Analysis of Valentine Twitter Data Kyle Witt, Veslava Ovendale, - - PowerPoint PPT Presentation

analysis of valentine twitter data
SMART_READER_LITE
LIVE PREVIEW

Analysis of Valentine Twitter Data Kyle Witt, Veslava Ovendale, - - PowerPoint PPT Presentation

Analysis of Valentine Twitter Data Kyle Witt, Veslava Ovendale, Arash Naderpour Introduction Problem: how can businesses utilize Valentine s Twitter data in their practices Research questions: Look into peoples attitudes towards


slide-1
SLIDE 1

Analysis of Valentine Twitter Data

Kyle Witt, Veslava Ovendale, Arash Naderpour

slide-2
SLIDE 2

Introduction

Problem: how can businesses utilize Valentine’ s Twitter data in their practices Research questions: Look into people’s attitudes towards Valentine’s Day and what does Valentine’s Day mean to people. Why: Draw insights that can help businesses make informed decisions. Relevant research:

  • Netbase performed sentiment analysis in February 2016
  • Bollen et al. investigate whether public mood as measured from large-scale collection of tweets

is correlated or even predictive of DJIA values.

slide-3
SLIDE 3

Use cases for Twitter data

  • Our Question From Businesses:

○ We asked a financial analyst with the Seattle-based consultancy Rainier Group LLC catering to the needs of various businesses (bakeries, grocery stores, wineries) and a data analyst with Zulily how they would utilize Twitter data?

  • Their Answers:

○ “We would totally care about how many people supported which retailer and also their location like country or city.” (Zulily) ○ “Also we would care what they ordered.” (Zulily) ○ “If I was a card company, I would want to know at what time, how many times, people are tweeting Sarcastic tweets v. Romantic tweets so as to make cards in different parts of the season.” (Rainier Group LLC)

slide-4
SLIDE 4

Data Collection

  • Python

○ Modified HCDE module

  • Database

○ MySQL

  • Duration

○ February 11th, 12:00AM CMT ○ February 18, 12:00AM CMT

  • Collected

○ Tweets ○ Users ○ Place IDs and geolocations

slide-5
SLIDE 5

Preparing Data

  • Separate Ads from Non-ads

○ Binary classifier ○ Manual coded training set

  • Sentiment Analysis

○ Vader ○ Positive and Negative ○ Intensity

  • Filter by US Time Zones
  • Term Frequency

○ Top 100 by sentiment ○ Custom list

slide-6
SLIDE 6

Final Dataset

slide-7
SLIDE 7

Data Analysis

Advertisement Vs Actual Tweets

slide-8
SLIDE 8

Sentiment Analysis of All Tweets

Data Analysis

slide-9
SLIDE 9

Changing Sentiments

Sentiment Analysis of All Tweets (Difference)

slide-10
SLIDE 10

What Valentine’s Day Means

Top Most 100 Frequent Positive Terms

slide-11
SLIDE 11

Analyzing Data

Top Most 100 Frequent Negative Terms

slide-12
SLIDE 12

Valentine’s Day meaning

Top Most 10 Frequent Positive and Negative Terms

slide-13
SLIDE 13

Thank you

Questions?