CSEIT183687 | Received : 20 July 2018 | Accepted : 05 August 2018 | July-August-2018 [ 3 (6) : 428-434 ]
International Journal of Scientific Research in Computer Science, Engineering and Information Technology
© 2018 IJSRCSEIT | Volume 3 | Issue 6 | ISSN : 2456-3307
428
Sentiment Analysis Using Parallel Computing Through GPU
Harshita Mandloi, Shraddha Masih
School of Computer Science and IT, DAVV, Indore, Madhya Pradesh, India
ABSTRACT
Parallel Computing is becoming important in the field of computer science and is proven as a high-performance
solution. Over the couple of years, GPU has gained an important place in the field of high-performance
computing. Social media is expanding at present and becoming important in society. Social network sites allow
users to communicate with people in the network by sharing posts, images, videos, status. The proposed system
gathers the information from the social media websites and performs the sentiment analysis on the social media
data using GPU. The work concentrates on recognizing the sentiment information from the text reviews and
using that to identify the items. The aim of this paper is to do analytics on social media data. Analysis is done on
the data using K Nearest Neighbor algorithms and Support Vector Machine algorithm on the GPU.
Keywords: Parallel Computing, GPU, Social media, K nearest neighbor, SVM
I. INTRODUCTION
Nowadays social media, blogs, and other media
produce a huge amount of data on the world wide
web. This huge amount of data contain a critical
opinion related to the information that can be used to
benefit business and other aspects of commercial and
scientific industries. Analysis of user sentiments can
help in a business decisions. Sentiments can be
categorized into 3 types:- Positive, Negative and
Neutral.
Graphics Processing Unit has shown better
computational performance as compared to the
current main-stream multicore Central Processing
Unit. Apart from graphics and multimedia processing,
high-performance GPUs are also mapped to take care
of General Purpose Computing.
Machine learning is the current trend in computing
that is focused on designing and developing
algorithms which allows computers to learn. It can
capture characteristics of interest in order to make a
prediction for a new data query. The gathered data is
considered as training data which illustrate the
relationships among the observed variables. Many
important patterns can be recognized after applying
the learning procedure. Supervised learning is the
data mining task of classifying data into labeled data.
This work is mainly focused on the classification
tools- Support Vector Machine and K Nearest
Neighbor. To improve the performance of this
algorithm GPU is used with CUDA. Parallel SVM
algorithm is implemented on GPU using CUDA
framework. The implementation of parallel SVM
achieves a great performance using GPU. This
software utilizes the parallelism in both data level
and task level to maximize the performance of the
GPU.
II. BACKGROUND
A. Parallel Computing- Parallel computing is
becoming important day by day. Dividing large task
into sub task and assigning these tasks to
multiprocessor and executing them concurrently is
called parallel processing. Parallel computing is used
for high performance computing in the field of data
analytics. For the high performance computing the