Diagnosing Parkinson’s Disease Using Voice Features Based On Deep Learning And Information Gain

Many studies have worked hard in recent years to advance the field of Parkinson’s disease early detection. Parkinson’s disease (PD) is a condition that affects the brain and impairs speech, movement, and thought. Previous research has indicated that the use of specific computer programmes, such as SVM (Support Vector Machine), NB (Naive Bayes), K-NN (K-Nearest Neighbour), and ANN (Artificial Neural Network), can be highly beneficial in the diagnosis of illnesses based on subjective sensations or feelings. Selecting crucial speech traits to aid diagnosis and eliminate problems has been suggested as a novel approach. However, the input data reveals the issue of class imbalances has an impact on both the performances of the classifiers and the findings in previous studies that employed SVM, K-NN, NB, and ANN for classification. ANN is said to provide great accuracy. NN (neural networks) need a lot of data to train or learn. They are unable to extrapolate from small training data sets. They fail to generalise successfully to fresh cases because they memorise the training data. This paper developed an improved model for detecting PD utilising a DL (deep learning) approach to solve this issue. Initial data equalization was carried out in this study using the SMOTE (Synthetic Minority Oversampling Technique). In order to normalise the input data in this study, min-max normalisation is used after data balancing. With the help of the information acquisition model, significant aspects are chosen. After the feature sets have been chosen, a classifier is used to identify Parkinson’s disease. Finally, in this study, speech features processed using WRNN (weighted recurrent neural network) and samples with PD detected. The suggested approach’s values of accuracy, precision, recovery, and f-scores were assessed where outcomes demonstrated that the suggested model outperformed other current models.