Matrix computations (3rd ed.)
Improving Business Process Quality through Exception Understanding, Prediction, and Prevention
Proceedings of the 27th International Conference on Very Large Data Bases
Computers in Industry - Special issue: Process/workflow mining
BRAID: stream mining through group lag correlations
Proceedings of the 2005 ACM SIGMOD international conference on Management of data
Executive information systems: their impact on executive decision making
Journal of Management Information Systems - Special issue: Organizational impact of group support systems, expert systems, and executive information systems
| Hi-index | 0.00 |
Within Business Intelligence (BI) systems, a Key Performance Indicator (KPI) is a measurement of how well the organization, or a specific individual or process within that organization, performs an operational, tactical, or strategic activity that is critical for the current and future success of that organization [1]. The leading indicators are one type of KPIs that present key drivers of business value, are predictors of future outcomes, and offer the organization the unique opportunity to positively effect, or properly plan for, the future. Therefore, effective leading indicators are critical to the success of any business organization. However, identifying leading indicators is often non-trivial. It may require months to collect requirements, standardizing definitions and rules, prioritizing metrics, and soliciting feedback, etc. Moreover, because the time shifts between the leading indicators and the corresponding affected lagging indicators are vague and often inconstant for variability of business concerns, the traditional approach depending on domain experts' experiences is labor-intensive and error-prone. In this paper, we propose a semi-automatic system with an iterative learning process for analyzing operational metrics, factoring out the key performance indicators (KPIs) and then further discovering leading indicators. Two case studies are also conducted by applying the proposed methods in the production printing domain. The proposed system has two key differentiations and novelties: (1) the semi-automatic framework simplifies many traditional labor-intensive and error-prone steps by using temporal data mining techniques combined with specific domain knowledge, thus enabling timely access to operational metrics, KPI analysis, and powerful leading indicator discovery; (2) an iterative learning methodology not only continuingly uncovers the "root" leading indicators, but also enables the flexibility and adaptability for metric updates and additional data collection points.