DATA ANALYSIS OF JOBERTY.RS DATA USING MACHINE LEARNING TECHNIQUES

Abstract

Employee satisfaction is essential for any organization to progress successfully and achieve its goals. Companies need to have insight into the attitudes and opinions of both their team and those from other companies. On the other hand, such information helps job seekers to discover the strengths and weaknesses of the companies they are interested in. In this paper, a data set of Serbian IT companies was created using reviews of employees, former employees, candidates, and other information from the Jobery.rs website. The analyzes performed on this set are divided into three tasks. In the first task, the most significant aspects of the company were determined while observing this task as a regression problem with the average rating of the company as the target label. The results indicate that the employer's flexibility and attitude towards the employees are more significant than used technologies and salary. Sentiment analysis of comments regarding the selection process was performed using the BERT model. It was concluded that HR and the technical interview have an equal influence on forming the impression about the selection process. Using the RandomForest model, a model was trained that performs prediction of the percentage of employees who think that the salary is fair, based on the evaluation of the benefits offered by the company. А score of 0.1 MSE was achieved. An analysis of the importance of the input features was performed on the trained model.