#探索性分析

import plotly.graph_objects as go
from plotly.subplots import make_subplots

# 计算离职和在职的数量
attrition_counts = train_df['Attrition'].value_counts()
labels = ['在职', '离职']
values = [attrition_counts[0], attrition_counts[1]]

# 定义配色
colors = ['#CDBBA7', '#C02B34']

fig = make_subplots(rows=1, cols=2, subplot_titles=('离职标签人数统计', '离职标签分布统计'),
                    specs=[[{"type": "bar"}, {"type": "pie"}]])

# 添加柱状图
fig.add_trace(
    go.Bar(x=labels, y=values, marker_color=colors, opacity=.8,
           text=values, textposition='outside', width=0.5, showlegend=False),
    row=1, col=1
)

# 添加饼状图
fig.add_trace(
    go.Pie(labels=labels, values=values, marker_colors=colors, opacity=.8, name='占比', hole=0.5, showlegend=True),
    row=1, col=2
)

# 更新布局
fig.update_layout(
    title_text='标签统计',
    plot_bgcolor='#F4F2F0', paper_bgcolor='#F4F2F0',  # 设置画布和图表的背景颜色
    width=1000, height=500,
)

# 显示图表
fig.show()

# 将分类变量转换为数值变量
train_df_encoded = pd.get_dummies(train_df)

# 计算相关性矩阵并移除 'Attrition' 自己
corr_with_label = train_df_encoded.corr()['Attrition'].drop('Attrition').sort_values(ascending=False)

# 创建条形图
fig = go.Figure()

fig.add_trace(go.Bar(
    x=corr_with_label.index,
    y=corr_with_label.values,
    marker_color=['#C02B34' if x > 0 else '#CDBBA7' for x in corr_with_label.values]
))

# 更新布局
fig.update_layout(
    title='Correlation of Features with Attrition',
    xaxis_title='Features',
    yaxis_title='Correlation Coefficient',
    plot_bgcolor='#F4F2F0',
    paper_bgcolor='#F4F2F0',
    height=800,
    width=1500
)

# 显示图表
fig.show()

import pandas as pd
import plotly.figure_factory as ff
import plotly.graph_objects as go

# 选择一个连续特征进行可视化，比如 'Age'
feature = 'Age'

# 按 'Attrition' 分组获取特征数据
no_attrition = train_df[train_df['Attrition'] == 0][feature].dropna()
yes_attrition = train_df[train_df['Attrition'] == 1][feature].dropna()

# 数据分组和标签
hist_data = [no_attrition, yes_attrition]
group_labels = ['No', 'Yes']
colors = ['#CDBBA7', '#C02B34']

# 使用 create_distplot 创建分布图
fig = ff.create_distplot(hist_data, group_labels, show_hist=True, show_rug=False, colors=colors)

# 更新布局
fig.update_layout(
    title_text=f'Distribution of {feature} by Attrition',
    xaxis_title_text=feature,
    yaxis_title_text='Density',
    plot_bgcolor='#F4F2F0',
    paper_bgcolor='#F4F2F0', width=1000, height=500
)

# 显示图表
fig.show()

# 选择一个离散特征，比如 'Department'
feature = 'Department'

# 计算每个部门的在职和离职人数
department_counts = train_df.groupby([feature, 'Attrition']).size().unstack(fill_value=0)
department_counts['Total'] = department_counts.sum(axis=1)
department_counts['Attrition Rate'] = department_counts[1] / department_counts['Total']

# 创建子图
fig = make_subplots(specs=[[{"secondary_y": True}]])

# 添加在职和离职的柱状图
fig.add_trace(
    go.Bar(name='No', x=department_counts.index, y=department_counts[0], marker_color='#CDBBA7'),
    secondary_y=False,
)

fig.add_trace(
    go.Bar(name='Yes', x=department_counts.index, y=department_counts[1], marker_color='#C02B34'),
    secondary_y=False,
)

# 添加离职率的折线图
fig.add_trace(
    go.Scatter(name='Attrition Rate', x=department_counts.index, y=department_counts['Attrition Rate'], mode='lines+markers', marker_color='#497B7A', opacity = 0.8),
    secondary_y=True,
)

# 更新布局
fig.update_layout(
    title_text=f'Distribution and Attrition Rate by {feature}',
    yaxis_title='Count',
    yaxis2_title='Attrition Rate',
    plot_bgcolor='#F4F2F0',
    paper_bgcolor='#F4F2F0',
    barmode='group'
)

fig.update_yaxes(showgrid=False, secondary_y=False)
fig.update_yaxes(showgrid=False, secondary_y=False)

# 显示图表
fig.show()

# 选择特征和标签
feature_categorical = 'MaritalStatus'
feature_continuous = 'DistanceFromHome'
target = 'Attrition'

# 创建箱线图
fig = px.box(train_df, y=feature_continuous, x=feature_categorical, color=target,
             points=False, hover_data=train_df.columns,
             color_discrete_map={0: '#CDBBA7', 1: '#C02B34'})

# 更新布局
fig.update_layout(
    title=f'Distribution of {feature_continuous} by {feature_categorical} and {target}',
    xaxis_title=feature_categorical,
    yaxis_title=feature_continuous,
    plot_bgcolor='#F4F2F0',
    paper_bgcolor='#F4F2F0', width=1000, height=500
)

# 显示图表
fig.show()