The purpose of this project is to develop a mobile determination system of potential drug in natural product based on Artificial Neural Network (ANN). The project is divided into 2 parts. The first part is to find the most suitable neural network to be implemented in the potential drug determination system. This project compares the performance among the conventional Multilayered Perceptron (MLP) network trained with Back Propagation (BP), Levenberg Marquardt (LM) and Bayesian Regularization (BR) learning algorithms and the Hybrid Multilayered Perceptron (HMLP) network trained with Modified Recursive Prediction Error (MRPE) learning algorithm. The performance analysis is carried out based on the determination accuracy. The HMLP network with the MRPE algorithms obtained the best performance with accuracy of 82.03% as compared to that of the MLP network trained with BP, LM and BR with accuracy 78.56%, 80.47% and 81.25% respectively. Thus, the HMLP network is implemented in the potential drug determination system. The second part integrates the determination system in web based. The user need to login to the web site and register as a new user. After verified the email and passwords, the user can use the potential drug determination system and view the result from result page. Overall, this project has successfully developed a mobile system which can determine the potential drug of the natural product based on ANN.
_____________________________________________________________________________________
Tujuan projek ini ialah membina sistem penentuan mobil bagi dadah berpotensi dalam produk semulajadi dengan rangkaian neural buatan. Projek ini boleh dibahagikan kepada dua bahagian. Bahagian pertama ialah mencari rangkaian neural buatan yang paling sesuai untuk diimplementasikan dalam sistem penentuan dadah berpotensi. Projek ini membandingkan prestasi rangkaian Perseptron Berbilang Lapisan (MLP) dengan pembelajaran Back Propagation (BP), Levenberg Marquardt (LM) dan Bayesian Regularization (BR) dan rangkaian Perseptron Berbilang Lapisan Hibrid (HMLP) dengan pembelajaran Modified Recursive Prediction Error (MRPE). Analisis prestasi adalah berdasarkan kejituan penentuan. Rangkaian HMLP dengan pembelajaran MRPE menghasilkan prestasi terbaik dengan kejituan 82.03% berbandingkan kejituan bagi rangkaian MLP yang dilatih dengan pembelajaran BP, LM dan BR iaitu masing-masing 78.56%, 80.47% dan 81.25%. Maka, rangkaian HMLP dipilih untuk diimplementasikan dalam sistem yang dibina. Bahagian kedua mengintegrasikan sistem penentuan dalam web based. Pengguna dikehendaki mendaftar masuk web itu dan mendaftar sebagai pengguna baru untul menggunakan sistem penentuan dadah berpotensi dalam produk semulajadi yang dibina. Kesimpulannya, projek ini telah berjaya membina satu sistem mobil yang boleh menentukan dadah berpotensi bagi produk semulajadi berdasarkan rangkaian neural buatan dengan kejituan yang tinggi.
mobile determination system of potential drug in natural product based on Artificial Neural Network (ANN).; drug determination system; Multilayered Perceptron (MLP) network trained with Back Propagation (BP), Levenberg Marquardt (LM) and Bayesian Regularization (BR) learning algorithms and the Hybrid Multilayered Perceptron (HMLP) network trained with Modified Recursive Prediction Error (MRPE) learning algorithm
Development of a simple mobile determination system for potential drug in natural product based on artificial neural network_Tan Bee Ling_E3_2009_875004428_00003095250_NI