Page 14 - Big Data Analytics for Intelligent Healthcare Management

4 CHAPTER 1 BIO-INSPIRED ALGORITHMS FOR BIG DATA ANALYTICS

(to store data), (2) extraction and cleaning (cleansing of data), (3) integration and aggregation
(compiling of required data), (4) modeling and analysis (study of data), and (5) data interpretation
(represent data in required form).

1.3 BIO-INSPIRED ALGORITHMS FOR BIG DATA ANALYTICS: A TAXONOMY
This section presents the existing literature of bio-inspired algorithms for big data analytics. The bio-
inspired algorithms for big data analytics are categorized into three categories: ecological, swarm-
based, and evolutionary. Fig. 1.5 shows the taxonomy of bio-inspired algorithms for big data analytics
along with focus of study (FoS).

1.3.1 EVOLUTIONARY ALGORITHMS
Kune et al. [10] proposed a genetic algorithm (GA) based data-aware family scheduling approach for
analytics of big data, which focuses on bandwidth utilization, computational resources, and data de-
pendencies. Moreover, the GA algorithm decoupled data and computational services are provided
as cloud services. The results demonstrate that the GA algorithm gives effective results in terms of
turnaround time because the GA algorithm processes data using parallel processing. Gandomi et al.
[11] proposed a multiobjective genetic programming (GP) algorithm-based approach for big data min-
ing, which is used to develop the concrete creep model to provide unbiased and accurate predictions.
The GP model works with high and normal strength. Elsayed and Sarker [12] proposed a differential
evolution (DE) algorithm-based big data analytics approach, which uses local search to increase the
exploitation capability of the DE algorithm. This approach optimizes the big data 2015 benchmark
problems with both multi- and single-objective problems but it exhibits large computational time. Ka-
shan et al. [13] proposed an evolutionary strategy (ES) algorithm-based big data analytics technique,
which processes data efficiently and accurately using parallel scheduling of cloud resources. Further,
the ES algorithm minimizes the execution time by partitioning a group of jobs into disjointed sets, in
which the same resources execute all the jobs in the same set.
Mafarja and Mirjalili [14] proposed a simulated annealing (SA) algorithm-based big data optimi-
zation technique, which uses the whale optimization algorithm (WOA) to architect various feature se-
lection approaches to reduce the manipulation by probing the most capable regions. The proposed
approach helps to improve the classification accuracy and selects the most useful features for catego-
rization tasks. Further, Barbu et al. [15] proposed an SA algorithm-based feature selection (SAFS) tech-
nique for big data learning and computer vision. Based on a criterion, the SAFS algorithm removes
variables and tightens a sparsity constraint, which reduces the problem size gradually during the iter-
ations and this makes it mainly fit for big data learning. Tayal and Singh [16] proposed big data an-
alytics based on the FSO and SA-based hybrid (FSOSAH) technique for a stochastic dynamic facility
layout-based multiobjective problem to manage data effectively. Saida et al. [17] proposed the cuckoo
search optimization (CO) algorithm-based big data analytics approach for clustering data. Further, dif-
ferent datasets from the UCI Machine Learning Repository are considered to validate the CO algorithm
through experimental results and these datasets perform better in terms of computational efficiency and
convergence stability.