Personalized-Cancer-Diagnosis

Using classical machine learning techniques for classifying the data into 9 classes which can be further used for cancer detection. Performed exploratory data analysis and univariate analysis to observe how ”gene”, its ”variation” and related ”texts” impacts in classifying. Used one hot encoding and response coding for texts and trained models using Naive Bayes, KNN, Logistic Regression, SVM, Random Forests, Stacking Models and Maximum Voting Classifier and then compared the performance of all to classify the data in nine classes. Achieved the best precision-recall, least percentage of miss-classification, and log-loss of 1.048 for Logistic Regression with balancing.