A fuzzy regression model for the prediction of oral cancer susceptibility

摘要: Precise and accurate predictive models are very important in cancer screening initiatives. The need for new approaches philosophies modeling prediction disease susceptibility studies influenced by the recent advances soft computing as well questionable accuracy inapplicability to individual of commonly used statistical analysis techniques. Soft especially fuzzy concept is highly suitable dealing with vague, ambiguous complex information. purpose this research develop a computer-prototype using linear regression dichotomous response variable general particular. objective thesis present development,testing validation an adaptive algorithm that can be binary variable. In machine learning developed estimate unknown dependency between set given input variables its corresponding which nature. Thus aim study predict outcome at group level specific. proposed model was experimented on oral data validated hypertension set. model’s performance measured based calibration discrimination abilities. performance, interpretation abilities selection ability were then against four including clinicians’ predictions, logic, neural networks logistic predictions. regression, network found have better compared clinicians logic model. IV compatible performances. Similarly, three capable finding ‘optimal’ predictor utilization different Both produced transparent equation associations predictors predicted clearly projected, feature lacking However, handling relationship independent dependent failed address. conclusion, good results obtained application suggest reasonable, desirable effective producing valid intelligent exploratory predicting