4)12th_L-1_PYTHON-PANDAS-I.pptx

PYTHON-PANDAS-I
STD-XII
INFORMATICS PRACTICES
LESSON-1
Order of some data structures
learnt so far:-
List
Dictionary
Numpy array ndarray
Series
Data Frame
list[1,’ab’,7.5] heterogeneous data(implicit indexing)
dictionary{1:’ab’,2:’cd’}
Ndarray[1,’ab’,2.5][‘1’,’ab’,’2,5’] homogeneous
data(implicit indexing)
Series[1,2,3] homogeneous data(explicit indexing+implicit)
Dataframeheterogeneous data
0 1
0 Mona 20
1 Gita 30
0 1
1 2
2 3
1D
2D
1)
2) Data Structures with egs
Mutable Immutable
List
Dictionary
Sets
Ndarray(size immutable)
Series(size immutable)
DataFrames
Tuple
Int
Float
Boolean
String
unicode
3)

1)HOW TO INSTALL THE pandas library in idle

7) open cmd
8) type cd press enter
9) type cd<space>paste the copied path here press enter
10) So now you are in the Scripts folder.
11) Type python –m pip install --upgrade pip (press enter)
12) Type pip install pandas
Pip- full form
preferred
installer
program
c

I. INTRODUCTION
II. USING PANDAS
III. WHY PANDAS?
IV. PANDAS DATA STRUCTURES
I- SERIES
A. CREATING SERIES
1) Creating empty series
2) Creating non empty series
i. Python sequence
ii. An ndarray
iii. Python dictionary
iv. Scalar value
3) Creating Series Objects-Additional
Functionality
i. Adding NaN in Series
ii. Specify the index and data for the
Series
iii. Specify the datatype with index and
data
iv. Using a mathematical func. For data of
a Series.
B. SERIES OBJECTS ATTRIBUTES
C. ACCESSING A SERIES OBJECTS AND
ELEMENTS
i. Access individual elements
ii. Slicing a series objects
D. OPERATIONS ON SERIES OBJECTS
i. Modify elements
ii. head() and tail()
iii. Vector Operations on Series objects
iv. Arithmetic on Series
v. Filtering Entries
vi. Sorting Series Values
E. DIFFERENCE BETWEEN NUMPY AND SERIES
OBJECTS
II -DataFrame
I -Series
SUMMARY-DATA SERIES()
Syntax ds=pd.Series(data,[index],[dtype])
did

I) INTRODUCTION
Pandas or Python Pandas-It is a python library for data analysis. Pan|da|s
which is a term for quantitative analysis for multi-dimensional structured data sets.
Data analysis- It is the process of evaluating big data sets using analytical and statistical tools to discover useful
information and conclusions to help in decision making.
The main author of Pandas is WesMcKinney
II) USING PANDAS
Pandas is an opensource BSD(Berkeley Software Distribution license) library built for Python Programming Language. It provides
high-performance, easy to use data structures and data-analysis tool.
To work with pandas you need to import
import pandas as pd
panel data
system
Def:-
Pandas has become a very popular choice
for data analysis.
Data analysis-refers to the process of
evaluating big data sets using analytical
and statistical tools so as to discover
useful information and conclusions to
support business decision-making.
Extra-
Pandas is a BSD licensed, open source package of Python which is popular for data science. It has been built
on the Numpy package. It offers powerful, flexible and expressive data structures that make the manipulation
of the data and make the analysis easier. One of the data structures available is the DataFrame. The Pandas
DataFrame can be seen as a table. In this data structure, data is organized into rows and columns, which
makes a two dimensional data structure. The size of this data structure is mutable and can be modified.
extra
BSD license imposes minimal
restrictions on the use and
distribution of software.

III) WHY PANDAS?
 It can read or write in many diff data formats(integers,float,double,etc)
It can calculate in all possible ways(across rows and columns)
It support reshaping the data into diff forms
It supports visualization using matplotlib and seaborn etc. libraries
more pg(2)
IV) PANDAS DATA STRUCTURES
 DataStructures-They refer to a specialized way of storing data so as to apply a specific type of
functionality on them.
A Data structure is a particular way of storing and organizing data in a computer to suit a specific purpose so that it can be accessed and
worked with in appropriate ways. ***
Depending on the requirement of the situation a data structure is decided for that situation
Two very basic data structures of Python
Def:-
I -Series
II- DataFrame
There are many more data structures such as panels but we are not studying these right now.

TWO BASIC STRUCTURES SERIES AND DATAFRAME
SERIES-1D data structure of Python Pandas
DataFrame- 2D data structure of Python Pandas
Series DataFrame Object
Q Diff between Series and DataFrames
row
index
(index)
row index(index)
data
data
Series DataFrames
Type Of Data Homogeneous, 1D Heterogenous,2D
Mutability Value Mutable
(elements value can change)
Value Mutable
(elements value can
change)
Size immutable(once created its size
cannot be changed). If you want to add
or drop elements, internally a new series
object will be created.
Size mutable(once created
its size can be
changed)You can add or
delete elements in an
existing df.
Object datatype explanation link-
https://stackoverflow.com/questions/21018654/st
rings-in-a-dataframe-but-dtype-is-object
So in Series and df the datatype string is called
object datatype(explanation in the link)
column index(columns)
?

Data Series(size immutability ka explanation) just for understanding not compulsory(ndarray also size immutable)
Same address of ds
Diff Address of the array
of the ds
same
diff
EXTRA
SLIDE

I) Series-
It is a pandas data structure that represents a one dimensional array-like object containing an array of data(of any numpy
datatype) and an associated array of data labels, called its index.
*It is 1D
*It has 2 main components
1) an array of actual data
2)an associated array of indexes or data labels.
1) Creating Empty Series Object
Syntax <series obj>=pandas.Series() # egds=pd.Series()
The above statement will create an empty Series type Objects with no values and having a default datatype as ‘float64’
Def:-
Index Data
0 21
1 22
2 23
3 24
Index Data
Jan 21
Feb 22
March 23
April 24
Index Data
A 21
B 22
C 23
D 24
A) CREATING SERIES OBJECTS

Prog1
<Series Object>=pd.Series(data=,[index=idx],[dtype])
data- i)Python sequencelist,tuple(set cannot be used here, it is
unordered and has no indexing at all)
ii)ndarray
iii)Python Dictionary
iv)scalar value
idx- sequence of numbers or labels of any valid numpy datatypes.
i) Specify data as Python sequence-
Syntax <Series Obj>=pd.Series(<any python sequence>)
Prog2
2) Creating non-Empty Series Objects- To create these objects you need to specify arguments for data and index as the foll
Syntax:-
You may or may not write data=
did

Prog3 pg(5)Eg1
Prog4 pg(5)Eg2
Prog5 pg(6)Eg3
Prog6 pg(6)Eg4
Prog7 pg(6)Eg5
Error
Only 1 seq allowed
string list
list
error
Renaming a series
object(extra)
The name of the
Series becomes its
index or column
name, if it is used
to form a
DataFrame.
Not like lists
s1=pd.Series(list(“hello”))
s1=pd.Series([‘h’,’e’,’l’,’l’,’o’])
or

Some ways of creating a numpy array
1)np.array(sequence)
2)np.arange(start,stop,step)works like range() of python
But allows to work with floating numbers
3)np.linspace(start,stop,number of elements) returns a float
4)np.tile([seq],number of times to tile)
Numpy array
A numpy array stores homogeneous data in continuous
fixed memory locations
For numpy array we need to write
import numpy as np
Creating Series using numpy arrays-
Egs-
float64
default is 50
array() does not work
with strings
Arr=np.array(“hello”)
absurd output

ii) Specify data as an ndarray-
Prog-8
Prog-9 pg(7)Eg6
Prog-10 pg(7)Eg7
Any
sequence
Tuple,set,
Dict
More egs
of
arange
Skipping 1

np.tile() works with any sequence
tuple
set
string
dictionary
Extra slide

iii) Specify data as a Python dictionary-
Prog-11 (keys of the dictionary become the index of the series object
and values of the dictionary become the data of the Series object)
Prog-12 pg(7) Eg-8
Indexes are not in the same order as given in the dictionary above

We can assign an explicit index to the series object using the index parameter to the Series().
If u assign an alpha numeric explicit index then the default numeric indexing also work.(0,1,2……)
If u assign a numeric explicit index then the default numeric indexing does not work
Eg1-explicit numeric index
Eg2-explicit non numeric index
The series can have elements with same index!!
Eg3
Eg4
Also
work
This default
numeric indexing
does not work
s[0]error
s[0]1

iv) Specify data as a scalar value:-
A scalar value is one unit of data can be either a number or a chunk of text.
Keep in mind that the length of the data and index should be same, but when you have a scalar or a single value as data, you
can have index which is more in length than the data. In that case the data will keep repeating to match the length of the
index.
If the data is a python sequence then the index length has to match the length of the data
eg:- Prog13:-
s1=pd.Series(["hello"],index=[1,2,3])
Error since [“hello”] is a
sequence(list) of length one.

Prog14- pg(8) Eg9
Prog15-pg(8) Eg10

3) Creating Series Objects-Additional Functionality
i) Specifying /Adding NaN values in a Series Object
Legal empty value is np.NaN. It is defined in the numpy module and hence you can use np.NaN to specify a missing
value (or use None)
Note :- None is python internal type which can be considered as an equivalent to Null. It is used to define a null value or no
value at all. Prog16While missing values are NaN in numerical arrays, they are None in object arrays.
ii) Specifying index(es) as well as data with Series()
◦ While creating Series we can provide the values and also the indexes.
◦ Both have to be sequences.
Syntax:- <Series Obj>=pandas.Series(data=None, index=None)
Eg
The datatype
of NaN is
float64
The datatype
of None is
NoneType

Extra-
NaN can be used as a numerical value on mathematical operations, while None cannot (or at least shouldn't). None is an internal Python type
( NoneType) and would be more like "inexistent" or "empty" than "numerically invalid" in this context.
Extra slide-np.NaN and None
error
No error
Arrays and series and dfs allow vectorized operations not lists
eg1
eg2
eg3
eg4 No error for
Series and df
with None
and an
arithmetic
operation

Prog17
Provide the same
number of
indices as values
in data array
error
Change order of
index and
data(works)
List comprehension

Prog-18 pg(10) Eg 11
iii) Specifying data type along with data and index-
Syntax- <Series Obj>=pd.Series(data=None, index=None, dtype=None)
*If u don’t specify a datatype, then pandas creates a Series with the nearest datatype to store the given
values. You can specify a datatype using Numpy datatype with dtype attribute.
Eg1
Read
None is the default value
for diff parameters taken in
case no value is provided
for a parameter

Platform specific
Error cause it belongs
to the numpy library
Platform independant
Numpy understands
that np.int is same
np.int32
Extra slide
intplatform specific
np.intsame as np.int32(platform independent)
np.int32platform independent

Prog19-
iv) Using a mathematical function/expression to create data array in Series()-
The series allows you to define a function or expression that can calculate values for data sequence
Syntax- <Series Obj>=pandas.Series(index=None,data=<function/expression>)

Prog20-
We can do vectorized operation on a numpy array(a*2 or a**2),so that this operation is applied on every element
of Numpy array.
But if we apply a similar operation on a python list then the result will be entirely diff.
Lets see….
Has to be an ndarray

Prog21-
*imp note-while creating a Series object, when u give an index array as a sequence then there is no compulsion for
the uniqueness of indexes. i,e you can have duplicate entries in the index array and Python will not raise any error.

Prog22-
No error even if index is
same
Indices need not be unique in panda series. This will only cause an error if
you perform an operation that requires unique indices.

Prog23-pg(12)Eg12
it will
give an
error
If
np.int32
error
Default
dt is
float64

When u create a series object all the information related to it is available through attributes. You can use these
attributes in the following format-
Syntax- <Series obj>.<attribute name>
1) <Series object>.index returns the index(axis labels) of the Series (sequence)
2) <Series object>.values returns an ndarray of the values (array)
3) <Series object>.dtype returns the datatype of each element of the series
4) <Series object>.shape returns the shape of the series in the form of a tuple(tuple)
5) <Series object>.itemsize size in bytes occupied by each element of the series(eg dtype is int 64,then itemsize will
return 8)(64bits/8)(the memory occupied by each element of the series object)
6) <Series object>.size returns the number of elements in the series object
7) <Series object>.nbytes (size*itemsize) returns the total memory occupied by the series object
8) <Series object>.hasnans returns True if the series object has any NaN value in it.
9) <Series object>.empty returns True if the series object is empty
10) <Series object>.count() returns a count of all the non-NaN values in the Series Object
11) <Series object>.ndim returns the number of dimensions of the data
12) len(<Series object>) returns the total number of elements in the Series Object including NaNs.
13) <Series object>.index.name Name of the index; can be used to assign a new name to the index (name of the rowindex in a df)
14) <Series object>.name returns or assigns name to series object (name of the col in a df)
B) SERIES OBJECTS ATTRIBUTES

(a)Retrieving Index Array and Data Array-
Prog24
Since index not
specified it will take
range(0 to 4) by
default.

Note-The name of the series is like
the name of a column when the
series adds into a df and vice versa(if
a col is extracted from a df its column
heading is the name of the series
object)
Name of series object-
Eg-
Eg-
Index name
of series obj
A=s.rename()temp
A=s.name=“abc”permanent
S=pd.Series(data,index,dtype,name)permanent

(d) Retrieving datatype(dtype) and itemsize-
For further programs use obj 2 and obj3 as follows
To know the datatype of each element of the series we use<obj name>.dtype
To know the number of bytes allocated to each element of the object we use<obj name>.itemsize
To know the type of object itself we use the type() method of python.
Prog25-
Depricated
else(8)

(e)Retrieving shape(including NaNs)
The shape tells us how many elements it contains including missing or empty values. Since there is only one axis in
Series is shown as (<n>,) where n is the number of elements in the object eg;-
Prog-26
*Series are always 1D.
(f) Retrieving dimensions, size and nbytes
Prog-27
Size x itemsize (4x8)=32 (3x8)=24

(g) Checking for emptiness and Presence of NaNs
Empty-It means, any of the axes are of length 0.If the data series has NaN or None it is not considered empty.
Prog28 Prog29:
*if you want to check if the series object has NaN, you can use len() to get the total number of elements
and<series>.count() to get the count of non-NaN values in series object.
Prog-30 Prog-31

Prog32-pg(15)Eg13 pg(17)
8x4=32 4x4=16

i)Accessing individual elements:-
To access individual elements you can give its index in square [ ] along with its name.
Syntax- <Series obj>[valid index]
Prog 33-
C) ACCESSING A SERIES OBJECT AND ITS ELEMENTS i. Access individual elements RTelement
ii. Slicing a series objects RT series object
Eg1- negative indexing in pandas
eg2
Negative indexing
works with
alphanumeric indexes
only

ii)Extracting Slices-
*imp-Slicing takes place position wise and not the index wise in a series object
Internally there is a position associated with element-first element gets position as 0,second element
gets position as 1 and so on.(Irrespective of their index labels, position always begins from 0)
A slice object is created from a Series Obj using the syntax-
syntax :-<object>[start:end:step]
but the start and end signify the position of the elements not their indexes.
The output of the slice is a series object.
0
1
2
0
1
2
3
If u use numeric values in slicing, then it is treated as position which goes till stop-1
If u use alphanumeric labels while slicing, then it includes the stop element.

Slices created, are views of the data series, so any change in the slice or main dataseries will reflect in both the
places- (shallow copy) Prog36- Eg14 pg(20)
Prog35-
Reminder-
While slicing a list, any change in one will not reflect in the other .
So slicing a list creates a True copy of the list.
List slicingTrue copy (list.copy(),list(),l[::])3 ways of true copy
Series slicingshallow copy
series
S8[‘A’:’B’]*100
or

i) Modifying elements of Series Object
Syntax:- <Series Object>[<index>]=<new data values>1 item
<Series Object>[start:stop]=<new data value>1 element
or exact no.of elements as the left side
Eg:-
D) OPERATIONS ON SERIES OBJECT Prog37
have become float implicitly

Prog38-Eg15 pg(21)
s13[2:4]???
or
It is
working

Renaming Indexing-
You can change indexes of Series object by assigning new index array to its index attribute
Syntax <Object>.index=<new index array>
Prog39
**Ensure that the size of the new index matches with
the existing index array size.
**remember that series are value mutable but size
immutable.

ii) head() and tail() functions-
head(<n>)-used to fetch first n rows from pandas object.
tail(<n>)- returns the last n rows from pandas object.
Syntax <pandas object>.head(<n>)
<pandas object>.tail(<n>)
If you don’t provide the n parameter then head will return the first 5 and tail will return the last 5 rows from the
series object.
Prog40- Prog41-pg(20)Eg16 pg(22)

iii)Vector Operations on Series Objects-
Vector operations mean that if you apply a function or expression then it is individually applied on each item of the
object.
-Series objects are built upon Numpy arrays(ndarrays),they also support vectorized operations just like ndarrays.
Prog42-
Original series will not change .
Don’t
copy
extra eg

iv) Arithmetic on Series objects
***
When you perform arithmetic operations on two series type objects
the data is aligned on the basis of matching index(called Data
Alignment in pandas objects) and then perform arithmetic
operations. For non overlapping indexes, the arithmetic operations
results as a NaN(Not a Number)
new-Pg(24)

Prog43 Eg17 pg(25)
Prog44 Eg18 pg(25)

v) Filtering entries- You can filter out entries from a Series objects using expressions that are of Boolean type.
Prog45-
When you apply a comparison operator directly on a pandas Series
object, then it works like a vectorized operation and applies this check on
each individual element of Series object and returns Boolean.
When you apply this check with the Series Object inside[ ] , you will
find that it returns filtered result containing only the values that
return True.

Prog46-Eg19 pg(26)
Prog 47-Eg20 pg(27)

vi)Sorting Series Values-
You can sort the values of a Series Object on the basis of values and indexes
Sorting on the basis of Values
Syntax- <Series Obj>.sort_values([ascending=True/False]) default-True
Prog48-
Temporary change
Or ds.sort_values(inplace=True)

Sorting on the basis of Index-(temporary change)
Syntax- <Series obj>.sort_index([ascending=True/False) default-True
Prog49-
Original
ds
Or ds.sort_index(ascending=False,inplace=True)
print(ds)

NDARRAYS SERIES OBJECT
1) We can perform arithmetic operations on arrays
only if the shape of the 2 arrays match, else we get
an error.
1) With Series objects, the data of the 2 Series
Objects is aligned as per matching indexes and
operations are performed on them and for non
matching indexes NaN is returned.
2) The indexes are always numeric starting from 0.
(Implicit indexing)
2) Series objects can have any type of indexes ,
including numbers(not necessarily starting from
zero), letters, labels strings etc.(explicit indexing is possible)
E) DIFFERENCE NUMPY ARRAYS AND SERIES OBJECTS-
3)Numpy arrays cannot have the same index
values for 2 items
4) Ndarrays can have any dimensions
3)Series can have the same labels for 2 items.
4)Series has only one dimension

2 D(dictionary)
Series Object vs 1D Numpy
Arraypg29 (pl learn)
Pg 29learn the entire page from the TB
properly

SOME ADDITIONAL OPERATIONS ON SERIES OBJECTS-
1)RE-INDEXING-(temporary change)(inplace does not work)
If you need to create a similar object with a different order of same indexes.
Syntax- <Series Object>=<Object>.reindex(<sequence with new order of the indexes>)
In this , the same data values and their indexes will be stored in the new object as per the defined order of
index in the reindex().
Prog-50
Note-If the reindex
consists of a new
index , we get a NaN
for that index but
not an error.

2) Dropping Entries from an axis-(temporary)
To remove an entry from a series Object we can use drop()
Syntax <Series Object>.drop(<row label>) removing 1 element
<Series Object>.drop(<list of row labels>) removing more than 1 element at one go
Prog-51
END OF SERIES
X X
or ds. drop(1,inplace=True)
Prog-52
What ever u can see
What ever u can see
You can also use del keyword of python to delete
an object-
Eg
Del does not allow
To delete multiple
Elements at one go
From a series object.
** del works with list slicing and
Individual elements

Q1.
Q2. Correct the error
Q3. Write the output of the following 2 codes-
Q4. Write the output-
DATA SERIES

ANSWERS-
A1. The number of index and elements in the ds obj don’t match
A2. range(0,4) or range(1,5)
A3.
A4.

II) DATA FRAME DATA STRUCTURE
DataFrame- It is a 2D labeled array like pandas data structure that stores an ordered
collection columns that can store data of diff. types.
A 2D array is an array having single-dimension arrays as its elements.
Eg-If no. of elements in an array a[7][9] is 7rowsx9 cols=63
MAJOR CHARACTERISTICS OF A DATAFRAME- pg(31)
1) It has 2 axes-a row index(axis0) and a column index(axis1)
2) Each value is identified with a row index and a column index. The row index is called index and the
column index is called columns.
3) The indexes can be letters, numbers or strings.
4) Columns can have data of a different types.
5) It is value mutable
6) It is size mutable
Def
NAME AGE RURAL URBAN
Abc 80 876 1123
Xyz 45 NaN 765
Pqr NaN 543 NaN
0
1
2
Row index
axis=0
column index
Axis=1
Data values
Missing values
series
dataframe Notice the col index

SUMMARY (DF)part
III) CREATING A DATA FRAME
Syntax- df=pd.DataFrame(<2D
datastructure>,[index],[columns])
1)Creating a df object from a 2-D Dictionary
2) Creating a DataFrame Object from a List of
Dictionaries/Lists(2D List)
3) Creating a DataFrame Object from a 2-D array
4) Creating a dataframe from a 2D dictionary having
values as series.
5) Creating a DataFrame object from another
DataFrame object-
IV) DATAFRAME ATTRIBUTES
V) SELECTING OR ACCESSING DATA-
◦ i) Selecting or accessing a column
◦ ii) Selecting/Accessing multiple columns
◦ iii)Selecting/Accessing a subset from a DF using Row/Col
names
◦ a) To access a row
◦ b) To access multiple rows
◦ c) To access subset cols
◦ d) To access a range of cols from a range of rows
◦ iv) Selecting rows/cols from a DF
a) Creating a dataframe from a 2D dictionary having values as lists/ndarrays/series
b) Creating a dataframe from a 2D dictionary having values as dictionaries.
v) Selecting or accessing Individual values
VI) ADDING/MODIFYING ROWS/COLS VALUES IN DF
i) Adding/modifying a column
ii) Adding/modifying a row
iii)Modifying a single cell
VII) DELETING/RENAMING COLUMNS/ROWS in a DF
i)Deleting rows/cols
a) Deleting a column using del
b) Deleting a row using drop
c) Deleting row/s col/s using drop
ii) Renaming rows/cols
VIII) MORE ON DF INDEXING-BOOLEAN INDEXING
i) Creating df with Boolean indexes
ii) Accessing rows from DF with Boolean Indexes
same

III) CREATING AND DISPLAYING A DATAFRAME
-A dataframe object can be created by passing data in a 2D format.
-We need to import both pandas and numpy
Syntax- <dataframe obj>=panda.DataFrame(<2D data structure>,[columns=<col sequence>],[index=<index
sequence>])
2D data structures could be made up of
i) 2D dictionary i.e. dictionaries having lists or dictionaries or ndarrays or Series objects etc.
ii) 2D ndarrays (NumPy Arrays)
iii) Series type Object
iv)Another df object.
1)Creating a df object from a 2-D Dictionary-
What is a 2-d dictionary?
1D
dictionary
2D
dictionary
No error
so
Df=pd.DataFrame() creates an empty df having no rows and no columns
gita

1)Creating a df object from a 2-D Dictionary-
A 2D dictionary is a dictionary having items as (key,value) where value part is a data structure of any type
-another dictionary
-an ndarray
-a Series Object
-a list
but the value part of the keys should have similar structure and equal length.
a) Creating a dataframe from a 2D dictionary having values as lists/ndarrays-
Prog-1 Create a df-------------------------------------------------------------------------
Marks Sport Student
70 Cricket Rahul
80 Badminton Neha
90 Football Mark
100 Athletics Smith
0
1
2
3
Not of equal length
Now of Equal
length
Note-if the 2d dictionary has values as lists and if the
length of the lists don’t match. You will get an error.

*Its index is assigned automatically 0 onwards and columns created from keys are placed in sorted order.
*keys of dictionary have become columns of df
* You can specify your own sequence for the index
Prog2-same df as prog1 but with indexes as I,II,III,IV
Not for me
Dictionary values as lists

The number of indexes should match with the length of the dictionary values else python will give an error
Prog-3 same df as Prog1 above but this time the
dictionary values will be ndarrays-
Note-If length of the inner nparray not same then error!!!

Prog3_a- Create the following df using dictionaries and its values as Series.
Note-
If we use a 2d dictionary to create a df
With its values as series and if the
length of the series is not equal, no
error, NaN’s are added wherever
required.
List
Ndarray
Dictionary
Series
Errors
No errors
If length is not the same

Prog-4 pg(34) Eg21
b) Creating a dataframe from a 2D dictionary having values as dictionaries-
Prog-5
2D dictionary with values
as lists.
Outer dictionary keys as columns
Inner dictionary keys as indexes
Note –if u are using a 2d dictionary with values as dictionaries to make a df and if the
length of the inner dictionary don’t match then no error, NaN will be put in that place.

Try it yourself- create the foll data frame using a 2D dictionaries with values as
1)Lists
2)Ndarrays
3)Dictionaries
1.
2.
3.

Prog-6 Create a df from a 2D dictionary, Sales, which stores the quarter-wise sales as inner dictionary for 2 years
Pg(34)Eg22
Prog7 pg(35 Eg23)
Eg-if length of the inner
dictionary values don’t match
df3.index
df3.columns
Summary-
Creating df with a 2d dictionary with values
1.List
2.Ndarray
3.Series
4.dictionary
If length not same error
If length not same error
If length not same no error
If length not same no error

2)Creating a DataFrame Object from a List of Dictionaries/Lists(2D List)no errors with diff lengths of inner seq
---List of Lists
Prog-8
---List of Dictionaries
The dictionary keys will become columns and the inner dictionary values will become rows.
Default
index
Eg(don’t
copy)
No error
Even if the length of the
inner list is not the same

If the dictionaries in the list have diff lengths, no error
Eg
Don’t copy
If the ndarrays in the list have diff lengths, no error

Prog-9 Prog10-pg(36)Eg24
Prog11-pg(37)Eg25
Prog12-pg(37)Eg26
If length of
dictionaries don’t
match,no error, a
NaN will be put in
that place.

Prog-11
Prog-12 wap to create a dataframe from a list containing 2 lists, each containing Target and Sales figures of 4 zonal offices. Give appropriate
row labels.
.

3) Creating a DataFrame Object from a 2D ndarray
You can also pass a 2D Numpy array having shape(<n>,<n>) to a DataFrame() to create a dataframe Object.
Prog14- Prog15
Default
index
and
columns
Give columns
Prog16-

-Ndarrays that are passed to DataFrame have same number of elements in each of the rows.
-If rows of ndarrays differ in length(if the number of elements in each row differ, then Python will create a single
column in the dataframe and the datatype in the column will be object.
Prog-18 pg(39) Eg27
Prog19 pg(39) Eg28
column and index from the user.
Same as
[1,2,3]

Prog-18 pg(39) Eg27
Prog19 pg(39) Eg28

4) Creating a DataFrame Object from a 2D Dictionary with values as Series Objects
You can create a DF obj by using multiple series objects. In a 2D dictionary, u can have the value parts as series
objects and then pass this dictionary as argument to create a DF object.
◦ Prog-20

Prog-21pg(40)Eg29
Arrays also allow vectorized operations

5) Creating a DataFrame object from another DataFrame object-
df1=pd.DataFrame(df) or df1=df -any change in one reflects the other.
Or df2=df.copy() -any change in one does not reflect the other.
Prog22-
*note-DF’s can also be created from
text/csv files.

IV) DATAFRAME OBJECT ATTRIBUTES slide31(attributes of ds)
When you create a DF obj, all information related to it is available through its attributes. You can use these
attributes in the following format
Syntax <DF obj>.<attribute name> TB pg(41)
Attribute Description
index The index row labels of the DataFrame(sequence)
columns The column labels of the Dataframe(sequence)
axes Returns a list representing both the axes(axis 0
and axis 1) of the Dataframe
dtypes Returns the dtypes of data in the DF (column
wise)
size Returns an int of the number of elements in the
df obj
shape Returns a tuple representing the dimensions of
the df
values Returns a numpy representation of the dataframe
empty Indicator whether DataFrame is empty
ndim Returns an int representing the number of
axes/array dimensions
T Transpose index and columns.

We will be using this df for all attribute programs-
(a)Retrieving various properties of a Df Object- dfn.index dfn.columns dfn.axes dfn.dtypes
Prog-23-
datatype is listed for
individual columns.
Another eg of dtypes
? Object dt

(b)Getting number of rows in a DF-len(df)
len(<Df object>) will return the number of rows in a dataframe or len(dfn.index) or dfn.shape[0] dfn-
Eg:- len(dfn)---3
(c)Getting count of non-Na values in DF
Like series ,u can use count() with a DF to
get the count of non-NaN values,but count with
a DF is a little elaborate-
i) If u don’t pass any argument or pass 0(default 0),then it returns count of non-NaN
values for each column.
Prog-24
ii) If u pass argument as 1,then it returns count of non-Na values for each row
Prog-25
i
dfn.count()-for each column
dfn.count(1)-for each row
eg

(d) Transposing a DF-Df.T
You can transpose a DF by swapping its indexes and columns by using attribute T
Prog-26
Prog-27 pg(43)Eg30
Weight Age Name
0 40 15 Rohit
1 50 17 Sahil
2 37 14 Rina

(e) Retrieving size, shape, no. of dimensions of the DF object-
dfn.size-returns the no. of elements in the df obj
dfn.shape-returns a tuple giving the no. of rows and columns in a tuple form
dfn.ndim-returns the no. of dimensions of the DF object as an int
Prog-28
(f) Numpy Representaion of DataFrame-
You can represent the values of a Df object in Numpy way using-
Prog-29

(g) Checking for empty df-
A df is said to be empty if its any axes(0 or 1) has no values
Having np.NaN does not mean empty

V) SELECTING OR ACCESSING DATA-
From a df you can extract or select desired rows and columns-
dtf5
i) Selecting/Accessing a column-
<df>.colname -> no single quotes here
<df>[‘colname’]

Prog-30-write the output of the following-
ii) Selecting/Accessing Multiple Columns(selective cols)
You can give a list of columns inside square brackets with df objects-
Syntax- <df obj>[[colname,colname,colname,……]]
Prog-31 Write the output-

Prog32-pg(46)Eg 31
iii) Selecting/Accessing a Subset from a DF using row/col names
.loc always begin from a row
.loc end row and end col are inclusive
Syntax-<DF>.loc[startrow:endrow,startcolumn:end column]
row column
Both end indexes are
inclusive for loc
loc always works with labels

(a) To access a row- just give the row label/name as
<df>.loc[<row label>,:] -best(prefer don’t miss the comma and colon)
<df>.loc[<row label>,]
<df>.loc[<row label>]
Prog33-Access the Delhi row in diff ways
(b) To access multiple rows-
<df>.loc[<start row>:<end row>,:]
Prog34 Display the rows of
Mumbai and Kolkata
dtf5.loc[[rowname,rowname,rowname]]
dtf5.loc[[‘Mumbai’,’Kolkata’]]
dtf5.loc[‘Delhi’ : , :]-then all rows will follow in the
output(entire df)

Prog35 write the output
(c) To access subset of columns-
Syntax- <df obj>.loc[ : ,<start column>:<end column>] -> don’t miss the colon and comma
Multiple columns
df[[col1,col2,col3,…….]]

(d) To access range of columns from a range of rows-
Prog-38 write the output
Prog-39 pg(48) Eg32

iv) Selecting rows/columns from a DataFrame-
Sometimes your df may not contain row and column labels or you may not remember them.in such cases you can
extract subset from dataframe using the row and column numeric position , but this time you will use iloc instead
of loc.
iloc means integer location
Syntax- <df>.iloc[start row index : end row index , start column index : end col index]
just like slicing
Both end indexes are
exclusive
df.iloc[]-works on position only only only
like
ds slicing also works on position only only
iloc[0:2,1:1] ?

V) Selecting/Accessing Individual value-
df[colname][rowname]
df.loc[rowname][colname] or df.loc[rowname,colname]
df.colname[rowname/row int pos] –TB(1)
Eg-
df.at[rowname,colname]/loc –TB(2)
df.iat[rowindex,colindex]/iloc -TB(3)
Eg-
at-access a single value for a row/column label pair
iat-access a single value for a row/column pair by integer position
only only only(iat works with integer position)
-------

VI) ADDING/MODIFYING ROW’S/COLUMN’S VALUES IN DATAFRAME
◦ i) Adding/modifying a column
◦ ii) Adding/modifying a row
◦ iii) Adding/modifying a single cell
i) Adding/modifying a column-
You can refer to a column in a df in multiple ways
Assigning a value to a column
 will modify it, if the column already exists
 will add a new column, if it does not exist already
Syntax- <df>.<colname>=<new value>
<df>[‘colname’]=<new value>
If the colname does not exist in the df, then new column with this name is added.
Prog42- Add a column Density to the dtf5-
Or dtf5.at[:,’Density’]=500
Or dtf5.loc[:,’Density’]=500
Or dtf5=dtf5.assign(Density=500)
Since a column Density does not exist already in the df a new column got added.
*now change the values of the density column-
dtf5
Other ways-
Can also be used (Density=[500,600,700,200])
temporary
Cant add a row or column
with iloc and iat.
If this 500 is not there then at[] gives an error

Prog42-continued
ii) Adding /Modifying a row-
Like columns, you can change or add row to a DF using at or loc attribute
as explained-using at or loc
Syntax- <df obj>.at[<rowname>,:]=<new value>
<df obj>.loc[<rowname>,:]=<new value>
If such a row does not exist then python adds a new row else edits its values
Prog43 Add a row Bangalore with value 1200 to dtf5
Note*the new sequence should have values for all the columns, else error
* note-The sequence which contains the values
of the new column must have values equal to
number of rows in the df, else pyhton will give
an error.
If one less value given then error
***rows cannot be added using iloc []or iat[]
If this <new value> is not there then at[] gives an error

Prog-44 pg(55)Eg 36
Should be 4 elements in the list
36

iii) Modifying a single cell-
You can use any method to access a single cell. Any method which allows you to access a single cell.
Eg- <DF>.iat[rowposition,colposition]=new value
<DF>.colname[row label/index]=new value
Prog 45- Change the value of population of Bangalore to 5555
x--------x(Topic)

VII) DELETING/RENAMING COLUMNS/ROWS
Python Pandas gives us 2 ways to delete rows and cols-
-del statement
-drop() function
To rename rows/cols
-rename() function
i) Deleting rows/columns in a DF-
(a) Delete a column use del-works with labels
Syntax- del<df obj>[‘colname’]
Prog 46-Delete the Density column from dtf5
Permanent change
del drop
Permanent change Temporary change
Allows to delete columns Allows to delete rows and columns
Allows to delete only 1 column at 1
time
Allows to delete 1 or more rows/cols at
1 time

(b) Delete a row use drop()----drop() works with labels
Syntax-
<df>.drop(label or sequence of labels)
Prog 47-Delete the rows of Mumbai and Delhi
Or dtf5.drop([0,1]) this can be used only when the dtf5 has numeric labels of 0 1 2 3…. Else error.
(c)Delete a row/col using drop()-
Syntax- <df>.drop([label/ sequences of labels ],[axis=0/1,inplace=False])
Eg-
Temporary change
default axis=0
To make a permanent change for drop, you can
use the inplace argument with drop

Prog 48 pg(57)Eg37
iii) Renaming rows/cols labels-
Syntax <df>.rename(index={change name dictionary},columns={change name
dictionary},[inplace=False])
Or <df>.rename({change name dictionary},[axis=0/1],[inplace=False])
If u want to rename row labels then use only index arg
If u want to rename columns labels then use only column arg
If u want to rename both then use both the arguments with dictionaries as {old name:new name}
inplace-default False (if inplace True then change happens in place and is permanent and None is returned)
37
In this method u can
change both at one go!!

Prog- 49 Make the following DF in 3 diff ways and change its row labels to A,B,C,D Rollno Name Marks
SecA 1 Rishi 97
SecB 2 Arun 98
SecC 3 Rohan 98
SecD 4 Soham 99

Rollno Name Marks
SecA 1 Rishi 97
SecB 2 Arun 98
SecC 3 Rohan 98
SecD 4 Soham 99
Prog50-Write a program to change the column name Rollno to Rno of
the following df
Prog51 pg(59) Eg38
Prog52 pg(60)Eg39

VIII)Selecting DataFrame Rows/Columns based on Boolean Conditions pg(50)
Sometimes we need to select rows/cols from a dataframe based on a condition, just the way you filtered the entries in series
objects.
When you compare a dataframe with a value then pandas executes the comparison condition for each element of the df and
returns a True/False accordingly for each element.
Prog-53
You can apply condition to individual columns or a range of values too
Prog-54
df
When condition is given on the entire
df, then it applies the condition on
each individual element o the df and
returns True and False for each
element of the df.
By giving a condition like this, has only given u a result as True or False.
But to extract a subset of the df for which the condition is True all u need to do is-
 Write the condition in [ ] next to the name of the df like-
Syntax-
<df>[condition]
Or
<df>.loc[condition]

Prog-55
Internally pandas checks the condition for each row and returns True or False. These truth values act as an index for the rows. The rows with
True index are returned.

Creating a New DF from a DataFrame -Shallow vs real copy pg(56)
Eg
Here copy=False by default so a shallow
copy is made
Shallow
copy
True,
Deep copy
df1=df.copy()

IX) MORE ON DF INDEXING-BOOLEAN INDEXING
Def:- Boolean indexing-means having Boolean values(True or False) or(1 or 0) as indexes of a df.
WHY?
In some cases you may need to divide our data in 2 subsets-True or False
Eg- School decided to have online classes and the schedule may look like
Day Classes
True Mon 6
False Tue 0
True Wed 3
False Thur 0
True Fri 8
-so we have 2 groups 1)True Rows
2)False Rows
This info is useful when we want to find out of when we
have online classes and when we don’t.
So Boolean indexing divide the df in 2 groups
i) CREATING DF WITH BOOLEAN INDEXING
Prog56- Create the df as above and name it as classdf
Don’t put single quotes
then it will become string
not boolean

ii) Accessing rows from df with Boolean indexes- my doubt
We need to make use of
<df>.loc[True]
<df>.loc[False]
<df>.loc[0]
<df>.loc[1]
Prog66- Write the output-
x-------------------------------------------x Pyhton Pandas-1 ends

Practical Questions-
Q1. Given a series which holds the area of some states in km2.Write code to find out the biggest and smallest
three areas from the given Series.
ds=pd.Series([100,20,30,44,272,65,222])
Q2. From the above series find out the areas which are more than 200km2.
Q3.Write a Program to create a series object with 6 random integers and having indexes as :[‘p’,’q’,’r’,’n’,’t’,’v’]
Q4. Write a program to create data series and then change the indexes of the Series object in any random order.
A1- A2-
A3- A4-
1,21for 1 to 20 can be given

H/W
Q5. WAP to Sort the values of a Series object s1 in ascending order of its values and store it into series object s2
Q6. WAP to Sort the values of a Series object s1 in descending order of its indexes and store it into series object s3
Q7. Given a Series object s4. WAP to change the values at its 2nd row(index1) and 3rd row to 8000
Q8. Given a Series object s5.WAP to calculate the cubes of the Series values.
Q9. Given a Series object s5.WAP to store the squares of the Series values in object s6. Display s6’s values which
are > 15.

Q10. WAP to display the number of rows and number of columns in DataFrame df.
Q11. WAP to display the number of rows and number of columns in DataFrame df without the shape attribute.
Q12. Given the df WAP to display the Weight of first and third rows.
df---- Age Name Weight
0 15 Arnav 42
1 22 Charles 75
2 35 Guru 66

Q13. Name the data structures of Python’s pandas library.
Q14.WAP to create a Series Object Temp1 that stores the temperatures of 7 days in it . Take any random 7
temperatures.
Q15. Make a series same as Q14. and save it in temp2.Index it with ‘Mon’,’Tue’……..
Series
DataFrame
Panel

Q18.Write a program to create three different series objects from the three columns of a DataFrame df.
Q19. Write a program to create three different series objects from the three rows of a DataFrame df.

Q20. create a Series from an ndarray which stores characters from ‘a’to ‘g’
Q21.create a Series that stores the table of number 5
Q22. Write a program to create a df that stores 2 columns which store the series objects of the previous 2
questions (20 and 21)
Take it as ds1

Q23- Create a df storing salesmen details(name, zone, sales) of five salesmen.
Q24-Three dictionaries store details of 3 employees as (empno, name). Write a program to create a dataframe
from these.
or

Q25.A list stores 3 dictionaries each storing(old price, new price, change) .wap to create a df from it.
Q26. Write code to extract first 10 rows from a dataframe called df using iloc()
df.iloc[0:10,:]
Or
df.iloc[0:10]

Q29- write the output of the following-
Ans-29 Q30.From the earlier df display:-
◦ 1)only row ‘a’ from df,df1,df2
◦
◦
◦ 2)add an empty columns ‘x’ to all the dfs.
◦ 3)display rows 0 and 1 from the three dfs
◦
Empty gives
false

4)12th_L-1_PYTHON-PANDAS-I.pptx

More Related Content

Similar to 4)12th_L-1_PYTHON-PANDAS-I.pptx (20)

Recently uploaded (20)

4)12th_L-1_PYTHON-PANDAS-I.pptx