Using Natural Language Processing and Data Mining for Forecasting Consumer Spending Through Social Media

Document Type

Article

Publication Date

Winter 2-14-2024

Abstract

With the considerable rise in social media users, the platforms are no longer limited to sharing blog posts and emotions and making online friends. It is currently used for quite larger purposes which have created what is called the Big Data. This research paper covers one of the most recent studies in forecasting, which is forecasting through social media. Social media forecasting is a very promising yet controversial technique that numerous researchers have been quite interested to investigate. With the Big Data challenges, social media forecasting requires a few advanced machine learning and social media analytic techniques such as: data mining and sentimental analysis. The objective of this research covers the study of forecasting consumer intentions and understanding whether it can be possible to make future predictions about consumers’ spending from the enormous data shared every day. Data used in the experimental procedures of this research was collected in two methods, public surveys, and data mining techniques. The survey included in this research was publicly shared for people on a social media platform to anonymously participate and share their experience. For data mining techniques, some algorithms are proposed such as Part-of-speech Tagging, word2vec and semantic vectors. Forecasting techniques are applied as Time series analysis and SARIMA. The results discuss the different errors with each method to conclude which is the most accurate forecasting technique in our model.

Link to Full Text

Share

COinS