Data mining approach for detecting key performance indicators

Abstract

Background and Objective: Key performances indicators (KPIs) are an integral part of business intelligence systems as the choice of KPIs are critical to success. This study aimed to propose a solution to detect KPIs from historical organizational data using data mining algorithms and analyzes the relation between factors that will affect the performance to help organizations execute their business strategy. This approach does not involve domain experts to identify or validate KPIs. Materials and Methods: Information gain algorithm implemented with Weka (InfoGainAttributeEval) used for feature selection to rank the attributes that affect the performance. Moreover, an improved FP-growth algorithm was used to find the correlation between attributes. Results: The KPIs detection model was tested using 6 years of banking data. To detect the non-performing loans KPI the model indicates strong correlation between non-performing loans and attributes such as accounts with low salaries, young clients and accounts with a monthly issuance of statements. Conclusion: The proposed KPI detection approach can make the process of selecting KPIs more efficient; it can be used as a method to determine the most appropriate KPIs. This model will enable decision makers to make timely and appropriate strategic decisions.