Mobile Application Development for Disease Diagnosis based on Symptoms using Machine Learning Techniques

In recent times most of the people die due to unavailability of proper diagnosis and medical services. In such cases the patient doesn't have quick access to medical services or face difficulties in using them. This could be very helpful in those remote regions where proper medical services are not available in hand or the patient is in need of immediate medical attention. Therefore, we created a self-assessment application which uses the power of machine learning to predict around 40 diseases, based on the symptoms of the disease. In self-assessment it is important that the prediction from the model stays fairly accurate. The existing machine learning techniques suffer from various drawbacks, including decreased accuracy, which is of primary importance in such cases. Our proposed system diagnoses diseases from the symptoms using the following steps. Firstly, data preprocessing was done and then the data was passed through various machine learning classification models, like K Nearest Neighbor, Support Vector Machine, Decision Tree, Random Forest Classifier, Naïve Bayes and Artificial Neural Network. After performing hyper parameter optimization, the probability vectors of all the models are added and then divided by the number of models. The number of diseases to be shown on application screen depends on the threshold values. The predicted disease(s) is received from the model and they are sent back to the application, where they are shown to the user. Our model proves to be the best when compared to the existing models.