e-ISSN: 2348-6848

International Journal of Research p-ISSN: 2348-795X

Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

Python and Machine Learning

Mrs. Smita Desai & Ms. Shreya Desai
1 Department of Computer Science, Bharatesh College of Computer Applications, Belagavi, Karnataka, In-
dia
2 Electronic and communications, Gogte Institute of Technology, Belagavi, Karnataka, India

Abstract—Python is gaining the popularity. level language. Python offers three main paradigm
In this paper we have tried to find the to its developers namely, object orientation, func-
characteristics of Python language that helps
tional and structured programming. Python has
it gain the attention of the programmers. This
paper is an initiative to review the role of multiple implementations, the most popular one is
Python in machine learning. Machine Cpython- a default implementation. Other python
learning is the most happening technology of
implementations are Jpython – scripted in Java,
todays'sworld.The main aim of Machine
Learning is to allow the computers learn PyPy – written in Rpython, Iron-Pyhton written in
automatically without human intervention and C#. These implementations work in the native lan-
adjust its actions accordingly. In this paper
guage they are written in but are capable of inte-
main focus is on the popularity of the Python
as a language preferred by the developers racting with other languages through the use of
forMachine Learning. We have included the modules. Most of these implementations are pen
statistics of other computer languages and source and free.
Python to support the popularity of Python in
machine learning. The main focus is on the Python is the most preferred language over other
topic “ Python - the ideal language for languages like Java, C and C++ by the program-
Machine Learning” mers.Most of the software development companies
Index Terms—Machine Learning with prefer Python as a programming language because
Python, Popular language for Machine
Learning, Python programming, Python then of its fewer coding lines and versatile features.
and now, Python implementations. Some applications of Python are:
 Image processing and Graphic design

1. INTRODUCTION  Game development

 Web frameworks and web applications
Python‘s role in Machine Learning is gaining  Enterprise applications
more and more attention. The main reason could be  Language development
the simplicity of python language and its rich set of
libraries. Machine Learning is an application of Artificial
Python is a popularly used general purpose high Intelligence (AI). It gives the ability to learn auto-

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4013

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

matically from the experience and improve accor-

dingly without being explicitly programmed. The  Simple Code
process of learning starts with the data or observa- Python‘s simple to learn syntax and excellent
tions. The iterative nature allows models to be able readability style makes it very easy for the begin-
to adapt the new data exposed to them independent- ners to understand and code.
ly.
 Integration Feature
Python integrates the Enterprise Application
2. MERITS AND DEMERITS OF PYTHON Integration which makes it easy to develop web
2.1 Advantages
services by invoking COM or COBRA components.
 Extensive Support Libraries
It also has powerful control capabilities as it calls
Most of the frequently used tasks are already
directly through C, C++ of Java via Jython. Python
scripted into its standard libraries making the Py-
also processes XML and other markup languages as
thon code petite.
it can run on all modern operating systems through
same byte code.
 Open Source and Community Develop-
ment
2.2 Disadvantages
Python is developed under OSI approved
 Low Speed
open source license, thus making it available free to
 Week Mobile Computing
use and distribute even for commercial pur-
 Under developed Database Access layers
pose.Python development is driven by community
which collaborates for its code through hosting con- 3. MACHINE LEARNING
ferences, mailing lists etc.
Machine learning is the technology that allows
machines to learn on its own from the past expe-
 Third Party Module Support
rience, data or examples. There has been significant
Third party modules present in Python Pack-
advances in machine learning due to the advance-
age Index (PyPI) makes it possible to interact with
ments in technology, easy and increased availability
most of the other languages and platforms.
of data and the computing power.
In addition to image processing and voice recog-
 User friendly Data Structure
nition systems which are machine learning driven
Python has built-in dictionary ad list data
systems, it holds the developments in diverse range
structure, which can be used to construct run time
of fields that includes education, health care and
data structures quickly.

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4014

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

many more. It could also support scientific ad-

vances by handling the large datasets.

3.1 Machine Learning Types

The three key branches of machine learning are:
 Supervised learning
In this type, the machine is trained with la-
belled data. The label categories each data points
into one or more groups, such as ‗fruits‘ or ‘vegeta-
Kaggle offer machine learning completions and
bles‘. The machine learns how this data is struc-
have polled their user base as to the tools and pro-
tured(known as training data) and uses this to pre-
gramming languages used by partisans in competi-
dict the categories of new or ‗test‘ data.
tion.

 Unsupervised learning
Unsupervised learning is learning without la-
bels. It aims to detect the characteristics that make
Kaggle‘s poll results in 2011 is as follows:
the data more or less similar to each other. For ex-
ample, creating clusters and assigning data to these
clusters.

 Reinforcement learning
Reinforcement learning lies between unsu-
pervised and supervised learning. It focuses on
learning from experience.

Following is the poll of languages by kdNuggets

Popular programming languages on Kaggle

While many machine learning algorithms have

been around for a long time, the ability to automati-
cally apply complex mathematical calculations to

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4015

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

big data – over and over, faster and faster is a recent  Ensemble modeling
development.
3.4 Who is Using Machine Learning?
Few examples of machine learning applications:
 The self-driving Google car Most industries working with large amounts of
data have recognized the importance of machine
 Online recommendation offers such as Ama-
learning technology. By gleaning insights from this
zon
data (often in real time) organizations are able to
 Knowing what customers are saying about
work more efficiently or gain advantage over com-
you on Twitter – Machine learning combined
petitors.
with linguistic rule creation
 Fraud detection
 Financial services
3.2 Why is Machine Learning Important? Banks and other businesses in the financial
The popularity of machine learning is due to the industry use machine learning technology for two
same factors that have made data mining and Baye- key purposes: to identify important insights in data,
sian analysis more popular than ever. Things like and prevent fraud. The insights can identify in-
growing volumes and varieties of available data, vestment opportunities, help investors know when
computational processing that is cheaper and more to trade. Data mining can also identify clients with
powerful, and affordable data storage. high risk profiles, use cyber surveillance to pinpoint
All of these makes me possible to quickly and warning signs of fraud.
automatically produce models that can analyze big-
 Government
ger, more complex data and deliver faster, more
Government agencies such as public safety
accurate results- even on a very large scale. Build-
and utilities have a particular need for machine
ing precise models, an organization has better
learning since hay have multiple sources of data
chance of identifying profitable opportunities or
that can be mined for insights. Analyzing sensor
avoiding unknown risks.
data, for example, identifies ways to increase effi-
ciency and save money. Machine learning can also
3.3 What is Required to Create Good Machine
help detect fraud and minimizing identity theft.
Learning Systems?
 Data preparation capabilities  Health care
 Algorithms – basic and advanced Machine learning is a fast growing trend in
 Automation and iterative processes the health care industry, wearable devices and sen-
 Scalability sors can use data asses a patient‘s health in real

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4016

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

time. The technology can also help medical exerts Python. PyBrain is implemented in Python, with the
analyze data to identify trends or red flags that may scientific library SciPy. Pybrain is designed to be
lead to improved diagnoses and treatment. able to connect various types of architectures and
algorithms. PyBrain provides a toolbox for super-
 Oil and gas
vised, unsupervised and reinforcement learning as
Finding new energy sources. Analyzing min-
well as black box and multi objective optimization.
erals in the ground. Predicting refinery sensor fail-
The library includes different types of training
ure. Streamlining oil distribution to make it more
algorithms, specialized data sets, trainable architec-
efficient and cost effective. The number of machine
tural components etc. Appropriate data handling
learning use cases for this industry is vast and still
tools have been developed for special applications
expanding.
reinforcement learning, handwriting recognition
applications.
 Marketing and Sales
Websites recommending items you might like
based on previous purchases are using machine lea- 4.1 What’s Unique in Pybrain

vening to analyze your buying history – and pro- PyBrain is short forPython-Based Reinforcement

mote other items you‘d be interested in. This ability Learning, Artificial Intelligence and Neural Net-

to capture data, analyze it and use it to personize a work Library. Amongst few machine learning libra-

shopping experience or implementing a marketing ries available, PyBrain aims to be very easy-to-use

campaign is the future of retail. modular library that offers flexibility and algo-
rithms for research.
 Transportation PyBrain contains algorithms for neural networks,
Analyzing data to identify patterns and trends for reinforcement learning (and the combination of
is key to the transportation industry, which relies on the two), for unsupervised learning and evolution.
making routes more efficient and predicting poten- The library is used around neural networks in the
tial problems to increase profitability. The data kernel and all of the training methods accept a
analysis and modeling aspects of machine learning neural network as the to-be-trained instance. This
are important tools to delivery companies, public makes PyBrain powerful tool for real-life tasks.
transportation and other transportation organiza-
tions. 4.2 Using PyBrain
PyBrain is open source and free to use for everyone
4. PYTHON’S PYBRAIN
(it is licensed under the BSD Software License).
PyBrain is a machine learning library written in We can download it and start using the algorithms

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4017

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

and modules in our project. ly.

5. USER FRIENDLY DATA STRUCTURES

Python has built-in list and dictionary data struc-

tures which can be used to construct fast runtime
data structures. Python also provides the option of
dynamic high-level data typing that reduces the
length of support code that is needed.

5.1 Python Pandas

The python package Pandas can help automate the
process of data inspection and handling. It proves
6. CONCLUSION
particularly useful for the early stages of data in-
In this paper the recent trends of research on Py-
spection and preprocessing.
thon‘s rich set of libraries used in machine learning
The Pandas package makes it possible to read
was reviewed. The focus was on the role of Python
data into a specialized data structure called data
as an upcoming machine learning language pre-
frame. The data frame is modeled after the CRAN-
ferred by the developers. In this paper we discussed
R data structure of the same name.
the importance, new trends and the vital role of ma-
The Pandas can be difficult to install as it has
chine learning in real life applications. This paper
number of dependencies that need to be correctly
reviews the Python libraries like PyBrain, Python
versioned. The installation procedures are easy to
Pandas and Python‘s Ensemble Package that are
follow and result in compatible installations of wide
used for machine learning project. We have also
variety of packages for data analysis and machine
discussed the simplicity and flexibility of Python
learning.
that makes it programmer‘s first choice. Though the
5.2 Python’s Ensemble Package research on Machine learning languages shows R as
The Python scikit-learn ensemble module houses a the first preferred language over Python, Python is
Radom Forest algorithm and a Gradient Boosting picking up the attention of the programmers. This
algorithm for regression problem. RandomFore- research paper on the Python‘s role as a program-
stRegressorhas several attributes, including the ming language in machine learning will help serve
trained trees that makeup the ensemble. The predict as a reference for researchers and machine learning
method will use the trained trees to make predic- programmers and also be an introduction for those
tions, so you will not generally access those direct- who are less familiar with the subject.

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4018

 e-ISSN: 2348-6848
International Journal of Research p-ISSN: 2348-795X
Available at Volume 04 Issue-17
https://edupediapublications.org/journals December 2017

REFERENCES [5] S.P. Bingulac, ―On the Compatibility of Adap-

[1] J.S. Bridle, ―Probabilistic Interpretation of Feed tive Controllers,‖ Proc. Fourth Ann. Allerton

forward Classification Network Outputs, with Rela- Conf. Circuits and Systems Theory, pp. 8-16,

tionships to Statistical Pattern Recognition,‖ Neuro- 1994. (Conference proceedings)

computing—Algorithms, Architectures and Applica- [6] H. Goto, Y. Hasegawa, and M. Tanaka, ―Effi-

tions, F. Fogelman-Soulie and J. Herault, eds., cient Scheduling Focusing on the Duality of

NATO ASI Series F68, Berlin: Springer-Verlag, pp. MPL Representation,‖ Proc. IEEE Symp. Com-

227-236, 1989. (Book style with paper title and edi- putational Intelligence in Scheduling(SCIS

tor) ’07), pp. 57-64, Apr. 2007,

[2] Mrs. Smita Desai, Miss. Shreya Desai ―Smart Ve- doi:10.1109/SCIS.2007.367670. (Conference

hicle Automation‖,IJCSMC, Vol. 6,Issue. 9, Sep- proceedings)

tember2017, pg.46 – 50 [7] J. Williams, ―Narrow-Band Analyzer,‖ PhD dis-

[3] W.-K. Chen, Linear Networks and Systems. sertation, Dept. of Electrical Eng., Harvard Un-

Belmont, Calif.: Wadsworth, pp. 123-135, iv., Cambridge, Mass., 1993. (Thesis or disserta-

1993. (Book style) tion)

[4] D.S. Coming and O.G. Staadt, "Velocity- [8] J.M.P. Martinez, R.B. Llavori, M.J.A. Cabo,

Aligned Discrete Oriented Polytopes for Dy- and T.B. Pedersen, "Integrating Data Ware-

namic Collision Detection," IEEE Trans. Vi- houses with Web Data: A Survey," IEEE

sualization and Computer Graphics, vol. 14, Trans. Knowledge and Data Eng., preprint, 21

no. 1, pp. 1-12, Jan/Feb 2008, Dec. 2007,

doi:10.1109/TVCG.2007.70405. (IEEE Transac- doi:10.1109/TKDE.2007.190746.(PrePrint)

tions )

Available online: https://edupediapublications.org/journals/index.php/IJR/ P a g e | 4019