Time Series & Projection

Time Series & Projection

1. Moving Average

   • Weighted average of consecutive points of a series
   • Accuracy: It performs poorly on the forecasting of certain data
   • Application: inventory control
   • Data required: minimum of two years of sales history and it takes no more than a day to achieve results
   • Cost: $0.005 and the calculation is possible without calculator

2. Exponential Smoothing

   • Similar to Moving Average. Recent data is given more weight.
   • Accuracy: It performs poorly on most cases but on ST, it inclines to better results
   • Application: Productory and inventory control
   • Data required: Minimum to two years of sales and short time frame to do
   • Cost: $0.005 and can be calculated without calculator

3. Box-Jenkins

   • Mathematical model that assigns smaller errors to history
   • Accuracy: It delivers excellent results on the ST but on LT tends to perform poorly
   • Application: Productory and inventory control, forecast cash balances
   • Data required: Same as Moving average but requires more history to improve the model
   • Cost: $10.00 and takes 1 - 2 days to do

4. X-11

   • Decomposes a time series into seasonals, trend cycles and irregular elements. Primarly used for detailed time series analysis
   • Accuracy: Works excellent on the ST but isn't accurate on the LT
   • Application: Tracking and warning, forecasts of the company
   • Data required: Minimum of 3 years history to start
   • Cost: $10.00 and takes 1 day to do

5. Trend projections

   • Cost: Varies with application and takes 1 day to do
   • Fits a trend line to mathematical equation and then projects it to the future. Variations: slope characteristic method, polynomials and logarithms
   • Accuracy: It delivers excellent results, ST and LT
   • Application: New product forecats
   • Data required: Minimum of five years annual data to start