Optimized Functional Link Neural Networks with Flower Pollination Algorithm for Classification of Breast Cancer and Diabetes
Muhammad Nouman Atta, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Shafiullah, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Esha Rahim, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Sana Salah-Uddin, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Abdullah Khan, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Arshad Khan, Institute of Computer Sciences and Information Technology (ICS/IT), The University of Agriculture, Peshawar - Pakistan.
Corresponding Author:
Muhammad Nouman Atta (mna.aupkp@gmail.com)
Abstract:
Functional Link Neural Network (FLNN) is a type of Artificial Neural Network (ANN) which is used for different classification tasks. The FLNN consist of a high order of input layer neuron without a hidden layer. The traditional FLNN models confronts local minima and slow convergence due to inappropriate initial weight values which degrade the classification performance. This paper address the issue by proposing an approach such as Flower Pollination Algorithm (FPA) to optimize FLNN weights. The FPA mimics the pollination process in nature, enabling FLNN to efficiently navigate the search space for optimal weight configurations. This paper optimized the FLNN with FPA to remove the local minima problem in classifying breast cancer and diabetes. Experiments are conducted using breast cancer and diabetes datasets from the UCI machine learning repository and evaluate the proposed approach’s performance in terms of accuracy and Mean Square Error (MSE). Comparative analysis with ANN, FLNN, and Accelerated Particle Swarm Optimization Functional Link Neural Network (APSOFLNN) algorithms exhibits the effectiveness of the proposed model with enhanced classification accuracy of 98.98% and 98.93%, and reduced Mean Square Error (MSE) of 0.0204 and 0.0214 for breast cancer and diabetes, respectively, having the potential to serve in medical diagnosis applications.
Keywords:
APSOFLNN; Classification; FLNN; FPA; Optimization.