Page 60 - Big Data Analytics for Intelligent Healthcare Management

52 CHAPTER 3 BIG DATA ANALYTICS IN HEALTHCARE: A CRITICAL ANALYSIS

Generated
reports
External healthcare
data
Different big
Transformation data platforms
of raw data and analytics
Raw into (e.g., hadoop, Big data
healthcare recognizable mapreduce, pig analytics Queries
data and
big data
format piglatin, hive
etc.)
Internal
healthcare data OLAP for
business
intelligence
Data mining for
knowledge
discovery

FIG. 3.5
Level 1 architectural framework for big data analytics in healthcare data.

formats such as ASCII/text, flat files, .csv, relational tables etc. [35]. It is clear that medical big data
analytics is different from traditional big data analytics. The main reason for this is the varying
nature of raw medical big data and also a heterogeneous source of generation. Because of this,
an extra preprocessing step is needed for data cleansing, data normalization, and data transforma-
tion. If it is the case of medical image big data, then special enhancement techniques (most pref-
erably those with fuzzy based techniques) should be adopted for better enhancement and hence a
better medical diagnosis.

3.6.2 DIFFERENT TOOLS USED IN BIG DATA ANALYTICS IN HEALTHCARE DATA
A number of different platforms and tools are used in big data analytics in healthcare data. Among
them, the ones [35] which are frequently used are briefly introduced below:
(a) The Hadoop Distributed File System (HDFS):
This is the prime data storage system used by Hadoop applications. Here a NameNode and
DataNode architecture is used to implement a DFS (distributed file system) that provides
high-performance access to data across highly scalable Hadoop clusters. It has the ability to rapidly
transfer data between different computer nodes. The complexity lies in the usage and this its main
drawback. Also, it is not suited for small data.
(b) MapReduce:
MapReduce is the central part of the Apache Hadoop software framework. It acts in two vital
roles: mapper and reducer. In the first role, it filters and parcels out work to various nodes
within the cluster or map and, in the second role, it systematizes and reduces the results from
each node into a consistent answer to a query. When tasks are executed, it tracks the processing of