Banyak penyelidikan telah dijalankan untuk mengkaji potensi penggunaan rangkaian neural buatan dalam pengawasan biologi kualiti air. Projek ini akan mengkaji kesesuaian penggunaan rangkaian neural buatan untuk membangunkan sistem penunjuk kualiti air sungai berdasarkan komposisi alga. Pengenalan tentang asas rangkaian neural dan seni bina beberapa rangkaian neural akan dibincangkan. Di dalam projek ini, beberapa rangkaian neural akan dibangunkan. Diantaranya ialah Perceptron Berbilang Lapisan (MLP), Perceptron Berbilang Lapisan Hibrid (HMLP) dan Fungsi Asas Jejarian (RBF). Rangkaian-rangkaian neural yang telah dibangunkan akan dibandingkan dari segi keringkasan seni bina dan kejituan dalam pengkelasan kualiti air sungai. Projek ini telah membuktikan bahawa rangkaian HMLP yang dilatih dengan algoritma pembelajaran MRPE menghasilkan kejituan yang paling tinggi jika dibandingkan dengan rangkaian MLP dan RBF. Kejituan rangkaian HMLP dalam pengkelasan kualiti air sungai adalah setinggi 90%. Suatu sistem pintar yang menggunakan rangkaian HMLP telah dibangunkan untuk tujuan pengkelasan kualiti air sungai berdasarkan komposisi alga. Sistem pintar yang dicadangkan ini mempunyai beberapa kelebihan seperti mesra pengguna, mempunyai kejituan yang tinggi dan menjimatkan masa.
______________________________________________________________________________________
Throughout the years, many researches have been conducted on the potential applications of Artificial Intelligence (AI) in the biological monitoring of river quality. This project will provide an overview regarding the feasibility of the application of neural networks for direct classification of river water quality based on algae composition. A brief introduction to neural networks and the suitability of neural network for use in river water quality determination will be investigated. In this project, several neural networks will be developed and their performance are compared to yield the most suitable network that will be used to model the classification system for determination of river water quality based on algae composition. Among the types of neural network that will be developed are Multilayer Perceptron network (MLP), Radial Basis Function (RBF) network and Hybrid Multilayer Perceptron (HMLP) network. This study proves that the HMLP network trained using the MRPE algorithm achieves the best performance as compared to the MLP and RBF network. The HMLP network produces 90% accuracy. In this study, an intelligent system is developed for the classification of river water quality using the HMLP network. The proposed system provides several advantages in terms of its applicability, high accuracy, user- friendliness and as well as yields faster results compared to conventional system.
Development of an intelligent system for river water quality classification based on algae composition_Fong Wai Mei_E3_2006_NI