客户流失预测系统", page_icon="📊", layout="wide" ) # 页面标题 st.title("📊 客户流失预测与行动建议系统") # 创建Agent实例 agent = ChurnPredictionAgent() # 侧边栏：客户信息输入 st.sidebar.header("客户信息输入") # 客户信息表单 with st.sidebar.form("customer_form"): # 基本信息 col1, col2 = st.columns(2) with col1: gender = st.selectbox("性别", ["Male", "Female"]) SeniorCitizen = st.selectbox("是否为老年人", [0, 1]) Partner = st.selectbox("是否有伴侣", ["Yes", "No"]) Dependents = st.selectbox("是否有家属", ["Yes", "No"]) tenure = st.number_input("在网时长（月）", min_value=0, max_value=100, value=12) with col2: PhoneService = st.selectbox("是否开通电话服务", ["Yes", "No"]) MultipleLines = st.selectbox("是否开通多条线路", ["Yes", "No", "No phone service"]) InternetService = st.selectbox("网络服务类型", ["DSL", "Fiber optic", "No"]) OnlineSecurity = st.selectbox("是否开通在线安全服务", ["Yes", "No", "No internet service"]) OnlineBackup = st.selectbox("是否开通在线备份服务", ["Yes", "No", "No internet service"]) # 服务信息 col3, col4 = st.columns(2) with col3: DeviceProtection = st.selectbox("是否开通设备保护服务", ["Yes", "No", "No internet service"]) TechSupport = st.selectbox("是否开通技术支持服务", ["Yes", "No", "No internet service"]) StreamingTV = st.selectbox("是否开通流媒体电视服务", ["Yes", "No", "No internet service"]) StreamingMovies = st.selectbox("是否开通流媒体电影服务", ["Yes", "No", "No internet service"]) with col4: Contract = st.selectbox("合同类型", ["Month-to-month", "One year", "Two year"]) PaperlessBilling = st.selectbox("是否使用无纸化账单", ["Yes", "No"]) PaymentMethod = st.selectbox("支付方式", [ "Electronic check", "Mailed check", "Bank transfer (automatic)", "Credit card (automatic)" ]) MonthlyCharges = st.number_input("月费用", min_value=0.0, max_value=200.0, value=50.0, step=0.01) TotalCharges = st.number_input("总费用", min_value=0.0, max_value=10000.0, value=600.0, step=0.01) # 提交按钮 submit_button = st.form_submit_button("🚀 预测流失风险") # 主内容区 if submit_button: # 创建CustomerData实例 customer_data = CustomerData( gender=gender, SeniorCitizen=SeniorCitizen, Partner=Partner, Dependents=Dependents, tenure=tenure, PhoneService=PhoneService, MultipleLines=MultipleLines, InternetService=InternetService, OnlineSecurity=OnlineSecurity, OnlineBackup=OnlineBackup, DeviceProtection=DeviceProtection, TechSupport=TechSupport, StreamingTV=StreamingTV, StreamingMovies=StreamingMovies, Contract=Contract, PaperlessBilling=PaperlessBilling, PaymentMethod=PaymentMethod, MonthlyCharges=MonthlyCharges, TotalCharges=TotalCharges ) # 使用ML预测工具 with st.spinner("🔄 正在预测流失风险..."): prediction_result = agent.predict_churn(customer_data) # 显示预测结果 st.header("📋 预测结果") col1, col2 = st.columns(2) with col1: st.subheader("客户基本信息") info_df = pd.DataFrame({ "属性": ["性别", "是否为老年人", "是否有伴侣", "是否有家属", "在网时长（月）"], "值": [gender, SeniorCitizen, Partner, Dependents, tenure] }) st.dataframe(info_df, use_container_width=True, hide_index=True) with col2: st.subheader("服务信息") service_df = pd.DataFrame({ "属性": ["合同类型", "网络服务类型", "支付方式", "月费用", "总费用"], "值": [Contract, InternetService, PaymentMethod, MonthlyCharges, TotalCharges] }) st.dataframe(service_df, use_container_width=True, hide_index=True) # 预测结果卡片 st.subheader("🎯 流失预测") col1, col2, col3 = st.columns(3) with col1: st.metric( label="预测结果", value="会流失" if prediction_result.prediction == 1 else "不会流失", delta="高风险" if prediction_result.prediction == 1 else "低风险", delta_color="inverse" ) with col2: st.metric( label="流失概率", value=f"{prediction_result.probability:.2%}", delta=f"{prediction_result.probability:.2%}", delta_color="inverse" ) with col3: st.metric( label="使用模型", value=prediction_result.model_used.upper(), delta="LightGBM", delta_color="off" ) # 行动建议 st.header("💡 行动建议") with st.spinner("🔄 正在生成行动建议..."): suggestions = agent.get_action_suggestions( customer_id="CUST-" + np.random.choice(1000, size=1)[0].astype(str), prediction=prediction_result.prediction, probability=prediction_result.probability, customer_data=customer_data ) # 显示行动建议 st.subheader("📋 个性化行动建议") for i, suggestion in enumerate(suggestions.suggestions, 1): with st.expander(f"建议 {i}"): st.write(suggestion) # 数据可视化 st.header("📊 数据可视化") # 流失概率仪表盘 st.subheader("流失概率仪表盘") # 创建仪表盘 col1, col2 = st.columns(2) with col1: # 流失概率图表（使用Streamlit内置的进度条） st.subheader(f"流失概率: {prediction_result.probability:.2%}") # 进度条显示流失概率 st.progress(prediction_result.probability, text=f"流失概率: {prediction_result.probability:.2%}") # 风险等级 if prediction_result.probability < 0.3: risk_level = "低风险" risk_color = "green" elif prediction_result.probability < 0.7: risk_level = "中风险" risk_color = "yellow" else: risk_level = "高风险" risk_color = "red" st.markdown(f"**风险等级**: {risk_level}", unsafe_allow_html=True) with col2: # 客户特征重要性 st.subheader("客户特征分析") # 示例特征重要性数据（实际应用中应从模型获取） feature_importance = { "合同类型": 0.25, "网络服务类型": 0.20, "在网时长": 0.15, "月费用": 0.12, "是否开通技术支持": 0.10, "支付方式": 0.08, "是否开通在线安全服务": 0.05, "是否有伴侣": 0.03, "是否有家属": 0.02 } feature_df = pd.DataFrame({ "特征": list(feature_importance.keys()), "重要性": list(feature_importance.values()) }).sort_values(by="重要性", ascending=False) st.bar_chart(feature_df.set_index("特征"), use_container_width=True, color="#1f77b4") # 系统信息 st.header("ℹ️ 系统信息") col1, col2 = st.columns(2) with col1: st.subheader("模型性能") st.markdown("- **模型类型**: LightGBM") st.markdown("- **ROC-AUC**: 0.8352") st.markdown("- **F1分数**: 0.5731") st.markdown("- **训练样本数**: 7043") with col2: st.subheader("系统功能") st.markdown("✅ 客户流失预测") st.markdown("✅ 个性化行动建议") st.markdown("✅ 数据可视化分析") st.markdown("✅ 交互式用户界面") else: # 初始页面 st.info("请在左侧填写客户信息，点击'🚀 预测流失风险'按钮开始预测") # 系统介绍 st.header("ℹ️ 系统介绍") st.markdown(""" 本系统基于机器学习和AI Agent技术，实现了客户流失预测与行动建议的闭环。 ### 系统功能 - **客户流失预测**: 使用LightGBM模型预测客户流失概率 - **个性化行动建议**: 根据客户特征生成可执行的行动建议 - **数据可视化分析**: 直观展示预测结果和客户特征重要性 ### 技术栈 - **机器学习**: LightGBM、Logistic Regression - **数据处理**: Polars、Pandas - **AI Agent**: Pydantic - **Web框架**: Streamlit ### 如何使用 1. 在左侧填写客户信息 2. 点击'🚀 预测流失风险'按钮 3. 查看预测结果和行动建议 4. 分析客户特征重要性 """)