Alexander Millward Undergraduate Dissertation 2014/15

Producing a Measure of Consumer Confidence Through Text Analysis of Twitter Data

Supervised by M.Hepple

Abstract

This report contains the review of literature, planned experimentation and progress to date of a project to produce a measure of UK consumer confidence derived from automated text processing, specifically sentiment analysis, of Twitter data. Textual data concerning the economy and its performance will be retrieved from a previously obtained sample of Twitter data through keyword searches and will be classified using a mixture of lexicon-based and machine learning techniques of sentiment analysis. Results will then be adjusted for comparison to existing survey-based indices of consumer confidence for the purpose of evaluation. The capacity of the generated ratio to forecast or match indices will be used to draw conclusions about potential practical use as an alternative.