rsmodel.py

import pandas as pd
import numpy as np
#import pickle

series = pd.read_csv("TV Series.csv")
series_data = pd.read_csv("TV Series.csv")

series_data=series_data.iloc[:10000]

#EDA
series_data = series_data.fillna('')
series_data.drop('Runtime',axis=1,inplace=True)

def clean_data(x):
    return str.lower(x.replace(" ", ""))

for i in series_data.columns:
    series_data[i] = series_data[i].apply(clean_data)

def create_soup(x):
    return x['Series Title']+ ' ' + x['Release Year'] + ' ' + x['Genre'] + ' ' +x['Rating']+' '+ x['Cast']+' '+ x['Synopsis']
series_data['soup'] = series_data.apply(create_soup, axis=1)

#tf-idf
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics.pairwise import cosine_similarity
count = CountVectorizer(stop_words='english')
count_matrix = count.fit_transform(series_data['soup'])
cosine_sim = cosine_similarity(count_matrix, count_matrix)

series_data=series_data.reset_index()
indices = pd.Series(series_data.index, index=series_data['Series Title'])

def get_recommendations_new(title):
    global cosine_sim
    title=title.replace(' ','').lower()
    idx = indices[title]

    # Get the pairwise similarity scores of all movies with that series
    sim_scores = list(enumerate(cosine_sim[idx]))

    # Sort the series based on the similarity scores
    sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True)

    # Get the scores of the 10 most similar series
    sim_scores = sim_scores[1:11]

    # Get the series indices
    movie_indices = [i[0] for i in sim_scores]

    # Return the top 10 most similar series
    result =  series['Series Title'].iloc[movie_indices]
    result = result.to_frame()
    result = result.set_index(np.arange(1,len(result)+1))
    return result


#pickle.dump(get_recommendations_new,open('rsmodel.pkl','wb'))