Authors:
Divya Ankam
and
Nizar Bouguila
Affiliation:
CIISE, Concordia University, Montreal and Canada
Keyword(s):
Compositional Data, Dirichlet Regression, Generalized Dirichlet, Market-shares, Financial Data Mining.
Related
Ontology
Subjects/Areas/Topics:
Applications of Expert Systems
;
Artificial Intelligence
;
Artificial Intelligence and Decision Support Systems
;
Case-Based Reasoning
;
Data Mining
;
Databases and Information Systems Integration
;
Enterprise Information Systems
;
Industrial Applications of Artificial Intelligence
;
Operational Research
;
Pattern Recognition
;
Sensor Networks
;
Signal Processing
;
Soft Computing
;
Symbolic Systems
;
Theory and Methods
Abstract:
We explore the idea that market-shares of any given company have a linear relationship with the number of times the company/product is searched for on the internet. This relationship is critical in deducing whether the funds spent by a firm on advertisements have been fruitful in increasing the market-share of the company. To deduce the expenditure on advertisement, we consider google-trends as a replacement resource. We propose a novel regression algorithm, generalized Dirichlet regression, to solve the resulting problem with information from three different information-technology fields: internet browsers, mobile phones and social networks. Our algorithm is compared to Dirichlet regression and ordinary-least-squares regression with compositional transformations. Our results show both the relationship between market-shares and google-trends, and the efficiency of generalized Dirichlet regression model.