GENETIC ALGORITHM (GA) OPTIMIZATION USING DIABETES EXPERIMENTAL DATA

International Journal of Fuzzy Logic Systems(IJFLS) Vol.7, No.3, October 2017
DOI : 10.5121/ijfls.2017.7302 15
GENETIC ALGORITHM (GA) OPTIMIZATION USING
DIABETES EXPERIMENTAL DATA
Ejiofor C. I1
&Laud Charles Ochei2
Department of Computer Science, University of Port-Harcourt, Port-Harcourt. Nigeria 1
Robert Gordon University, Aberdeen, United Kingdom,2
Abstract
Diabetes comprises of noisy features. This feature hampers classification and prediction for Artificial
Intelligence (AI) system. The optimization of diabetes dataset using Genetic Algorithm (GA) exploring its
fundamentals identifies the focus of this research. The dataset obtained from Biostat comprising of random
samples (fifteen: 15) and parameter variables five: cholestrol, high-density lipoprotein, age, height and
weight was used for the optimization. The simulation was Matrix Laboratory (MATLB). The optimized
dataset was validated using standard optimization equation resulting in percentage score of Forty-one
(41%) percent. This dataset will be using in classifying fuzzy system
Keywords
Genetic Algorithm, Diabetes.
1.INTRODUCTION
Diabetes is a condition where the body fails to utilize the ingested glucose properly. Diabetes is
caused by deficiency in insulin production or failure of the body to response to insulin production
(Ananya, 2017). Excess glucose overtime within the blood stream can result in eye, kidney and
nerve damage. It can also inflame heart disease, stroke and even amputation.
Diabetes has been identified as the fastest growing long term disease that affects millions of
people worldwide(MedlinePlus, 2017). The statisticindeed has been alarming across the globe. In
2013 it was estimated that over 382 million people worldwide were suffering from diabetes. In
the United Kingdom, 2million persons have been identified as suffering from diabetes with 750,
000 unaware of their current illness. In the United States 25.8 million people or 8.3% of its
population have been identified as diabetes sufferer with70 million remaining undiagnosed. In
2010, about 1.9 million new cases of diabetes were also identified in United State with a future
prediction of 1 in every 3 Americans having the chance of experiencing diabetes by 2050
(Ananya, 2013).
Diabetes can be categorized into: Type I and Type II. Type I diabetes is also known as insulin-
dependent diabetes mellitus and is usually associated with younger adolescence.This form of
diabetes is dueto the inability of the body to produce insulin as a result of autoimmune disorder
destroying the pancreases and eliminating the chance of insulin production. Approximately 10%
of diabetes diagnoses are associated with Type I.Type II diabetes; also known as adult-onset or
noninsulin-dependent diabetes. This form of diabetes exists due to failure of the pancreas to
produce enough insulin to metabolize glucose which is usually associated with
aging.Approximately 90% of all cases of diabetes worldwide are associated with Type II
(Medlinplus, 2017).

16
Diabetes symptoms vary from frequent urination, intense thirst and hunger, weight gain, unusual
weight loss, fatigue, male sexual dysfunction, numbness and tingling in hands and feet
(MedlinPlus, 2017 and Ananya, 2017).
Treatment and diagnosis of diabetes usually are complex with physician depending on patient’s
symptoms in collaboration with Age, weight, height, family history and the contributing factor of
alcoholism and lack of physical exercise in identifying type I or II diabetes. While the diagnoses
have been consistent over time, the complication experienced by novel physician and the non-
availability of experienced expert has fostered numerous Artificial Intelligence (AI) models.
Although these model has been employed for the prediction of diabetes built to complement the
conventional approaches of physician-patient interaction (Mehdi et al., 2012; Meysam and
Mahdi, 2016)these models, possibly have suffer for inaccuracies due to noisy training sample
which has hampered training cases
This research paper explores genetic algorithm optimization technique for preprocessing diabetes
datasetidentifying change variation within the dataset. The variation in change will be explored
using standard optimization equation.
2.GENETIC ALGORITHM (GA) OVERVIEW
Genetic Algorithm (GA) considered a search and optimization technique based on the adaptation
of natural selection process and Evolutionary Algorithms (EA) has found its mark in Artificial
intelligence problem domains. GA provides a framework for solving both constrained and
unconstrained optimization problems based on a natural selection process which mimics
biological evolution (Akbari, 2010). GA is used in arriving at optimal solutions; solutions
directing the optimization to the best possible area (Eiben, 1994) through the modification of
population of individual solutions.
GA usually creates an optimization process using an initial population. This population
encompasses solution seen as candidates with each candidate’s solution possess initial possible
solutions. In each generation, the fitness of each individual are usually expressed and examined
using the objective function created in most cases based on adaptation, user examine,
experimental design and trial error (Son et al., 2016). The fittest individual are stochastically
selected from the current population with each genome modified based on genetic operator. This
modification creates new generation which are repeatedly evaluated. Commonly, the algorithm
terminates when either a maximum number of generations has been produced, or a satisfactory
fitness level has been reached for the population (Harik et al., 2006).
GA population size depends largely on the problem domain. This population usually combined
numerous possible solutions which are generated randomly forming the search or state space.
These solutions are usually attuned toward better solutions (Taherdangkoo et al., 2012).
Successive generation solutions are combined with preceding generation to breed a new
generation. Individual solutions are selected through a fitness-based process,
where fitter solutions are typically more likely to be selected. Novel solutions are produced using
parent solution in breeding pool of successive generation by producing children solution (Coffin
and Robert, 2008). The optimization process usually terminate based on minimum criteria
satisfaction, allocated budget and highest fitness value (Echegoyen et al., 2012).
Solution encoding, objective function identification, identified solution, application of operator
and termination are indeed the fundamental components of GA (Coffin and Robert, 2008; Akbari,
2010 Echegoyen et al., 2012).

17
3.METHODOLOGY:DIABETES DATASET OPTIMIZATION USING MATRIX
LABORATORY (MATLAB)
The dataset for simulation was obtained from Biostat. This data served as the experimental data
for optimization. The dataset comprises of fifteen samples spread across five decisionvariables:
cholestrol, high-density lipoprotein, age, height and weight. Table 3.1 depicts the diabetes
simulation data.
Table 3.1: Un-Optimized Diabetes Dataset

18
3.1MATLAB GA SIMULATIONS
Figure 3.1: GA-Simulation Chart
The chart of Figure3.1 captures the fundamentals of diabetes GA simulation. It shows clearly that
five input variable were utilized in collaboration with fifteen samples. The population type for
simulation was explored as double while the function was uniformly created. The chart also
depicts the weighted score for each column parameter input. The simulations chart also capture
best fitness value, best individual value, genealogy, score and selection.

19
Figure 3.2: Generation 1 (initial) GA-Simulation chart
Figure 3.2 provides the first generation simulation identifying fitness value; with the best fitness
value at generation one shown as 343and the over means of 15 samples identified as 506. The
individual generation is identified as initial. The selection function picture the probability open to
each sample in selecting prospecting children for mating within the next generation, with the fifth
individual having the highest section probability of 6. The current best individual provides the
best individual score per generation while the fitness score for each individual can be identified
successively for each generation. For this generation an average fitness score less than 600 was
initially maintained.

20
Figure 3.3: Generation 15 (final) GA-Simulation chart
Figure 3.3 provides the final generation simulation identifying fitness value; with the best fitness
value at generation fifteen shown as 307 and the over means of 15 samples identified as 301. The
individual generation is identified as fifteen. The selection function picture the probability open to
each sample in selecting prospecting children for mating within the next generation, with the fifth
individual having the highest section probability of 7. The current best individual provides the
best individual score per generation while the fitness score for each individual can be identified
successively for each generation. For this generation an average fitness score above 300 was
maintained.

21
Table 3.2: Optimized Diabetes Dataset
Table 3.2 provides the change obtained succeeding optimization. The values cut across
cholesterol, high density lipoporotein, Age, height and weight and showed the variation in data
changed compared to previous data value appearing on table 3.1.
4.VALIDATION OF OPTIMIZED DIABETES DATASET
Validation provides a definite proof in ascertaining the variation in change and determining
percentage optimization. It measures how much the fundamental components have been
optimized. Equation 4.1 determines these changes.
Where
R0= summation of fitness values of optimized dataset
M0=summation of fitness values of the non-optimized dataset
Table 4.2, captures the non–optimized and the optimized dataset the dataset are exemplified from
case 1-10 and 110.
Case Cholesterol
High-
Density
Lipoprotein AGE Height Weight
Weighted
Score Status
Confusion
Matrix
Case 1 78 12 67 67 119 343 Type II TP
Case 2 78 12 67 67 119 343 Type II FP
Case 3 78 36 27 67 119 327 Type II TP
Case 4 78 12 40 59 121 310 Type II TP
Case 5 78 12 36 67 119 312 Type II TP
Case 6 79 13 27 62 119 300 Type II TN
Case 7 78 12 27 66 119 302 Type II TN
Case 8 78 12 67 66 119 342 Type II TN
Case 9 78 12 45 62 119 316 Type II TP
Case 10 78 12 37 69 119 315 Type II TP
Case 11 78 12 27 67 119 303 Type II TP
Case 12 78 12 29 64 119 302 Type II TN
Case 13 78 12 34 67 119 310 Type II TN
Case 14 78 12 36 59 119 303 Type II TP
Case 15 78 12 29 63 119 300 Type II TN

22
Table 4.3: Non- Optimized and Optimized fitness Values
Degree of Optimization (DoO) = [4728 - 8017]/8017
= 0.4102
= 0.4102 * 100
= 41.1%
The degree of optimization shows clearly that 41% variation change has occurred within the
dataset that has been optimized. The graph of figure 4.1 graphical depicts this percentage
change.This variation in change has improved the given dataset and subsequently eliminated
noisy features.

23
Figure 4.1: Degree of Change for Diabetes data
The graph of figure 4.1 provides a percentage change of 40%. This change is perceived as the
stall change value. The point at which change optimization on the same dataset could be
stochastic possible as it run into infinite
5. CONCLUSION
This research has established the usefulness of genetic Algorithm as an optimization tool in
ascertaining optimal samples for used for prediction. The data obtained from biostat have been
optimized using an appropriate objective function and associated fitness values. Matric laboratory
interface provided simulation interfaces captured variation change within the dataset. The
validation using standard optimization equation captured an optimization change of 41%. This
dataset will be utilized probably in provided classification for fuzzy system.
REFERENCES
[1] Akbari Z. (2010). "A multilevel evolutionary algorithm for optimizing numerical functions" IJIEC 2
(2011): 419–430
[2] Ananya (2017), What is Diabetes, retrieved online from https://www.news-medical.net/health/What-
is-Diabetes.aspx
[3] Coffin, D.; S., Robert E. (2008). "Linkage Learning in Estimation of Distribution Algorithms".
Linkage in Evolutionary Computation. Springer Berlin Heidelberg: 141–156. doi:10.1007/978-3-540-
85068-7_7.
[4] Eiben, A. E. et al (1994). Genetic algorithms with multi-parent recombination, PPSN III: Proceedings
of the International Conference on Evolutionary Computation. The Third Conference on Parallel
Problem Solving from Nature: 78–87. ISBN 3-540-58484-6.

24
[5] Echegoyen, C.; Mendiburu, A. Santana, R.; Lozano, J. A. (2012). "On the Taxonomy of Optimization
Problems under Estimation of Distribution Algorithms". Evolutionary Computation. 21 (3): 471–495.
ISSN 1063-6560. doi:10.1162/EVCO_a_00095.
[6] Harik G. R.; Lobo, F. G.; Sastry, K. (2006), Linkage Learning via Probabilistic Modeling in the
Extended Compact Genetic Algorithm (ECGA),Scalable Optimization via Probabilistic Modeling
[7] Springer Berlin Heidelberg: 39–61. doi:10.1007/978-3-540-34954-9_3.
[8] MedlinePlus (2017), Diabetes, retrieved online from http:// www.medlineplus.com
[9] Mehdi K., Saeede E. and Jamshid P. (2012), Diagnosing Diabetes Type II Using a Soft Intelligent
Binary Classification Model, Review of Bioinformatics and Biometrics (RBB) Volume 1 Issue 1,
December 2012 9-23.
[10] Meysam J., and Mahdi M. (2016), Comparison of Predictive Models for the Early Diagnosis of
Diabetes, Kournal of Health Information Research, Vol 22(2), Pp.95-100
[11] Son D. D., Kazem A. and Romeo M. (2016), Maximsing Performance of Genetic Algorithm Solver in
Matlab, Advance online publication:, Pp. 1-9
[12] Taherdangkoo, M.; Paziresh, M.; Yazdi, M.; Bagheri, M. H. (2012). An efficient algorithm for
function optimization: modified stem cells algorithm, Central European Journal of Engineering. 3 (1):
36–50. doi:10.2478/s13531-012-0047-8.

GENETIC ALGORITHM (GA) OPTIMIZATION USING DIABETES EXPERIMENTAL DATA

More Related Content

What's hot (14)

Similar to GENETIC ALGORITHM (GA) OPTIMIZATION USING DIABETES EXPERIMENTAL DATA (20)

More from Wireilla (20)

Recently uploaded (20)

GENETIC ALGORITHM (GA) OPTIMIZATION USING DIABETES EXPERIMENTAL DATA