| Summary, etc |
Typhoid fever, also called enteric fever, is caused by Salmonella enterica serovarTyphi, a gram-negative<br/>bacterium. Estimates for 2000 suggest that around 21.5 million infections and 200 thousand fatalities due<br/>to typhoid fever are reported worldwide every year. Typhoid fever and paratyphoid fever, especially<br/>Among Nigerian kids and adolescents, keep being significant causes of disease and death. The aim of this<br/>study is to use Techniques for Data Mining to develop a Typhoid risk predictive model in Nigerians using<br/>relevant risk factors.<br/>Historical data on the distribution of typhoid risk among participants were gathered using questionnaires<br/>after medical professionals identified linked typhoid risk variables. The predictive model for typhoid risk<br/>was developed using the algorithm for decision trees to define and account for variables linked to typhoid<br/>risk The Waikato Environment for Knowledge Analysis (WEKA) software – a suite of machine learning<br/>algorithms was used to develop the predictive model as the simulation environment. Holdout and 10-fold<br/>cross-validation techniques were used to evaluate the performance of the algorithms. The data sets were<br/>therefore subject to 10-fold cross validation using the two (2) chosen decision trees learning algorithms,<br/>namely: C4.5 implemented as the WEKA J48 algorithm and the Naïve Bayes algorithm.<br/>The result of the performance evaluation of the C4.5 and naïve Bayes’ algorithms are presented in Table 4.3.<br/>The true positive rate which gave a description of the proportion of actual cases that was correctly predicted<br/>which showed values of 0.783, 0.519, 0.722 and 0.619 respectively for no, low, moderate and high risk cases<br/>by the C4.5 decision trees algorithm and 0.739, 0.556, 0.611 and 0.667 for the naïve Bayes classifier.<br/>The study presented a predictive model of typhoid risk using relevant risk factors selected from a predefined<br/>set of typhoid risk factors in Nigerians using the C4.5 decision trees algorithm that outperformed the<br/>performance of the classification of the naïve Bayes. A stronger understanding of the connection between<br/>the characteristics appropriate to typhoid risk was suggested following the creation of the forecast model<br/>for typhoid risk classification. The model can also be incorporated into the current Health Information<br/>System (HIS) that captures and manages clinical data that can be supplied to the predictive model of<br/>typhoid risk classification, thus enhancing clinical choices influencing typhoid risk and evaluating clinical<br/>data that affects typhoid risk from distant places in real time.<br/> |
