dec 16 semester 3 powerpoint presentation
TRANSCRIPT
Sentiment Analysis Costa Coffee and Starbucks
N.Ire and Mainland U.K.
By David Bourke
Summary of Presentation•What is Sentiment Analysis. Twitter• 3 Different types of Twitter Searches.•Rapidminer Extensions.•Collecting the Data. Using Geo Location•Cleaning The Data. •Predictive Modelling. Polarity.•Results (1)Word Lists. (2)Charts. Polarity & Gender •Q and A.
Twitter Searches• Twitter Search Data is Pulled.
• Twitter Streaming Data is pushed. 2% - 40%
• Twitter Firehose Data is pushed. Guaranteed 100%
Rapidminer Extensions• Search Twitter
• Text Processing
• Analyse Sentiment Aylien.com (3rd Party)
• Extract Gender Namsor.com (3rd Party)
Collecting The Data Geo Location Points.
• Magherfelt N.Ire
• Falkirk• Leeds• Norwich Airport• Shrewsbury• Stonehouse• Exeter• London
Cleaning The Data
• Remove Duplicates• Tokenize (Non Letters) • Transform Cases ( CAPITAL LETTERS to lower case)• Replace # @ Link• Stop Words (and) (the)• Stemming (Reading Read (ing)
Predictive Modelling Methods Used
• Naïve Bayes• K NN• Decision Tree• Random Forest• Deep Leaning
Most Common WordsCosta Starbucks
• Costa Starbucks• Christmas Christmas• Hot Chocolate Hot Chocolate• Ginger Bread Hot Fudge• Cup Cup
Summary of Twitter Data
• Costa Coffee 1184 Starbucks 8457
• Starbucks has 8 times more Twitter activity than Costa Coffee.
• Starbucks has very small amount of Negative Twitter Data.
Q & A