International Journal of Civil Engineering and Technology (ijciet)

3. RELATED WORK 
Big data deals with large amount of data which constructs unstructured data in the which are 
collected from various sources which may consists of different format like plain text, image, 
speech, video and more which includes major components of big data to handle [1] [2].
Using unstructured data which reveals very important interrelationships which are very 
difficult to determine the way managing unstructured data will be big point at issue. The 
conversion to get structured data from unstructured data is considered in order to analyze and 
for better management of data [3]. 
The wide ranging of neural networks is also called as deep learning or deep neural 
networks (DNN) which as great impact in processing large volume of unstructured data which 

---

Analysis of Big Data with Neural Network
http://iaeme.com/Home/journal/IJCIET
214 
[e-mail] 
ranges in processing the data of speech recognition, visual object classification, information 
retrieval and natural language data. Deep neural network is the greatest tool recognized for 
processing big data [4] [5].
Big data along with neural networks which build up with two portions called training 
phase and operational phase. Training phase is obtained for specific task which makes use of 
neural networks which demands in storing large amount of memory to store the data and to 
compute the results which demands large amount of time consumption as well which leads to 
large scale of neural networks [6].
The performance of neural networks is been classified in the form of accuracy and the 
performance id predicted according to the accuracy levels of the data along with the increase 
in the range of neural network [7]. In order to proceeding huge assortment of data along with 
large volume data which involves neural networks and big data analytics we need more 
amount of computational power along with memory consumption as primary impact to 
process and evaluate the data which should emerge less cost as well[8]. 
There exists a survey which explore the computational power of GPU’s and for 
demonstrating an effective GPU’s for training phase which comes with the concept of big 
data analytics and large scale neural networks with DNN’s along with recurrent neural 
network[9][10].
Recurrent neural networks plays vital role for storing the information and capturing long 
range dependencies between the input data and it is a peculiar sample of neural network with 
which enables recurrent connection for RNN. However there exists large amount of 
computational complexities and difficulties which exists in training the data[11].
Understanding recurrent neural networks in terms of internal dynamics which as receive 
more and more attention the world of big data and neural networks. Most of the research has 
been proved in the concept of recurrent neural networks which provides awareness in terms of 
visualization properties which inherit internal state properties of RNN’s[12]. 
With this approach we come across two varieties of sequential networks which includes 
simple recurrent neural networks (SRR’s) for more efficient extension of GPU’s to gated 
recurrent units(GRU’s) which provide a huge research in recurrent neural networks along 
with the computational properties to manage the data in a very effective manner[13][14].