This tutorial would help you to learn Data Science with Python by examples. It is designed for beginners who want to get started with Data Science in Python. Python is an open source language and it is widely used as a high-level programming language for general-purpose programming. It has gained high popularity in data science world. In the PyPL Popularity of Programming language index, Python leads with a 29 percent share. In advanced analytics and predictive analytics market, it is ranked among top 3 programming languages for advanced analytics.

10 divided by 3 is 3.33

remainder after 10 divided by 3 is 1

It is used to assign a value to the declared variable. For e.g.

You can select multiple elements from a list using the following method

Suppose you want python to assume

!pip install pandas

!pip uninstall pandas

!pip show pandas

!pip list

!pip install --upgrade pandas --user

s1[0]

It is equivalent to data.frame in R. It is a 2-dimensional data structure that can store data of different data types such as characters, integers, floating point values, factors. Those who are well-conversant with MS Excel, they can think of data frame as Excel Spreadsheet.

To keep a single variable, you can write in any of the following three methods -

We can remove variables by using df.drop() function. See the example below -

To summarize or explore data, you can submit the command below.

To summarise all the

To select only a particular variable, you can write the following code -

We can use astype() function to make id as a categorical variable.

In the code below, we rename the variable rank to '

In python, we can write R-style model formula

The output looks like below -

To avoid multi-collinearity, you can set one of the category as reference category and leave it while creating dummy variables. In the script below, we are leaving first category.

Data Science with Python Tutorial |

Table of Contents

## Introduction

Python is widely used and very popular for a variety of software engineering tasks such as website development, cloud-architecture, back-end etc. It is equally popular in data science world. In advanced analytics world, there has been several debates on R vs. Python. There are some areas such as number of libraries for statistical analysis, where R wins over Python but Python is catching up very fast.With popularity of big data and data science, Python has become first programming language of data scientists.
There are several reasons to learn Python. Some of them are as follows -

- Python runs well in automating various steps of a predictive model.
- Python has awesome robust libraries for machine learning, natural language processing, deep learning, big data and artificial Intelligence.
- Python wins over R when it comes to deploying machine learning models in production.
- It can be easily integrated with big data frameworks such as Spark and Hadoop.
- Python has a great online community support.

Do you know these sites are developed in Python?

- YouTube
- Dropbox
- Disqus

## Python 2 vs. 3

Google yields thousands of articles on this topic. Some bloggers opposed and some in favor of 2.7. If you filter your search criteria and look for only recent articles, you would find Python 2 is no longer supported by the Python Software Foundation. Hence it does not make any sense to learn 2.7 if you start learning it today. Python 3 supports all the packages. Python 3 is cleaner and faster. It is a language for the future. It fixed major issues with versions of Python 2 series. Python 3 was first released in year 2008. It has been 12 years releasing robust versions of Python 3 series. You should go for latest version of Python 3.## How to install Python?

There are two ways to download and install Python**Download Anaconda.**It comes with Python software along with preinstalled popular libraries.- Download
**Python**from its official website. You have to manually install libraries.

Recommended :Go for first option and download anaconda. It saves a lot of time in learning and coding Python

Coding Environments

Anaconda comes with two popular IDE :
- Jupyter (Ipython) Notebook
- Spyder

**Spyder**. It is like RStudio for Python. It gives an environment wherein writing python code is user-friendly. If you are a SAS User, you can think of it as SAS Enterprise Guide / SAS Studio. It comes with a syntax editor where you can write programs. It has a console to check each and every line of code. Under the 'Variable explorer', you can access your created data files and function.

**I highly recommend Spyder!**

Spyder - Python Coding Environment |

**Jupyter (Ipython) Notebook**Jupyter is equivalent to markdown in R. It is useful when you need to present your work to others or when you need to create step by step project report as it can combine code, output, words, and graphics.

Spyder Shortcut keys

The following is a list of some useful spyder shortcut keys which makes you more productive.
**Press F5**to run the entire script**Press F9**to run selection or line- Press
**Ctrl+ 1**to comment / uncomment - Go to front of function and then
**press Ctrl + I**to see documentation of the function **Run**%reset -f to clean workspace**Ctrl + Left click on object**to see source code**Ctrl+Enter**executes the current cell.**Shift+Enter**executes the current cell and advances the cursor to the next cell

List of arithmetic operators with examples

Arithmetic Operators | Operation | Example |
---|---|---|

+ | Addition | 10 + 2 = 12 |

– | Subtraction | 10 – 2 = 8 |

* | Multiplication | 10 * 2 = 20 |

/ | Division | 10 / 2 = 5.0 |

% | Modulus (Remainder) | 10 % 3 = 1 |

** | Power | 10 ** 2 = 100 |

// | Floor | 17 // 3 = 5 |

(x + (d-1)) // d | Ceiling | (17 +(3-1)) // 3 = 6 |

## Basic programs in Python

Example 1

#Basics x = 10 y = 3 print("10 divided by 3 is", x/y) print("remainder after 10 divided by 3 is", x%y)

**Result :**

10 divided by 3 is 3.33

remainder after 10 divided by 3 is 1

Example 2

x = 100 x > 80 and x <=95 x > 35 or x < 60

x > 80 and x <=95 Out[45]: False

x > 35 or x < 60 Out[46]: True

Comparison, Logical and Assignment Operators

Comparison & Logical Operators | Description | Example |
---|---|---|

> | Greater than | 5 > 3 returns True |

< | Less than | 5 < 3 returns False |

>= | Greater than or equal to | 5 >= 3 returns True |

<= | Less than or equal to | 5 <= 3 return False |

== | Equal to | 5 == 3 returns False |

!= | Not equal to | 5 != 3 returns True |

and | Check both the conditions | x > 18 and x <=35 |

or | If atleast one condition hold True | x > 35 or x < 60 |

not | Opposite of Condition | not(x>7) |

**Assignment Operators**

It is used to assign a value to the declared variable. For e.g.

**x += 25 means x = x+25**.

x = 100 y = 10 x += y print(x)

print(x) 110In this case, x+=y implies x=x+y which is x = 100+ 10.

Similarly,you can use x-=y, x*=y and x /=y

## Python Data Structures

In every programming language, it is important to understand the data structures. Following are some data structures used in Python.1. List

It is a sequence of multiple values. It allows us to store different types of data such as integer, float, string etc. See the examples of list below. First one is an integer list containing only integer. Second one is string list containing only string values. Third one is mixed list containing integer, string and float values.- x = [1, 2, 3, 4, 5]
- y = [‘A’, ‘O’, ‘G’, ‘M’]
- z = [‘A’, 4, 5.1, ‘M’]

Get List Item

We can extract list item using Indexes. **Index starts from 0 and end with (number of elements-1).**

x = [1, 2, 3, 4, 5] x[0] x[1] x[4] x[-1] x[-2]

x[0] Out[68]: 1 x[1] Out[69]: 2 x[4] Out[70]: 5 x[-1] Out[71]: 5 x[-2] Out[72]: 4

**x[0]**picks first element from list.

**Negative sign**tells Python to search list item from right to left.

**x[-1]**selects the last element from list.

You can select multiple elements from a list using the following method

x[:3] returns[1, 2, 3]

2. Tuple

A tuple is similar to a list in the sense that it is a sequence of elements. The difference between list and tuple are as follows -- A tuple cannot be changed once constructed whereas list can be modified.
- A tuple is created by placing comma-separated values inside parentheses
**( )**. Whereas, list is created inside square brackets**[ ]**

Examples

K = (1,2,3)

State = ('Delhi','Maharashtra','Karnataka')

Perform for loop on Tuple

for i in State:

print(i)

Delhi Maharashtra KarnatakaDetailed Tutorial : Python Data Structures

Functions

Like **print()**, you can create your own custom function. It is also called user-defined functions. It helps you in automating the repetitive task and calling reusable code in easier way.

Rules to define a function

- Function starts with
**def**keyword followed by function name and**( )** - Function body starts with a colon (:) and is indented
- The keyword
**return**ends a function andgive value of previous expression.

def sum_fun(a, b): result = a + b return result

z = sum_fun(10, 15)

**Result :**z = 25

Suppose you want python to assume

**0 as default value**if no value is specified for parameter b.

def sum_fun(a,In the above function, b is set to be 0 if no value is provided for parameter b. It does not mean no other value than 0 can be set here. It can also be used asb=0): result = a + b return result z = sum_fun(10)

**z = sum_fun(10, 15)**

Python Conditional Statements

Conditional statements are commonly used in coding. It is IF ELSE statements. It can be read like : " if a condition holds true, then execute something. Else execute something else"
**Note :**The if and else statements ends with a colon

**:**

Example

k = 27 if k%5 == 0: print('Multiple of 5') else: print('Not a Multiple of 5')

**Result :**Not a Multiple of 5

## List of popular packages (comparison with R)

Some of the leading packages in Python along with equivalent libraries in R are as follows-**pandas**. For data manipulation and data wrangling. A collections of functions to understand and explore data. It is counterpart of**dplyr**and**reshape2**packages in R.**NumPy**. For numerical computing. It's a package for efficient array computations. It allows us to do some operations on an entire column or table in one line. It is roughly approximate to**Rcpp**package in R which eliminates the limitation of slow speed in R.**Numpy Tutorial****Scipy**.For mathematical and scientific functions such asintegration, interpolation, signal processing, linear algebra, statistics, etc. It is built on Numpy.**Scikit-learn.**A collection of machine learning algorithms. It is built on Numpy and Scipy. It can perform all the techniques that can be done in R using**glm, knn, randomForest, rpart, e1071**packages.**Matplotlib.**For data visualization. It's a leading package for graphics in Python. It is equivalent to**ggplot2**package in R.**Statsmodels.**For statistical and predictive modeling. It includes various functions to explore data and generate descriptive and predictive analytics. It allows users to run descriptive statistics, methods to impute missing values, statistical tests and take table output to HTML format.**pandasql.**It allows SQL users to write SQL queries in Python. It is very helpful for people who loves writing SQL queries to manipulate data. It is equivalent to**sqldf**package in R.

*Maximum of the above packages are already preinstalled in Spyder.*

Comparison of Python and R Packages by Data Mining Task

Task | Python Package | R Package |
---|---|---|

IDE | Rodeo / Spyder | Rstudio |

Data Manipulation | pandas | dplyr and reshape2 |

Machine Learning | Scikit-learn | glm, knn, randomForest, rpart, e1071 |

Data Visualization | ggplot + seaborn + bokeh | ggplot2 |

Character Functions | Built-In Functions | stringr |

Reproducibility | Jupyter | Knitr |

SQL Queries | pandasql | sqldf |

Working with Dates | datetime | lubridate |

Web Scraping | beautifulsoup | rvest |

## Popular python commands

The commands below would help you to install and update new and existing packages. Let's say, you want to install / uninstall**pandas**package.

Run these commands from IPython console window. Don't forget to add ! before pip otherwise it would return syntax error.

**Install Package**

!pip install pandas

**Uninstall Package**

!pip uninstall pandas

**Show Information about Installed Package**

!pip show pandas

**List of Installed Packages**

!pip list

**Upgrade a package**

!pip install --upgrade pandas --user

How to import a package

There are multiple ways to import a package in Python. It is important to understand the difference between these styles.
**1. import pandas as pd**
It imports the package **pandas **under the alias **pd**. A function DataFrame in package pandas is then submitted with **pd.DataFrame**.

**2. import pandas**

It imports the package without using alias but here the function DataFrame is submitted with full package name **pandas.DataFrame**

**3. from pandas import * **

It imports the whole package and the function DataFrame is executed simply by typing DataFrame. It sometimes creates confusion when same function name exists in more than one package.

Pandas Data Structures - Series and DataFrame

In pandas package, there are two data structures - series and dataframe. These structures are explained below in detail -
1. Series

It is a one-dimensional array. You can access individual elements of a series using position. It's similar to vector in R.
*In the example below, we are generating 5 random values.*

import pandas as pd import numpy as np s1 = pd.Series(np.random.randn(5)) s1

0 -2.412015 1 -0.451752 2 1.174207 3 0.766348 4 -0.361815 dtype: float64

Extract first and second value

You can get a particular element of a series using index value. See the examples below -s1[0]

-2.412015s1[1]

-0.451752s1[:3]

0 -2.412015 1 -0.451752 2 1.174207

2. DataFrame

It is equivalent to data.frame in R. It is a 2-dimensional data structure that can store data of different data types such as characters, integers, floating point values, factors. Those who are well-conversant with MS Excel, they can think of data frame as Excel Spreadsheet.

Comparison of Data Type in Python and Pandas

The following table shows how Python and pandas package stores data.Data Type | Pandas | Standard Python |
---|---|---|

For character variable | object | string |

For categorical variable | category | - |

For Numeric variable without decimals | int64 | int |

Numeric characters with decimals | float64 | float |

For date time variables | datetime64 | - |

## Important Pandas Functions (vs. R functions)

The table below shows comparison of pandas functions with R functions for various data wrangling and manipulation tasks. It would help you to memorize pandas functions. It's a very handy information for programmers who are new to Python. It includes solutions for most of the frequently used data exploration tasks.Functions | R | Python (pandas package) |
---|---|---|

Installing a package | install.packages('name') | !pip install name |

Loading a package | library(name) | import name as other_name |

Checking working directory | getwd() | import os os.getcwd() |

Setting working directory | setwd() | os.chdir() |

List files in a directory | dir() | os.listdir() |

Remove an object | rm('name') | del object |

Select Variables | select(df, x1, x2) | df[['x1', 'x2']] |

Drop Variables | select(df, -(x1:x2)) | df.drop(['x1', 'x2'], axis = 1) |

Filter Data | filter(df, x1 >= 100) | df.query('x1 >= 100') |

Structure of a DataFrame | str(df) | df.info() |

Summarize dataframe | summary(df) | df.describe() |

Get row names of dataframe "df" | rownames(df) | df.index |

Get column names | colnames(df) | df.columns |

View Top N rows | head(df,N) | df.head(N) |

View Bottom N rows | tail(df,N) | df.tail(N) |

Get dimension of data frame | dim(df) | df.shape |

Get number of rows | nrow(df) | df.shape[0] |

Get number of columns | ncol(df) | df.shape[1] |

Length of data frame | length(df) | len(df) |

Get random 3 rows from dataframe | sample_n(df, 3) | df.sample(n=3) |

Get random 10% rows | sample_frac(df, 0.1) | df.sample(frac=0.1) |

Check Missing Values | is.na(df$x) | pd.isnull(df.x) |

Sorting | arrange(df, x1, x2) | df.sort_values(['x1', 'x2']) |

Rename Variables | rename(df, newvar = x1) | df.rename(columns={'x1': 'newvar'}) |

## Examples - Data analysis with Pandas

1. Import Required Packages

You can import required packages using **import**statement. In the syntax below, we are asking Python to import numpy and pandas package. The 'as' is used to alias package name.

import numpy as np

import pandas as pd

2. Build DataFrame

We can build dataframe using **DataFrame()**function of pandas package.

mydata = {'productcode': ['AA', 'AA', 'AA', 'BB', 'BB', 'BB'], sales': [1010, 1025.2, 1404.2, 1251.7, 1160, 1604.8], cost' : [1020, 1625.2, 1204, 1003.7, 1020, 1124]} df = pd.DataFrame(mydata)In this dataframe, we have three variables - productcode, sales, cost.

Sample DataFrame |

To import data from CSV file

You can use read_csv() function from pandas package to get data into python from CSV file.
mydata= pd.read_csv("C:\\Users\\Deepanshu\\Documents\\file1.csv")Make sure you use

**double backslash**when specifying path of CSV file. Alternatively, you can use forward slash to mention file path inside read_csv() function.

Detailed Tutorial : Import Data in Python

3. To see number of rows and columns

You can run the command below to find out number of rows and columns.df.shape

**Result :**(6, 3). It means 6 rows and 3 columns.

4. To view first 3 rows

The df.head(N) function can be used to check out first some N rows.df.head(3)

cost productcode sales 0 1020.0 AA 1010.0 1 1625.2 AA 1025.2 2 1204.0 AA 1404.2

5. Select or Drop Variables

To keep a single variable, you can write in any of the following three methods -

df.productcodeTo select variable by column position, you can use

df["productcode"]

df.loc[: , "productcode"]

**df.iloc**function. In the example below, we are selecting second column. Column Index starts from 0. Hence, 1 refers to second column.

df.iloc[: , 1]We can keep multiple variables by specifying desired variables inside [ ]. Also, we can make use of df.loc() function.

df[["productcode", "cost"]]

df.loc[ : , ["productcode", "cost"]]

**Drop Variable**

We can remove variables by using df.drop() function. See the example below -

df2 = df.drop(['sales'], axis = 1)

6. To summarize data frame

To summarize or explore data, you can submit the command below.

df.describe()

cost sales count 6.000000 6.00000 mean 1166.150000 1242.65000 std 237.926793 230.46669 min 1003.700000 1010.00000 25% 1020.000000 1058.90000 50% 1072.000000 1205.85000 75% 1184.000000 1366.07500 max 1625.200000 1604.80000

To summarise all the

**character variables**, you can use the following script.

df.describe(include=['O'])Similarly, you can use

**df.describe(include=['float64'])**to view summary of all the numeric variables with decimals.

To select only a particular variable, you can write the following code -

df.productcode.describe()

OR

df["productcode"].describe()

count 6 unique 2 top BB freq 3 Name: productcode, dtype: object

7. To calculate summary statistics

We can manually find out summary statistics such as count, mean, median by using commands belowdf.sales.mean() df.sales.median() df.sales.count() df.sales.min() df.sales.max()

8. Filter Data

Suppose you are asked to apply condition - productcode is equal to "AA" and sales greater than or equal to 1250.
df1 = df[(df.productcode == "AA") & (df.sales >= 1250)]It can also be written like :

df1 = df.query('(productcode == "AA") & (sales >= 1250)')In the second query, we

**do not need**to specify DataFrame along with variable name.

9. Sort Data

In the code below, we are arrange data in ascending order by sales.df.sort_values(['sales'])

10. Group By : Summary by Grouping Variable

Like SQL GROUP BY, you want to summarize continuous variable by classification variable. In this case, we are calculating average sale and cost by product code.df.groupby(df.productcode).mean()

cost sales productcode AA 1283.066667 1146.466667 BB 1049.233333 1338.833333Instead of summarising for multiple variable, you can run it for a single variable i.e. sales. Submit the following script.

df["sales"].groupby(df.productcode).mean()

11. Define Categorical Variable

Let's create a classification variable - id which contains only 3 unique values - 1/2/3.df0 = pd.DataFrame({'id': [1, 1, 2, 3, 1, 2, 2]})

*Let's define as a categorical variable.*

We can use astype() function to make id as a categorical variable.

df0.id = df0["id"].astype('category')Summarize this classification variable to check descriptive statistics.

df0.describe()

id count 7 unique 3 top 2 freq 3

Frequency Distribution

You can calculate frequency distribution of a categorical variable. It is one of the method to explore a categorical variable.df['productcode'].value_counts()

BB 3 AA 3

12. Generate Histogram

Histogram is one of the method to check distribution of a continuous variable. In the figure shown below, there are two values for variable 'sales' in range 1000-1100. In the remaining intervals, there is only a single value. In this case, there are only 5 values. If you have a large dataset, you can plot histogram to identify outliers in a continuous variable.
df['sales'].hist()

Histogram |

13. BoxPlot

Boxplot is a method to visualize continuous or numeric variable. It shows minimum, Q1, Q2, Q3, IQR, maximum value in a single graph.
df.boxplot(column='sales')

BoxPlot |

Detailed Tutorial :

Data Analysis with Pandas TutorialData Science using Python - Examples

In this section, we cover how to perform data mining and machine learning algorithms with Python. sklearn is the most frequently used library for running data mining and machine learning algorithms. We will also cover statsmodels library for regression techniques. statsmodels library generates formattable output which can be used further in project report and presentation.1. Install the required libraries

Import the following libraries before reading or exploring data#Import required libraries

import pandas as pd

import statsmodels.api as sm

import numpy as np

2. Download and import data into Python

With the use of python library, we can easily get data from web into python.# Read data from web

df = pd.read_csv("https://stats.idre.ucla.edu/stat/data/binary.csv")

Variables Type Descriptiongre Continuous Graduate Record Exam score gpa Continuous Grade Point Average rank Categorical Prestige of the undergraduate institution admit Binary Admission in graduate school

**The binary variable admit is a target variable.**

3. Explore Data

Let's explore data. We'll answer the following questions -- How many rows and columns in the data file?
- What are the distribution of variables?
- Check if any outlier(s)
- If outlier(s), treat them
- Check if any missing value(s)
- Impute Missing values (if any)

# See no. of rows and columns

df.shape

**Result :**400 rows and 4 columns

In the code below, we rename the variable rank to '

**position**' as rank is already a function in python.

# rename rank columnSummarize and plot all the columns.

df = df.rename(columns={'rank': 'position'})

# Summarize

df.describe()

# plot all of the columns

df.hist()

Categorical variable Analysis

It is important to check the frequency distribution of categorical variable. It helps to answer the question whether data is skewed.# Summarize

df.position.value_counts(ascending=True)

1 61 4 67 3 121 2 151

Generating Crosstab

By looking at cross tabulation report, we can check whether we have enough number of events against each unique values of categorical variable.pd.crosstab(df['admit'], df['position'])

position 1 2 3 4 admit 0 28 97 93 55 1 33 54 28 12

Number of Missing Values

We can write a simple loop to figure out the number of blank values in all variables in a dataset.for i in list(df.columns) :In this case, there are no missing values in the dataset.

k = sum(pd.isnull(df[i]))

print(i, k)

## Data Science with Python

In this section we covered how we can build predictive model using common statistical modeling techniques. It also includes data processing steps prior to model development and validation.### Logistic Regression

Logistic Regression is a special type of regression where target variable is categorical in nature and independent variables be discrete or continuous. In this post, we will demonstrate only**binary logistic regression**which takes only binary values in target variable. Unlike linear regression, logistic regression model returns probability of target variable.It assumes binomial distribution of dependent variable. In other words, it belongs to binomial family.

In python, we can write R-style model formula

**y ~ x1+ x2+ x3**using

**patsy**and

**statsmodels**libraries. In the formula, we need to define variable 'position' as a categorical variable by mentioning it inside capital C(). You can also define reference category using

**reference=**option.

#Reference CategoryIt returns two datasets - X and y. The dataset 'y' contains variable admit which is a target variable. The other dataset 'X' contains Intercept (constant value), dummy variables for Treatment, gre and gpa. Since 4 is set as a reference category, it will be 0 against all the three dummy variables. See sample below -

from patsy import dmatrices, Treatment

y, X = dmatrices('admit ~ gre + gpa + C(position, Treatment(reference=4))', df, return_type = 'dataframe')

P P_1 P_2 P_3 3 0 0 1 3 0 0 1 1 1 0 0 4 0 0 0 4 0 0 0 2 0 1 0

Split Data into two parts

80% of data goes to training dataset which is used for building model and 20% goes to test dataset which would be used for validating the model.from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

Build Logistic Regression Model

By default, the **regression without formula style**does not include intercept. To include it, we already have added

**intercept**in

**X_train**which would be used as a predictor

**.**

#Fit Logit model logit = sm.Logit(y_train, X_train) result = logit.fit() #Summary of Logistic regression model result.summary() result.params

Logit Regression Results ============================================================================== Dep. Variable: admit No. Observations: 320 Model: Logit Df Residuals: 315 Method: MLE Df Model: 4 Date: Sat, 20 May 2017 Pseudo R-squ.: 0.03399 Time: 19:57:24 Log-Likelihood: -193.49 converged: True LL-Null: -200.30 LLR p-value: 0.008627 ======================================================================================= coef std err z P|z| [95.0% Conf. Int.] --------------------------------------------------------------------------------------- C(position)[T.1] 1.4933 0.440 3.392 0.001 0.630 2.356 C(position)[T.2] 0.6771 0.373 1.813 0.070 -0.055 1.409 C(position)[T.3] 0.1071 0.410 0.261 0.794 -0.696 0.910 gre 0.0005 0.001 0.442 0.659 -0.002 0.003 gpa 0.4613 0.214 -2.152 0.031 -0.881 -0.041 ======================================================================================

Confusion Matrix and Odd Ratio

Odd ratio is exponential value of parameter estimates.#Confusion Matrix

result.pred_table()

#Odd Ratio

np.exp(result.params)

Prediction on Test Data

In this step, we take estimates of logit model which was built on training data and then later apply it into test data.
#prediction on test data

y_pred = result.predict(X_test)

Calculate Area under Curve (ROC)

# AUC on test data

false_positive_rate, true_positive_rate, thresholds = roc_curve(y_test, y_pred)

auc(false_positive_rate, true_positive_rate)

**Result :**AUC =0.6763

Calculate Accuracy Score

accuracy_score([ 1 if p > 0.5 else 0 for p in y_pred ], y_test)

### Decision Tree

Decision trees can have a target variable continuous or categorical. When it is continuous, it is called regression tree. And when it is categorical, it is called classification tree. It selects a variable at each step that best splits the set of values. There are several algorithms to find best split. Some of them are Gini, Entropy, C4.5, Chi-Square. There are several advantages of decision tree. It is simple to use and easy to understand. It requires a very few data preparation steps. It can handle mixed data - both categorical and continuous variables. In terms of speed, it is a very fast algorithm.#Drop Intercept from predictors for tree algorithms X_train = X_train.drop(['Intercept'], axis = 1) X_test = X_test.drop(['Intercept'], axis = 1) #Decision Tree from sklearn.tree import DecisionTreeClassifier model_tree = DecisionTreeClassifier(max_depth=7) #Fit the model: model_tree.fit(X_train,y_train) #Make predictions on test set predictions_tree = model_tree.predict_proba(X_test) #AUC false_positive_rate, true_positive_rate, thresholds = roc_curve(y_test, predictions_tree[:,1]) auc(false_positive_rate, true_positive_rate)

**Result :**AUC = 0.664

Important Note

Feature engineering plays an important role in building predictive models. In the above case, we have not performed variable selection. We can also select best parameters by using grid search fine tuning technique.

### Random Forest

Decision Tree has limitation of overfitting which implies it does not generalize pattern. It is very sensitive to a small change in training data. To overcome this problem, random forest comes into picture. It grows a large number of trees on randomised data. It selects random number of variables to grow each tree. It is more robust algorithm than decision tree. It is one of the most popular machine learning algorithm. It is commonly used in data science competitions. It is always ranked in top 5 algorithms. It has become a part of every data science toolkit.#Random Forest from sklearn.ensemble import RandomForestClassifier model_rf = RandomForestClassifier(n_estimators=100, max_depth=7) #Fit the model: target = y_train['admit'] model_rf.fit(X_train,target) #Make predictions on test set predictions_rf = model_rf.predict_proba(X_test) #AUC false_positive_rate, true_positive_rate, thresholds = roc_curve(y_test, predictions_rf[:,1]) auc(false_positive_rate, true_positive_rate) #Variable Importance importances = pd.Series(model_rf.feature_importances_, index=X_train.columns).sort_values(ascending=False) print(importances) importances.plot.bar()

**Result :**AUC = 0.6974

### Grid Search - Hyper Parameter Tuning

The sklearn library makes hyper-parameters tuning very easy. It is a strategy to select the best parameters for an algorithm. In scikit-learn they are passed as arguments to the constructor of the estimator classes. For example, max_features in randomforest. alpha for lasso.from sklearn.model_selection import GridSearchCV rf = RandomForestClassifier() target = y_train['admit'] param_grid = { 'n_estimators': [100, 200, 300], 'max_features': ['sqrt', 3, 4] } CV_rfc = GridSearchCV(estimator=rf , param_grid=param_grid, cv= 5, scoring='roc_auc') CV_rfc.fit(X_train,target) #Parameters with Scores CV_rfc.grid_scores_ #Best Parameters CV_rfc.best_params_ CV_rfc.best_estimator_ #Make predictions on test set predictions_rf = CV_rfc.predict_proba(X_test) #AUC false_positive_rate, true_positive_rate, thresholds = roc_curve(y_test, predictions_rf[:,1]) auc(false_positive_rate, true_positive_rate)

### Cross Validation

# Cross Validation from sklearn.linear_model import LogisticRegression from sklearn.model_selection import cross_val_predict,cross_val_score target = y['admit'] prediction_logit = cross_val_predict(LogisticRegression(), X, target, cv=10, method='predict_proba') #AUC cross_val_score(LogisticRegression(fit_intercept = False), X, target, cv=10, scoring='roc_auc')

### Preprocessing Steps

1.**The machine learning package sklearn requires all categorical variables in numeric form**. Hence, we need to convert all character/categorical variables to be numeric. This can be accomplished using the following script. In sklearn, there is already a function for this step.

from sklearn.preprocessing import LabelEncoder def ConverttoNumeric(df): cols = list(df.select_dtypes(include=['category','object'])) le = LabelEncoder() for i in cols: try: df[i] = le.fit_transform(df[i]) except: print('Error in Variable :'+i) return df ConverttoNumeric(df)

Encoding |

2. Create Dummy Variables

Suppose you want to convert categorical variables into dummy variables. It is different to the previous example as it creates dummy variables instead of convert it in numeric form.productcode_dummy = pd.get_dummies(df["productcode"]) df2 = pd.concat([df, productcode_dummy], axis=1)

The output looks like below -

AA BB 0 1 0 1 1 0 2 1 0 3 0 1 4 0 1 5 0 1

Create k-1 Categories

To avoid multi-collinearity, you can set one of the category as reference category and leave it while creating dummy variables. In the script below, we are leaving first category.

productcode_dummy = pd.get_dummies(df["productcode"], prefix='pcode', drop_first=True) df2 = pd.concat([df, productcode_dummy], axis=1)

3. Impute Missing Values

Imputing missing values is an important step of predictive modeling. In many algorithms, if missing values are not filled, it removes complete row. If data contains a lot of missing values, it can lead to huge data loss. There are multiple ways to impute missing values. Some of the common techniques - to replace missing value with mean/median/zero. It makes sense to replace missing value with 0 when 0 signifies meaningful. For example, whether customer holds a credit card product.
Fill missing values of a particular variable

# fill missing values with 0 df['var1'] = df['var1'].fillna(0) # fill missing values with mean df['var1'] = df['var1'].fillna(df['var1'].mean())

Apply imputation to the whole dataset

from sklearn.preprocessing import Imputer # Set an imputer object mean_imputer = Imputer(missing_values='NaN', strategy='mean', axis=0) # Train the imputor mean_imputer = mean_imputer.fit(df) # Apply imputation df_new = mean_imputer.transform(df.values)

4. Outlier Treatment

There are many ways to handle or treat outliers (or extreme values). Some of the methods are as follows -- Cap extreme values at 95th / 99th percentile depending on distribution
- Apply log transformation of variables. See below the implementation of log transformation in Python.

import numpy as np

df['var1'] = np.log(df['var1'])

5. Standardization

In some algorithms, it is required to standardize variables before running the actual algorithm. Standardization refers to the process of making mean of variable zero and unit variance (standard deviation).#load dataset dataset = load_boston() predictors = dataset.data target = dataset.target df = pd.DataFrame(predictors, columns = dataset.feature_names) #Apply Standardization from sklearn.preprocessing import StandardScaler k = StandardScaler() df2 = k.fit_transform(df)

End Notes

Next Step - Practice, practice and practice. Download free public data sets from Kaggle / UCLA websites and try to play around with data and generate insights from it with pandas package and build statistical models using sklearn package. I hope you would find this tutorial helpful. I tried to cover all the important topics which beginner must know about Python. Once completion of this tutorial, you can flaunt you know how to program it in Python and you can implement machine learning algorithms using sklearn package.
Hi, excelent tutorial!!! I'm mostly a user of R but want to learn python. The thing is i work a lot with spatial data: spatial relationships (spdep), interpolation (kriging with gstat or multilevel B-Splines with MBA) etc.; and then machine learning methods with the data that comes from spatial features.

ReplyDeleteI understand that the ML cappabilities are already in Pythoon but i'm worried about the spatial workflow, can you give me some insights on this?

Thanks,

Great blog!

Thanks for developing this. For first time after few attempts, I can start working with Python!

ReplyDeleteGlad you found it helpful. Cheers!

DeleteHi Deepanshu. Can i have your contact number please. I want to talk regarding the courses.

DeleteThanks. Some things come late in the tutorial (like the np loading) but it is a good overview.

ReplyDeleteExcelent! I appreciate the comparison between R and Python commands! Very useful!

ReplyDeleteVery well written article!

ReplyDeleteHi.

ReplyDeleteI am using Pythin 3.6.

y, X = dmatrices("admit ~ gre + gpa + C(position, Treatment(reference=4))", df, return_type = 'dataframe')

This code generate this error

C, including its class ClassRegistry, has been deprecated since SymPy

1.0. It will be last supported in SymPy version 1.0. Use direct

imports from the defining module instead. See

https://github.com/sympy/sympy/issues/9371 for more info.

.

.

.

TypeError: 'bool' object is not callable

How can I handle this ?

Thank you

Very useful tutorial, lucidly presented

ReplyDelete⏳⛳🏳🚩🏁⛖🔛✖➕➖➗⁉⚌⚲

ReplyDeleteThanks for an amazing introduction to Python.

ReplyDeleteThank you for this interesting tutorial.

ReplyDeleteNicely written.. Thanks

ReplyDeleteThank you for the tutorial. Bookmarked this so I can learn to use what you find essential when using the Pandas package.

ReplyDeleteExcellent resources to get hands on quick with Python

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteHey very nice blog!!

ReplyDeleteI enjoy reading through your article post, I wanted to write a little comment to support you and wish you a good continuation. All the best for all your blogging efforts. The way of explanation about the comparison between R and Python is nice.

Appreciate the recommendation! Let me try it out.

Keep working ,great job!

Very useful tutorial, lucidly presented

ReplyDeleteI didn’t know about that Jupyter thing, thank you for information. Spyder Shortcut Keys are quite useful too, but I think what most important in programming is to know the top. For example, the top programming languages 2018 can help you to know all the possibilities and the most convenient keys. I use the website to know what works now and came with some new ideas. If you can be really fast you won’t succeed.

ReplyDeletehi while i am running

ReplyDeleteimport pandas as pd

s1 = pd.Series(np.random.randn(5))

s1

Its gives out an error as "np is not defined".

can you please rectify?

you also need to submit "import numpy as np" before pd.series()

DeleteHey there, you have to import numpy as well

Deletene pulse ultra for beginners

ReplyDeletesir..my doubt is that "Do we need to worry about removal of variables based on multicollinearity or the sklearn will take care of it automatically"

ReplyDeleteWe need to handle multicollinearity issue. Sklearn package would not take care of it automatically.

Deletegreat one!

ReplyDeleteI enjoyed reading through your post, The way of explanation about the comparison between R and Python is nice.

ReplyDeleteKeep working ,great job!

ReplyDeleteI enjoyed reading through your post, The way of explanation about the comparison between R and Python is nice.

Keep working ,great job!

Excellent tutorial.....

ReplyDeleteGreat

ReplyDeleteHello, I have browsed most of your posts. This post is probably where I got the most useful information for my research. Thanks for posting, maybe we can see more on this. Are you aware of any other websites on this subject. Visit here

ReplyDeleteBro Listendata is my fav blog for datascience. I have already learnt R using your tutorials, now I am learning Python. I am extremely thankful to you for such amazing content.

ReplyDeleteGod bless you and Lots of love!!!!!

This comment has been removed by a blog administrator.

ReplyDeletedont judje mi spelling

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteCool stuff you have, and you keep overhaul every one of us.

ReplyDeleteExcelR Solutions

More Info

ReplyDeleteI have express a few of the articles on your website now, and I really like your style of blogging. I added it to my favorite’s blog site list and will be checking back soon…

Modification as per my observation in Logistic regression

ReplyDelete(1) The other dataset 'X' contains Intercept (constant value), dummy variables for Treatment, gre and gpa. It may may right to modify this statement like

The other dataset 'X' contains Intercept (constant value), gre, gpa and dummy variables for Treatment

(2) In logistic regression these two module need to import

from sklearn.metrics import roc_curve

from sklearn.metrics import auc

This comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteWhat a really awesome post this is. Truly, one of the best posts I've ever witnessed to see in my whole life. Wow, just keep it up.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteTruly impressive. I cannot thank you enough.

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteThanks for the guide.

ReplyDeleteNice post.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteI'm really thankful that I read this. It's extremely valuable and quite informative and I truly learned a great deal from it.

ReplyDeleteThank you for this article. I learn Python now and it is very helpful for me

ReplyDeleteThanks for Sharing

ReplyDeleteThis will be really helpful, thanks for sharing. will be waiting for more.

ReplyDeleteThis comment has been removed by a blog administrator.

ReplyDeleteThanks for such a great post and the review, I am totally impressed! Keep stuff like this coming. ExcelR Data Analyst Course

ReplyDeleteGreat content useful for all the candidates of Data Science training who want to kick start these career in Data Science training field.

ReplyDeleteim really loved this post and informations.

ReplyDeleteThanks

this content more valuable then any other video.

ReplyDeleteThanks

This comment has been removed by a blog administrator.

ReplyDeleteHey, thank you for sharing this very informative piece of content .

ReplyDeleteThanks for sharing such a valuable blog. I appreciated your effort in this blog. Keep sharing some more articles.

ReplyDeleteGreat post i must say and thanks for the information. Education is definitely a sticky subject. However, is still among the leading topics of our time. I appreciate your post and look forward to more.

ReplyDeleteData Science Course in Bangalore

I want to leave a little comment to support and wish you the best of luck.we wish you the best of luck in all your blogging enedevors.

ReplyDeletedata science course fees in bangalore

i am glad to discover this page : i have to thank you for the time i spent on this especially great reading !! i really liked each part and also bookmarked you for new information on your site.

ReplyDeletecyber security training in bangalore

You appear to know a lot about this. It’s like you read my thoughts!, Thanks for writing this.

ReplyDeleteData Science Course in Vadodara

Your style is really unique in comparison to other people I have read stuff from. I appreciate you for posting when you have the opportunity, Guess I will just bookmark this site.

ReplyDeleteData Science Course in Nagpur

nice article and very informative.

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteExcellent post,Keep sharing such type of wonderful post..

ReplyDeleteVery informative and well written post.

ReplyDeleteNice blog and informative content.

ReplyDeleteI really enjoy simply reading all of your weblogs. Simply wanted to inform you that you have people like me who appreciate your work. Definitely a great post. Hats off to you! The information that you have provided is very helpful.

ReplyDeletedata scientist training and placement in hyderabad

This could be an awesome point to start learning python. I'm mostly over php/javascript but it would be a personal challenge for me to start digging it. Thanks for sharing!

ReplyDeleteThanks for sharing such a valuable blog. I appreciated your effort in this blog. Keep sharing some more articles.

ReplyDeletethanks for sharing this

ReplyDeleteThis comment has been removed by the author.

ReplyDeleteI am a new user of this site, so here I saw several articles and posts published on this site, I am more interested in some of them, will provide more information on these topics in future articles.

ReplyDeletedata science course in london

Amazingly by and large very interesting post. I was looking for such an information and thoroughly enjoyed examining this one. Keep posting. An obligation of appreciation is all together for sharing.data science training in kolhapur

ReplyDeleteThe blog is informative and very useful therefore, I would like to thank you for your effort in writing this article.

ReplyDeleteData Analytics Course in Lucknow

Extremely overall quite fascinating post. I was searching for this sort of data and delighted in perusing this one. Continue posting. A debt of gratitude is in order for sharing.data science course in kolhapur

ReplyDeleteNice blog, informative and knowledgeable content. Thanks for sharing with us.

ReplyDeleteData Science Course in Hyderabad

I'm genuinely getting a charge out of scrutinizing your richly formed articles. Apparently you consume a huge load of energy and time on your blog. I have bookmarked it and I am expecting scrutinizing new articles. Continue to do amazing.data science course in ghaziabad

ReplyDeleteThis post of you is looking so helpfull in getting the knwoledge or help in enhace the knowledge keep on posting like this thanks

ReplyDeleteLink:-https://delhicoursestraininginstitute.com/top-best-data-science-training-institute-in-delhi/

Thanks for Very useful article.

ReplyDeletepower bi training

Thanks for sharing informative post.

ReplyDeletepower bi course

Extremely overall quite fascinating post. I was searching for this sort of data and delighted in perusing this one. Continue posting.

ReplyDeleteNice article, thanks for sharing.

ReplyDeletepower bi course

I love to recommend you Where can crawl Exciting Products latest Jackets, Coats and Vests Click Here Big Boss Leather Jacket

ReplyDeleteExtremely overall quite fascinating post. I was searching for this sort of data and delighted in perusing this one. Continue posting. A debt of gratitude is in order for sharing.data analytics course in warangal

ReplyDeleteThis is a wonderful inspiring article. I am practically satisfied with your great work. You have really put together extremely helpful data. Keep it up.. Are you planning to visit Kenya?For this, you need to fill the Kenya Visa Application Form and pay the kenya e visa fees online.

ReplyDeleteI want to leave a little comment to support and wish you the best of luck.we wish you the best of luck in all your blogging enedevors

ReplyDeletebusiness analytics course in varanasi

Website is so easy to use – I am impressed with it. Thank you for sharing.

ReplyDeleteDevOps Course

Amazing knowledge and I like to share this kind of information with my friends and hope they like it they why I do

ReplyDeletefull stack web development course

I want to leave a little comment to support and wish you the best of luck.we wish you the best of luck in all your blogging enedevors

ReplyDeletedata science course in varanasi

Nice content, so much convincing and easy to grasp. Thank you for sharing.

ReplyDeleteSAP analytics cloud training

Nice post of you all the information shared here is to the point and clear words expressed here keep on posting this type of content thanks

ReplyDeleteLink:-https://delhicoursestraininginstitute.com/top-best-data-science-training-institute-in-delhi/

Extremely overall quite fascinating post. I was searching for this sort of data and delighted in perusing this one. Continue posting. A debt of gratitude is in order for sharing.data science training in warangal

ReplyDeleteAmazingly by and large very interesting post. I was looking for such an information and thoroughly enjoyed examining this one. Keep posting. An obligation of appreciation is all together for sharing.data science training in gwalior

ReplyDeleteYour Site is very nice, and it's very helping us this post is unique and interesting, thank you for sharing this awesome information. and visit our blog site also ........ โจ๊กเกอร์

ReplyDeleteโจ๊กเกอร์

DeleteAmazingly by and large very interesting post. I was looking for such an information and thoroughly enjoyed examining this one. Keep posting. An obligation of appreciation is all together for sharing.data science training in kolhapur

ReplyDeleteThis is a wonderful article, Given so much info in it, These type of articles keeps the users interest in the website, and keep on sharing more ... good luck. Data Science Course in Dehradun

ReplyDeleteI want to leave a little comment to support and wish you the best of luck.we wish you the best of luck in all your blogging enedevors

ReplyDeletedata science course in trivandrum

you have done a great job. I will definitely dig it and personally recommend to my friends. Thank you for sharing konveksi kaos

ReplyDelete