import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.metrics import mean_squared_error, accuracy_score
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier

sns.set(style="whitegrid")
plt.rcParams['figure.figsize'] = (10, 6)

data = pd.read_csv('/content/Facebook Metrics of Cosmetic Brand.csv')
data.head()

!pip install pingouin
!pip install simpy

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pingouin as pg
import simpy
import random
import joblib
from scipy import stats
from scipy.stats import shapiro, f_oneway, pearsonr, chi2_contingency, ttest_ind
from scipy.fft import fft
from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.model_selection import train_test_split, GridSearchCV, cross_val_score
from sklearn.metrics import mean_squared_error, r2_score, accuracy_score, confusion_matrix, classification_report
from sklearn.linear_model import LogisticRegression, LinearRegression
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor, AdaBoostRegressor, VotingRegressor
from sklearn.utils import resample
from sklearn.impute import SimpleImputer
from sklearn.inspection import PartialDependenceDisplay
from statsmodels.tsa.seasonal import seasonal_decompose
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.stats.outliers_influence import variance_inflation_factor
from statsmodels.tsa.stattools import ccf

from pandas.plotting import autocorrelation_plot, lag_plot
import warnings

warnings.filterwarnings('ignore', category=FutureWarning)
warnings.filterwarnings('ignore', category=UserWarning)
warnings.filterwarnings('ignore', category=RuntimeWarning)
warnings.filterwarnings('ignore', category=DeprecationWarning)
warnings.filterwarnings('ignore', category=ImportWarning)
warnings.filterwarnings('ignore', category=SyntaxWarning)
warnings.filterwarnings('ignore', category=PendingDeprecationWarning)
warnings.filterwarnings('ignore', category=ResourceWarning)

sns.set(style='whitegrid')
plt.rcParams['figure.figsize'] = (12, 8)

data = pd.read_csv('/content/Facebook Metrics of Cosmetic Brand.csv')

print("Sample of dataset:")
display(data.head())

print(f"Dataset shape: {data.shape}")

print(f"Columns in the dataset: {data.columns.tolist()}")

print("\nDataset Information:")
data.info()

print("\nSummary Statistics:")
display(data.describe())

print("\nSummary Statistics for Categorical Columns:")
categorical_columns = data.select_dtypes(include=['object']).columns
display(data[categorical_columns].describe())

print("\nSummary Statistics for Cetegorical Columns:")
categorical_columns = data.select_dtypes(include=['object']).columns
display(data[categorical_columns].describe())

duplicate_rows = data.duplicated().sum()
print(f"\nNumber of duplicate rows: {duplicate_rows}")

print("\nUnique values in each column:")
for column in data.columns:
  unique_values = data[column].nunique()
  print(f"{column}: {unique_values} unique values")

print("\nDistribution of uniquye values in categorical columns:")
for column in categorical_columns:
  value_counts = data[column].value_counts()
  print(f"\n{column} distribution")
  print(value_counts)

print("\nSkewness of numerical columns:")
numerical_columns = data.select_dtypes(include=[np.number]).columns
skewness = data[numerical_columns].skew()
print(skewness)

print("\nKutosis of numerical columns:")
kurtosis = data[numerical_columns].kurtosis()
print(kurtosis)

print("\nPairwise correlatoin of numerical features:")
pairwise_corr = data[numerical_columns].corr()
display(pairwise_corr)

print("\nHighly correlated feature pairs:")
threshold = 0.8
high_corr_pairs = [(i, j, pairwise_corr.loc[i, j]) for i in pairwise_corr.columns for j in pairwise_corr.columns if i != j and abs(pairwise_corr.loc[i, j]) > threshold]
for i, j, corr_value in high_corr_pairs:
  print(f"Correlation between {i} and {j}: {corr_value:.2f}")

print("\nVariance Inflation Factor (VIF) analysis for multicollinearity:")
vif_data = pd.DataFrame()
vif_data["features"] = numerical_columns
vif_data["VIF"] = [variance_inflation_factor(data[numerical_columns].fillna(0).values, i) for i in range(len(numerical_columns))]
display(vif_data)

print("\nShapiro-Wilk test for normality of numerical columns:")
for col in numerical_columns:
  stat, p = shapiro(data[col].dropna())
  print(f"Shapiro-Wilk test for {col}: Statistics={stat:.3f}, p={p:.3f}")
  if p > 0.05:
        print(f"The {col} distribution looks normal (fail to reject H0)\n")
  else:
        print(f"The {col} distribution does not look normal (reject H0)\n")

print("\nANOVA test for interaction between categorical and numerical features:")
for cat_col in categorical_columns:
  for num_col in numerical_columns:
    groups = [data[num_col][data[cat_col] == cat]
  for cat in data[cat_col].unique()]
  f_stat, p_val = f_oneway(*groups)
  print(f"ANOVA test for interaction between {cat_col} and {num_col}: F-statistic={f_stat:.3f}, p-value={p_val:.3f}")
  if p_val < 0.05:
    print(f"Significant interaction detected between {cat_col} and {num_col}\n")
  else:
    print(f"No significant interaction detected between {cat_col} and {num_col}")

print("\nMissing Values in Each Column:")
missing_values = data.isnull().sum()
missing_percentage = data.isnull().mean() * 100
missing_data = pd.DataFrame({
    'Missing Values': missing_values,
    'Percentage': missing_percentage
})
display(missing_data)

plt.figure(figsize=(12, 8))
sns.heatmap(data.isnull(), cbar=False, cmap='viridis')
plt.title('Missing Data Heatmap')
plt.show()

threshold = 30
columns_with_missing_above_threshold = missing_data[missing_data['Percentage'] > threshold].index.tolist()
print(f"\nColumns with more than {threshold}% missing values:")
print(columns_with_missing_above_threshold)

data_cleaned = data.drop(columns = columns_with_missing_above_threshold)
print(f"\nShape of data after dropping columns with > {threshold}% missing values: {data_cleaned.shape}")

numerical_columns = data_cleaned.select_dtypes(include=[np.number]).columns
data_cleaned[numerical_columns] = data_cleaned[numerical_columns].fillna(data_cleaned[numerical_columns].median())

categorical_columns = data_cleaned.select_dtypes(include=['object']).columns
for column in categorical_columns:
  data_cleaned[column].fillna(data_cleaned[column].mode()[0], inplace=True)

print("\nMissing Values After Imputation:")
display(data_cleaned.isnull().sum())

print("\nDistribution of 'Type' column:")
type_counts = data['Type'].value_counts()
display(type_counts)

plt.figure(figsize=(10, 6))
sns.countplot(x='Type', data=data, palette='Set3')
plt.title('Distribution of Post Types')
plt.xlabel('Type of Post')
plt.ylabel('Count')
plt.show()

print("\nDistribution of 'Category' column:")
category_counts = data['Category'].value_counts
display(category_counts)

plt.figure(figsize=(10, 6))
sns.countplot(x='Category', data=data, palette='Set2')
plt.title('Distribution of Post Categories')
plt.xlabel('Category of Post')
plt.ylabel('Count')
plt.show()

print("\nDistribution of 'Paid' column:")
paid_counts = data['Paid'].value_counts()
display(paid_counts)

plt.figure(figsize=(10, 6))
sns.countplot(x='Paid', data=data, palette='Set1')
plt.title('Distribution of Paid vs Non-Paid Posts')
plt.xlabel('Paid (1 = Yes, 0 = No)')
plt.ylabel('Count')
plt.show()

print("\nCross-tabulation of 'Type' and 'Paid' columns:")
type_paid_crosstab = pd.crosstab(data['Type'], data['Paid'])
display(type_paid_crosstab)

type_paid_crosstab.plot(kind='bar', stacked=True, colormap='coolwarm')
plt.title('Stacked Bar Plot of Post Type vs Paid Status')
plt.xlabel('Type of Post')
plt.ylabel('Count')
plt.legend(title='Paid', loc='upper right')
plt.show()

print("\nCross-tabulation of 'Category' and 'Paid' columns:")
category_paid_crosstab = pd.crosstab(data['Category'], data['Paid'])
display(category_paid_crosstab)

category_paid_crosstab.plot(kind='bar', stacked=True, colormap='viridis')
plt.title('Stacked Bar Plot of Post Catgory vs Paid Status')
plt.xlabel('Category of Post')
plt.ylabel('Count')
plt.legend(title='Paid', loc='upper right')
plt.show()

numerical_metrics = ['like', 'comment', 'share']

for metric in numerical_metrics:
  plt.figure(figsize=(18, 6))
  plt.subplot(1, 3, 1)
  sns.boxplot(x='Type', y=metric, data=data, palette='Set3')
  plt.title(f'Distribution of {metric} by Post Type')

  plt.subplot(1, 3, 2)
  sns.boxplot(x='Category', y=metric, data=data, palette='Set2')
  plt.title(f'Distribution of {metric} by Post Category')

  plt.subplot(1, 3, 3)
  sns.boxplot(x='Paid', y=metric, data=data, palette='Set1')
  plt.title(f'Distribution of {metric} by Paid Status')

  plt.tight_layout()
  plt.show()

for metric in numerical_metrics:
  plt.figure(figsize=(18, 6))
  plt.subplot(1, 3, 1)
  sns.violinplot(x='Type', y=metric, data=data, palette='coolwarm', inner='quartile')
  plt.title(f'Violin Plot of {metric} by Post Type')
  plt.subplot(1, 3, 2)
  sns.violinplot(x='Category', y=metric, data=data, palette='viridis', inner='quartile')
  plt.title(f'Violin Plot of {metric} by Post Category')

  plt.subplot(1, 3, 3)
  sns.violinplot(x='Paid', y=metric, data=data, palette='magma', inner='quartile')
  plt.title(f'Violin Plof of {metric} by Paid Status')

  plt.tight_layout()
  plt.show()

from scipy.stats import chi2_contingency

categorical_pairs = [('Type', 'Paid'), ('Category', 'Paid'), ('Type', 'Category')]
print("\nChi-Square Test for Independence between Categorical Variables:")
for pair in categorical_pairs:
  contingency_table = pd.crosstab(data[pair[0]], data[pair[1]])
  chi2, p, dof, expected = chi2_contingency(contingency_table)

  print(f"Chi-Square Test between {pair[0]} and {pair[1]}:")
  print(f"Chi2 = {chi2:.2f}, p-value = {p:.3f}")
  if p < 0.05:
    print(f"There is a significant association between {pair[0]} and {pair[1]}.\n")
  else:
    print(f"No significant association between {pair[0]} and {pair[1]}.\n")