Framework for Feature Selection in Health Assessment Systems
Anomaly detection in health assessment systems has gained much attention in the recent past. Various feature selection techniques have been proposed for successful anomaly detection. However, these methods do not cater for the need to select features in health assessment systems. Most of the present techniques are data dependent and do not offer an option for incorporating domain information. This paper proposes a novel domain knowledge-driven feature selection framework named domain-driven selective wrapping (DSW) that can help in the selection of a correlated feature subset. The proposed framework uses an expert’s domain knowledge for the selection of subsets. The framework uses a custom-designed logic-driven anomaly detection block (LDAB) as a wrapper. The experiment results show that the proposed framework is able to select feature subsets more efficiently than traditional sequential selection methods and is very successful in detecting anomalies.
KeywordsFeatures Health assessment Smart questionnaire
The work presented in this paper is supported by the SIEF STEM+ funding. Fan Dong is the recipient of a SIEF STEM+ Business Fellowship.
- 9.Langley, P.: Selection of relevant features in machine learning. In: Proceedings of the AAAI Fall Symposium on Relevance, pp. 140–144 (1994)Google Scholar
- 21.Zhao, Z., Liu, H.: Semi-supervised feature selection via spectral analysis. In: SDM (2007)Google Scholar