Efficient Feature Extraction Based on Optimized Gated Recurrent Unit Recurrent Neural Network for Automatic Thyroid Prediction

Abstract

In this paper, develop Efficient Feature Extraction Based Recurrent Neural Network (EFERNN). Initially, the databases are gathered from the open-source system. After that, the pre-processing technique is developed for correcting missing values by the normalization technique of min-max normalization. The pre-processed data is utilized for feature extraction by using feature extraction techniques such as Two-Level Feature Extraction (TLFE) techniques. In level1, the ranked filter feature set technique is utilized to rank the features based on doctor recommendations. In order to execute the label-driven validation, ranking measures are used. In level 2, features are ranked and selected using a variety of metrics, including info gain, gain ratio, chi-square, and relief. In level 2, the effective features are chosen from the feature set using a fuzzy-based composite measure. In order to categorise thyroid disease from the databases, the Optimized Gated Recurrent Unit - Recurrent Neural Network (GRU-RNN) is used. In the GRU-RNN, the weight is selected with the assistance of the COOT Optimization Algorithm. The suggested method is put into practise in MATLAB, and its effectiveness is assessed by taking into account statistical measurements like kappa, accuracy, precision, recall, sensitivity, specificity and F Measure. To validate the proposed technique, it is compared with conventional techniques such as Deep Belief Neural Network (DBN). Artificial Neural Network (ANN) and Support Vector Machine (SVM).

COOT-Optimized Real-Time Drowsiness Detection using GRU and Enhanced Deep Belief Networks for Advanced Driver Safety

“…By integrating GRU into the framework, COOT can effectively capture the dynamic temporal patterns inherent in drowsinessrelated data, such as fluctuations in eye closure duration or changes in facial expressions over time. This temporal awareness enables COOT to discern subtle variations indicative of drowsiness more accurately, thus improving the overall detection performance [14]. On the other hand, EDBN is adept at learning hierarchical representations of raw sensor data.…”