Management of filariasis using prediction rules derived from data mining

Area of Science:

• * Public Health Entomology
• * Vector-Borne Disease Control
• * Data Mining and Predictive Analytics

Background:

• * Bancroftian filariasis is a significant public health concern in India, primarily transmitted by the mosquito vector Culex quinquefasciatus.
• * Effective control strategies rely on accurate forecasting of mosquito vector populations.
• * Existing methods for prediction may not fully leverage complex entomological and environmental data.

Purpose of the Study:

• * To demonstrate the application of Classification and Regression Trees (CART) for controlling Culex quinquefasciatus, the vector for bancroftian filariasis.
• * To derive prediction rules for mosquito abundance using entomological, meteorological, and socio-economic data.
• * To assess the utility of CART in forecasting vector densities.

Main Methods:

• * Utilized a comprehensive database on filariasis and CART software (Salford Systems Inc., USA).
• * Categorized baseline entomological data into mosquito abundance, meteorological factors, and socio-economic details.
• * Employed predictor variables including temperature, rainfall, humidity, wind speed, and house type to predict the target variable (month).

Main Results:

• * CART successfully ranked predictor variables based on their influence on mosquito abundance and seasonal distribution.
• * The model identified key factors influencing vector populations, enabling the derivation of prediction rules.
• * The approach demonstrated its capability to forecast vector densities.

Conclusions:

• * CART is a valuable tool for forecasting mosquito vector densities, crucial for the control of bancroftian filariasis.
• * The predictive model aids in proactive disease management by anticipating vector population fluctuations.
• * This data-driven approach supports evidence-based interventions for filariasis control programs in India.