Page 276 - Big Data Analytics for Intelligent Healthcare Management

270 CHAPTER 11 KIDNEY-INSPIRED ALGORITHM AND FUZZY CLUSTERING

that will possibly present a solution for a specified optimization problem. The balance between inten-
sification and diversification has a substantial result on the competence of a metaheuristic. It can be an
incentive for initiating the KA (kidney-inspired algorithm).
The KA, which is a new population-based metaheuristic optimization algorithm inspired by the
kidney process in the human body, is an extremely contemporary optimization algorithm developed
by Jaddi et al. [24] in 2017. The KA performs additional efficiently when compared to the current evo-
lutionary algorithms. However, KA was meant for constraint search spaces. This algorithm was
completely based on the filtration, reabsorption, secretion, as well as excretion processes that take place
in the kidneys of a human body. According to the original KA, the solutions will be sieved in an amount
that is premeditated based on intent roles of all explorations in the existing population of each repe-
tition. The sifted solutions as the improved resolutions are stimulated to FB (filtered blood) and the
remaining are transported to waste, which is represented by the bad solutions. This is a recreation
of the glomerular filtration procedure in the kidney. The results that are not used will be reassessed
in the iteration if they assure the rate of filtration. If not, this can be excluded from the waste solution,
thus replicating the excretion and reabsorbing functions of the kidney. Subsequent to assignment of all
the clarifications in the solutions, the greatest of them is ranked and filtration products and waste blood
are combined to form a novel population and the rate of filtration can be modernized.
This KA was inspired by the kidney process in a human body and it is a population-based algorithm.
In the kidney, urine configuration consists of four steps: (1) filtration, (2) reabsorption, (3) secretion,
and (4) excretion. In filtration, it absorbs the transmission of both the solutes and water into the tubules
from the blood in the kidneys. The association of functional solutes and water against the tubules and
reverse into the blood is reabsorption. In the process of secretion, the tubules emit additional and harm-
ful matter into the tubules. Finally, in the excretion process, the waste matter produced during the pre-
vious three steps departs the body by means of urine. The above-mentioned four steps were taken into
consideration in the projected traditional KA.
KA begins with a solution of solutes and water elements (solutions or particles). At each iteration,
the percolated solutes rely on a percolation rate that is based on objective values of all the solutes. The
percolated solutes are processed to FB and the rest are moved to (W) waste. The above steps replicate
the glomerular filtration procedure in the kidneys. Absorption, excretion, and secretion are the remain-
ing three steps in the filtration process of the kidney. A solute allotted to W is again absorbed if it is to be
part of FB after pertaining the reabsorption operator, if not it is evacuated from the waste. Additionally,
a solution in FB is secreted if it is not upgraded to the bad solution in FB. W and FB are combined to be
the novel population and the filtration charge is reorganized. In such an algorithm, the invention of a
reabsorption operator and a novel explanation is premeditated based on the existing solution and the
finest solution set-up so far. Here, diversification is accomplished through the process of filtration and
intensification is presented by the novel solution reabsorption and generation process.
Solving data mining problems (especially clustering problems) has always been a tedious task for
all researchers, due to the unsupervised nature. There are so many efficient and robust techniques de-
veloped to handle the nonlinearity in the data irrespective of the nature and attributes of the data. At the
same time, there is always scope to develop the models with the latest developments in the form of
algorithms/programs that are able to perform/aid the model better than the earlier versions. Although
a number of techniques have been developed to solve clustering problems with both K-means (Clus-
tering Algorithm) and Fuzzy C-Means (FCM) Clustering Algorithm, issues such as initial trapping,
slow convergence, higher execution time, algorithmic complexity, etc. still need to be solved. Apart