On Mar 13, 11:56 am,
d...@autobox.com wrote:
On Mar 13, 2:23 pm, "socraticquest" <socraticqu...@hotmail.com> wrote:
On Mar 12, 3:14 pm, "socraticquest" <socraticqu...@hotmail.com> wrote:
Hello,
I'm seeking software that applies true-random numbers inforecasting.
Basically, this software applies true random numbers in analyzing
frequency patterns to make projections e.g., sales figures that have
no seasonal-related trends.
I had misplaced the reference to the download software that applies
true random numbers in frequency analysis. For what I remember,
forecasts are quite accurate.
Thank-you
To make a long story short. The software I'm seeking probably uses
random-sampling of data in a time series analysis. e.g., I have two-
years (100-weeks of sales data). I'm attempting to forecast the 101th
week of sales; where the previous two-years of data have no seasonal-
trend patterns.
Random-sampling analyzes the two years of data (a pseudo-frequency
analysis of sorts) and produces forecasts based on sample sales-
figures that appear most frequently.
So you are ASSUMING the following :
1. There is no relationship between successive weeks.
2. There are No Outliers or unusual values in the historical Series
3. There have been NO MOVEMENTS or SHIFTS IN THE MEAN for successive
weeks in the historical Series
4. You are unaware of any EVENTS that may have occurred and may have
had an affect on certain historical weeks
5. You are unaware of any possible set of PREDICTOR/CAUSAL/SUPPORTING/
HELPING/RIGHT-HAND SIDE/INPUT series that may have had an affect.
6. There is no relationship between weeks 52 period apart
7.The observed VARIANCE i.e. the variation around the local mean is
known to be invariant or constant over the historical data.
and now you want to use the most popular value or that value which has
arisen most frequently as the CONSTANT FORECAST for the next n weeks ?
Is this so ?
Not quite! With random-sampling of 100 weeks of sales results, the
most favorable projections to make are often NOT the more frequent
sales patterns. In fact, the favorable projections are around (even
below) the average of occurences in sales data.
From what I'm inferring, it seems this software makes forecasts based
on the percentage of different random-samplings corresponding to sales-
figures i.e.,the one or two sales-figures that have highest percentage
of different random-samplings (in a pseudo-frequency analysis) are
supposedly favorable to forecast.
Sales over two-years have been somewhat erratic; as this is a two-year
old (new) venture. Seasonal trends have been irrelevent.
Dave Reilly
Automatic Forecasting
Systemshttp://www.autobox.com
215-675-0652- Hide quoted text -
- Show quoted text -- Hide quoted text -
- Show quoted text -- Hide quoted text -
- Show quoted text -
