leks39

#importing word cloud image from directory
mask = np.array(Image.open("C:\\Users\\ofagb\\Downloads\\GOT.jpg"))
plt.figure(figsize = (18,12))
plt.imshow(mask, interpolation='bilinear') 
plt.axis('off') 
plt.show() 

#create word cloud
masked_wordcloud = WordCloud(background_color = 'white', contour_color = 'black', contour_width = 3,
                             min_font_size = 8, max_words = 300, mask = mask)

leks39

#Calculating text polarity using TextBlob
def polarity(Processed_Text):
    return TextBlob(Processed_Text).sentiment.polarity
    
#Add polarity score to df
hotd_df2['Polarity'] = hotd_df2['Processed_Text'].apply(polarity)
hotd_df2.head()

#Creating polarity sentiment column
def p_sentiment(label):

leks39

#first step in text processing is to taking out stopwords from the text using NLTK stop_words method
#stop words are common words in english that add little or no value in NLP
stop_words = list(stopwords.words('english'))

#user defined common words that do not add value in the context of the show will also be removed
hotd_stop_words = ['house','GOT','thrones','lady','lord','ser','gameofthrones', 'hotd',
                   'premiere','watching','got','episode','serie','houseofthedragon',
                   'fan','hbo','hbomax','finale', 'end', 'getting', 'today', 'week','watch',
                   'stream', 'houseofdragon', 'houseofthedragonhbo','houseofthedragonep']

leks39

#selecting main characters from HOTD show
hotd_chars = ['rhaenyra', 'daemon','alicent', 'otto', 'viserys', 'rhaenys','vaemond', 'lucerys', 'jacaerys', 'lyonel',
                   'corlys','aegon', 'aemond', 'larys','harwin','laenor', 'laena', 'criston', 'mysaria', 'helaena']
                   
#replacing mispellings and nicknames of character names with actual names
hotd_df['Text'] = hotd_df['Text'].str.replace('rhaneyra','rhaenyra')
hotd_df['Text'] = hotd_df['Text'].str.replace('rhaneyras','rhaenyra')
hotd_df['Text'] = hotd_df['Text'].str.replace('rhaneyratargaryen','rhaenyra')
hotd_df['Text'] = hotd_df['Text'].str.replace('visery','viserys')
hotd_df['Text'] = hotd_df['Text'].str.replace('viseryss','viserys')

leks39

##GETTING TOP TWEET LOCATIONS
#replace empty values with NaN - *this affects the location column
hotd = hotd.replace(r'^\s*$', np.nan, regex=True)
hotd.head()

#filling null values with next value using bfill method - replaces Nan with next value 
hotd_df = hotd.fillna(method = 'bfill')
hotd_df.head()

#visualization for top 10 tweet locations

leks39

#hotd1_df - October 2022 tweets - Episode 7 to 10
maxTweets = 200000
# Creating list to append tweet data to
tweets_list = []

# Using TwitterSearchScraper to scrape data and append tweets to list
for i,tweet in enumerate(sntwitter.TwitterSearchScraper
                         ('#HouseoftheDragon since:2022-08-21 until:2022-10-24 lang:"en"').get_items()):
    if i>maxTweets:
        break

leks39

#import libraries

#core python packages for data analysis and arithmetics
import pandas as pd
import numpy as np
pd.set_option('max_colwidth', None)

#vizualization packages
import seaborn as sns
import matplotlib.pyplot as plt

leks39

#Genre Distribution
Genre_Distr = result.Genre_Name.value_counts()
Genre_Distr

#Visualisating Genre distribution
sns.countplot(y = 'Genre_Name', data = result, palette='viridis', 
              order = result['Genre_Name'].value_counts().index).set(title='Genre Distribution')

leks39

#Splitting songs into 5 genres
kmeans = KMeans(n_clusters = 5)
label = kmeans.fit_predict(pca_scores)
unique_labels = np.unique(label)

#Plotting clusters
for i in unique_labels:
    plt.scatter(pca_scores[label==i,0], pca_scores[label==i,1], label=i, s=80)
    
plt.legend()

leks39

#standardization
scaler = StandardScaler()
audio_features2_std = scaler.fit_transform(audio_features2)

#fitting standardised data using PCA
pca = PCA()
pca.fit(audio_features2_std)

#variance generated by each feature 
pca.explained_variance_ratio_