Entrepreneurship in the Digital Economy - Application of machine learning in text mining and profile matching

Abstract

A wide variety of job search platforms and services are available to the public. Some are specialised to one sector, while others have distinctive added features, such as a company ranking system or salary projections. However, none of these uses a holistic user profile that includes not only hard and soft skills but also personality traits and professional goals. This is a market gap which can be closed by creating a new platform business utilising text mining and machine learning approaches. The goal of this thesis is to create a proof of concept for the algorithm and evaluate whether the idea would create additional value for the customer. The development of the business plan and obtaining a wider view of the market demands are topics of a different thesis. To test the business idea, a user experience and user interface prototype and a matching algorithm prototype were created. The purpose of the matching algorithm is to observe whether a search engine that operates with many key words (10+) would provide a better match than the existing solutions. The prototypes were created, and a lab experiment was conducted to evaluate their performance. Multiple approaches for information retrieval were evaluated, such as transfer learning, named entity recognition, lookup search and naïve Bayes. Each one was tested and it was decided to develop the algorithm using the lookup search and the naïve Bayes....