Prediction of Parkinson’s Disease : A comparative analysis of supervised machine learning algorithms using voice and speech signal data

Sammanfattning

Background: People irrespective of race and place, get affected by Parkinson’sDisease (PD). Approximately, 1% of the world’s population gets affected by thisdisease after reaching the age of 60. PD is a neural disorder that causes uncontrollable shaking of the legs or whole body. People with this disease might also developsymptoms such as insomnia, memory-related issues, depression, and changes in behavior. Parkinson’s Disease PD is a neurodegenerative disorder that progressivelyaffects both motor skills and other functions. Detecting it early is essential for better disease management. One of the earliest signs of PD is changes in speech, asindividuals often experience alterations in their voice, pronunciation, and fluency. Itis very important to predict the chance of falling victim to the disease by taking afew attributes into account. This can be done with the help of machine learningtechniques. Machine learning algorithms such as Logistic Regression (LR), RandomForest (RF), and K Nearest Neighbours (KNN) were used to predict if someone isat risk of getting diagnosed with this disorder. Objectives: The objective of this research work is to build machine learning modelsby training machine learning algorithms and to find out which algorithm is the mostaccurate in predicting Parkinson’s Disease. Methods: Two research methodologies were used in this research. We have used Literature review and Experimentation to complete the objectives. We have reviewedmany existing research papers published on PD and its prediction using machinelearning. Finally, we have used experimentation to build three machine learningmodels by training their respective algorithms with a data set collected from theweb. Results: By performing a Literature review, algorithms like LR, RF, and KNNwere selected. In the experimentation part, models have been built by training thealgorithms with the data set "Parkinson’s Disease" and KNN is the best-performingalgorithm with the highest accuracy. Conclusions: Three machine learning models, LR, RF, and KNN were built bytraining their respective algorithms using the data set "Parkinson’s Disease". Thisresearch has concluded by saying that KNN is the best-performing algorithm as itachieved the highest accuracy in the prediction of PD.