Project 1
Best Model - Naive Bayes:
Error Rate 0.079
Accuracy Rate 0.921
True Positives 0.328
True Negatives 0.968
Naive Bayes has highest true negative prediction, which would make it a good model to help alert people who are writing reviews that are unhelpful. If a review is flagged, then the user could be shown some tips to improve their review.
Project 2
Best Model - Ridge Regression:
Variance Training: 0.83
Variance Test: 0.76
Would be interesting to create some sort of online tool for people changing industries or cities to see what they should use as a starting point of negotiation for their new salary.
Project 3
Number of Clusters: 120
Obvious use is a recommendation system… another idea could be grouping for price comparisons?