A Dense Network Model for Outlier Prediction Using Learning Approaches

Abstract

There are various sub-categories in outlier prediction and the investigators show less attention to related domains like outliers in audio recognition, video recognition, music recognition, etc. However, this research is specific to medical data analysis. It specifically concentrates on predicting the outliers from the medical database. Here, feature mapping and representation are achieved by adopting stacked LSTM-based CNN. The extracted features are fed as an input to the Linear Support Vector Machine () is used for classification purposes. Based on the analysis, it is known that there is a strong correlation between the features related to an individual's emotions. It can be analyzed in both a static and dynamic manner. Adopting both learning approaches is done to boost the drawbacks of one another. The statistical analysis is done with MATLAB 2016a environment where metrics like ROC, MCC, AUC, correlation co-efficiency, and prediction accuracy are evaluated and compared to existing approaches like standard CNN, standard SVM, logistic regression, multi-layer perceptrons, and so on. The anticipated learning model shows superior outcomes, and more concentration is provided to select an emotion recognition dataset connected with all the sub-domains.