Intelligent Data Analysis and Predictive Models for Regional Labor Markets

Abstract

The digitizing economies call for new methods in studying social-economic phenomena that are often short-lived and for which no pre-identified set of indicators had been developed. In our paper we demonstrate how intelligent analysis of online data can supplement the use of more traditional methods, such as the ones relying on official statistical reports or sample surveys. We outline benefits and disadvantages for each group of methods, and also identify some challenges in joining the data obtained from the diverse sources – particularly, the classification of the data per industry sectors. The data that we used for building ARIMA models were obtained with our dedicated labor market monitoring software system, operating from 2011 and currently containing 10+ million unique data records for vacancies and resumes. We found that for average wages the official statistics data can be approximated (error 7.82%) and possibly refined by the wage levels that companies offer in the openly posted vacancies ads. Further, we constructed predictive models for the employees’ demand by the companies and found positive and negative influences (Lag −2, Lag −3 and Lag −4) for several industry sectors for which online data had been collected. The data from the identified groups can be used as leading indicators to predict situation on the labor markets.