A Robust Approach based on CNN-LSTM Network for the identification of diabetic retinopathy from Fundus Images

Arumugam, Sajeev Ram; Devi, E. Anna; Rajeshram, V.; Balakrishna, R.; Karuppasamy, Sankar Ganesh; Kumar, S. Vinoth

doi:10.1109/icesic53714.2022.9783570

Cited by 6 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In engineering, they're commonly used for design optimization, power system scheduling [26], and drone path planning [27]. In data science, they prove valuable for tasks like feature selection [28], clustering [29], network analysis , and fine-tuning machine learning model parameters [30]. Meanwhile, in finance, they find application in portfolio optimization and risk assessment.…”

Section: Introductionmentioning

confidence: 99%

Multi-strategy Enhanced Artificial Rabbits Optimization for Prediction of Grades in Tourism Service Communication Courses

Jia,

2024

Preprint

View full text Add to dashboard Cite

Predicting students' grades through their classroom behavior is a longstanding concern in education. Recently, artificial intelligence has shown remarkable potential in this area. In this paper, the Artificial Rabbits Optimization Algorithm (ARO) is chosen to enhance the predictor's capabilities. ARO is a recently proposed and popular metaheuristic algorithm known for its simple and straightforward structure. However, like other metaheuristic algorithms, ARO often falls into local optima and, as iterations increase, the convergence speed slows down, leading to lower convergence accuracy. To address this issue, we introduce a Multi-Strategy Enhanced Artificial Rabbits Optimization Algorithm (MEARO). In MEARO, we first employ a Nonlinear exploration and exploitation transition factor (NL) to improve the balance between exploration and exploitation in ARO. we employ a Stochastic Dynamic Centroid Backward Learning approach (SOBL) to improve both the quality and diversity of the population. This ensures a broader optimization of the search area and boosts the chances of locating the global optimum. Lastly, we incorporate a Dynamic Changing Step Length Development strategy to enhance the randomness and development capability of ARO. To confirm the efficiency of MEARO, we compared its performance with eight other sophisticated algorithms using the CEC2017 benchmark. Our findings indicate that MEARO outperforms the other algorithms we tested. Furthermore, we optimized two critical parameters of the Kernel Extreme Learning Machine (KELM) using the MEARO algorithm, boosting its classification performance. Moreover, experimental results on the collected student performance dataset show that the KELM model optimized by MEARO outperforms other benchmarked models in terms of various metrics. Finally, we also find that interest in the course, frequency of classroom discussion, and access to extra knowledge and information related to the course are significant factors affecting performance.

show abstract