MGT 3501 Lecture Notes - Lecture 13: Randomness
Document Summary
Easy to use and very commonly use. Main idea: prediction of the future depends mostly on the most recent observation, and the latest forecast. Forecast of the period before the current time period (f) Acts as a weight for the user to select the importance of the past data. Uses very little storage space for data. When a trend and seasonality is present: Exponential smoothing with trend and seasonality component. Can have all three, two of those three, but at least one. If a trend is in the data, moving average is not the one to start with. If there"s randomness in the data, linear regression shouldn"t be the one to start it. Can be up and down, but over time, there"s an increase. Can see a peak every november, but still a bit random in between, but overall trending down. First thing to do is to graph.