For the JHU Data Science Capstone project, the slides describe the aims, workings, and user interface of an application for Switfkey-style word prediction.

For the first milestone report for the Data Science Specialization Capstone, this report outlines the data cleaning and shows an exploratory analysis of n-grams with n up to 5. A first simple predictive modeling strategy is also given.

This Rpubs presentation is a pitch for the Shiny app at https://stargaser.shinyapps.io/queryWise/ . Astronomy enthusiasts among the public can now plot infrared data from the Wide-Field Infrared Surveyor!

We analyze the Storm Events Database from the National Weather Service (NWS), described at http://www.ncdc.noaa.gov/stormevents/details.jsp?type=eventtype. Our goals are to assess which events have taken the highest toll on population health in terms of injuries and fatalities, and to determine which events cause the most harm to the economy as measured by damage to property and crops. The database includes events from 1950 to 2011, but includes complete records from 1996 onwards; since we are assessing the impact of types of events, we limit our analysis to the 1996 to 2011 timeframe. We normalize the reported weather events to a standard set of reference categories provided by the NWS with the database. Our analysis finds that excessive heat was associated with the most fatalities, followed by tornados, flooding, lightning, rip currents and high wind events. Tornadoes, flooding, excessive heat, high winds and lightning were the most typical causes of weather-related injuries. Wind events, floods, hurricanes, lightning and tornados caused the most property damage; these events plus drought led to the worst crop damage.