Optimized Hybrid Deep Neural Networks for Efficient Alzheimer's Disease Classification

Artificial Neural Networks (ANNs) tuned with least square optimization PSO are used in the present training to take on the problem of Alzheimer's disease prediction. Several studies have demonstrated the reliability of ANN-based techniques in estimating LSM. However, most ANN training methods face significant challenges, such as low learning rates and being unable to move beyond the indigenous minimal amount. The visibility of ANN is improved by resquare optimization techniques algorithms (ROA) such as the PSO container. Typical practitioners don't undertake any sort of success planning to identify network design or pertinent components before applying PSO prototypes to ANN training. This led us to focus on a mock-up ANN mixing request for a fuzzy-based forecast in the present investigation. After the Alzheimer's disease was discovered, a substantial quantity of data was collected for the ICA-SVM-ANN and AD-ICA-PSO-ROA-ANN network models. The total number of records collected was 9467 for preparation and 45175 for testing. Exercise and difficult datasets were created using these data. The PSO algorithm's settings (such as system limits and loads) were kept at their optimal levels to guarantee the highest possible return on investment. One numerical catalogue, A method known as Real Root-Mean-Squared Error (RRMSE) was utilised in order to determine the predicted outcomes of a model for both the training data set and the testing data set. Therefore, the combined replicas gave a good performance, but the hybrid ANN model may be more effective than ANN, as measured by the scoring system developed. In addition, the inferential results showed that the AD-ROA-PSO-ANN prototype performed better than both the ANN and mixed Neural models at the same time.