12. Results SentimentAnalysis
Based on the study of Ting,Tsang, &H.C (2011), the data is used forthe establishment of
training data as much as 30% and70% as test data.
13. Text Category
I am a ram guy but I totally appreciate vintage sports cars and that 1950s
corvette is totally amazing!
POSITIVE
when will gm step up and do the right thing and recall the defective dash
in the 07-14 trucks? I have bought many gm trucks and am honestly
ashamed that you cannot stand behind a product better than this.
NEGATIVE
Textareclassifiedintothreecategory:Positive(+)andNegative(-)
Classification
16. Results TopicModelling
To map the detailed topic, the author separate sentiment analysis result into2 files (positive andnegative). Neutral classified text is
not used since it contains a lotof data noises (e.g. advertisement, no sentiment words, etc.), except for Instagram, the authorwill
focus on neutral, because it has relatively faramount between positive, negative andneutral, so we can analyse the neutral data to
findout the wordorfrequently used hashtags. Then list the most probableterms within topics to summarize the topic.
17. 1) The first dominant positive topics in the dataset are about the free rides, promo. We
interpret this topic as an information in which most of the users tweeted about the promotion of
Uber,thepublicisreactingpositivelytowardthepromotion.
2) Second dominant positive topic in the dataset is about the driver compliment, the
users express their feelings on the usage of the services. Most of the customer is satisfied with
theservice,thiscanbe interpretedfromthewordlove,good, drivers.
Twitter-Uber:Positive
18. 1) There is a negative sentiment that influences the customer opinion. Based on the visualization
above, the first dominant negative topic is about sexual-harassment is probably occurred by the
news that is stating that there is a sexual-harassment towards Uber’s employee. These issues are
revealed by former Uber engineer. then it is followed by the claim of the Uber CEO who conduct the
investigation.
2) The second negative sentiment on the dataset is related to the driver. Even though some people
satisfied with the service, there are several complaints on its service, some ofthe drivers have been
reported driving while he/she is drunk. The customer is unsatisfied with the fact that the driver
tends to talk with the customer they expect to remain silent or calm, this is referred to driver talk
shitstop.
Twitter-Uber:Negative
20. The world’s most valuable resource
is no longer oil, but data.
www.economist.com
“
BIG DATA & SOCIAL COMPUTING LABORATORY 2017
SCHOOL OF ECONOMICS AND BUSINESS
TELKOM UNIVERSITY