notebook.community

Edit and run



In [1]:

    
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt# Load data
from sklearn import linear_model
from sklearn.cluster import KMeans as km
import statsmodels.api as sm

%matplotlib inline
city_data = pd.read_csv('USData_ClassProject1.csv')



In [16]:

    
x = city_data.PercentWhite
y = city_data.MedianHouseholdIncome 
x = sm.add_constant(x)

clf = linear_model.LinearRegression()
clf.fit(x,y)
y_hat = clf.predict(x)

print "R^2: ", clf.score(x,y)

def cluster(X):
    clf.fit(X)
    return clf.predict(X)

clf,X = km(3), city_data[['PercentWhite','MedianHouseholdIncome']]

X.PercentWhite = (X.PercentWhite - np.mean(X.PercentWhite))/np.std(X.PercentWhite)
X.MedianHouseholdIncome = (X.MedianHouseholdIncome - np.mean(X.MedianHouseholdIncome))/np.std(X.MedianHouseholdIncome)

plt.scatter(city_data.PercentWhite, city_data.MedianHouseholdIncome, c=cluster(X), s=50)
plt.plot(x.PercentWhite, y_hat, 'b', alpha=0.9,color="white")
plt.ylabel('Median Household Income')
plt.xlabel('Percent White')









    



R^2:  0.0318199157966






    Out[16]:





<matplotlib.text.Text at 0x10e846b10>



In [ ]:



In [ ]:

    
X.PercentWhite = (X.PercentWhite - np.mean(X.PercentWhite))/np.std(X.PercentWhite)
X.MedianHouseholdIncome = (X.MedianHouseholdIncome - np.mean(X.MedianHouseholdIncome))/np.std(X.MedianHouseholdIncome)

plt.scatter(rich_people.PercentWhite, rich_people.MedianHouseholdIncome, c=cluster(X), s=50)
plt.ylabel('Median Household Income')
plt.xlabel('Percent White')



In [ ]:

    
print "Mean: " + str(np.mean(city_data.PercentWhite))

print "Median: " + str(np.median(city_data.PercentWhite))

print "Median Population: " + str(np.median(city_data.TotalPopulation))

print "Mean Popluation: " + str(np.mean(city_data.TotalPopulation))



In [ ]:

    
rich_people = city_data[['City','State','MedianHouseholdIncome','PercentWhite', 'TotalPopulation']][city_data.MedianHouseholdIncome >= 150000]
print "Mean white: " + str(np.mean(rich_people.PercentWhite))
print "Median white: " + str(np.median(rich_people.PercentWhite))
rich_people



In [ ]:

    
x = rich_people.PercentWhite
y = rich_people.MedianHouseholdIncome 
x = sm.add_constant(x)

clf = linear_model.LinearRegression()
clf.fit(x,y)

y_hat = clf.predict(x)

plt.scatter(rich_people.PercentWhite,rich_people.MedianHouseholdIncome,color="Black")
plt.plot(x.PercentWhite, y_hat, 'b', alpha=0.9)

print "R^2: ", clf.score(x,y)

Clustering

Used clustering of white people vs Median Household Income

Without Normalization

Clusters are visualy weird. Because income is much more differnt in magnitude then percent of anything with a max of 100.



In [ ]:

    
def cluster(X):
    clf.fit(X)
    return clf.predict(X)

clf,X = km(2), rich_people[['PercentWhite','MedianHouseholdIncome']]


plt.scatter(rich_people.PercentWhite, rich_people.MedianHouseholdIncome, c=cluster(X), s=50)

plt.ylabel('Median Household Income')
plt.xlabel('Percent White')

With Normalization

It makes much more sense. So use normalization!!!!



In [ ]:

    
X.PercentWhite = (X.PercentWhite - np.mean(X.PercentWhite))/np.std(X.PercentWhite)
X.MedianHouseholdIncome = (X.MedianHouseholdIncome - np.mean(X.MedianHouseholdIncome))/np.std(X.MedianHouseholdIncome)

plt.scatter(rich_people.PercentWhite, rich_people.MedianHouseholdIncome, c=cluster(X), s=50)
plt.ylabel('Median Household Income')
plt.xlabel('Percent White')



In [ ]:

    
well_off_people = city_data[['City','MedianHouseholdIncome','PercentWhite', 'TotalPopulation']][city_data.MedianHouseholdIncome >= 100000]
print "Mean white: " + str(np.mean(well_off_people.PercentWhite))
print "Median white: " + str(np.median(well_off_people.PercentWhite))
well_off_people



In [ ]:

    
x = well_off_people.PercentWhite
y = well_off_people.MedianHouseholdIncome 
x = sm.add_constant(x)

clf = linear_model.LinearRegression()
clf.fit(x,y)

y_hat = clf.predict(x)

plt.scatter(well_off_people.PercentWhite,well_off_people.MedianHouseholdIncome,color="Black")
plt.plot(x.PercentWhite, y_hat, 'b', alpha=0.9)

print "R^2: ", clf.score(x,y)



In [ ]:

    
very_white_people = city_data[['City','MedianHouseholdIncome','PercentWhite', 'TotalPopulation']][city_data.PercentWhite >= 95]
print "Mean income: " + str(np.mean(very_white_people.MedianHouseholdIncome))
print "Median income: " + str(np.median(very_white_people.MedianHouseholdIncome))

very_white_people



In [ ]:

    
x = very_white_people.PercentWhite
y = very_white_people.MedianHouseholdIncome 
x = sm.add_constant(x)

clf = linear_model.LinearRegression()
clf.fit(x,y)

y_hat = clf.predict(x)

plt.scatter(very_white_people.PercentWhite,very_white_people.MedianHouseholdIncome,color="Black")
plt.plot(x.PercentWhite, y_hat, 'b', alpha=0.9)

print "R^2: ", clf.score(x,y)



In [ ]:

    
X = city_data.PercentHispanic 
y = city_data.EdHighSchoolPercent
x = sm.add_constant(X)


clf = linear_model.LinearRegression()
clf.fit(x,y)

y_hat = clf.predict(x)

plt.scatter(X,city_data.EdHighSchoolPercent,color="Black")
plt.plot(x.PercentHispanic, y_hat, 'b', alpha=0.9)

plt.xlabel("PercentHispanic percent")
plt.ylabel("Highschool percent")



print "R^2: ", clf.score(x,y)



In [ ]:

    
mean_of_city_percentage_of_people_with_no_collage = np.mean(city_data.EdHighSchoolPercent) + np.mean(city_data.EdElementaryPercent)


mean_of_city_percentage_of_people_with_no_collage

X = city_data.TotalPopulation 
y = city_data.EdHighSchoolPercent
x = sm.add_constant(X)


clf = linear_model.LinearRegression()
clf.fit(x,y)

y_hat = clf.predict(x)

plt.scatter(X,city_data.EdHighSchoolPercent,color="Black")
plt.plot(x.TotalPopulation, y_hat, 'b', alpha=0.9)

plt.xlabel("Total percent")
plt.ylabel("Highschool percent")



print "R^2: ", clf.score(x,y)