import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestRegressor
from sklearn.tree import DecisionTreeRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

# Load a social media dataset simulating multichannel post engagement
np.random.seed(42)
n_posts = 500
views = np.random.randint(100, 100000, n_posts)
df = pd.DataFrame({
    'post_id': range(1, n_posts+1),
    'platform': np.random.choice(['YouTube','Instagram','TikTok'], n_posts),
    'date': pd.date_range('2023-01-01', periods=n_posts, freq='6h'),
    'views': views,
    'likes': (views * np.random.uniform(0.02, 0.15, n_posts)).astype(int),
    'comments': (views * np.random.uniform(0.001, 0.05, n_posts)).astype(int),
    'shares': (views * np.random.uniform(0.001, 0.03, n_posts)).astype(int)
})
print(df.shape)
print(df.head(3))

df['engagement_rate'] = (df['likes'] + df['comments'] + df['shares']) / df['views'] * 100
print(df[['platform', 'views', 'likes', 'comments', 'shares', 'engagement_rate']].head())

top5 = df.sort_values('engagement_rate', ascending=False).head(5)
print(top5[['post_id', 'platform', 'views', 'engagement_rate']])

plt.figure(figsize=(8,5))
sns.boxplot(x='platform', y='engagement_rate', data=df)
plt.title('Engagement Rate Distribution by Platform')
plt.ylabel('Engagement Rate (%)')
plt.xlabel('Platform')
plt.show()

corr = df[['views', 'likes', 'comments', 'shares', 'engagement_rate']].corr()
plt.figure(figsize=(6,5))
sns.heatmap(corr, annot=True, cmap='coolwarm')
plt.title('Correlation Matrix of Engagement Metrics')
plt.show()

df['platform_code'] = df['platform'].map({'YouTube':0, 'Instagram':1, 'TikTok':2})
print(df[['platform', 'platform_code']].head())

features = ['views', 'likes', 'comments', 'shares', 'platform_code']
X = df[features]
y = df['engagement_rate']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
tree = DecisionTreeRegressor(random_state=42, max_depth=4)
tree.fit(X_train, y_train)
y_pred = tree.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
r2 = r2_score(y_test, y_pred)
print('MSE:', round(mse, 2), '| R2:', round(r2, 3))

rf = RandomForestRegressor(random_state=42, n_estimators=100)
rf.fit(X_train, y_train)
importances = rf.feature_importances_
forest_importance = pd.Series(importances, index=features)
forest_importance.sort_values(ascending=True).plot(kind='barh', color='navy')
plt.title('Random Forest Feature Importance')
plt.xlabel('Importance Score')
plt.show()

from sklearn.inspection import PartialDependenceDisplay
top_feature = forest_importance.idxmax()
PartialDependenceDisplay.from_estimator(rf, X_test, [top_feature],
                                       feature_names=features,
                                       grid_resolution=30)
plt.title(f'Partial Dependence: {top_feature} vs Engagement Rate')
plt.show()

df_missing = df.copy()
df_missing.loc[df_missing.sample(frac=0.05, random_state=42).index, 'comments'] = np.nan
print('Missing comments:', df_missing["comments"].isna().sum())
df_missing['comments'] = df_missing['comments'].fillna(0)
print('After fillna, missing comments:', df_missing['comments'].isna().sum())

# Wrong: sums engagement across all posts, losing date info
agg_wrong = df.groupby('platform').sum(numeric_only=True)['engagement_rate']
print('Aggregation error example:\n', agg_wrong)
# Correct: calculates average engagement per post for each platform
agg_correct = df.groupby('platform')['engagement_rate'].mean()
print('Correct aggregation:\n', agg_correct)

trend = df.groupby(df['date'].dt.to_period('M'))['engagement_rate'].mean()
trend.plot(marker='o', figsize=(9,4))
plt.title('Monthly Engagement Rate Trend')
plt.ylabel('Avg Engagement Rate (%)')
plt.xlabel('Month')
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

mean_engagement = df['engagement_rate'].mean()
topfactor = forest_importance.idxmax()
tip = f'Focus on improving {topfactor}: it has the biggest impact on engagement! Your average engagement rate is {mean_engagement:.2f}%.'
print(tip)