Statistical Predictions or Just Magic?

At the end of my graduate career at the University of Toronto, I joined a statistics competition hosted by the Statistical Society of Canada with my besties from my Master's program. We chose the Canadian disaster project. As I was reading through it again today, I'd like to post the original work I did for my group. Of course, there was a lot of data manipulation to get into my time series format. As for exploratory analysis, flood was the most frequent and the most expensive disaster in Canada. Ontario tends to have the most flood occurrences than any other provinces in Canada.

I first used the ARIMA, but it wasn't predicting anything. Thus, I turned my attention to Holt-Winters Exponential Smoothing Method on R. This method showed that the rainy season for the next 10 years starting from 2015 is between April and June across Canada. It also showed a gradual decline in the number of floods, which was optimistic. The predicted costs and the number of evacuations also followed a seasonal pattern within the rainy season for the next 10 years.

As a result, I suggested to the Government of Canada to be aware of flooding season between April and June of every year and be flexible to relocate resources including finance and personnel.