I had a hard time narrowing down options for a topic because I was not sure where to start looking. I went onto the tensorflow website and explored all the tutorials available to see if any ideas stuck out and image segmentations looked interesting. Then I went onto Kaggle and started lookinig through public datasets to see what was on the first few pages. I found two that I really liked, one with images of fruits to be identified or a set of wines that had chemical characteristics that could be used to predict the quality of the wine. After some thought and downloading the two datasets, I decided on centering my final project on the wine dataset because it the file was a more manageable size for the python programs I am using.
For this project, I found a study that used this data, both in 2015 and 2009. The study from 2009 used more advanced methods to what we have seen in class through things like neural networks and the latter used different methods called ‘fuzzy’ models. I was intrigued to see if what we have learned in class could actually compare to the ‘fuzzy’ models. I think it would be interesting to use a neural network model first, and if it turns out the model cannot predict wine quality very well, then switch to or add a secondary model to compare to such as random forest.