We are developing a system for the identification of the areas at risk and for the surveillance of some important tick borne diseases (Lyme disease and TBE) recorded in Trentino region (Italian Alps) since 1990. The system is based on the integration of an on-line Geographical Information System (GIS) and Machine Learning tools. It includes a WEBGIS service, which allows data notification through Internet into a data base recording system dedicated to the on-line storage of a series of epidemiological and field data records. The system is used for a fast and accurate geo-location of epidemiological data and for the production of risk maps. On this data, machine Learning techniques as bagging, boosting of tree-based classifiers and a cost-sensitive variant were applied to produce predictive classification models of tick distribution. Predictor variables are obtained from GIS layers (e.g. vegetation type, altitude, exposure, slope, host density, geological substratum, etc.). The procedure has been applied to produce digital maps on the distribution of I. ricinus and to predict Lyme disease risk in Trentino. Work in progress includes the use of satellite data combined with intensive field data collection for the prediction of TBE risk
Mapping tick borne diseases risk in Trentino, Italian Alps
Merler, Stefano;Furlanello, Cesare;Rizzoli, Annamaria;
2002-01-01
Abstract
We are developing a system for the identification of the areas at risk and for the surveillance of some important tick borne diseases (Lyme disease and TBE) recorded in Trentino region (Italian Alps) since 1990. The system is based on the integration of an on-line Geographical Information System (GIS) and Machine Learning tools. It includes a WEBGIS service, which allows data notification through Internet into a data base recording system dedicated to the on-line storage of a series of epidemiological and field data records. The system is used for a fast and accurate geo-location of epidemiological data and for the production of risk maps. On this data, machine Learning techniques as bagging, boosting of tree-based classifiers and a cost-sensitive variant were applied to produce predictive classification models of tick distribution. Predictor variables are obtained from GIS layers (e.g. vegetation type, altitude, exposure, slope, host density, geological substratum, etc.). The procedure has been applied to produce digital maps on the distribution of I. ricinus and to predict Lyme disease risk in Trentino. Work in progress includes the use of satellite data combined with intensive field data collection for the prediction of TBE riskI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
