To analyze the sentiments of people by analyzing their tweets about the presidential candidates and classify them into positive, negative or neutral sentiments. The tweets here refer to the ones from the 2012 Presidential elections between the two candidates – Barrack Obama and Mitt Romney. The result of classification will give us a possible outcome of the election.
Test Data Description: The given test data was in an Excel spreadsheet by the name –‘Project2_Testing.xlsx’. It consisted about 600 tweets with their respective class labels. The respective tweets were in separate sheets under the names –‘Obama-test’ and ‘Romney-test’. The experimental results are as follows:
RESULTS FOR OBAMA:
Overall Accuracy: 54.6%
Positive Negative Neutral
Precision 0.7333 0.6667 0.3402
Recall 0.2374 0.6905 0.6055
F-Score 0.3587 0.6784 0.4356
Confusion Matrix:
Positive Negative Neutral
Positive 33 48 58
Negative 8 174 70
Neutral 4 39 66
RESULTS FOR ROMNEY:
Overall Accuracy: 56.6%
Positive Negative Neutral
Precision 0.6 0.7791 0.2688
Recall 0.3391 0.6643 0.5376
F-Score 0.4333 0.7172 0.3584
Confusion Matrix:
Positive Negative Neutral
Positive 39 29 47
Negative 9 194 89
Neutral 17 26 50