Performance Rankings

Best Accuracy scores. All from Part 2 unless noted:

Part 2 Accuracy
60.42% (cleaned just punctuation, removed smileys from training, lowered min count frequency to 2, increased LogReg training iterations, then set custom pos/neg probability thresholds)
60.26% (no punctuation cleaning, but removed smileys, also added "i love" tweets to training, lowered min count frequency to 3, increased LogReg training iterations, then set custom pos/neg probability thresholds)
59.84%
59.12%
59.04%
58.83%
58.24%
58.22%
58.01%
57.21%
56.34%
56.31%
56.23%
56.04%
55.91%
55.41%
54.40%