Fuzzy Logic Approach to Predict the Outcome of Tuberculosis Treatment Course Destination

摘要

Tuberculosis (TB) treatment with patient supervision and support as an element of global plan to stop TB designed by World Health Organization in 2006 requires prediction of patient treatment course destination to determine how intensive should be the level of supplying services and supports in DOTS (directly-observed treatment, short course). This study was aimed to develop a model using fuzzy logic technique to forecast TB cases treatment course destination. The five given outcomes included getting cured, completion treatment courses, quite the treatment course, fail in treatment, and dead. 16 parameters verified by former studies were applied as predictors. The data set with 9672 Iranian patients were divided as training to build a model and testing datasets to check the predictive ability of fuzzy model. Using principal component analysis 5 clusters of variables were extracted. 14 inputs (x) categorized in 5 identified components and based on the expert's knowledge 50 fuzzy sets were developed. For each fuzzy set (A) triangular membership function was determined μ_A(x): X → [0,1] presenting the degree to which x is an element of set A. Predictive model was developed by learning from given historical datasets, based on an integration of simplified fuzzy technique and recursive least square learning algorithm. After applying testing set to developed model by training set, the gained mean absolute percentage error (MAPE) was 1.258. To sum up, fuzzy logic is an acceptable technique with easy to understand output to predict outcomes of tuberculosis treatment course destination.