doi:10.4102/http://www.sajhrm.co.za sajhrm.v11i1.449
Page 8 of 10
Also, the number of variables that have a strong correlation with the target variable (promotion score) were decreased to five variables as follows:
• education level or degree (repeated in 11 rules)
• examination score (repeated in 8 rules)
• interview score (repeated in 8 rules)
• province of employment or job location (repeated in 6 rules)
• years of experience (repeated in two rules).
Final research model
The summary of the foregone steps is depicted in Figure 2. It is worthwhile to note that in the final of the eleven-step approach presented earlier, stresses on the repeatability of the data mining process and the benefit of the discovered knowledge are in redoing of the steps. This way, improvements can be made in data mining results and rules creation in each iteration of the exam. Furthermore, data mining will be performed in a more guided manner.
FIGURE 2: Knowledge discovery model from database of entrance examinations with data mining.
Define the problem and targets
Database recognition
Define the target variables and independent variables
Delete incomplete records
Repair missing data
Delete extra variables
Change variables
Combine variables
Conversion variables
Classification the independent variables
Possible models design
Run models with Training data
Select trees with upper accuracy
Test models with examination data
Final selection trees
Do you accept
chosen trees?
No
Yes
Rules extracting from chosen trees
Interpretation rules by HRM experts
Usage knowledge perquisite for personnel selection in future
Rules
evaluation
Knowledge discovery
Continuous improvement and data mining repetition for future