Computer Hardware Dataset Analysis - UCI

In [2]:

    

import numpy as np
import pandas as pd
%pylab inline
pylab.style.use('ggplot')
import seaborn as sns


    

    




Populating the interactive namespace from numpy and matplotlib

In [3]:

    

url = 'https://archive.ics.uci.edu/ml/machine-learning-databases/cpu-performance/machine.data'
data = pd.read_csv(url, header=None)


    

In [20]:

    

data.head()


    

    
Out[20]:






  

    

      

      
VENDOR
      
MYCT
      
MMIN
      
MMAX
      
CACH
      
CHMIN
      
CHMAX
      
PRP
    
  
  

    

      
0
      
adviser
      
125
      
256
      
6000
      
256
      
16
      
128
      
198
    
    

      
1
      
amdahl
      
29
      
8000
      
32000
      
32
      
8
      
32
      
269
    
    

      
2
      
amdahl
      
29
      
8000
      
32000
      
32
      
8
      
32
      
220
    
    

      
3
      
amdahl
      
29
      
8000
      
32000
      
32
      
8
      
32
      
172
    
    

      
4
      
amdahl
      
29
      
8000
      
16000
      
32
      
8
      
16
      
132
    
  

In [5]:

    

data.columns = ['VENDOR', 'MODEL', 'MYCT', 'MMIN', 
                 'MMAX', 'CACH', 'CHMIN', 'CHMAX', 'PRP', 'ERP']


    

In [6]:

    

# Drop the ERP column - this is an estimate
data = data.drop('ERP', axis=1)


    

In [7]:

    

data.VENDOR.value_counts().plot(kind='barh')


    

    
Out[7]:





<matplotlib.axes._subplots.AxesSubplot at 0x1b5d8934978>






    

In [8]:

    

# Drop the model column as well
data = data.drop('MODEL', axis=1)


    

In [9]:

    

feature_names = data.columns.drop('VENDOR')

for fname in feature_names:
    _ = pylab.figure()
    _ = data.loc[:, fname].plot(kind='hist', title=fname)


    

In [10]:

    

_, axes = pylab.subplots(6, figsize=(10, 21))

n_columns = data.columns.drop(['VENDOR', 'PRP'])

for i, fname in enumerate(n_columns):
    sns.regplot(x=fname, y='PRP', data=data, ax=axes[i])
    
pylab.tight_layout()


    

In [11]:

    

corrs = data.loc[:, n_columns].corrwith(data.loc[:, 'PRP'])
corrs.plot(kind='barh')


    

    
Out[11]:





<matplotlib.axes._subplots.AxesSubplot at 0x1b5da716da0>






    

In [12]:

    

f_corrs = data.loc[:, n_columns].corr()
sns.heatmap(f_corrs, annot=True)


    

    
Out[12]:





<matplotlib.axes._subplots.AxesSubplot at 0x1b5da25def0>






    

In [13]:

    

import statsmodels.formula.api as sm


    

In [14]:

    

model = sm.ols(formula='PRP ~ MMAX + MMIN + CACH + CHMAX', data=data)
result = model.fit()
result.summary()


    

    
Out[14]:





OLS Regression Results

  
Dep. Variable:
           
PRP
       
  R-squared:         
 
   0.860


  
Model:
                   
OLS
       
  Adj. R-squared:    
 
   0.857


  
Method:
             
Least Squares
  
  F-statistic:       
 
   312.1


  
Date:
             
Sun, 06 Aug 2017
 
  Prob (F-statistic):
 
9.84e-86


  
Time:
                 
09:32:36
     
  Log-Likelihood:    
 
 -1152.7


  
No. Observations:
      
   209
      
  AIC:               
 
   2315.


  
Df Residuals:
          
   204
      
  BIC:               
 
   2332.


  
Df Model:
              
     4
      
                     
     
 
   


  
Covariance Type:
      
nonrobust
    
                     
     
 
   




      
         
coef
     
std err
      
t
      
P>|t|
  
[0.025
    
0.975]
  


  
Intercept
 
  -40.9794
 
    6.042
 
   -6.783
 
 0.000
 
  -52.891
 
  -29.068


  
MMAX
      
    0.0053
 
    0.001
 
    8.264
 
 0.000
 
    0.004
 
    0.007


  
MMIN
      
    0.0149
 
    0.002
 
    8.202
 
 0.000
 
    0.011
 
    0.018


  
CACH
      
    0.5873
 
    0.136
 
    4.326
 
 0.000
 
    0.320
 
    0.855


  
CHMAX
     
    1.4359
 
    0.211
 
    6.807
 
 0.000
 
    1.020
 
    1.852




  
Omnibus:
       
105.696
 
  Durbin-Watson:     
 
   1.143
 


  
Prob(Omnibus):
 
 0.000
  
  Jarque-Bera (JB):  
 
1234.809
 


  
Skew:
          
 1.613
  
  Prob(JB):          
 
7.32e-269


  
Kurtosis:
      
14.463
  
  Cond. No.          
 
2.46e+04
 

In [15]:

    

from sklearn.model_selection import KFold
from sklearn.metrics import r2_score


    

In [26]:

    

n_splits = 3
fold = KFold(n_splits=n_splits, shuffle=True)

scores = []

for train_idx, test_idx in fold.split(data):
    model = sm.ols(formula='PRP ~ MMAX + MMIN + CACH + CHMAX', data=data.loc[train_idx])
    result = model.fit()
    test_features = data.loc[test_idx].drop('PRP', axis=1)
    
    predictions = result.predict(test_features)
    actual = data.loc[test_idx, 'PRP']
    
    score = r2_score(actual, predictions)
    scores.append(score)
    
scores = pd.Series(scores)


    

In [27]:

    

scores.plot(kind='bar')


    

    
Out[27]:





<matplotlib.axes._subplots.AxesSubplot at 0x1b5da3fa9b0>