Fetal Weight Prediction Using Neural Ensemble Based C4.5 For Low Birth Weight Fetuses

Loo, Kien Lim (2010) Fetal Weight Prediction Using Neural Ensemble Based C4.5 For Low Birth Weight Fetuses. Project Report. UTeM, Melaka,Malaysia. (Submitted)

	PDF (24 Pages) Fetal_Weight_Prediction_Using_Neural_Ensemble_Based_C4.5_For_Low_Birth_Weight_Fetuses_Loo_Kien_Lim_R858.L66_2010_-_24_Pages.pdf - Submitted Version Download (4MB)
	PDF (Full Text) Fetal_Weight_Prediction_Using_Neural_Ensemble_Based_C4.5_For_Low_Birth_Weight_Fetuses_Loo_Kien_Lim_R858.L66_2010.pdf - Submitted Version Restricted to Registered users only Download (30MB)

Abstract

With the continually increasing of infants' mortality rates and birth defect rates, Low Birth Weight (LBW) is of concern for everyone. A woman's behavior during pregnancy such as smoking habits can greatly alter the chance of carrying infant in abnormal term. LBW fetus is baby with a weight of less than 2,500 grams at birth. Characteristics of baby with LBW include born prematurely and small body for their age. Because of the problems with the placenta and the mother's health particularly during pregnancy, the infant has less time to grow and gain weight in the mother's uterus. It can also caused by environmental factors and inheritance genetics. Consequences, LBW infants face a higher risk of death within the first year of life and have higher rates of disability and disease than other normal infants. Hence, the purpose of this project is to purpose a classification technique, Neural ensemble Based C4.5 (NeC4.5) on prediction of LBW fetuses through developing a classification application while compare the result analysis with Decision Tree C4.5 and RBF Neural Network. This thesis describes the analysis, design of NeC4.5 and result analysis. Basically, this application involves three modules which is data loading, data classification and result analysis. Object Oriented Analysis and Design has been chosen as a methodology for this project and will be implemented along the application development process. After carried lOxlO cross-validation of classification for NeC4.5, result analysis in term of mean Accuracy and mean Fmeasure shown that NeC4.5 has a best overall performance than the other two techniques due to it is a hybrid technique where its generalization ability can be better than Decision Tree C4.5 and its good comprehensibility. Hence, NeC4.5 is a good classifier towards solving LBW problem. However, the execution time performance still needs to improve in the future.

Actions (login required)

View Item