Using Genetic Algorithms to Optimize the Selection of Cost Drivers in Activity-based Costing

Abstract

In this paper, we address a cost-drivers optimization (CDO) problem in which two separate but interrelated decisions (i.e. the number of cost drivers needed and which cost drivers to use) are considered. It is desirable to have (1) an optimal selection of cost drivers in order to provide better indication of product costs and (2) an optimal number of cost drivers in order to avoid excessive control costs and to minimize information costs associated with data collection, storage and processing. The objective of the CDO problem is to balance savings in information costs with loss of accuracy. We propose an heuristic procedure based on genetic algorithms as an alternative with the potential to address more generalized objective functions. Genetic algorithms represent an innovative and promising heuristic approach which does produce results superior to published alternatives. The development and implementation of the algorithm is supported with the literature review and comparative analysis. We also comment on the complexity and experimental design issues for addressing large and practical problems.

Predicting earnings management through machine learning ensemble classifiers

“…Primarily, an initial population is generated followed by the steps depicted in Figure 3 until the final threshold (fitted individual) is achieved (Nair et al, 2010; Paul et al, 2014). Previous research shows the strength of GA as a wrapper FS approach and also as an optimization algorithm in different contexts such as credit scoring classification (Koutanaei et al, 2015), financial distress prediction problem (Fallahpour et al, 2017), credit card fraud detection (Duman & Sahin, 2011), choice of cost drivers in activity‐based costing (Levitan & Gupta, 1996), and financial statement fraud detection (Alden et al, 2012). Figure 3 depicts the different steps of the GA procedure.…”

A conceptual model to minimise operational cost for free shuttle buses: a case study in Macau

“…Then, Bahad and Balachandran (1993) provided an optimisation model that balanced savings in information processing costs with loss of accuracy and showed how to determine the number of drivers and identified the representative cost of drivers. Levitan and Gupta (1996) used genetic algorithms to optimise the selection of drivers in ABC and addressed a cost-drivers optimisation (CDO) problem in which two separate but interrelated decisions were considered. Additionally, Kim and Han (2003) applied a hybrid genetic algorithm and neural network approach in ABC.…”

Retour sur l'imputation des charges indirectes en comptabilité de gestion : comment bien spécifier les activités et leurs inducteurs ?

“…Chromosome 4 1 1 0 1 0 60 Chromosome 5 0 1 1 1 0 40 Chromosome 6 1 1 1 0 1 70 Ces étapes sont répétées jusqu'à ce que l'algorithme converge vers une solution quasi optimale. Levitan et Gupta (1996) ont proposé l'algorithme génétique comme alternative aux fonctions objectifs développées par les approches heuristiques. Utilisant des données issues de cas présents dans la littérature, ils concluent que l'utilisation d'un algorithme génétique non seulement réduit les coûts d'information grâce à la sélection d'un nombre inférieur d'inducteurs, mais produit également une fonction de coût mieux spécifiée, même en présence d'un nombre plus petit d'inducteurs.…”