sentiment analysis on public comments data
TRANSCRIPT
CONSUMER SENTIMENT ON MIDRANGE SEDANS
FARHAN HABIB
SATHISH KUMAR PATTURAJ
TAO LIU
BACKGROUND•About 1000 Customers’ reviews for the four best selling small compact sedans in US from KBB.com
•Four small compact sedans are Honda Civic, Toyota Corolla, Chevrolet Cruze and Volkswagen Jetta
•Our goal is to find what customers like and dislike based on their reviews
CONSUMER REVIEW FROM KBB.COM
DATASET
• Created a data set from the reviews of the four sedans from the Kelly Blue Book website: http://www.kbb.com/
TOPIC ANALYSIS• Extract main points from 1000s of comments
• Insights on consumer feelings about the brand, product.
TOPIC ANALYSIS
Pros
CONS
TOPIC INFERENCES
•Customers are more happy on mileage and looks on all the 4 cars.
•But worried about the road noise and the absence of power back seat.
SENTIMENT ANALYSIS
• Performed using R
• Compared the words in the comments to the words in the positive and negative word files and assigned appropriate sentiment scores to the comments.
SENTIMENT HISTOGRAM USING R
SENTIMENT SCORE INTERPRETATIONS
•Used 1-way anova to determine the best model in the segment according to segment.
OVERALL
POSITIVE
NEGATIVE
SENTIMENT ANALYSIS
Inferences
• Jetta’s customers are either completely satisfied or completely dissatisfied.
•Corolla and Cruze – stable
•Civic has low sentiment
CORRESPONDENCE ANALYSIS
•Analyzed to find which cons topic is associated with more negative sentiment scores.
CORRESPONDENCE ANALYSIS(ALL CARS-CONS)
CONS
CORRESPONDENCE ANALYSIS
Inferences
•Topics 2 and 4 are more negatively viewed than other cons topics
•More road noise and cheap interiors
•Manufacturers need to focus on these defects immediately to be ahead of competition.
THE END