Video advertisement mining for predicting revenue using random forest
Abstract
Shaken by the threat of financial crisis in 2008, industries began to work on the topic of predictive analytics to efficiently control inventory levels and minimize revenue risks. In this third-generation age of web-connected data, organizations emphasized the importance of data science and leveraged the data mining techniques for gaining a competitive edge. Consider the features of Web 3.0, where semantic-oriented interaction between humans and computers can offer a tailored service or product to meet consumers' needs by means of learning their preferences. In this study, we concentrate on the area of marketing science to demonstrate the correlation between TV commercial advertisements and sales achievement. Through different data mining and machine-learning methods, this research will come up with one concrete and complete predictive framework to clarify the effects of word of mouth by using open data sources from YouTube. The uniqueness of this predictive model is that we adopt the sentiment analysis as one of our predictors. This research offers a preliminary study on unstructured marketing data for further business use.
Degree
M.S.
Advisors
Springer, Purdue University.
Subject Area
Business administration|Information science|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.