Improving Data Quality for Better Control of Aedes-Borne Disease Risk

Abstract

Visual larval survey of container habitats is conducted as a routine mission of the Department of Disease Control (DDC), Thailand. To facilitate this, DDC has deployed a mobile application, namely TanRabad SURVEY, throughout the country since 2016. Here, each inspected place with its intrinsic place type and building names must be initially input via natural language to proceed the larval survey data collection. Upon a survey completion, the appropriate larval indices (e.g. House Index (HI), Container Index (CI) and Breteau Index (BI)) are automatically calculated. HI and BI are for villages, while CI is for other inspected places. These larval indices are then applied as factors for the vector control management. However, about 21% of inspected places stored in TanRabad database are found with inappropriate place types. These poor place types result in the procurement of inapplicable larval indices and hence ineffective vector control management. Ideally, the quality of place types can potentially be improved once their poorness is notified to users. This paper has thus proposed a novel and comprehensive place type quality assessment technique, namely pAssessor, with respect to buildings textually and variously defined for places. Specifically, pAssessor is driven by the building-place ontology, building semantic selection, boosted features, learned building-place relations and probability values of all place types. The experimental results showed that the efficiency of pAssessor in assessing the quality of place types is greater than 87.5%.