Using Probabilistic Models for Missing Data Prediction in Network Industries Performance Measurement Systems
Kristjan Kuhi, Kati Kõrbe Kaare, Ott Koppel · Procedia Engineering · 2015
The vast development of information and communication technologies has created new possibilities to acquire and analyze data to take performance measurement systems to next level. Most commonly performance measurement has been known as a financial management tool. Sophisticated new technologies have made it possible to collect continuous real-time data and enabled to start designing and implementing nonfinancial performance measurement systems. Most network industries are undertakings of dominant position and therefore subjects to strict supervision. For the authorities to fulfill their regulatory functions, precise monitoring and systemized feedback on the performance of network industries is essential. The problem lies in non-complete data in terms of missing, faulty or delayed values which might lead to incorrect management decisions. The objective of this paper is to explore the use of mathematical models for missing data prediction in performance measurement systems. Applying deterministic models hide the uncertainty of the value state therefore with higher likelihood false diagnoses occur. Authors propose probabilistic models because likelihood based methods for missing data calculation are able to take into account different parameters and time aspect in a single model to convey more trustworthy estimates in performance measurement systems than traditional methods.