Comparing Machine Learning and Traditional Statistical Methods for Gold Price Prediction

Gold is as a chemical element with the symbol AU, which belongs to the metal chemical element group. However, this thesis examines gold from a financial rather than a chemical perspective. We can assume that gold was the first instrument used to store the value and for other exchanging purposes. Ever since, gold is an instrument that is still widely used to store value, its demand shifts when economic recessions or instability is expected in the future, which affects the price. It is still considered financial instrument that relatively stable and not volatile. Countries might use this instrument to prepare for upcoming recession or to simply store its cash resources in gold, hence its price plays important role when it comes to countries’ political or economic decisions. Knowing gold’s exact future price therefore plays crucial role for investment decisions. Knowing the exact future price is of course impossible, but having a reliable estimate may be very useful for policymakers or investors. Machine Learning and Traditional Statistical methods are well-suited for estimating future gold price. Traditional Statistical methods are grounded by statistical and mathematical theory, machine learning models can identify non-linear patterns. By conducting predictive analysis using both type of models, this thesis comprehensively explains selected predictive models, identifies their pros and cons and compares performance of selected models