Machine Learning Reccomendation System

This project was created for my masters year disseration. The following is from the abscract.

Increasingly machine learning is being used across many industries. This is due to machine learning’s ability to process vast amounts of data (such as user information like search history and content interactions) and then offer personalised data to the user based on these factors. One industry that could benefit from machine learning, is within the field of online video game distribution. The largest platform for this, Steam, does not currently use machine learning for its recommendation systems. (At the time of this projects creation, steam does use machine learning now) This project aims to prove the effectiveness of this method within this industry by developing a recommendation model with machine learning and having participants evaluate its effectiveness in order to come to a conclusion whether it is better than current methods. The results of the tests proved that machine learning models do in fact aide video game distribution platforms in user retention and the discoverability of games. This was shown in the results, as an overwhelming majority of participants agreed that recommendations were accurate and felt the recommendations would be persuasive enough for them to buy the recommended games without doing further research for themselves. One thing presented that could be built upon after the conclusion of the project, was that a more thorough comparison of the older methods could be made, with more plans for users to test both recommendations side by side. Overall though, the project was a success.
Download dissertation to learn more about the project
Download sourcecode