import pandas as pd
import numpy as np
import openml
import warnings
warnings.filterwarnings('ignore')

# Example 1: Load a customer satisfaction dataset with Likert scale ratings
dataset = openml.datasets.get_dataset(42178)
df, _, _, _ = dataset.get_data(dataset_format='dataframe')
print(df.shape)
print(df.head(3))

# Example 3: Count responses for a Likert-scale question (e.g., OnlineSecurity)
likert_counts = df['OnlineSecurity'].value_counts(dropna=False)
print(likert_counts)

# Example 4: Calculate the percentage distribution for a Likert-scale question
likert_percent = df['OnlineSecurity'].value_counts(normalize=True, dropna=False) * 100
print(likert_percent.round(2))

# Example 5: Visualize Likert-scale ratings (bar plot)
import matplotlib.pyplot as plt
likert_percent.plot(kind='bar', color='skyblue')
plt.title('Online Security - Survey Responses (%)')
plt.ylabel('Percentage')
plt.xlabel('Response')
plt.show()

# Example 6: Calculate mean satisfaction rating for a numeric scale (e.g., MonthlyCharges)
mean_charge = df['MonthlyCharges'].mean()
print(f'Average Monthly Charges: ${mean_charge:.2f}')

# Example 7: Segment Likert-scale ratings by customer demographics (e.g., gender)
grouped = df.groupby('gender')['OnlineSecurity'].value_counts(normalize=True).unstack().fillna(0) * 100
print(grouped.round(1))

# Example 8: Cross-tabulate two categorical variables (OnlineSecurity and Churn)
ct = pd.crosstab(df['OnlineSecurity'], df['Churn'], normalize='columns') * 100
print(ct.round(1))

# Example 9: Create a satisfaction index from several Likert-scale columns
cols = ['OnlineSecurity', 'TechSupport', 'DeviceProtection']
df['satisfaction_index'] = df[cols].applymap(lambda x: 1 if x == 'Yes' else 0).mean(axis=1)
print(df[['satisfaction_index']].head())

# Example 10: Trend analysis for satisfaction index (by tenure in years)
df['tenure_years'] = (df['tenure'] // 12).astype(int)
trend = df.groupby('tenure_years')['satisfaction_index'].mean()
print(trend)

# Example 11: Advanced - Filter for survey non-response / missing data
missing = df[cols].isnull().sum()
print('Missing responses in key satisfaction columns:')
print(missing)

# Example 12: Advanced - Handle missing values by filling with 'No Response'
df_filled = df.copy()
df_filled[cols] = df_filled[cols].fillna('No Response')
print(df_filled[cols].head())

# Example 13: Advanced - Misinterpreted Likert scales: reverse coding
df['Feedback_Quality'] = np.random.choice([1,2,3,4,5], size=len(df))
df['Quality_reversed'] = 6 - df['Feedback_Quality']
print(df[['Feedback_Quality','Quality_reversed']].head())

# Example 14: Error Handling - Wrong groupby column name
try:
    df.groupby('gendre')['OnlineSecurity'].mean()
except Exception as e:
    print('Error:', e)

# Example 15: Error Handling - Interpreting NPS scores as Likert scale
nps_df = pd.DataFrame({'NPS_Score': np.random.randint(0, 11, 100)})
print(nps_df['NPS_Score'].describe())
print('Mean NPS: Not directly interpretable as Net Promoter Score, use the correct classification!')

# Example 16: Best Practice - Calculate NPS breakdown
nps_df['Type'] = np.where(nps_df['NPS_Score'] >= 9, 'Promoter',
                      np.where(nps_df['NPS_Score'] <= 6, 'Detractor', 'Passive'))
nps_counts = nps_df['Type'].value_counts(normalize=True) * 100
print('NPS Classification (%):')
print(nps_counts.round(1))

# Example 17: Best Practice - Segmentation by SeniorCitizen and satisfaction
seg = df.groupby('SeniorCitizen')['satisfaction_index'].mean()
print('Average satisfaction index by senior citizen status:')
print(seg.round(3))

# Example 18: Best Practice - Multi-way cross-tab (Churn by InternetService and satisfaction)
df['satisfaction_bin'] = pd.cut(df['satisfaction_index'], bins=[-0.1,0.33,0.66,1], labels=['Low','Medium','High'])
crosstab = pd.crosstab([df['InternetService'], df['satisfaction_bin']], df['Churn'], normalize='columns') * 100
print(crosstab.round(1))

# Example 19: Trend chart - satisfaction index by tenure
trend.plot(marker='o')
plt.xlabel('Tenure (years)')
plt.ylabel('Avg Satisfaction Index')
plt.title('Satisfaction Trend by Customer Tenure')
plt.show()

# Example 20: Tiny end-to-end market research workflow summary
def survey_analysis_workflow(df):
    # Step 1: Check missing values
    missing = df[cols].isnull().sum().sum()
    # Step 2: Compute satisfaction index
    satisfaction = df[cols].applymap(lambda x: 1 if x=='Yes' else 0).mean(axis=1)
    # Step 3: Split by churn
    churned = satisfaction[df['Churn']=='Yes'].mean()
    not_churned = satisfaction[df['Churn']=='No'].mean()
    print(f'Total missing responses: {missing}')
    print(f'Churned Customers Avg Satisfaction: {churned:.3f}')
    print(f'Non-Churned Customers Avg Satisfaction: {not_churned:.3f}')
    if churned < not_churned:
        print('Insight: Lower satisfaction is linked to higher churn. Prioritize improvements!')
    else:
        print('No strong satisfaction-churn link found.')

survey_analysis_workflow(df)