import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from wordcloud import WordCloud, STOPWORDS
import warnings
warnings.filterwarnings('ignore')

# Load a sample open-ended customer feedback dataset
df = pd.DataFrame({'CustomerID':[1,2,3,4,5],
    'Feedback':['Great service and friendly staff',
               'Delivery was slow and packaging was poor',
               'Excellent quality, will buy again',
               'Customer support needs improvement',
               'Good value for money']})
print(df.shape)
print(df.head())

# See feedback comments quickly
for idx, row in df.iterrows():
    print(f'Customer {row.CustomerID}: "{row.Feedback}"')

# Word frequency counts: simple example
from collections import Counter
feedback_text = ' '.join(df['Feedback']).lower()
words = feedback_text.split()
stopwords = set(STOPWORDS)
filtered_words = [w for w in words if w not in stopwords]
word_counts = Counter(filtered_words)
print(word_counts.most_common(7))

# Basic word cloud visualization
wc = WordCloud(width=480, height=240, background_color='white', stopwords=STOPWORDS).generate(feedback_text)
plt.figure(figsize=(8,4))
plt.imshow(wc, interpolation='bilinear')
plt.axis('off')
plt.title('Most Common Words in Customer Feedback')
plt.show()

# Add a longer feedback set for richer analysis
more_data = [
    'Staff did not help me with my question',
    'Fast shipping and clean packaging',
    'I found the website confusing',
    'Always a great deal at this store',
    'Delivery timing could be better',
    'Returns process was simple',
    'The product broke after one week',
    'Excellent value and quick response from support'
]
more_ids = range(6, 14)
df2 = pd.DataFrame({'CustomerID': list(more_ids), 'Feedback': more_data})
df = pd.concat([df, df2], ignore_index=True)
print(df.shape)
print(df.tail())

# Create a word cloud from the larger feedback set
feedback_text = ' '.join(df['Feedback']).lower()
wc = WordCloud(width=520, height=260, background_color='white', stopwords=STOPWORDS, max_words=30).generate(feedback_text)
plt.figure(figsize=(10,5))
plt.imshow(wc, interpolation='bilinear')
plt.axis('off')
plt.title('Customer Feedback: Top 30 Words')
plt.show()

# Bar plot of most common feedback words
from collections import Counter
words = feedback_text.split()
filtered_words = [w for w in words if w not in stopwords]
top_words = Counter(filtered_words).most_common(10)
words, counts = zip(*top_words)
plt.figure(figsize=(8,4))
sns.barplot(x=list(counts), y=list(words), palette='Blues_r')
plt.xlabel('Count')
plt.title('Top 10 Words in Customer Feedback')
plt.tight_layout()
plt.show()

# Adding sentiment: quick manual tagging
def tag_sentiment(text):
    positives = ['great', 'excellent', 'good', 'fast', 'clean', 'value', 'simple', 'quick', 'help']
    negatives = ['poor', 'slow', 'broke', 'confusing', 'did not', 'needs', 'not']
    text_lower = text.lower()
    if any(word in text_lower for word in positives):
        return 'Positive'
    elif any(word in text_lower for word in negatives):
        return 'Negative'
    else:
        return 'Neutral'
df['Sentiment'] = df['Feedback'].apply(tag_sentiment)
print(df[['Feedback', 'Sentiment']].head(8))

# Visualize distribution of sentiments
sent_counts = df['Sentiment'].value_counts()
sns.barplot(x=sent_counts.index, y=sent_counts.values, palette='viridis')
plt.xlabel('Sentiment')
plt.ylabel('Number of Comments')
plt.title('Feedback Sentiment Distribution')
plt.tight_layout()
plt.show()

# Feedback themes by sentiment: intermediate pattern
for sentiment in ['Positive','Negative','Neutral']:
    subset = df[df['Sentiment']==sentiment]
    if len(subset):
        subset_text = ' '.join(subset['Feedback']).lower()
        sub_wc = WordCloud(width=420, height=190, background_color='white', stopwords=STOPWORDS, max_words=10).generate(subset_text)
        plt.figure(figsize=(6,3))
        plt.imshow(sub_wc, interpolation='bilinear')
        plt.axis('off')
        plt.title(f'{sentiment} Feedback: Top Words')
        plt.show()

# Connecting text feedback to scores: setup NPS example
np.random.seed(42)
nps_scores = np.random.randint(0, 11, size=len(df))
df['NPS_Score'] = nps_scores
print(df[['Feedback','NPS_Score']].head())

# Visualize average NPS by sentiment
nps_by_sent = df.groupby('Sentiment')['NPS_Score'].mean()
sns.barplot(x=nps_by_sent.index, y=nps_by_sent.values, palette='mako')
plt.ylabel('Average NPS Score')
plt.title('Average NPS by Feedback Sentiment')
plt.tight_layout()
plt.show()

# Intermediate: Find and visualize top trigrams in feedback
from sklearn.feature_extraction.text import CountVectorizer
vectorizer = CountVectorizer(stop_words='english', ngram_range=(3,3))
X = vectorizer.fit_transform(df['Feedback'])
sum_words = X.sum(axis=0)
words_freq = [(word, sum_words[0, idx]) for word, idx in vectorizer.vocabulary_.items()]
words_freq = sorted(words_freq, key=lambda x: x[1], reverse=True)[:7]
trigrams, counts = zip(*words_freq) if words_freq else ([], [])
plt.figure(figsize=(10,3))
sns.barplot(x=list(counts), y=list(trigrams), palette='flare')
plt.xlabel('Count')
plt.title('Most Common Trigrams in Feedback')
plt.tight_layout()
plt.show()

# Advanced: Connect text feedback to customer demographics
ages = np.random.randint(18, 65, len(df))
regions = np.random.choice(['North', 'South', 'East', 'West'], len(df))
df['Age'] = ages
df['Region'] = regions
plt.figure(figsize=(7,4))
sns.boxplot(x='Region', y='NPS_Score', data=df, palette='pastel')
plt.title('NPS Score Distribution by Region')
plt.show()

# Advanced: Topic separation using simple keyword matching
df['Topic'] = df['Feedback'].str.contains('delivery|shipping|packaging|timing', case=False).map({True: 'Delivery', False: 'Other'})
topic_counts = df.groupby(['Topic','Sentiment']).size().unstack(fill_value=0)
topic_counts.plot(kind='bar', stacked=True, color=['green','red','grey'], figsize=(7,4))
plt.title('Topic vs Sentiment in Feedback')
plt.ylabel('Number of Comments')
plt.xticks(rotation=0)
plt.tight_layout()
plt.show()

# Error handling 1: Missing comment fills
df_with_missing = df.copy()
df_with_missing.loc[2,'Feedback'] = np.nan
df_with_missing['Feedback_filled'] = df_with_missing['Feedback'].fillna('no feedback provided')
print(df_with_missing[['CustomerID','Feedback','Feedback_filled']].head(5))

# Error handling 2: Incorrect groupings
try:
    # Wrong: Group by non-existing column
    test = df.groupby('Sent').size()
except Exception as e:
    print('Grouping error:', e)

# Error handling 3: Misinterpreting sentiment and NPS
df['Likert_Score'] = np.where(df['NPS_Score'] >= 8, 'Agree', np.where(df['NPS_Score'] <= 3, 'Disagree', 'Neutral'))
crosstab = pd.crosstab(df['Sentiment'], df['Likert_Score'])
print(crosstab)

# Pattern: Index construction (complaints index)
df['Is_Complaint'] = df['Feedback'].str.contains('broke|did not|needs|poor|confusing|not help', case=False)
complaint_index = df['Is_Complaint'].mean()
print(f'Complaint Index: {complaint_index:.2%} of comments are complaints')

# Pattern: Follow-up deep dive on frequent complaint
complaints = df[df['Is_Complaint']]
print('Sample complaints:')
print(complaints['Feedback'].tolist()[:5])