import streamlit as st
import pandas as pd
import numpy as np
import joblib
import os
import time
import matplotlib.pyplot as plt
import matplotlib as mpl
import matplotlib.font_manager as fm
import seaborn as sns
from pyspark.sql import SparkSession
from pyspark.ml.feature import VectorAssembler, StandardScaler
from pyspark.ml.classification import Logistic
Regression, DecisionTreeClassifier, Random
ForestClassifier
from pyspark.ml.evaluation import BinaryClassificationEvaluator
from pyspark.ml.tuning import ParamGridBuilder, CrossValidator
from pyspark.sql.functions import when, col
from sklearn.metrics import classification_report, confusion_matrix
import warnings
import dask.dataframe as dd
from dask.diagnostics import ProgressBar
from dask_ml.preprocessing import StandardScaler as DaskStandardScaler
import tempfile
warnings.filterwarnings("ignore")
plt.rcParams[&#
39;font.sans-serif&#
39;] = [&#
39;SimHei&#
39;]
plt.rcParams[&#
39;axes.unicode_minus&#
39;] = False
# 页面设置
st.set_page_config(
page_title="单宽转融用户预测系统",
page_icon="📶",
layout="wide",
initial_sidebar_state="expanded"
)
# 自定义CSS样式
st.markdown("""
<style>
.stApp {
background:
linear-gradient(1
35deg, #f5f7fa 0%, #e4edf5 100%);
font-family: &#
39;Helvetica Neue&#
39;, Arial, sans-serif;
}
.header {
background:
linear-gradient(90deg, #2c
3e50 0%, #4a6491 100%);
color: white;
padding: 1.5rem;
border-radius: 0.75rem;
box-shadow: 0 4px 12px rgba(0,0,0,0.1);
margin-bottom: 2rem;
}
.card {
background: white;
border-radius: 0.75rem;
padding: 1.5rem;
margin-bottom: 1.5rem;
box-shadow: 0 4px 12px rgba(0,0,0,0.08);
transition: trans
form 0.
3s ease;
}
.card:hover {
trans
form: translateY(-5px);
box-shadow: 0 6px 16px rgba(0,0,0,0.12);
}
.stButton button {
background:
linear-gradient(90deg, #
3498db 0%, #1a5276 100%) !important;
color: white !important;
border: none !important;
border-radius: 0.5rem;
padding: 0.75rem 1.5rem;
font-size: 1rem;
font-weight: 600;
transition: all 0.
3s ease;
width: 100%;
}
.stButton button:hover {
trans
form: scale(1.05);
box-shadow: 0 4px 8px rgba(52, 152, 219, 0.4);
}
.feature-box {
background:
linear-gradient(1
35deg, #e
3f2fd 0%, #bbdefb 100%);
border-radius: 0.75rem;
padding: 1.5rem;
margin-bottom: 1.5rem;
}
.result-box {
background:
linear-gradient(1
35deg, #e8f5e9 0%, #c8e6c9 100%);
border-radius: 0.75rem;
padding: 1.5rem;
margin-top: 1.5rem;
}
.model-box {
background:
linear-gradient(1
35deg, #fff
3e0 0%, #ffe0b2 100%);
border-radius: 0.75rem;
padding: 1.5rem;
margin-top: 1.5rem;
}
.stProgress > div > div > div {
background:
linear-gradient(90deg, #2ecc71 0%, #27ae60 100%) !important;
}
.metric-card {
background: white;
border-radius: 0.75rem;
padding: 1rem;
text-align: center;
box-shadow: 0 4px 8px rgba(0,0,0,0.06);
}
.metric-value {
font-size: 1.8rem;
font-weight: 700;
color: #2c
3e50;
}
.metric-label {
font-size: 0.9rem;
color: #7f8c8d;
margin-top: 0.5rem;
}
.highlight {
background:
linear-gradient(90deg, #ffeb
3b 0%, #fbc02d 100%);
padding: 0.2rem 0.5rem;
border-radius: 0.25rem;
font-weight: 600;
}
.stDataFrame {
border-radius: 0.75rem;
box-shadow: 0 4px 8px rgba(0,0,0,0.06);
}
.risk-high {
background-color: #ffcdd2 !important;
color: #c62828 !important;
font-weight: 700;
}
.risk-medium {
background-color: #fff9c4 !important;
color: #f57f17 !important;
font-weight: 600;
}
.risk-low {
background-color: #c8e6c9 !important;
color: #
388e
3c !important;
}
</style>
""", unsafe_allow_html=True)
def preprocess_data(ddf):
"""
使用Dask进行数据预处理，支持大数据处理
参数:
ddf (dask.DataFrame): 原始数据
返回:
processed_ddf (dask.DataFrame): 处理后的数据
feature_cols (list): 特征列名列表
"""
# 创建副本以避免修改原始数据
processed_ddf = ddf.copy()
# 1. 删除无意义特征
drop_cols = [&#
39;BIL_MONTH&#
39;, &#
39;ASSET_ROW_ID&#
39;, &#
39;CCUST_ROW_ID&#
39;, &#
39;BELONG_CITY&#
39;,
&#
39;MKT_CHANNEL_NAME&#
39;, &#
39;MKT_CHANNEL_SUB_NAME&#
39;, &#
39;PREPARE_FLG&#
39;,
&#
39;SERV_START_DT&#
39;, &#
39;COMB_STAT_NAME&#
39;, &#
39;FIBER_ACCESS_CATEGORY&#
39;]
# 检查并删除存在的列
existing_cols = [col
for col in drop_cols if col in processed_ddf.columns]
processed_ddf = processed_ddf.drop(columns=existing_cols)
# 2. 处理缺失值
# 数值型特征用均值填充
numeric_cols = processed_ddf.select_dtypes(include=[np.number]).columns.tolist()
if &#
39;is_rh_next&#
39; in numeric_cols:
numeric_cols.remove(&#
39;is_rh_next&#
39;)
# 计算均值（Dask需要先persist）
with ProgressBar():
means = processed_ddf[numeric_cols].mean().compute()
# 填充缺失值
for col in numeric_cols:
processed_ddf[col] = processed_ddf[col].fillna(means[col])
# 类别型特征用"Unknown"填充
object_cols = processed_ddf.select_dtypes(include=[&#
39;object&#
39;]).columns.tolist()
for col in object_cols:
processed_ddf[col] = processed_ddf[col].fillna("Unknown")
#
3. 离散特征编码
# 对二元特征进行简单映射
binary_cols = [&#
39;IF_YHTS&#
39;, &#
39;is_kdts&#
39;, &#
39;is_itv_up&#
39;, &#
39;is_mobile_up&#
39;, &#
39;if_zzzw_up&#
39;]
for col in binary_cols:
if col in processed_ddf.columns:
processed_ddf[col] = processed_ddf[col].map({&#
39;否&#
39;: 0, &#
39;是&#
39;: 1, 0: 0, 1: 1, &#
39;Unknown&#
39;: -1})
# 对性别进行映射
if &#
39;GENDER&#
39; in processed_ddf.columns:
gender_mapping = {&#
39;男&#
39;: 0, &#
39;女&#
39;: 1, &#
39;Unknown&#
39;: -1}
processed_ddf[&#
39;GENDER&#
39;] = processed_ddf[&#
39;GENDER&#
39;].map(gender_mapping)
# 4. 用户星级映射
if &#
39;MKT_STAR_GRADE_NAME&#
39; in processed_ddf.columns:
star_mapping = {
&#
39;五星级&#
39;: 5,
&#
39;四星级&#
39;: 4,
&#
39;三星级&#
39;:
3,
&#
39;二星级&#
39;: 2,
&#
39;一星级&#
39;: 1,
&#
39;Unknown&#
39;: 0
}
processed_ddf[&#
39;MKT_STAR_GRADE_NAME&#
39;] = processed_ddf[&#
39;MKT_STAR_GRADE_NAME&#
39;].map(star_mapping)
# 5. 特征工程
# 计算消费比率（套餐价格/出账金额）
if &#
39;PROM_AMT&#
39; in processed_ddf.columns and &#
39;STMT_AMT&#
39; in processed_ddf.columns:
processed_ddf[&#
39;CONSUMPTION_RATIO&#
39;] = processed_ddf[&#
39;PROM_AMT&#
39;] / (processed_ddf[&#
39;STMT_AMT&#
39;] + 1)
# 计算流量使用密度（下载流量/在网天数）
if &#
39;DWN_VOL&#
39; in processed_ddf.columns and &#
39;ONLINE_
DAY&#
39; in processed_ddf.columns:
processed_ddf[&#
39;TRAFFIC_DENSITY&#
39;] = processed_ddf[&#
39;DWN_VOL&#
39;] / (processed_ddf[&#
39;ONLINE_
DAY&#
39;] + 1)
# 是否有终端设备
if &#
39;TERM_CNT&#
39; in processed_ddf.columns:
processed_ddf[&#
39;HAS_TERMINAL&#
39;] = (processed_ddf[&#
39;TERM_CNT&#
39;] > 0).astype(int)
# 6. 标准化处理
scaler = DaskStandardScaler()
numeric_cols_
for_scaling = list(set(numeric_cols) - set([&#
39;is_rh_next&#
39;]))
if len(numeric_cols_
for_scaling) > 0:
processed_ddf[numeric_cols_
for_scaling] = scaler.fit_trans
form(processed_ddf[numeric_cols_
for_scaling])
# 保存特征列
feature_cols = [col
for col in processed_ddf.columns if col != &#
39;is_rh_next&#
39;]
return processed_ddf, feature_cols, means, numeric_cols_
for_scaling, scaler
def create_spark_session():
"""创建或获取现有的Spark会话"""
return SparkSession.builder \
.appName("SingleToMelt
UserPrediction") \
.config("spark.sql.shuffle.partitions", "8") \
.config("spark.driver.memory", "8g") \
.config("spark.executor.memory", "8g") \
.getOrCreate()
def train_
models(spark_df, feature_cols):
"""
使用Spark训练多个模型并评估性能
参数:
spark_df (pyspark.sql.DataFrame): 处理后的数据
feature_cols (list): 特征列名列表
返回:
results (dict): 包含训练好的模型及其性能指标
"""
# 初始化Spark会话
spark = create_spark_session()
# 将特征列组合为特征向量
assembler = VectorAssembler(inputCols=feature_cols, outputCol="rawFeatures")
assembled_df = assembler.trans
form(spark_df)
# 标准化特征
scaler = StandardScaler(inputCol="rawFeatures", outputCol="features")
scaler_model = scaler.fit(assembled_df)
scaled_df = scaler_model.trans
form(assembled_df)
# 划分训练集和测试集
train_df, test_df = scaled_df.randomSplit([0.8, 0.2], seed=42)
# 定义评估器
lr = Logistic
Regression(featuresCol="features", labelCol="is_rh_next")
dt = DecisionTreeClassifier(featuresCol="features", labelCol="is_rh_next")
rf = Random
ForestClassifier(featuresCol="features", labelCol="is_rh_next", numTrees=10)
# 定义参数网格
lr_param_grid = ParamGridBuilder() \
.addGrid(lr.regParam, [0.01, 0.1]) \
.addGrid(lr.elasticNetParam, [0.0, 0.5]) \
.build()
dt_param_grid = ParamGridBuilder() \
.addGrid(dt.maxDepth, [5, 10]) \
.addGrid(dt.minInstancesPerNode, [10, 20]) \
.build()
rf_param_grid = ParamGridBuilder() \
.addGrid(rf.numTrees, [10, 20]) \
.addGrid(rf.maxDepth, [5, 10]) \
.build()
# 定义交叉验证器
evaluator = BinaryClassificationEvaluator(labelCol="is_rh_next", metricName="areaUnderROC")
lr_cv = CrossValidator(estimator=lr,
estimatorParamMaps=lr_param_grid,
evaluator=evaluator,
numFolds=
3)
dt_cv = CrossValidator(estimator=dt,
estimatorParamMaps=dt_param_grid,
evaluator=evaluator,
numFolds=
3)
rf_cv = CrossValidator(estimator=rf,
estimatorParamMaps=rf_param_grid,
evaluator=evaluator,
numFolds=
3)
# 训练模型
results = {}
# 逻辑回归
with st.spinner("正在训练逻辑回归模型..."):
lr_model = lr_cv.fit(train_df)
lr_predictions = lr_model.trans
form(test_df)
lr_auc = evaluator.evaluate(lr_predictions)
lr_accuracy = lr_predictions.filter(lr_predictions.is_rh_next == lr_predictions.prediction).count() / test_df.count()
results["logistic_
regression"] = {
"model": lr_model,
"auc": lr_auc,
"accuracy": lr_accuracy,
"best_params": lr_model.bestModel._java_obj.parent().extractParamMap()
}
# 决策树
with st.spinner("正在训练决策树模型..."):
dt_model = dt_cv.fit(train_df)
dt_predictions = dt_model.trans
form(test_df)
dt_auc = evaluator.evaluate(dt_predictions)
dt_accuracy = dt_predictions.filter(dt_predictions.is_rh_next == dt_predictions.prediction).count() / test_df.count()
results["decision_tree"] = {
"model": dt_model,
"auc": dt_auc,
"accuracy": dt_accuracy,
"best_params": dt_model.bestModel._java_obj.parent().extractParamMap(),
"feature_importances": dt_model.bestModel.featureImportances.toArray().tolist()
}
# 随机森林
with st.spinner("正在训练随机森林模型..."):
rf_model = rf_cv.fit(train_df)
rf_predictions = rf_model.trans
form(test_df)
rf_auc = evaluator.evaluate(rf_predictions)
rf_accuracy = rf_predictions.filter(rf_predictions.is_rh_next == rf_predictions.prediction).count() / test_df.count()
results["random_
forest"] = {
"model": rf_model,
"auc": rf_auc,
"accuracy": rf_accuracy,
"best_params": rf_model.bestModel._java_obj.parent().extractParamMap(),
"feature_importances": rf_model.bestModel.featureImportances.toArray().tolist()
}
return results
# 标题区域
st.markdown("""
<div class="header">
<h1 style=&#
39;text-align: center; margin: 0;&#
39;>单宽转融用户预测系统</h1>
<p style=&#
39;text-align: center; margin: 0.5rem 0 0; font-size: 1.1rem;&#
39;>基于大数据挖掘的精准营销分析平台</p>
</div>
""", unsafe_allow_html=True)
# 页面布局
col1, col2 = st.columns([1, 1.5])
# 左侧区域 - 图片和简介
with col1:
st.markdown("""
<div class="card">
<h
3 style=&#
39;text-align: center; color: #2c
3e50;&#
39;>精准营销系统</h
3>
<p style=&#
39;text-align: center;&#
39;>利用先进数据挖掘技术识别潜在融合套餐用户</p>
</div>
""", unsafe_allow_html=True)
# 使用在线图片作为占位符
st.image("https://images.unsplash.com/photo-1550751822256-00808c92fc8d?ixlib=rb-4.0.
3&ixid=M
3wxMjA
3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%
3D%
3D&auto=
format&fit=crop&w=1200&q=80",
caption="精准营销示意图",
use_column_width=True)
st.markdown("""
<div class="feature-box">
<h4>📈 系统功能</h4>
<ul>
<li>用户转化预测</li>
<li>多模型对比分析</li>
<li>特征重要性分析</li>
<li>可视化数据洞察</li>
</ul>
</div>
""", unsafe_allow_html=True)
# 右侧区域 - 功能选择
with col2:
st.markdown("""
<div class="card">
<h
3 style=&#
39;color: #2c
3e50;&#
39;>请选择操作类型</h
3>
<p>您可以选择训练新模型或查看现有模型分析结果</p>
</div>
""", unsafe_allow_html=True)
# 功能选择
option = st.radio("",
["🚀 训练新模型 - 使用新数据训练预测模型",
"🔍 模型分析 - 查看现有模型的分析结果"],
index=0,
label_visibility="hidden")
# 模型训练部分
if "训练新模型" in option:
st.markdown("""
<div class="model-box">
<h4>模型训练</h4>
<p>上传训练数据并训练新的预测模型</p>
</div>
""", unsafe_allow_html=True)
# 上传训练数据
train_file = st.file_uploader("上传训练数据 (CSV格式)", type=["csv"], accept_
multiple_files=False)
if train_file is not None:
try:
# 使用Dask读取大文件
with tempfile.TemporaryDirectory() as tmpdir:
tmp_path = os.path.join(tmpdir, "large_file.csv")
with open(tmp_path, "wb") as f:
f.write(train_file.read())
# 分块读取配置
chunksize = 10**6 # 1MB每块
raw_ddf = dd.read_csv(tmp_path, blocksize=chunksize,
assume_missing=True,
dtype={&#
39;is_rh_next&#
39;: &#
39;int8&#
39;})
# 显示数据预览
with st.expander("数据预览", expanded=True):
st.dataframe(raw_ddf.head(1000, npartitions=1).compute())
col1, col2 = st.columns(2)
col1.metric("总样本数", f"{raw_ddf.shape[0].compute():,}")
col2.metric("特征数量", raw_ddf.shape[1].compute() - 1)
# 检查目标变量是否存在
if &#
39;is_rh_next&#
39; not in raw_ddf.columns:
st.warning("⚠️ 注意：未找到目标变量 &#
39;is_rh_next&#
39;")
# 数据预处理按钮
if st.button("开始数据预处理",
use_container_width=True):
with st.spinner("正在进行数据预处理，请稍候..."):
processed_ddf, feature_cols, means, numeric_cols_
for_scaling, scaler = preprocess_data(raw_ddf)
# 保存预处理参数
preprocessor_params = {
&#
39;means&#
39;: means,
&#
39;numeric_cols_
for_scaling&#
39;: numeric_cols_
for_scaling,
&#
39;scaler&#
39;: scaler,
&#
39;feature_cols&#
39;: feature_cols
}
joblib.dump(preprocessor_params, &#
39;preprocessor_params.pkl&#
39;)
# 保存处理后的数据
processed_ddf.to_csv(&#
39;processed_data_*.csv&#
39;, index=False)
st.success("✅ 数据预处理完成！")
# 显示处理后的数据统计
st.subheader("数据质量检查")
col1, col2 = st.columns(2)
col1.write("缺失值统计:")
col1.write(processed_df.isnull().sum()[processed_df.isnull().sum() > 0])
col2.write("异常值检测:")
for col in processed_df[feature_cols]:
q1 = processed_df[col].quantile(0.25)
q
3 = processed_df[col].quantile(0.75)
iqr = q
3 - q1
outlier_count = ((processed_df[col] < (q1 - 1.5 * iqr)) |
(processed_df[col] > (q
3 + 1.5 * iqr))).sum()
if outlier_count > 0:
col2.write(f"{col}: {outlier_count} 个异常值")
# 可视化关键特征分布
st.subheader("关键特征分布")
fig, axes = plt.subplots(2, 2, figsize=(12, 10))
sns.histplot(processed_df[&#
39;AGE&#
39;], ax=axes[0, 0], kde=True)
sns.histplot(processed_df[&#
39;ONLINE_
DAY&#
39;], ax=axes[0, 1], kde=True)
sns.histplot(processed_df[&#
39;PROM_AMT&#
39;], ax=axes[1, 0], kde=True)
sns.histplot(processed_df[&#
39;DWN_VOL&#
39;], ax=axes[1, 1], kde=True)
plt.tight_layout()
st.pyplot(fig)
# 目标变量分布
st.subheader("目标变量分布")
fig, ax = plt.subplots(figsize=(6, 4))
sns.countplot(x=&#
39;is_rh_next&#
39;, data=processed_df, ax=ax)
ax.set_xlabel("是否转化 (0=未转化, 1=转化)")
ax.set_ylabel("用户数量")
ax.set_title("用户转化分布")
st.pyplot(fig)
# 特征与目标变量相关性
st.subheader("特征与转化的相关性")
correlation = processed_df[feature_cols + [&#
39;is_rh_next&#
39;]].corr()[&#
39;is_rh_next&#
39;].sort_values(ascending=False)
fig, ax = plt.subplots(figsize=(10, 6))
sns.barplot(x=correlation.values, y=correlation.index, ax=ax)
ax.set_title("特征与转化的相关性")
st.pyplot(fig)
# 模型训练按钮
if st.button("开始模型训练",
use_container_width=True):
if not os.path.exists(&#
39;processed_data.csv&#
39;):
st.error("请先进行数据预处理")
else:
# 加载处理后的数据
processed_df = pd.read_csv(&#
39;processed_data.csv&#
39;)
preprocessor_params = joblib.load(&#
39;preprocessor_params.pkl&#
39;)
feature_cols = preprocessor_params[&#
39;feature_cols&#
39;]
# 转换为Spark DataFrame
spark = create_spark_session()
spark_df = spark.createDataFrame(processed_df)
# 训练模型
with st.spinner("正在训练模型，请耐心等待..."):
results = train_
models(spark_df, feature_cols)
# 保存模型结果
joblib.dump(results, &#
39;model_results.pkl&#
39;)
st.success("🎉 模型训练完成！")
# 显示模型比较
st.subheader("模型性能对比")
model_per
formance = pd.DataFrame({
&#
39;模型&#
39;: [&#
39;逻辑回归&#
39;, &#
39;决策树&#
39;, &#
39;随机森林&#
39;],
&#
39;准确率&#
39;: [results[&#
39;logistic_
regression&#
39;][&#
39;accuracy&#
39;],
results[&#
39;decision_tree&#
39;][&#
39;accuracy&#
39;],
results[&#
39;random_
forest&#
39;][&#
39;accuracy&#
39;]],
&#
39;AUC&#
39;: [results[&#
39;logistic_
regression&#
39;][&#
39;auc&#
39;],
results[&#
39;decision_tree&#
39;][&#
39;auc&#
39;],
results[&#
39;random_
forest&#
39;][&#
39;auc&#
39;]]
}).sort_values(&#
39;AUC&#
39;, ascending=False)
st.table(model_per
formance.style.
format({
&#
39;准确率&#
39;: &#
39;{:.2%}&#
39;,
&#
39;AUC&#
39;: &#
39;{:.4f}&#
39;
}))
# 最佳模型特征重要性
best_model_name = model_per
formance.iloc[0][&#
39;模型&#
39;]
model_map = {
&#
39;逻辑回归&#
39;: &#
39;logistic_
regression&#
39;,
&#
39;决策树&#
39;: &#
39;decision_tree&#
39;,
&#
39;随机森林&#
39;: &#
39;random_
forest&#
39;
}
best_model_key = model_map[best_model_name]
best_model = results[best_model_key][&#
39;model&#
39;].bestModel
st.subheader(f"最佳模型 ({best_model_name}) 分析")
if best_model_key in [&#
39;decision_tree&#
39;, &#
39;random_
forest&#
39;]:
feature_importances = results[best_model_key][&#
39;feature_importances&#
39;]
importance_df = pd.DataFrame({
&#
39;特征&#
39;: feature_cols,
&#
39;重要性&#
39;: feature_importances
}).sort_values(&#
39;重要性&#
39;, ascending=False).head(10)
fig, ax = plt.subplots(figsize=(10, 6))
sns.barplot(x=&#
39;重要性&#
39;, y=&#
39;特征&#
39;, data=importance_df, ax=ax)
ax.set_title(&#
39;Top 10 重要特征&#
39;)
st.pyplot(fig)
# 显示最佳模型参数
st.subheader("最佳模型参数")
params = results[best_model_key][&#
39;best_params&#
39;]
param_table = pd.DataFrame({
&#
39;参数&#
39;: [str(param.name)
for param in params.keys()],
&#
39;值&#
39;: [str(value)
for value in params.values()]
})
st.table(param_table)
except Exception as e:
st.error(f"数据处理错误: {str(e)}")
# 模型分析部分
else:
st.markdown("""
<div class="model-box">
<h4>模型分析</h4>
<p>查看已有模型的详细分析结果</p>
</div>
""", unsafe_allow_html=True)
if not os.path.exists(&#
39;model_results.pkl&#
39;):
st.info("ℹ️ 当前没有可用模型。请先进行模型训练以生成分析报告。")
# 显示可执行的操作
col1, col2 = st.columns(2)
with col1:
st.markdown("""
<div class="card">
<h4>🛠️ 数据准备</h4>
<p>确保您已准备好符合要求的数据文件，包含所有必要的特征字段。</p>
</div>
""", unsafe_allow_html=True)
with col2:
st.markdown("""
<div class="card">
<h4>🧠 模型训练</h4>
<p>切换到“训练新模型”选项卡，上传您的数据并开始训练过程。</p>
</div>
""", unsafe_allow_html=True)
else:
# 加载模型结果
results = joblib.load(&#
39;model_results.pkl&#
39;)
preprocessor_params = joblib.load(&#
39;preprocessor_params.pkl&#
39;)
feature_cols = preprocessor_params[&#
39;feature_cols&#
39;]
# 模型选择
model_choice = st.selectbox(
"选择要分析的模型",
("逻辑回归", "决策树", "随机森林")
)
model_map = {
&#
39;逻辑回归&#
39;: &#
39;logistic_
regression&#
39;,
&#
39;决策树&#
39;: &#
39;decision_tree&#
39;,
&#
39;随机森林&#
39;: &#
39;random_
forest&#
39;
}
model_key = model_map[model_choice]
# 显示模型基本信息
model_info = results[model_key]
st.markdown(f"""
<div class="card">
<h
3>{model_choice}</h
3>
<p><strong>AUC得分:</strong> {model_info[&#
39;auc&#
39;]:.4f}</p>
<p><strong>准确率:</strong> {model_info[&#
39;accuracy&#
39;]:.2%}</p>
</div>
""", unsafe_allow_html=True)
# 显示参数详情
with st.expander("模型参数详情", expanded=False):
params = model_info[&#
39;best_params&#
39;]
param_table = pd.DataFrame({
&#
39;参数&#
39;: [str(param.name)
for param in params.keys()],
&#
39;值&#
39;: [str(value)
for value in params.values()]
})
st.table(param_table)
# 如果存在特征重要性信息
if model_key in [&#
39;decision_tree&#
39;, &#
39;random_
forest&#
39;]:
feature_importances = model_info[&#
39;feature_importances&#
39;]
importance_df = pd.DataFrame({
&#
39;特征&#
39;: feature_cols,
&#
39;重要性&#
39;: feature_importances
}).sort_values(&#
39;重要性&#
39;, ascending=False)
st.subheader("特征重要性分析")
# Top 10 重要特征
top_features = importance_df.head(10)
fig, ax = plt.subplots(figsize=(10, 6))
sns.barplot(x=&#
39;重要性&#
39;, y=&#
39;特征&#
39;, data=top_features, ax=ax)
ax.set_title(&#
39;Top 10 重要特征&#
39;)
st.pyplot(fig)
# 所有特征的重要性分布
fig, ax = plt.subplots(figsize=(10, 6))
sns.histplot(importance_df[&#
39;重要性&#
39;], bins=20, ax=ax)
ax.set_title(&#
39;特征重要性分布&#
39;)
st.pyplot(fig)
# 显示具体数值
st.write("特征重要性详细数据:")
st.dataframe(importance_df.style.background_gradient(subset=[&#
39;重要性&#
39;], cmap=&#
39;viridis&#
39;))
# 模型比较
st.subheader("与其他模型的对比")
model_per
formance = pd.DataFrame({
&#
39;模型&#
39;: [&#
39;逻辑回归&#
39;, &#
39;决策树&#
39;, &#
39;随机森林&#
39;],
&#
39;准确率&#
39;: [results[&#
39;logistic_
regression&#
39;][&#
39;accuracy&#
39;],
results[&#
39;decision_tree&#
39;][&#
39;accuracy&#
39;],
results[&#
39;random_
forest&#
39;][&#
39;accuracy&#
39;]],
&#
39;AUC&#
39;: [results[&#
39;logistic_
regression&#
39;][&#
39;auc&#
39;],
results[&#
39;decision_tree&#
39;][&#
39;auc&#
39;],
results[&#
39;random_
forest&#
39;][&#
39;auc&#
39;]]
}).sort_values(&#
39;AUC&#
39;, ascending=False)
fig, ax = plt.subplots(figsize=(10, 6))
model_per
formance.set_index(&#
39;模型&#
39;)[[&#
39;AUC&#
39;, &#
39;准确率&#
39;]].plot(kind=&#
39;bar&#
39;, ax=ax)
ax.set_title(&#
39;模型性能对比&#
39;)
ax.set_ylabel(&#
39;评分&#
39;)
plt.xticks(rotation=0)
st.pyplot(fig)
st.table(model_per
formance.style.
format({
&#
39;准确率&#
39;: &#
39;{:.2%}&#
39;,
&#
39;AUC&#
39;: &#
39;{:.4f}&#
39;
}).apply(lambda x: [&#
39;background: lightgreen&#
39; if x.name == model_per
formance.index[0] else &#
39;&#
39;
for _ in x]))
# 页脚
st.markdown("---")
st.markdown("""
<div style="text-align: center; color: #7f8c8d; font-size: 0.9rem; padding: 1rem;">
© 202
3 单宽转融用户预测系统 | 22
310
3027
3 基于Streamlit和Spark开发
</div>
""", unsafe_allow_html=True)
修改上述代码使其能读取大于200MB的csv文件，并给出完整修改后代码
不随自变量的变化而变化
扫一扫
1693
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
