In this RMarkdown document is compared the naive and seasonal methods.

We perform forecasting on monthly wine sales series using naive and seasonal methods.

How to partition a monthly time series using R.

I choose the implementation of the gradient boosting (ensemble learning) knows as XGBoost. The advantage of this technique over linear regression model is the ability to treat with missing data and outliers. Before to fit the model, I had to clean and transform the data. My new approach works better than the linear regression model, achieving a RMSE of 0.17235 (an improvement of 0.11619).

I fit a linear regression model between living area square feet and the sale price. The election of the variable living square feet is based on previous articles which mention the size is a key factor in housing market. I ignored houses with more 4000 living area square feet (outliers), since linear regression model is sensitive to outliers.

Data Science Capstone for the Data Science Specialization from Johns Hopkins University.

Milestone report in the Data Science Capstone.

In this report we aim to describe the most harmful and economic consequences from severe weather events in the United States between the years 1993 and 2011. We obtained the database from national oceanic and atmospheric administration website. The events in the database start in the year 1950 and end in November 2011. We decided to take the data since 1993 because of a lack of good records in previous years. We found that most harmful events and events with greatest economic consequences have low occurence.