Since combining the data of each month produced awful results, I decided not to do that but this lead to another problem. As I said in the previous article, Google Trend only offers data in months when choosing a large time range so this meant that I had to get the data in small sections and then combine them all together to obtain the values of Google Trend daily. Google Trend only provides daily data when the time range doesn’t exceed 6 to 8 months. After combining all of the data together, I obtained the datasets of Google Trend that contained daily data from 2013 up until 2018. I also did the same cleaning procedures to the previous data with some extra things and then I ran the Decision Tree Algorithm and these were the results:
A huge improvement in the accuracy scores but param model’s accuracy score is actually lower than the naive model which is okay. The param model doesn’t always guarantee an improvement in results.
Cryptocurrency 