American Journal of Computational and Applied Mathematics 2013, 3(3): 186-194
DOI: 10.5923/j.ajcam.20130303.06
Second Degree Chance Constraints with Lognormal
Random Variables – An Application to Fisher’s
Discriminant Function for Separation of Populations
Vas kar Sarkar
1,*
, Kripasindhu Chaudhuri
2
, Rathindra Nath Mukherjee
3
1
Department of Mathematics, Aryabhatta Institute of Engineering and Management Durgapur, Panagarh, Dist.- Burdwan, Pin - 713148,
West Bengal, India
2
Department of Mathematics, Jadavpur University, Jadavpur, Kolkata, 700032, West Bengal, India
3
Department of Mathematics, Burdwan University, Burdwan, West Bengal, India
Abstract In this paper, we have discussed a transformation procedure of the second-degree chance constraints to the
deterministic constraints for mathematical programming problems having general second degree chance constraints with
lognormal random variables. We have used geometric inequality for this transformation. The transformed deterministic
problem having non-linear constraints and linear or non-linear objective function can be solved using non-linear
programming algorithm. Also we have applied this model to Fisher’s discriminant function for separation of populations. A
numerical simulation have been considered along with a graphical representation of the reduced solution region.
Keywords Stochastic Programming, Chance Constraint Programming, Lognormal Random Variable, Geometric
Inequality, Deterministic Reduction, Non-linear Programming, Discriminant Function and Separation of Population
1. Introduction
Stochastic or probabilistic programming[13] deals with
the situations where some or all the parameters of the
mathematical programming problem are described by
stochastic or random variables rather than by deterministic
ones. Several models have been presented in the field of
stochastic programming[21]. Two major approaches to
stochastic programming[13,14] are recognized as:
1. Chance constrained programming
2. Two-stage programming
The chance constrained programming (CCP) can be used
to solve problems involving chance constraints, i.e.,
constraints having finite probability of being violated. Its
main feature is that the resulting decision ensures the
probability of satisfying the constraints, i.e. the confidence
level of being feasible. Thus, using CCP, the relationship
between profitability and reliability can be quantified. The
use of probabilistic constraints was initially introduced by
Charnes et al.[6,7], while an early use of probabilistic
programming in environmental economics is by Maler[17].
Since then, a number of environmental management case
studies use probabilistic constraints[5,10,11,19,22,23]. Also
the CCP, in recent years, has been generalized in several
* Corresponding author:
sarkar_vaskar@rediffmail.com (Vaskar Sarkar)
Published online at http://journal.sapub.org/ajcam
Copyright © 2013 Scientific & Academic Publishing. All Rights Reserved
directions and has various applications[21].
Many authors studied and developed this problem for the
parameters having normal distribution, uniform distribution
and exponential distribution, where the chance constraints
are linear[3, 4, 21]. But lognormal distribution (with two
parameters) has a significant role in human and ecological
risk assessment for many reasons, of which the main three
reasons are (i) many physical, chemical, biological,
toxicological and statistical processes tend to create random
variables that follow lognormal distributions[12], (ii)
lognormal distributions are self-replicating under
multiplication and division, i.e. products and quotients of
lognormal random variables themselves follow lognormal
distributions[1, 9] and (iii) when the conditions of the
Central Limit Theorem hold[16], the mathematical process
of multiplying a series of random variables will produce a
new random variable (the product), which tends (in limit) to
be lognormal in character, regardless of the distribution from
which the input variables arise[2].
Also we may consider the fact that many environmental
variables are non-negative means that they are generated by
a skewed distribution. Several skewed probability models
have been used to describe environmental data, including the
Poisson, negative binomial, Weibull, gamma, exponential
and the lognormal. Among these distributions, the lognormal
has been the most widely applied[18]. In the year 1996,
Cooper et.al.[8] noted the importance of using skewed
distributions, to represent environmental variables within
mathematical programming and lognormal distribution was