Hello World! This repo was made for the purpose of STAT 325 This repo includes Week 10 Files which was a test run for the repo. The rest of the files are capstone files for the project. Lots of parameters were set very low in the algorithms due to computational power limitations (Sorry for that).
For the capstone project:
The dataset provided has high dimensions. The target variable for prediction is the Failure.binary which automatically makes this a classification problem. Logloss was then used instead of RMSE which was for regression models.
Model 1: Here we train basic ensemble models using different machine learning strategies: Bagging, Stacking, Support Vector Machines.
Model 2: Moving things higher, we use the Deep Neural Network.
Model 3: Going blind with the data, we dont consider the target variable and instead do unsupervised learning with KNN, Hierchical Mode, and Model-based Clustering