🎨 Seaborn for CA Professionals

fig, ax = plt.subplots(figsize=(12, 6))
departments = ['Finance', 'Operations', 'Sales', 'HR', 'IT']
grades = ['Junior', 'Mid', 'Senior', 'Lead']
colors = ['#4472C4', '#ED7D31', '#A9D18E', '#FF5757']
# ... manually group data by dept and grade
# ... manually draw each boxplot with ax.boxplot()
# ... manually set flierprops, whiskerprops, medianprops
# ... manually add legend
# ... manually format axes

sns.boxplot(data=employees, x='Department', y='Annual_CTC', hue='Grade', palette='Set2')

┌──────────────────────────────────────────────────────────────────────────┐
│    Seaborn — statistical visualization, beautiful defaults               │ ← You are here
├──────────────────────────────────────────────────────────────────────────┤
│    Matplotlib — complete, low-level chart control                        │ ← Module 4
├──────────────────────────────────────────────────────────────────────────┤
│    NumPy / Pandas — the data underneath                                  │ ← Modules 2 & 3
└──────────────────────────────────────────────────────────────────────────┘

fig, axes = plt.subplots(1, 2)
sns.boxplot(data=df, x='Dept', y='Salary', ax=axes[0])   # goes in left subplot
sns.histplot(data=df, x='Amount', ax=axes[1])              # goes in right subplot

g = sns.pairplot(df)           # g is a PairGrid object
g.fig.suptitle('Pair Plot')    # access Figure like this

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

# --- Demonstrating axes-level vs figure-level ---

# Build a small sample
np.random.seed(42)
sample = pd.DataFrame({
    'Revenue': np.random.normal(1500, 300, 50),
    'Profit':  np.random.normal(300, 80, 50),
    'Quarter': np.random.choice(['Q1','Q2','Q3','Q4'], 50)
})

# AXES-LEVEL: can be placed in a subplot grid using ax=
fig, axes = plt.subplots(1, 2, figsize=(12, 5))
fig.suptitle('Axes-level Functions — placed in specific subplots', fontsize=12, fontweight='bold')

sns.boxplot(data=sample, x='Quarter', y='Revenue', ax=axes[0], palette='Blues_d')
axes[0].set_title('Box plot (axes-level) → ax=axes[0]')

sns.histplot(data=sample, x='Profit', kde=True, ax=axes[1], color='coral')
axes[1].set_title('Histplot (axes-level) → ax=axes[1]')

plt.tight_layout()
plt.show()

# FIGURE-LEVEL: creates its own figure, returns a Grid object
g = sns.pairplot(sample, hue='Quarter', height=2.5, plot_kws={'alpha': 0.5})
g.fig.suptitle('pairplot (figure-level) → creates its own Figure', y=1.02, fontsize=11)
plt.show()

print("Axes-level return type:", type(axes[0]))  # AxesSubplot
print("Figure-level return type:", type(g))       # PairGrid

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/4191714505.py:20: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=sample, x='Quarter', y='Revenue', ax=axes[0], palette='Blues_d')

Axes-level return type: <class 'matplotlib.axes._axes.Axes'>
Figure-level return type: <class 'seaborn.axisgrid.PairGrid'>

Month  | Revenue  | COGS    | Net_Profit
-------|----------|---------|----------
Apr    | 1250000  | 750000  | 197000
May    | 1380000  | 828000  | 229000
Jun    | 1120000  | 672000  | 185000

Month  | Metric      | Value
-------|-------------|--------
Apr    | Revenue     | 1250000
Apr    | COGS        | 750000
Apr    | Net_Profit  | 197000
May    | Revenue     | 1380000
...

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

# --- Wide format (typical of what you'd import from Excel) ---
wide_df = pd.DataFrame({
    'Month':      ['Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec','Jan','Feb','Mar'],
    'Revenue':    [1250, 1380, 1120, 1450, 1620, 1780, 1550, 1420, 1680, 1910, 2030, 2150],
    'COGS':       [ 750,  828,  672,  870,  972, 1068,  930,  852, 1008, 1146, 1218, 1290],
    'Net_Profit': [ 197,  229,  185,  242,  285,  319,  267,  240,  299,  341,  369,  392],
})

print("Wide format (from Excel):")
print(wide_df.head(3))

# --- Convert to Long (tidy) format using pd.melt() ---
long_df = wide_df.melt(
    id_vars=['Month'],              # columns to keep as-is
    value_vars=['Revenue', 'COGS', 'Net_Profit'],   # columns to "unpivot"
    var_name='Metric',              # new column name for the former column headers
    value_name='Amount'             # new column name for the values
)

# Fix month order for charts (Apr → Mar, not alphabetical)
month_order = ['Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec','Jan','Feb','Mar']
long_df['Month'] = pd.Categorical(long_df['Month'], categories=month_order, ordered=True)
long_df = long_df.sort_values('Month')

print("\nLong (tidy) format after melt():")
print(long_df.head(9).to_string(index=False))

# --- Seaborn works beautifully with long format ---
fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# One line draws all 3 metrics, coloured automatically
sns.lineplot(data=long_df, x='Month', y='Amount', hue='Metric', ax=axes[0],
             palette={'Revenue':'steelblue','COGS':'tomato','Net_Profit':'seagreen'},
             marker='o', linewidth=2)
axes[0].set_title('3 Metrics in 1 Line of Seaborn Code\n(hue=\'Metric\' does all the grouping)',
                  fontsize=10, fontweight='bold')
axes[0].set_ylabel('NPR  Thousands')
axes[0].tick_params(axis='x', rotation=45)

# Bar chart split by metric
sns.barplot(data=long_df, x='Month', y='Amount', hue='Metric', ax=axes[1],
            palette={'Revenue':'steelblue','COGS':'tomato','Net_Profit':'seagreen'})
axes[1].set_title('Grouped Bar Chart from Tidy Data', fontsize=10, fontweight='bold')
axes[1].set_ylabel('NPR  Thousands')
axes[1].tick_params(axis='x', rotation=45)
axes[1].legend(fontsize=8)

plt.tight_layout()
plt.show()

Wide format (from Excel):
  Month  Revenue  COGS  Net_Profit
0   Apr     1250   750         197
1   May     1380   828         229
2   Jun     1120   672         185

Long (tidy) format after melt():
Month     Metric  Amount
  Apr    Revenue    1250
  Apr Net_Profit     197
  Apr       COGS     750
  May    Revenue    1380
  May Net_Profit     229
  May       COGS     828
  Jun    Revenue    1120
  Jun Net_Profit     185
  Jun       COGS     672

sns.set_theme(style='whitegrid', palette='deep')

with sns.axes_style('darkgrid'):
    sns.histplot(...)   # only this chart uses darkgrid

# For profit/loss charts — diverging palette centred at zero
sns.heatmap(data, cmap='RdYlGn', center=0)

# For variance (positive=good, negative=bad)
palette = {True: 'seagreen', False: 'tomato'}

# For categories (departments, expense heads)
sns.set_palette('deep')   # or 'colorblind' for accessible reports

# Install if needed
# !pip install seaborn matplotlib pandas numpy

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

%matplotlib inline
plt.rcParams['figure.dpi'] = 100

print('Seaborn version:', sns.__version__)
print('Ready!')

Seaborn version: 0.13.2
Ready!

# ==== Build a rich financial dataset for this notebook ====
np.random.seed(42)

n = 200   # 200 employee records

departments = np.random.choice(['Finance', 'Operations', 'Sales', 'HR', 'IT'], n,
                                p=[0.20, 0.25, 0.30, 0.10, 0.15])
grades      = np.random.choice(['Junior', 'Mid', 'Senior', 'Lead'], n, p=[0.35, 0.30, 0.25, 0.10])

base_salary = {'Junior': 280000, 'Mid': 480000, 'Senior': 720000, 'Lead': 1080000}
salaries = np.array([base_salary[g] * np.random.uniform(0.85, 1.25) for g in grades])

experience  = np.array([{'Junior': 1.5, 'Mid': 4.0, 'Senior': 8.0, 'Lead': 12.0}[g] +
                         np.random.uniform(-1, 2) for g in grades]).clip(0, 20)
performance = np.clip(np.random.normal(3.2, 0.8, n), 1, 5).round(1)
bonus_pct   = np.clip(performance * 2.5 + np.random.uniform(-2, 3, n), 5, 25).round(1)
bonus_amt   = salaries * bonus_pct / 100

employees = pd.DataFrame({
    'Department': departments,
    'Grade':      grades,
    'Annual_CTC': salaries.round(0).astype(int),
    'Experience': experience.round(1),
    'Perf_Score': performance,
    'Bonus_Pct':  bonus_pct,
    'Bonus_Amt':  bonus_amt.round(0).astype(int),
})

# Invoice dataset
np.random.seed(7)
m = 300
invoice_df = pd.DataFrame({
    'Amount':    np.random.exponential(75000, m).clip(5000, 600000).round(-3).astype(int),
    'Category':  np.random.choice(['Materials', 'Services', 'Utilities', 'Travel', 'Misc'], m,
                                    p=[0.30, 0.25, 0.15, 0.20, 0.10]),
    'Month':     np.random.choice(['Apr','May','Jun','Jul','Aug','Sep',
                                    'Oct','Nov','Dec','Jan','Feb','Mar'], m),
    'VAT_Rate': np.random.choice([0, 13], m, p=[0.15, 0.85]),
    'Status':    np.random.choice(['Paid', 'Pending', 'Overdue'], m, p=[0.65, 0.25, 0.10]),
})

print(f'Employee dataset: {employees.shape[0]} records')
print(f'Invoice dataset:  {invoice_df.shape[0]} records')
print()
print(employees.head())

Employee dataset: 200 records
Invoice dataset:  300 records

   Department   Grade  Annual_CTC  Experience  Perf_Score  Bonus_Pct  \
0  Operations     Mid      427800         3.5         2.6        9.2   
1          IT  Junior      339086         1.3         3.6       11.0   
2       Sales  Junior      294588         1.0         4.7       14.7   
3       Sales  Senior      850020         7.3         4.3       10.5   
4     Finance     Mid      469450         3.4         4.5       13.1

   Bonus_Amt  
0      39358  
1      37299  
2      43304  
3      89252  
4      61498

# Available themes: darkgrid, whitegrid, dark, white, ticks
# Available palettes: deep, muted, bright, pastel, dark, colorblind, Blues, rocket, etc.

themes = ['darkgrid', 'whitegrid', 'white', 'ticks']

fig, axes = plt.subplots(1, 4, figsize=(16, 4))
fig.suptitle('Seaborn Theme Comparison', fontsize=14, fontweight='bold')

sample_x = ['Q1 (Shrawan-Ashwin)', 'Q2 (Kartik-Poush)', 'Q3 (Magh-Chaitra)', 'Q4 (Baishakh-Ashadh)']
sample_y = [3800, 4200, 4650, 6090]

for ax, theme in zip(axes, themes):
    with sns.axes_style(theme):
        ax.bar(sample_x, sample_y, color=sns.color_palette('Blues_d', 4))
        ax.set_title(f"style='{theme}'", fontsize=10)
        ax.set_ylabel('NPR  Thousands')

plt.tight_layout()
plt.show()

# Set a default theme for the rest of this notebook
sns.set_theme(style='whitegrid', palette='deep')
print("Default theme set to 'whitegrid' with 'deep' palette")

Default theme set to 'whitegrid' with 'deep' palette

# histplot — distribution of invoice amounts
fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# Simple histogram
sns.histplot(data=invoice_df, x='Amount', bins=25, color='steelblue',
             kde=True, ax=axes[0])
axes[0].set_title('Invoice Amount Distribution\n(Histogram + KDE)', fontsize=12, fontweight='bold')
axes[0].set_xlabel('Invoice Amount (NPR )')
axes[0].xaxis.set_major_formatter(plt.FuncFormatter(lambda x, _: f'NPR {x/1000:.0f}K'))

# Histogram by Category
sns.histplot(data=invoice_df, x='Amount', hue='Category', bins=20,
             multiple='stack', ax=axes[1])
axes[1].set_title('Invoice Amount by Category\n(Stacked Histogram)', fontsize=12, fontweight='bold')
axes[1].set_xlabel('Invoice Amount (NPR )')
axes[1].xaxis.set_major_formatter(plt.FuncFormatter(lambda x, _: f'NPR {x/1000:.0f}K'))

plt.tight_layout()
plt.show()

# kdeplot — Salary distribution by Grade (smooth density curves)
fig, ax = plt.subplots(figsize=(12, 6))

for grade in ['Junior', 'Mid', 'Senior', 'Lead']:
    subset = employees[employees['Grade'] == grade]['Annual_CTC']
    sns.kdeplot(subset / 100000, label=grade, fill=True, alpha=0.25, linewidth=2, ax=ax)

ax.set_title('Salary Distribution by Grade Level', fontsize=14, fontweight='bold')
ax.set_xlabel('Annual CTC (NPR  Lakhs)')
ax.set_ylabel('Density')
ax.legend(title='Grade', fontsize=10)
ax.xaxis.set_major_formatter(plt.FuncFormatter(lambda x, _: f'NPR {x:.0f}L'))

plt.tight_layout()
plt.show()

print('Average CTC by Grade:')
print(employees.groupby('Grade')['Annual_CTC'].mean().sort_values().apply(lambda x: f'NPR {x/100000:.1f} Lakhs'))

Average CTC by Grade:
Grade
Junior     NPR 3.0 Lakhs
Mid        NPR 5.0 Lakhs
Senior     NPR 7.6 Lakhs
Lead      NPR 11.4 Lakhs
Name: Annual_CTC, dtype: object

         Outlier ●
              |
   Whisker ───┤ Upper fence (Q3 + 1.5×IQR)
              │
   Q3    ┌────┤ 75th percentile
         │    │
   Median├────┤ 50th percentile (orange line)
         │    │
   Q1    └────┤ 25th percentile
              │
   Whisker ───┤ Lower fence (Q1 - 1.5×IQR)
              |
         Outlier ●

# Box plot — Invoice amounts by category (flags outliers automatically)
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

# Invoice amounts by expense category
sns.boxplot(data=invoice_df, x='Category', y='Amount',
            palette='Set2', ax=axes[0], width=0.5,
            flierprops=dict(marker='o', color='red', markersize=6, alpha=0.6))
axes[0].set_title('Invoice Amounts by Category\n(Red dots = Outliers)', fontsize=12, fontweight='bold')
axes[0].set_ylabel('Invoice Amount (NPR )')
axes[0].set_xlabel('')
axes[0].yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/1000:.0f}K'))
axes[0].tick_params(axis='x', rotation=15)

# Employee salary by department
sns.boxplot(data=employees, x='Department', y='Annual_CTC',
            palette='husl', ax=axes[1], width=0.5,
            flierprops=dict(marker='D', color='red', markersize=5, alpha=0.7))
axes[1].set_title('Salary by Department\n(Diamond = Outliers)', fontsize=12, fontweight='bold')
axes[1].set_ylabel('Annual CTC (NPR )')
axes[1].set_xlabel('')
axes[1].yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.1f}L'))
axes[1].tick_params(axis='x', rotation=15)

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/888267004.py:5: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=invoice_df, x='Category', y='Amount',
/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/888267004.py:15: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=employees, x='Department', y='Annual_CTC',

# Box plot with stripplot overlaid — shows individual data points
fig, ax = plt.subplots(figsize=(12, 6))

sns.boxplot(data=employees, x='Grade', y='Annual_CTC',
            order=['Junior', 'Mid', 'Senior', 'Lead'],
            palette=['#AED6F1','#85C1E9','#5DADE2','#2E86C1'],
            width=0.45, ax=ax,
            flierprops=dict(marker='', alpha=0))

sns.stripplot(data=employees, x='Grade', y='Annual_CTC',
              order=['Junior', 'Mid', 'Senior', 'Lead'],
              color='dimgray', alpha=0.4, size=4, jitter=True, ax=ax)

ax.set_title('Salary Distribution by Grade\n(Box + Individual Data Points)', fontsize=13, fontweight='bold')
ax.set_ylabel('Annual CTC (NPR )')
ax.set_xlabel('Employee Grade')
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.1f}L'))

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/660531211.py:4: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=employees, x='Grade', y='Annual_CTC',

# Violin plot — salary distribution by department
fig, ax = plt.subplots(figsize=(12, 6))

sns.violinplot(data=employees, x='Department', y='Annual_CTC',
               palette='pastel', inner='quartile', ax=ax,
               order=['Finance', 'IT', 'Operations', 'Sales', 'HR'])

ax.set_title('Salary Distribution by Department\n(Width = Density of employees at that salary level)',
             fontsize=12, fontweight='bold')
ax.set_ylabel('Annual CTC (NPR )')
ax.set_xlabel('Department')
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.0f}L'))

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/3679463453.py:4: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.violinplot(data=employees, x='Department', y='Annual_CTC',

# barplot — Average salary by Department (with confidence interval)
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

# Average CTC by department
dept_order = employees.groupby('Department')['Annual_CTC'].mean().sort_values(ascending=False).index
sns.barplot(data=employees, x='Department', y='Annual_CTC',
            order=dept_order, palette='Blues_d',
            estimator='mean', errorbar='ci', ax=axes[0])
axes[0].set_title('Average Salary by Department\n(Error bars = 95% Confidence Interval)', fontsize=11, fontweight='bold')
axes[0].set_ylabel('Average Annual CTC (NPR )')
axes[0].set_xlabel('')
axes[0].yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.1f}L'))

# Average bonus by Grade
grade_order = ['Junior', 'Mid', 'Senior', 'Lead']
sns.barplot(data=employees, x='Grade', y='Bonus_Pct',
            order=grade_order, palette='Oranges_d',
            ax=axes[1], errorbar='sd')
axes[1].set_title('Average Bonus % by Grade\n(Error bars = Std Deviation)', fontsize=11, fontweight='bold')
axes[1].set_ylabel('Bonus (%)')
axes[1].set_xlabel('')

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/3403452037.py:6: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.barplot(data=employees, x='Department', y='Annual_CTC',
/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/3403452037.py:16: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.barplot(data=employees, x='Grade', y='Bonus_Pct',

# countplot — how many invoices by status and category?
fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# Count of invoices by status
sns.countplot(data=invoice_df, x='Status',
              palette={'Paid': 'seagreen', 'Pending': 'steelblue', 'Overdue': 'tomato'},
              order=['Paid', 'Pending', 'Overdue'], ax=axes[0])
axes[0].set_title('Invoice Count by Status', fontsize=12, fontweight='bold')
axes[0].set_ylabel('Number of Invoices')
axes[0].set_xlabel('')
for p in axes[0].patches:
    axes[0].annotate(f'{int(p.get_height())}', (p.get_x() + p.get_width()/2, p.get_height() + 2),
                     ha='center', fontsize=11, fontweight='bold')

# Count by category and status (hue)
sns.countplot(data=invoice_df, x='Category', hue='Status',
              palette={'Paid': 'seagreen', 'Pending': 'steelblue', 'Overdue': 'tomato'},
              ax=axes[1])
axes[1].set_title('Invoice Count by Category & Status', fontsize=12, fontweight='bold')
axes[1].set_ylabel('Number of Invoices')
axes[1].set_xlabel('')
axes[1].legend(title='Status')
axes[1].tick_params(axis='x', rotation=15)

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/4118093210.py:5: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.countplot(data=invoice_df, x='Status',

# Correlation heatmap for employee data
numeric_cols = employees[['Annual_CTC', 'Experience', 'Perf_Score', 'Bonus_Pct', 'Bonus_Amt']]
corr_matrix  = numeric_cols.corr().round(2)

fig, ax = plt.subplots(figsize=(9, 7))

sns.heatmap(
    corr_matrix,
    annot=True,           # Show correlation values in each cell
    fmt='.2f',            # 2 decimal places
    cmap='RdYlGn',        # Red (negative) → Yellow (zero) → Green (positive)
    center=0,             # Centre the colour scale at 0
    square=True,
    linewidths=1,
    linecolor='white',
    ax=ax,
    cbar_kws={'label': 'Correlation Coefficient'}
)

ax.set_title('Correlation Matrix — Employee Metrics', fontsize=14, fontweight='bold', pad=15)
ax.set_xticklabels(ax.get_xticklabels(), rotation=30, ha='right')

plt.tight_layout()
plt.show()

# Key insights
print('Key Insights from Correlation Matrix:')
r_exp   = corr_matrix.loc['Annual_CTC', 'Experience']
r_perf  = corr_matrix.loc['Annual_CTC', 'Perf_Score']
r_bns   = corr_matrix.loc['Bonus_Pct',  'Perf_Score']
lbl_exp  = 'Strong +ve' if abs(r_exp)  > 0.5 else 'Moderate'
lbl_perf = 'Strong +ve' if abs(r_perf) > 0.5 else 'Moderate'
print(f'  CTC vs Experience:  {r_exp}  — {lbl_exp}')
print(f'  Bonus% vs Perf:     {corr_matrix.loc["Bonus_Pct","Perf_Score"]}  — Performance drives bonus')

Key Insights from Correlation Matrix:
  CTC vs Experience:  0.94  — Strong +ve
  Bonus% vs Perf:     0.79  — Performance drives bonus

# Monthly revenue heatmap — very common in business analysis
# Simulate 3 years of monthly revenue data
np.random.seed(5)
years = [2023, 2024, 2025]
months_list = ['Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec', 'Jan', 'Feb', 'Mar']

base    = 1200
trend   = 80    # ~80K growth per month over years
monthly_rev = {
    2023: [1050, 1100, 900, 1150, 1280, 1400, 1220, 1120, 1380, 1580, 1720, 1850],
    2024: [1250, 1380, 1120, 1450, 1620, 1780, 1550, 1420, 1680, 1910, 2030, 2150],
    2025: [1480, 1590, 1350, 1700, 1850, 1980, 1820, 1700, 1950, 2200, 2350, 2480],
}

rev_df = pd.DataFrame(monthly_rev, index=months_list)

fig, ax = plt.subplots(figsize=(10, 7))

sns.heatmap(rev_df, annot=True, fmt=',.0f', cmap='YlOrRd',
            linewidths=0.5, linecolor='white', ax=ax,
            cbar_kws={'label': 'NPR  Thousands'})

ax.set_title('Monthly Revenue Heatmap — 3 Year Trend\n(NPR Thousands)', fontsize=13, fontweight='bold')
ax.set_xlabel('Financial Year')
ax.set_ylabel('Month (Apr = FY Start)')

plt.tight_layout()
plt.show()

# Pair plot — all relationships in employee data at once
# Color-coded by Grade
pair_data = employees[['Annual_CTC', 'Experience', 'Perf_Score', 'Bonus_Pct', 'Grade']]

g = sns.pairplot(
    pair_data,
    hue='Grade',
    hue_order=['Junior', 'Mid', 'Senior', 'Lead'],
    palette='Set1',
    diag_kind='kde',     # Diagonal = KDE distribution
    plot_kws={'alpha': 0.5, 's': 30},
    corner=True          # Only lower triangle (avoids duplication)
)

g.fig.suptitle('Employee Metrics — Pair Plot by Grade', y=1.02, fontsize=13, fontweight='bold')
plt.show()

# regplot — does experience predict salary?
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

# Plot 1: Experience vs CTC
sns.regplot(data=employees, x='Experience', y='Annual_CTC',
            scatter_kws={'alpha': 0.4, 's': 30, 'color': 'steelblue'},
            line_kws={'color': 'navy', 'linewidth': 2},
            ax=axes[0])
axes[0].set_title('Experience vs Salary\n(Does experience predict CTC?)', fontsize=11, fontweight='bold')
axes[0].set_xlabel('Years of Experience')
axes[0].set_ylabel('Annual CTC (NPR )')
axes[0].yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.1f}L'))

corr_exp_ctc = employees[['Experience', 'Annual_CTC']].corr().iloc[0, 1]
axes[0].text(0.05, 0.92, f'Correlation: {corr_exp_ctc:.2f}', transform=axes[0].transAxes,
             bbox=dict(boxstyle='round', facecolor='lightyellow'))

# Plot 2: Performance Score vs Bonus %
sns.regplot(data=employees, x='Perf_Score', y='Bonus_Pct',
            scatter_kws={'alpha': 0.4, 's': 30, 'color': 'coral'},
            line_kws={'color': 'darkred', 'linewidth': 2},
            ax=axes[1])
axes[1].set_title('Performance Score vs Bonus %\n(Is bonus linked to performance?)', fontsize=11, fontweight='bold')
axes[1].set_xlabel('Performance Score (1–5)')
axes[1].set_ylabel('Bonus (%)')

corr_perf_bonus = employees[['Perf_Score', 'Bonus_Pct']].corr().iloc[0, 1]
axes[1].text(0.05, 0.92, f'Correlation: {corr_perf_bonus:.2f}', transform=axes[1].transAxes,
             bbox=dict(boxstyle='round', facecolor='lightyellow'))

plt.tight_layout()
plt.show()

# lmplot — regression by subgroup (Grade)
g = sns.lmplot(
    data=employees,
    x='Experience',
    y='Annual_CTC',
    hue='Grade',
    hue_order=['Junior', 'Mid', 'Senior', 'Lead'],
    palette='Set1',
    scatter_kws={'alpha': 0.4, 's': 30},
    height=6, aspect=1.5
)

g.set_axis_labels('Years of Experience', 'Annual CTC (NPR )')
g.fig.suptitle('Experience vs Salary by Grade\n(Regression lines show within-grade trends)',
               y=1.02, fontsize=12, fontweight='bold')

# Format y-axis
for ax in g.axes.flat:
    ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.0f}L'))

plt.tight_layout()
plt.show()

# Invoice amount distribution — one histogram per expense category
g = sns.FacetGrid(
    invoice_df,
    col='Category',
    col_wrap=3,          # 3 charts per row
    height=3.5,
    aspect=1.2,
    sharey=False         # Allow different y-axis scales per chart
)

g.map(sns.histplot, 'Amount', bins=15, color='steelblue', kde=True)

for ax in g.axes.flat:
    ax.xaxis.set_major_formatter(plt.FuncFormatter(lambda x, _: f'NPR {x/1000:.0f}K'))
    ax.tick_params(axis='x', rotation=30)

g.set_axis_labels('Invoice Amount', 'Count')
g.set_titles(col_template='{col_name}', fontweight='bold')
g.fig.suptitle('Invoice Amount Distribution by Expense Category', y=1.02, fontsize=13, fontweight='bold')

plt.tight_layout()
plt.show()

# Comprehensive HR Analytics Dashboard
fig = plt.figure(figsize=(16, 12))
fig.suptitle('HR Analytics Dashboard — FY 2081-82 BS', fontsize=16, fontweight='bold', y=0.98)

# Layout using gridspec
gs = fig.add_gridspec(2, 3, hspace=0.4, wspace=0.35)

# Chart 1: Salary by Department (box plot)
ax1 = fig.add_subplot(gs[0, 0:2])
dept_order = employees.groupby('Department')['Annual_CTC'].median().sort_values(ascending=False).index
sns.boxplot(data=employees, x='Department', y='Annual_CTC',
            order=dept_order, palette='Set2', ax=ax1, width=0.5)
ax1.set_title('Salary Distribution by Department', fontweight='bold')
ax1.set_ylabel('Annual CTC (NPR )')
ax1.set_xlabel('')
ax1.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/100000:.0f}L'))

# Chart 2: Headcount by Grade (count plot)
ax2 = fig.add_subplot(gs[0, 2])
grade_counts = employees['Grade'].value_counts().reindex(['Junior','Mid','Senior','Lead'])
ax2.barh(grade_counts.index, grade_counts.values,
         color=sns.color_palette('Blues_d', 4))
for i, v in enumerate(grade_counts.values):
    ax2.text(v + 0.5, i, str(v), va='center', fontsize=10)
ax2.set_title('Headcount by Grade', fontweight='bold')
ax2.set_xlabel('Number of Employees')

# Chart 3: Performance distribution
ax3 = fig.add_subplot(gs[1, 0])
sns.histplot(data=employees, x='Perf_Score', bins=15, kde=True, color='coral', ax=ax3)
ax3.axvline(employees['Perf_Score'].mean(), color='navy', linestyle='--',
             label=f"Mean: {employees['Perf_Score'].mean():.1f}")
ax3.set_title('Performance Score Distribution', fontweight='bold')
ax3.set_xlabel('Performance Score (1–5)')
ax3.legend()

# Chart 4: Bonus % vs Perf regression
ax4 = fig.add_subplot(gs[1, 1])
sns.regplot(data=employees, x='Perf_Score', y='Bonus_Pct',
            scatter_kws={'alpha': 0.35, 's': 25, 'color': 'steelblue'},
            line_kws={'color': 'navy'}, ax=ax4)
ax4.set_title('Performance vs Bonus %', fontweight='bold')
ax4.set_xlabel('Performance Score')
ax4.set_ylabel('Bonus (%)')

# Chart 5: Dept cost breakdown (stacked bar)
ax5 = fig.add_subplot(gs[1, 2])
dept_cost = employees.groupby('Department')[['Annual_CTC', 'Bonus_Amt']].sum() / 1e7  # In NPR  Crores
dept_cost.plot(kind='barh', stacked=True, ax=ax5,
               color=['steelblue', 'coral'], width=0.6)
ax5.set_title('Total Cost by Department\n(NPR  Crores)', fontweight='bold')
ax5.set_xlabel('NPR  Crores')
ax5.set_ylabel('')
ax5.legend(['Base CTC', 'Bonus'], fontsize=8)

plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/2995762040.py:11: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=employees, x='Department', y='Annual_CTC',

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

# (Re-run the data setup cell at the top if invoice_df is not available)

# Write your analysis here

# Write your dashboard here

# SOLUTION — Exercise 1: Invoice Outlier Analysis
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

# (Rebuild dataset if needed)
np.random.seed(7)
m = 300
invoice_df = pd.DataFrame({
    'Amount':   np.random.exponential(75000, m).clip(5000, 600000).round(-3).astype(int),
    'Category': np.random.choice(['Materials','Services','Utilities','Travel','Misc'], m, p=[0.30,0.25,0.15,0.20,0.10]),
    'VAT_Rate': np.random.choice([0, 13], m, p=[0.15, 0.85]),
    'Status':   np.random.choice(['Paid','Pending','Overdue'], m, p=[0.65, 0.25, 0.10]),
})

fig, axes = plt.subplots(1, 2, figsize=(15, 6))

# Box plot by GST rate
sns.boxplot(data=invoice_df, x='VAT_Rate', y='Amount', palette='coolwarm', ax=axes[0],
            flierprops=dict(marker='o', color='red', markersize=5, alpha=0.7))
axes[0].set_title('Invoice Amounts by GST Rate Slab\n(Red = Outliers)', fontweight='bold')
axes[0].set_xlabel('GST Rate (%)')
axes[0].set_ylabel('Invoice Amount (NPR )')
axes[0].yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f'NPR {y/1000:.0f}K'))

# Heatmap: Avg Amount by Category & Status
pivot = invoice_df.pivot_table(values='Amount', index='Category', columns='Status', aggfunc='mean').round(0)
sns.heatmap(pivot, annot=True, fmt=',.0f', cmap='Blues', linewidths=0.5, ax=axes[1],
            cbar_kws={'label': 'Avg Invoice Amount (NPR )'})
axes[1].set_title('Avg Invoice Amount\nby Category & Status', fontweight='bold')

plt.tight_layout()
plt.show()

# Identify outliers
Q1, Q3 = invoice_df['Amount'].quantile([0.25, 0.75])
IQR    = Q3 - Q1
upper  = Q3 + 1.5 * IQR
outliers = invoice_df[invoice_df['Amount'] > upper]
print(f'Outlier threshold: NPR {upper:,.0f}')
print(f'Outlier count: {len(outliers)} ({len(outliers)/len(invoice_df)*100:.1f}% of invoices)')
print(f'Total outlier value: NPR {outliers["Amount"].sum():,.0f}')

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/81329947.py:20: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.boxplot(data=invoice_df, x='VAT_Rate', y='Amount', palette='coolwarm', ax=axes[0],

Outlier threshold: NPR 231,375
Outlier count: 11 (3.7% of invoices)
Total outlier value: NPR 3,075,000

# SOLUTION — Exercise 2: Department-wise Dashboard
import seaborn as sns, matplotlib.pyplot as plt, pandas as pd, numpy as np

# Rebuild employees if needed (copy from data setup cell)
np.random.seed(42)
n = 200
departments = np.random.choice(['Finance','Operations','Sales','HR','IT'], n, p=[0.20,0.25,0.30,0.10,0.15])
grades      = np.random.choice(['Junior','Mid','Senior','Lead'], n, p=[0.35,0.30,0.25,0.10])
base_salary = {'Junior':480000,'Mid':720000,'Senior':1080000,'Lead':1560000}
salaries    = np.array([base_salary[g] * np.random.uniform(0.85, 1.25) for g in grades])
experience  = np.array([{'Junior':1.5,'Mid':4.0,'Senior':8.0,'Lead':12.0}[g]+np.random.uniform(-1,2) for g in grades]).clip(0,20)
performance = np.clip(np.random.normal(3.2, 0.8, n), 1, 5).round(1)
bonus_pct   = np.clip(performance*2.5+np.random.uniform(-2,3,n), 5, 25).round(1)
bonus_amt   = salaries * bonus_pct / 100
employees   = pd.DataFrame({'Department':departments,'Grade':grades,'Annual_CTC':salaries.round(0).astype(int),
                             'Experience':experience.round(1),'Perf_Score':performance,
                             'Bonus_Pct':bonus_pct,'Bonus_Amt':bonus_amt.round(0).astype(int)})

fig, axes = plt.subplots(2, 2, figsize=(16, 12))
fig.suptitle('HR Dashboard — Department-wise Analysis', fontsize=15, fontweight='bold')

# 1. Violin — salary by department
sns.violinplot(data=employees, x='Department', y='Annual_CTC', palette='pastel', ax=axes[0,0], inner='quartile')
axes[0,0].set_title('Salary Distribution by Department', fontweight='bold')
axes[0,0].set_ylabel('Annual CTC (NPR )')
axes[0,0].set_xlabel('')
axes[0,0].yaxis.set_major_formatter(plt.FuncFormatter(lambda y,_: f'NPR {y/100000:.0f}L'))

# 2. Bar — avg performance score by dept
dept_perf = employees.groupby('Department')['Perf_Score'].mean().sort_values(ascending=False)
axes[0,1].bar(dept_perf.index, dept_perf.values, color=sns.color_palette('Greens_d', len(dept_perf)))
axes[0,1].axhline(employees['Perf_Score'].mean(), color='red', linestyle='--', linewidth=1.5, label='Overall Avg')
axes[0,1].set_title('Avg Performance Score by Department', fontweight='bold')
axes[0,1].set_ylabel('Performance Score')
axes[0,1].set_ylim(0, 5)
axes[0,1].legend()

# 3. Scatter — Experience vs Bonus
for dept in employees['Department'].unique():
    sub = employees[employees['Department'] == dept]
    axes[1,0].scatter(sub['Experience'], sub['Bonus_Amt']/100000, label=dept, alpha=0.5, s=30)
axes[1,0].set_title('Experience vs Bonus Amount by Department', fontweight='bold')
axes[1,0].set_xlabel('Years of Experience')
axes[1,0].set_ylabel('Bonus Amount (NPR  Lakhs)')
axes[1,0].legend(fontsize=8, markerscale=1.5)

# 4. Heatmap — avg salary by Dept × Grade
heat_data = employees.pivot_table(values='Annual_CTC', index='Department', columns='Grade',
                                   columns_sort=False,
                                   aggfunc='mean').reindex(columns=['Junior','Mid','Senior','Lead']) / 100000
sns.heatmap(heat_data.round(1), annot=True, fmt='.1f', cmap='YlOrRd', ax=axes[1,1],
            cbar_kws={'label': 'Avg CTC (NPR  Lakhs)'}, linewidths=0.5)
axes[1,1].set_title('Avg Salary (NPR L) by Dept × Grade', fontweight='bold')
axes[1,1].set_ylabel('')

plt.tight_layout()
plt.show()

/var/folders/kp/mytrwq8157q9s1tmw7r79fkm0000gn/T/ipykernel_11210/3553043638.py:23: FutureWarning:

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.violinplot(data=employees, x='Department', y='Annual_CTC', palette='pastel', ax=axes[0,0], inner='quartile')



---------------------------------------------------------------------------

TypeError                                 Traceback (most recent call last)

Cell In[23], line 48
     45 axes[1,0].legend(fontsize=8, markerscale=1.5)
     47 # 4. Heatmap — avg salary by Dept × Grade
---> 48 heat_data = employees.pivot_table(values='Annual_CTC', index='Department', columns='Grade',
     49                                    columns_sort=False,
     50                                    aggfunc='mean').reindex(columns=['Junior','Mid','Senior','Lead']) / 100000
     51 sns.heatmap(heat_data.round(1), annot=True, fmt='.1f', cmap='YlOrRd', ax=axes[1,1],
     52             cbar_kws={'label': 'Avg CTC (NPR  Lakhs)'}, linewidths=0.5)
     53 axes[1,1].set_title('Avg Salary (NPR L) by Dept × Grade', fontweight='bold')


TypeError: DataFrame.pivot_table() got an unexpected keyword argument 'columns_sort'