This procedure aims to generate decision (classification) rules based on information about previously employed staff—people who have thoroughly learned the specifics of the work performed. Employees are one of the most important resources of any enterprise. On one hand, there is a belief that an employee well-matched to the specifics of the job position, i.e., possessing knowledge, experience, and appropriate predispositions, has a better chance of finding professional satisfaction. On the other hand, creating appropriate working conditions for an employee, and thus providing proper motivation, is—according to W.E. Deming’s principles—one of the essential conditions for continuous company improvement and increasing work quality. Finding employees on the market who meet the requirements and can handle the duties awaiting them is neither a trivial nor a cheap task. In many cases, employee selection is entrusted to external HR consulting firms that professionally handle recruitment. In such cases, the employer’s expectations are identified, and then job advertisements characterizing the position and the profile of the sought-after candidates are published in the media. After receiving applications, recruiters perform an initial selection (discarding “unsuitable” candidates), and selected candidates are interviewed for detailed, mainly psychological, verification and assessment. In practice, however, regardless of whether the recruitment is internal (carried out by the employer’s HR department) or external (carried out by a consulting firm), the selection itself is burdened with a very high risk resulting from the inability to predict both the abilities and the actual determination of employees to carry out the tasks entrusted to them. According to the authors of the article, this situation is mainly the result of insufficient use of the informational potential hidden in historical data, i.e., information about the work history of previously employed staff in positions similar or identical to the vacant ones. Considering the logistics market as a typical example of a service sector where the human factor plays a significant role in achieving the intended goal, the problem of selecting the right employees becomes exceptionally important.
The article presents an innovative employee selection procedure based on historical data mining using rough set theory. This procedure aims to generate decision (classification) rules based on information about previously employed staff—people who have thoroughly learned the specifics of the work performed. This classification can be carried out based on seniority or work performance. Knowing basic information about employees (for example, based on an analysis of HR data) makes it possible to find a set of interrelated employee characteristics that determine their assignment to an appropriate group (e.g., most useful, useful, useless). Among the benefits of this approach, one should mention:
• the possibility of constructing a properly profiled job advertisement for sought-after candidates
• the construction of an initial questionnaire for candidates
• the possibility of assessing and recommending candidates (through the application of decision rules) for the final stage of recruitment.
The proposed procedure is therefore a proposal to support the recruitment process, not to replace it entirely. The course of the employee recruitment procedure is presented using the example of a Poznań-based company operating in the postal and courier services industry, where the sought-after candidates are to be hired for the position of driver-courier. In this case, the classification of already employed staff was carried out based on seniority, as the problem of the analyzed enterprise turned out to be a very high employee turnover, periodically transforming into staff shortages dangerous for the execution of courier tasks.
General procedure scheme.
The essence of the proposed procedure is extracting knowledge about already employed staff (historical data mining) and generating decision rules that support the selection of candidates (prognostic part). Historical data about currently employed staff allow for the identification of those features that determine their usefulness and effectiveness at work. The ability to generalize such knowledge in the form of decision rules allows, in turn, to use this knowledge at the stage of selecting candidates for similar positions. The employee selection procedure consists of a series of actions divided into 3 stages (Figure 1). The first stage is related to obtaining information about currently working or former employees. Its direct result is a set of the most important characteristics for assessing an employee in a given job position. The second stage aims to generate decision rules, which constitute, on one hand, a generalization of knowledge about employees, and on the other, a “pattern” for the later classification of candidates. The result of this stage is a set of decision rules and an assessment of the reliability of these rules. If the set of rules is not reliable enough, for example due to the information possessed, it is necessary to return to stage 1 and iteratively repeat both stages until a satisfactory set of rules is obtained. The third stage aims to classify candidates based on the results of stages 1 and 2. The result of this stage is a questionnaire filled out by candidates, and consequently, the classification of these applications, a list of people recommended for further interviews. Methodologically, the foundations of the procedure are based on the rough set theory (RST), proposed by Z. Pawlak in the 1990s, and subsequently developed by S. Greco, B. Matarazzo, and R. Słowiński. A review and classification of the main trends of RST can be found in the work of P. Sawicki. Operationally, the employee selection procedure is supported by a computer application called 4eMka (Figure 1), prepared by a team of specialists from the Institute of Computing Science at the Poznań University of Technology.