A Comparative Study of Performances of Various Classification Algorithms for Predicting Salary Classes of Employees

In knowledge based industry, compensation planning is a key strategic area for growth and success. In order to retain high performance employees, optimum salary offer is essential. Determining such salary figures, based on various information about a current employee or a prospective employee, is a challenge that corporations face very frequently. Although HR managers typically tackle such salary prediction and negotiation issues in consultation with relevant department-level managers, any automated system with such capability would be of great help for them. Given the attributes of an employee (current or prospective), which includes her demographic profile along with other information such as qualification, performance level etc. , several wellknown classification algorithms can be used for the prediction of the salary class. But unfortunately, such details of employee data of any corporation are generally not available in public for performance evaluation of classification algorithms. In this paper, this limitation is overcome to some extent by using a public database (UCI census dataset) which have most of the attributes available for a segment of population for salary prediction. Although the data used in this experimental paper is not directly related to salary prediction of employees within an organization, but it can be extrapolated to be used in the former scenario as the tuples in UCI census dataset include employer type as an attribute. This analysis found that among five classification algorithms, decision tree and Bayesian belief network performs better than other three algorithms, namely, naïve Bayes, support vector machine and neural network. The software used for running these algorithms is WEKA which is a well-known university tool for machine learning. Keywords— Human Capital, Human Resource, Compensation Management, Salary Prediction, Knowledge Based Industry, Classification Algorithms, Machine Learning, Naïve-Bayes, Bayesian Belief Network, Support Vector Machine, SVM, Decision Tree, Neural Network, Backpropagation