Publication: Pengkelasan sel kanser pangkal rahim kepada sel normal dan tidak normal menggunakan analisis pembezalayan dan rangkaian neural
Loading...
Date
2006-05-01
Authors
Saidin, Mohammad Norrish
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The topic of this project is classification of cervical cells into normal and abnormal using 2 group discriminant analysis and neural network. The type of the
neural network is multilayed perceptron (MLP) network using software MATLAB® 6.5 and discriminant analysis using software SPSS® 13.0. The system is built to classify some certain data into two classes, which are normal or abnormal cells. Data are using for this project is nucleus size, cytoplasm size, nucleus grey level and cytoplasm grey level. The data are separated into two sets; training data set and testing data set. There are 128 data in training data set and 72 data in testing data set. The neural network is trained using two types of learning algorithms, which is Levenberg-Marquardt and Back Propagation. The optimum value of epoch and hidden nodes for each learning algorithm are determined based on the highest accuracy obtained during training phases. For discriminant analysis, training data are used to simulate to obtain accuracy and cut-off point. From the result, the neural network and disriminant analysis show the 100% accuracy. As a conclusion, the neural network and discriminant analysis has high capability to classify the cervical cells into normal and abnormal.