Employees reviews classification and evaluation (ERCE) model using supervised machine learning approaches

Abstract

The employees, as stakeholders of the organization, can contribute to the development and productiveness of the organization. In regards to satisfaction/dissatisfaction, the opinion of employees can perform dual rule. Firstly, it supports the organization to plan future strategies and enhance their yield; secondly, it can be helpful for aspirants in seeking their best choice. In this concern, we have classified the reviews of employees using two different modules. In the first module, we have experimented on ratings of reviews; then in the second module, the textual part of the reviews is used to classify employees as satisfied/unsatisfied. After that, the reasonable outcomes of both approaches are unified for the final prediction of reported reviews as proper/improper. For this purpose, we have implemented a purely supervised machine learning approach. The performance of state of the art classifiers along with TF-IDF (Term frequency-Inverse document frequency) and BoW (Bag-of word) is analyzed in the text module. In this comparison, ETC (Extra tree classifier) performed best in terms of accuracy in both modules. It shows 100% accuracy with rating and 79% accuracy with the textual part. Ultimately, we have implemented AND gate for the evaluation of proper/improper reviews. The results of AND gate evaluate that 76% of the reviews of employees are reported as properly and 24% are reported as improperly in the used dataset.