Our primary hurdle arose when integrating our sentiment analysis tool. We encountered multiple errors while querying the News API, initially receiving insufficient articles and struggling to ensure headline relevance. To overcome these challenges, we delved into in-depth research on News API endpoints and adjusted our URL accordingly. Additionally, we encountered difficulties with the country tag functionality. However, we devised a workaround by modifying our search query to explicitly include the country name, enabling us to obtain accurate sentiment analysis for targeted regions.

We encountered another challenge when we initially sought to construct our own dataset by leveraging multiple APIs. However, we soon realized that many of these APIs lacked the specific metrics essential for our analysis. Consequently, we shifted our focus towards finding existing datasets. The OECD emerged as an invaluable resource in this endeavor, offering a wealth of free databases covering numerous countries. Despite this resource proving its usefulness it was still challenging to find the specific metrics we needed and the only way to accomplish this was by combing through several websites until we found multiple appropriate datasets.

Finally, we also experienced difficulties with our model. Specifically, our model did not have a high accuracy and we were stumped initially as to why this could be the case. We found the most likely reason was that our model did not take into account enough features. This prompted us to find other relevant features. We then used gold per ounce prices over time and government national debt over time. After implementing these features we still found that our model was not producing the accuracy score that we wanted and we realized this could be because of our randomized grid search. We then removed this and found our model performed much better and we hypertuned our parameters after and achieved our best least MSE of 0.000497.