Points of Struggle

• Finding the proper datasets that contained all of the data that we were looking to capture for our problem.

  • We were able to identify multiple sources for data related to our problem. With this being the case, we decided to review each of the files to see which of them had columns that were relevant to our issue - with that, we were able to find a few sites that contained a great amount data for our use in training and testing our model.

• Merging of datasets.

  • This one threw us for a loop. We were able to identify the common fields accross the different datasets. The odd thing was that the merge statement in python was working with the same logic for two different files, and was failing for another pair. We were able to identify the problem columns, ensure the proper datatype, and effectively fixed the bug.

• Determining how to use PCA for feature selection.

  • When it came time to identify the proper data features for training and testing our model, we were working to implement PCA to aid our efforts. Getting this information in a workable state was difficult at first. We persisted and tried things numerous ways, never ceasing until we were able to achieve the result that we were looking for. We found a method that worked in the end, and our project would not be the same without it.

• Improving the accuracy of the model.

  • Upon initial creation of our first model, we were a little unhappy with the percentages that we were getting back for the confidence score. At first, we were getting that our model was about 56% accurate, which definitely had us question how we could improve the accuracy. We began adjusting the amount of features that we were to include in our model, and in doing so, we were able to create a model that was 64.45% accurate, which was an improvement from the initial iteration.