| inuggupx | Date: Sunday, 2013-08-25, 8:32 AM | Message # 1 |
|
Private
Group: Users
Messages: 2
Status: Offline
| Regression model for forecasting sales
It starts slow and slowly grows over time. cleaning it once a using linear regression, But if all data points are thought to be, The RMSE is in your order of 120 days, Which is very unacceptable.
If I only include the sales throughout the last 365 days (Last 40 or so data guidelines), The RMSE drops to a more suitable 10 days.
Is there a better way to do this without arbitrarily discarding data? everybody there's a small seasonal factor, Where sales appear slower inside a months. How can I include things like that into the regression?
1 the outcome points are $(go out with, identity)$, Where $ID$ is a number assigned incrementally each and every sale, And $Date$ is our date the sale was completed. for the some sales complete faster, Lower ID's can find yourself with slightly later dates.
I would ditch the cumulative approach and instead construct a scatterplot of sales vs time (Days or months). Your x axis most likely time (days, weeks, Months, consequently on.), or your yaxis will be sales made within that unit of time.
I think what you can find is that your data IS actually suited to a linear regression model. Since the slope of your cumulative sales is advancing, That means that you are insanely putting more sales per unit of time (y/x) As the time (by) also increases. This will me significantly to see if you do what I've described above.
Typically in such incidents people <a href=http://0120480514.com/price.html>0120480514.com/price.html</a> use Exponential Smoothing prediction methods such as HoltWinters prediction. The main advantage is that they are quite easy to calculate in Excel and they offer something akin to nonparametric regression upon which we can make prediction for next few intervals. It is also important that one may calculate standard error of prediction estimates, But for this you would be better of using specialized procedures, Like HoltWinters obtained in R.
Don't fret to use R, As it can be simply merged with Excel and is free. See this post for instance.
While I totally accept as true with Nick Adams, That you can do at least one regression on this data too, Be warned <a href=http://teibunsifucoidanya.com/japan.html>ヴィトン かばんteibunsifucoidanya.com/japan.html</a> that it's a time series, So you violate the usual assumption that the data points are independent from each other. So don't be tempted to calculate any sort of standard errors or relevance based on regression models. Model selection is done by identifying the main model by examinining it's autocorrelative structure. occasionally the data might be independent over time then one could use a simple regression model with time as a predictor. In most cases there is autodependence and then one has to identify a fitting model. Assuming any model is always just crazy. Early approaches assumed types of models for pedagogical reasons. The correct approach is to identify a proper ARIMA model that incorporates sufficient autoregressive memory and include any level shifts, Local time tastes, <a href=http://www.homeopatiktedavi.com/tote.html>www.homeopatiktedavi.com/tote.html</a> Seasonal pulses necessary to adequately describe the data. Pursue on the net for "intelligent arima softare" to help you. Try the available solutions on your data and measure forecast accuracy from a number of origins. Most <a href=http://celebrityrules.net/fake.html>uggブーツ偽物</a> software vendors have demo versions.
<a href=http://bbs.hkeec.org/home.php?mod=space&uid=75730>ugg</a><a href=http://mymaptowellness.com/forum/index.php?topic=3305.msg3305#msg3305>louis vuitton</a><a href=http://123.srworld.info/viewthread.php?tid=816521&extra=page%3D1&frombbs=1>ugg</a>
|
| |
|
|
