PANDAS in Machine Learning

• Dataframe
• Indexing and Selection
• Descriptive Statistics
• Handling missing data
• Reading and Writing files

Dataframe

# Import pandas
import pandas as pd

# Using python dictionary prepare data
data = {'Name':['Bill','Dan','Tony','Mark'],
         'Age':[28,29,31,27],   
         'Salary':[2000,2500,2100,2200]}
         
obj = pd.DataFrame(data)

# Prints the dataframe in tabular form
print obj   
'''
Output: 
      
   Age  Name  Salary
0   28  Bill    2000
1   29   Dan    2500
2   31  Tony    2100
3   27  Mark    2200
'''


# Prints the data of a specific column
print obj['Age']   
print obj.Age
'''
Output:

0    28
1    29
2    31
3    27
Name: Age, dtype: int64
'''

# Prints the columns of the dataframe
print obj.columns
'''
Output:

Index([u'Age', u'Name', u'Salary']
            , dtype='object')
'''

# Returns the data as a 2d array
print obj.values
'''
Output:

[[28L 'Bill' 2000L]
 [29L 'Dan' 2500L]
 [31L 'Tony' 2100L]
 [27L 'Mark' 2200L]]
'''

# Drops a specific row
print obj.drop(1)
'''
Output:

   Age  Name  Salary
0   28  Bill    2000
2   31  Tony    2100
3   27  Mark    2200
'''

# Drops a specific column
print obj.drop('Age',axis=1)
'''
Output:

   Name  Salary
0  Bill    2000
1   Dan    2500
2  Tony    2100
3  Mark    2200
'''

Indexing and Selection

# Gives specified indexing to your rows
data = pd.DataFrame(data, 
    index=['emp1','emp2','emp3','emp4'])

print data
'''
Output:

      Age  Name  Salary
emp1   28  Bill    2000
emp2   29   Dan    2500
emp3   31  Tony    2100
emp4   27  Mark    2200
'''

# Indexing format
# data[rows,columns]

# 1. Column selection

# If we know the column name
print data[['Age','Name']]
print data.loc[:,['Age','Name']]
print data.ix[:,['Age','Name']]

# If we don't know the column name
print data.iloc[:,[0,1]]

'''
Output:

      Age  Name
emp1   28  Bill
emp2   29   Dan
emp3   31  Tony
emp4   27  Mark
'''

# 2. Row selection

# If we know the index name
# prints indexes 0,1,2
print data.loc['emp1':'emp3',:]
'''
Output:

      Age  Name  Salary
emp1   28  Bill    2000
emp2   29   Dan    2500
emp3   31  Tony    2100
'''
# If we don't know index name
# prints indexes 0 and 1 excludes the last one
print data.iloc[0:2,:]
'''
Output:

      Age  Name  Salary
emp1   28  Bill    2000
emp2   29   Dan    2500
'''

# 3. Mixed selection

print data.loc['emp1':'emp3',['Name','Age']]
'''
Output:

      Name  Age
emp1  Bill   28
emp2   Dan   29
emp3  Tony   31
'''

# prints row no. 0,2,3 and column no. 0 and 2 
print data.iloc[[0,2,3],[0,2]]
'''
Output:
      Age  Salary
emp1   28    2000
emp3   31    2100
emp4   27    2200
'''

# Filtering data

# prints Name and Salary of those emloyees
# whose age is greater than 28
print data.loc[data.Age > 28,['Name','Salary']]
'''
Output:

      Name  Salary
emp2  Dan    2500
emp3  Tony   2100
'''

Descriptive Statistics

# prints multiple statistics
print data.describe()
'''
Output:

             Age      Salary
count   4.000000     4.00000
mean   28.750000  2200.00000
std     1.707825   216.02469
min    27.000000  2000.00000
25%    27.750000  2075.00000
50%    28.500000  2150.00000
75%    29.500000  2275.00000
max    31.000000  2500.00000
'''

# prints mean of salaries
print data.loc[:,'Salary'].mean()
'''
Output:

2200.0
'''

# prints minimum age of employee
print data.loc[:,'Age'].min()
'''
Output:

27
'''

Handling missing data

data = pd.DataFrame([[2.3,3.3,NaN],
      [7.5,NaN,9.8],[NaN,2.2,6.8],
      [5.6,9.2,7.4],[NaN,NaN,NaN]])

# 1. Filtering missing data

# Drop rows with null values
print data.dropna()
'''
Output:

     0    1    2
3  5.6  9.2  7.4
'''

# Drop row with all null values
print data.dropna(how='all')
'''
Output:

     0    1    2
0  2.3  3.3  NaN
1  7.5  NaN  9.8
2  NaN  2.2  6.8
3  5.6  9.2  7.4
'''

# 2. Filling missing data

# Fill null values with 0
print data.fillna(0)
'''
Output:

     0    1    2
0  2.3  3.3  0.0
1  7.5  0.0  9.8
2  0.0  2.2  6.8
3  5.6  9.2  7.4
4  0.0  0.0  0.0
'''

# Fill null values with mean
print data.fillna(mean(data))
'''
Output:
    
          0    1    2
0  2.300000  3.3  8.0
1  7.500000  4.9  9.8
2  5.133333  2.2  6.8
3  5.600000  9.2  7.4
4  5.133333  4.9  8.0
'''

# Null values for specific column
print data.fillna({0:2.5, 1:3.0, 2:5.5})
'''
Output:

     0    1    2
0  2.3  3.3  5.5
1  7.5  3.0  9.8
2  2.5  2.2  6.8
3  5.6  9.2  7.4
4  2.5  3.0  5.5
'''

Reading and Writing files

# Reads csv file
data = pd.read_csv('data.csv')

# Prints top 5 rows
print data.head()

'''
File Type     Reader

CSV           read_csv
JSON          read_json 
MS Excel      read_excel
SQL           read_sql
HTML          read_html
'''

# Saves the dataframe to a csv file
pd.DataFrame(data).to_csv('myfile.csv')