Recommender_system / recomender_system_cluster_based.py

added notes for this duplication

a2f4653 over 3 years ago

7.24 kB

# -*- coding: utf-8 -*-
"""recomender2.ipynb
Automatically generated by Colaboratory.
Original file is located at
    https://colab.research.google.com/drive/1WTIRumzmdqNSB2Z2dIv3r2LkSN5CdoU5
"""
!pip install transformers
from google.colab import drive
drive.mount('/content/drive')

import os
os.environ['KAGGLE_CONFIG_DIR'] = "/content/drive/MyDrive/kaggle"

# Commented out IPython magic to ensure Python compatibility.
# %ls

import pandas as pd
import numpy as np
from matplotlib import pyplot as plt
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score
from sklearn.preprocessing import scale, StandardScaler, normalize
from yellowbrick.cluster import KElbowVisualizer
from sklearn.manifold import TSNE
import seaborn as sns
import random

df1=pd.read_csv('/content/drive/MyDrive/kaggle/ratings.csv')
df2=pd.read_csv('/content/drive/MyDrive/kaggle/movies_metadata.csv')
df3=pd.read_csv('/content/drive/MyDrive/kaggle/ratings_small.csv')

df2

df2.shape

df2.columns

df2.info()

df2.isna().sum()

data=df2[['revenue','runtime','vote_average','vote_count','title']]

data

data=data.fillna(0)

#scale the data

scaler = StandardScaler()
data_scaled = scaler.fit_transform(data.iloc[:, :-1])
data_scaled

inertia = []
silhouette_scores = []
k_values = range(2, 10)
for k in k_values:
    kmeans = KMeans(n_clusters=k, random_state=3,n_init=4)
    kmeans.fit(data_scaled)
    #center=kmeans.cluster_centers
    inertia.append(kmeans.inertia_)
    silhouette_scores.append(silhouette_score(data_scaled, kmeans.labels_))
    print(silhouette_scores)

plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)
plt.plot(k_values, inertia, marker='o')
plt.xlabel('Number of clusters (k)')
plt.ylabel('Inertia')
plt.title('Inertia vs. Number of Clusters')

plt.subplot(1, 2, 2)
plt.plot(k_values, silhouette_scores, marker='o')
plt.xlabel('Number of clusters (k)')
plt.ylabel('Silhouette Score')
plt.title('Silhouette Score vs. Number of Clusters')

plt.tight_layout()
plt.show()

model = KMeans(n_init=8)
visualiser = KElbowVisualizer(model, k=(2,12))
visualiser.fit(data_scaled)
visualiser.show()

kmeans_mod = KMeans(n_clusters=5, n_init=10)
kmeans_mod.fit(data_scaled)

labels =kmeans_mod.fit_predict(data_scaled)
data['cluster']=labels
data=data[['revenue','runtime','vote_average','vote_count','cluster','title']]
data

data['cluster'].value_counts()

tsne = TSNE(n_components=2, random_state=42)
X_tsne = tsne.fit_transform(data_scaled)

plt.figure(figsize=(8,6))
plt.scatter(X_tsne[:, 0], X_tsne[:, 1], c=kmeans.labels_, cmap='viridis')
plt.xlabel('t-SNE Dimension 1')
plt.ylabel('t-SNE Dimension 2')
plt.title('Clustering Results with K-Means (k=5)'.format(5))
plt.colorbar(label='Cluster')
plt.show()

cdata=data[['cluster','title']]
cdata

cluster0=[]
cluster1=[]
cluster2=[]
cluster3=[]
cluster4=[]

filmname=''
for i in cdata.iloc():
     if i[0]==0:
      filmname=''
      filmname=i[1]
      #print(filmname)
      cluster0.append(filmname)
     elif i[0]==1:
        filmname=''
        filmname=i[1]
        cluster1.append(filmname)
     elif i[0]==2:
        filmname=''
        filmname=i[1]
        cluster2.append(filmname)
     if i[0]==3:
        filmname=''
        filmname=i[1]
        cluster3.append(filmname)
     if i[0]==4:
        filmname=''
        filmname=i[1]
        cluster4.append(filmname)
print(len(cluster0))
print(len(cluster1))
print(len(cluster2))
print(len(cluster3))
print(len(cluster4))


print((cluster0))
print((cluster1))
print((cluster2))
print((cluster3))
print((cluster4))

data=data.iloc[:, :-1]
data

from sklearn.decomposition import PCA
pca_mod = PCA(n_components=3)
# fit the PCA model to the (scaled) features
pca_mod.fit(data)

# plot the cumulative variance explained across all the components
# start the explained variance at 0
exp_variance = np.insert(pca_mod.explained_variance_ratio_, 0, 0)

# plot the explained variance
figure = plt.figure(figsize=(8, 5))

plt.plot(np.cumsum(exp_variance))
plt.xticks(range(0,6))
plt.xlabel("Number of principal components (PCs)", size=13)
plt.ylabel("% cumulative variance explained", size=13)
plt.title("Cumulative variance explained", size=15)

plt.show()
np.cumsum(exp_variance)

pca_mod = PCA(n_components=3)

data_pca = pd.DataFrame(data=pca_mod.fit_transform(data_scaled),
                        columns=['PC1', 'PC2', 'PC3'])

# view data frame of PCs
data_pca.head()

# combine data for plotting
data.reset_index(drop=True, inplace=True)
data_pca.reset_index(drop=True, inplace=True)
plot_data = pd.concat([data_pca, data['cluster']], axis=1)

# visualise the clusters by plotting the PCAs in a scatter/pair plot,
# with points colored by cluster membership
sns.pairplot(data=plot_data,
             hue='cluster',
             palette='Set1',
             corner=True)

cluster=[cluster0,cluster1,cluster2,cluster3,cluster4]

def recomender(input1,input2,input3):

  encod_input1=[]
  encod_input2=[]
  encod_input3=[]


  encod_input1.append(cluster0.count(input1))
  encod_input1.append(cluster1.count(input1))
  encod_input1.append(cluster2.count(input1))
  encod_input1.append(cluster3.count(input1))
  encod_input1.append(cluster4.count(input1))



  encod_input2.append(cluster0.count(input2))
  encod_input2.append(cluster1.count(input2))
  encod_input2.append(cluster2.count(input2))
  encod_input2.append(cluster3.count(input2))
  encod_input2.append(cluster4.count(input2))



  encod_input3.append(cluster0.count(input3))
  encod_input3.append(cluster1.count(input3))
  encod_input3.append(cluster2.count(input3))
  encod_input3.append(cluster3.count(input3))
  encod_input3.append(cluster4.count(input3))


  return encod_input1,encod_input2,encod_input3


def recomender2(a):
  s1=0
  s2=0
  s3=0
  for i in range(0,5):
    if a[0][i]!=0:
        s1=i
        #print(f'The first film is in cluster {i+1}')
    if a[1][i]!=0:
       s2=i
       #print(f'The second film is in cluster {i+1}')
    if a[2][i]!=0:
       s3=i
       #print(f'The third film is in cluster {i+1}')
  return s1, s2, s3



def recom(b):
  if b[0]==b[1]==b[2]:
    return random.choice(cluster[b[1]])
  if b[0]==b[1]!=b[2]:
    return random.choice(cluster[b[1]])
  if b[0]!=b[1]==b[2]:
     return random.choice(cluster[b[1]])
  if b[0]==b[2]!=b[1]:
     return random.choice(cluster[b[2]])
  if  b[0]!=b[2]!=b[1]:
     return random.choice(cluster[b[0]])

def original_recom(input1,input2,input3):
  return recom(recomender2(recomender(input1,input2,input3)))

a=recomender('The Ballroom','Scooby-Doo! and the Reluctant Werewolf','Last Summer')
a

b=recomender2(a)
print(b)

recom(b)

cluster4.count("Trial")

!pip install gradio

import numpy as np
import gradio as gr


def original_recom(input1,input2,input3):
  return recom(recomender2(recomender(input1,input2,input3)))
with gr.Blocks() as demo:
    gr.Markdown("Flip text or image files using this demo.")

    input1 = gr.Textbox(placeholder='first film name...')

    input2 = gr.Textbox(placeholder='second film name...')

    input3 = gr.Textbox(placeholder='third film name...')

    text_output=gr.Textbox( )


    text_button = gr.Button("Submit")
    text_button.click(original_recom, inputs=[input1,input2,input3], outputs=text_output)
demo.launch()

	# -- coding: utf-8 --
	"""recomender2.ipynb

	Automatically generated by Colaboratory.

	Original file is located at
	https://colab.research.google.com/drive/1WTIRumzmdqNSB2Z2dIv3r2LkSN5CdoU5
	"""
	!pip install transformers
	from google.colab import drive
	drive.mount('/content/drive')

	import os
	os.environ['KAGGLE_CONFIG_DIR'] = "/content/drive/MyDrive/kaggle"

	# Commented out IPython magic to ensure Python compatibility.
	# %ls

	import pandas as pd
	import numpy as np
	from matplotlib import pyplot as plt
	from sklearn.cluster import KMeans
	from sklearn.metrics import silhouette_score
	from sklearn.preprocessing import scale, StandardScaler, normalize
	from yellowbrick.cluster import KElbowVisualizer
	from sklearn.manifold import TSNE
	import seaborn as sns
	import random

	df1=pd.read_csv('/content/drive/MyDrive/kaggle/ratings.csv')
	df2=pd.read_csv('/content/drive/MyDrive/kaggle/movies_metadata.csv')
	df3=pd.read_csv('/content/drive/MyDrive/kaggle/ratings_small.csv')

	df2

	df2.shape

	df2.columns

	df2.info()

	df2.isna().sum()

	data=df2[['revenue','runtime','vote_average','vote_count','title']]

	data

	data=data.fillna(0)

	#scale the data

	scaler = StandardScaler()
	data_scaled = scaler.fit_transform(data.iloc[:, :-1])
	data_scaled

	inertia = []
	silhouette_scores = []
	k_values = range(2, 10)
	for k in k_values:
	kmeans = KMeans(n_clusters=k, random_state=3,n_init=4)
	kmeans.fit(data_scaled)
	#center=kmeans.cluster_centers
	inertia.append(kmeans.inertia_)
	silhouette_scores.append(silhouette_score(data_scaled, kmeans.labels_))
	print(silhouette_scores)

	plt.figure(figsize=(12, 4))
	plt.subplot(1, 2, 1)
	plt.plot(k_values, inertia, marker='o')
	plt.xlabel('Number of clusters (k)')
	plt.ylabel('Inertia')
	plt.title('Inertia vs. Number of Clusters')

	plt.subplot(1, 2, 2)
	plt.plot(k_values, silhouette_scores, marker='o')
	plt.xlabel('Number of clusters (k)')
	plt.ylabel('Silhouette Score')
	plt.title('Silhouette Score vs. Number of Clusters')

	plt.tight_layout()
	plt.show()

	model = KMeans(n_init=8)
	visualiser = KElbowVisualizer(model, k=(2,12))
	visualiser.fit(data_scaled)
	visualiser.show()

	kmeans_mod = KMeans(n_clusters=5, n_init=10)
	kmeans_mod.fit(data_scaled)

	labels =kmeans_mod.fit_predict(data_scaled)
	data['cluster']=labels
	data=data[['revenue','runtime','vote_average','vote_count','cluster','title']]
	data

	data['cluster'].value_counts()

	tsne = TSNE(n_components=2, random_state=42)
	X_tsne = tsne.fit_transform(data_scaled)

	plt.figure(figsize=(8,6))
	plt.scatter(X_tsne[:, 0], X_tsne[:, 1], c=kmeans.labels_, cmap='viridis')
	plt.xlabel('t-SNE Dimension 1')
	plt.ylabel('t-SNE Dimension 2')
	plt.title('Clustering Results with K-Means (k=5)'.format(5))
	plt.colorbar(label='Cluster')
	plt.show()

	cdata=data[['cluster','title']]
	cdata

	cluster0=[]
	cluster1=[]
	cluster2=[]
	cluster3=[]
	cluster4=[]

	filmname=''
	for i in cdata.iloc():
	if i[0]==0:
	filmname=''
	filmname=i[1]
	#print(filmname)
	cluster0.append(filmname)
	elif i[0]==1:
	filmname=''
	filmname=i[1]
	cluster1.append(filmname)
	elif i[0]==2:
	filmname=''
	filmname=i[1]
	cluster2.append(filmname)
	if i[0]==3:
	filmname=''
	filmname=i[1]
	cluster3.append(filmname)
	if i[0]==4:
	filmname=''
	filmname=i[1]
	cluster4.append(filmname)
	print(len(cluster0))
	print(len(cluster1))
	print(len(cluster2))
	print(len(cluster3))
	print(len(cluster4))


	print((cluster0))
	print((cluster1))
	print((cluster2))
	print((cluster3))
	print((cluster4))

	data=data.iloc[:, :-1]
	data

	from sklearn.decomposition import PCA
	pca_mod = PCA(n_components=3)
	# fit the PCA model to the (scaled) features
	pca_mod.fit(data)

	# plot the cumulative variance explained across all the components
	# start the explained variance at 0
	exp_variance = np.insert(pca_mod.explained_variance_ratio_, 0, 0)

	# plot the explained variance
	figure = plt.figure(figsize=(8, 5))

	plt.plot(np.cumsum(exp_variance))
	plt.xticks(range(0,6))
	plt.xlabel("Number of principal components (PCs)", size=13)
	plt.ylabel("% cumulative variance explained", size=13)
	plt.title("Cumulative variance explained", size=15)

	plt.show()
	np.cumsum(exp_variance)

	pca_mod = PCA(n_components=3)

	data_pca = pd.DataFrame(data=pca_mod.fit_transform(data_scaled),
	columns=['PC1', 'PC2', 'PC3'])

	# view data frame of PCs
	data_pca.head()

	# combine data for plotting
	data.reset_index(drop=True, inplace=True)
	data_pca.reset_index(drop=True, inplace=True)
	plot_data = pd.concat([data_pca, data['cluster']], axis=1)

	# visualise the clusters by plotting the PCAs in a scatter/pair plot,
	# with points colored by cluster membership
	sns.pairplot(data=plot_data,
	hue='cluster',
	palette='Set1',
	corner=True)

	cluster=[cluster0,cluster1,cluster2,cluster3,cluster4]

	def recomender(input1,input2,input3):

	encod_input1=[]
	encod_input2=[]
	encod_input3=[]


	encod_input1.append(cluster0.count(input1))
	encod_input1.append(cluster1.count(input1))
	encod_input1.append(cluster2.count(input1))
	encod_input1.append(cluster3.count(input1))
	encod_input1.append(cluster4.count(input1))



	encod_input2.append(cluster0.count(input2))
	encod_input2.append(cluster1.count(input2))
	encod_input2.append(cluster2.count(input2))
	encod_input2.append(cluster3.count(input2))
	encod_input2.append(cluster4.count(input2))



	encod_input3.append(cluster0.count(input3))
	encod_input3.append(cluster1.count(input3))
	encod_input3.append(cluster2.count(input3))
	encod_input3.append(cluster3.count(input3))
	encod_input3.append(cluster4.count(input3))


	return encod_input1,encod_input2,encod_input3


	def recomender2(a):
	s1=0
	s2=0
	s3=0
	for i in range(0,5):
	if a[0][i]!=0:
	s1=i
	#print(f'The first film is in cluster {i+1}')
	if a[1][i]!=0:
	s2=i
	#print(f'The second film is in cluster {i+1}')
	if a[2][i]!=0:
	s3=i
	#print(f'The third film is in cluster {i+1}')
	return s1, s2, s3



	def recom(b):
	if b[0]==b[1]==b[2]:
	return random.choice(cluster[b[1]])
	if b[0]==b[1]!=b[2]:
	return random.choice(cluster[b[1]])
	if b[0]!=b[1]==b[2]:
	return random.choice(cluster[b[1]])
	if b[0]==b[2]!=b[1]:
	return random.choice(cluster[b[2]])
	if b[0]!=b[2]!=b[1]:
	return random.choice(cluster[b[0]])

	def original_recom(input1,input2,input3):
	return recom(recomender2(recomender(input1,input2,input3)))

	a=recomender('The Ballroom','Scooby-Doo! and the Reluctant Werewolf','Last Summer')
	a

	b=recomender2(a)
	print(b)

	recom(b)

	cluster4.count("Trial")

	!pip install gradio

	import numpy as np
	import gradio as gr


	def original_recom(input1,input2,input3):
	return recom(recomender2(recomender(input1,input2,input3)))
	with gr.Blocks() as demo:
	gr.Markdown("Flip text or image files using this demo.")

	input1 = gr.Textbox(placeholder='first film name...')

	input2 = gr.Textbox(placeholder='second film name...')

	input3 = gr.Textbox(placeholder='third film name...')

	text_output=gr.Textbox( )


	text_button = gr.Button("Submit")
	text_button.click(original_recom, inputs=[input1,input2,input3], outputs=text_output)
	demo.launch()