We need someone who is very experienced in applying different forecasting and regression methods/algorithms to multiple time series of data collected in a daily frequency and forecast (or regress) future values of the same variable(s) in a daily frequency. Also the resource needs to be skilled in Java so that the forecasting/regression algorithms or processes can be translated into Java functions which can be integrated into an existing Java application.
The forecasting needs to be PRECISE so that we can be rest assured that there will be not more than a 3% variance between the daily forecasted values produced by the functions created by the resource and the actual values after they are eventually collected for these variables when the future business day comes to pass.
The time series of data that needs to be foretasted/regressed displays monthly, weekly (day of the week) and seasonality trends. Also multiple forecasting/regression algorithms will need to be coded in order to accommodate different sets of time series data. In certain cases we may have anywhere from 30 days to 1 year worth of historical time series data (daily frequency), so may be regression algorithms may be more appropriate. In other cases we may have more than 2 years worth of time series data (daily frequency) so a forecasting algorithm like the ''Holtz Winter'' algorithm may be more appropriate.
This is an urgent project and may result in an ongoing permanent position.