Cats - Unsupervised!

import pandas as pd
import numpy as np
from pathlib import Path
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
from sklearn import metrics
from scipy.spatial.distance import cdist

# loading the data
data_folder = Path("E:/Downloads")
file_to_open = data_folder / "Cat_personality_data.csv"
df = pd.read_csv(file_to_open)
df.head()

# cleaning up names
names = []
for i in df.columns[0:1]:
    #print(i[14:])
    names.append(i[14:])
for i in df.columns[1:9]:
    #print(i[13:])
    names.append(i[13:])
for i in df.columns[9:-2]:
    #print(i[14:])
    names.append(i[14:])
for i in df.columns[-2:]:
    #print(i)
    names.append(i)
df.columns=names
# visualizing the results
df.head()

df=df.iloc[:,:-2] # we drop gender and nationality for now

# entering it into k-means clustering
X = df.values
kmeans = KMeans(
    n_clusters=5,
    random_state=0).fit(X)

kmeans.labels_ #these are the clusters it came up with

array([2, 3, 2, ..., 0, 4, 4])

df['cluster']=kmeans.labels_

# thanks to: 
# https://pythonprogramminglanguage.com/kmeans-elbow-method/ 
# k means determine k
distortions = []
K = range(1,20)
for k in K:
    kmeanModel = KMeans(n_clusters=k).fit(X)
    kmeanModel.fit(X)
    distortions.append(sum(np.min(cdist(X, kmeanModel.cluster_centers_, 'euclidean'), axis=1)) / X.shape[0])
# Plot the elbow
plt.plot(K, distortions, 'bx-')
plt.xlabel('k')
plt.ylabel('Distortion')
plt.title('The Elbow Method showing the optimal k')
plt.show()

# repeating what we did above with clusters = 7 
X = df.values
kmeans = KMeans(
    n_clusters=7,
    random_state=0).fit(X)
df['cluster']=kmeans.labels_

cluster_attributes=df.groupby(['cluster']).mean()

zero_cluster=cluster_attributes[cluster_attributes.index==0].T

# Get top 3 and bottom 3 for one cluster
zero_cluster=cluster_attributes[cluster_attributes.index==0].T
sorted_zero_cluster=zero_cluster.sort_values(by=0)
top_sorted_zero_cluster=sorted_zero_cluster.iloc[[49,50,51]]
bottom_sorted_zero_cluster=sorted_zero_cluster.iloc[[0,1,2]]

bottom_sorted_curr_cluster.index.tolist()

['Aggressive_to_people', 'Irritable', 'Tense']

bottom_sorted_zero_cluster

# Repeat for each cluster:
cluster_output = []
top_attributes = []
bot_attributes = []
for i in range(7):
    curr_cluster=cluster_attributes[cluster_attributes.index==i].T
    sorted_curr_cluster=curr_cluster.sort_values(by=i)
    top_sorted_curr_cluster=sorted_curr_cluster.iloc[[49,50,51]]
    bottom_sorted_curr_cluster=sorted_curr_cluster.iloc[[0,1,2]]
    cluster_output.append([i,i,i])
    top_attributes.append(top_sorted_curr_cluster.index.tolist())
    bot_attributes.append(bottom_sorted_curr_cluster.index.tolist())
    #print("cluster ",i," top traits:")
    #print(top_sorted_curr_cluster)
    #print("cluster ",i," bottom traits:")
    #print(bottom_sorted_curr_cluster)

top_df = pd.DataFrame(data=top_attributes, columns=['highest', 'second highest', 'third highest'])

bot_df = pd.DataFrame(data=bot_attributes,columns=['lowest', 'second lowest', 'third lowest'])

results_df=top_df.join(bot_df, rsuffix='_bot')

results_df.T

For those same categories, Fearful_of_people is low. Makes sense.

More interesting perhaps, are the "smart cats". The ravenclaws, if you will.
Cluster 0 is still affectionate, but is also rated highly for Vigilance and Smarts.
Cluster 2 has actually lost affection as one of their top attributes, and has "Suspicious"
Cluster 3 is highest rated as inquisitive, as well as self assured.

Finally, poor cluster 4. They're Predictable, Insecure, and Suspicious.

df.groupby('cluster').count()

#all cats included
df.loc[:,['Friendly_to_people']].hist(bins=7)

array([[<matplotlib.axes._subplots.AxesSubplot object at 0x00000000159E7358>]],
      dtype=object)

# just one of the friendlies
one=df[df.cluster==1]
one.loc[:,['Friendly_to_people']].hist(bins=7)

array([[<matplotlib.axes._subplots.AxesSubplot object at 0x0000000015D74748>]],
      dtype=object)

# the insecure cats
one=df[df.cluster==4]
one.loc[:,['Friendly_to_people']].hist(bins=7)

array([[<matplotlib.axes._subplots.AxesSubplot object at 0x0000000015DB6828>]],
      dtype=object)