Hi, I am working for a company that would like to use Aanalysis Services for predictive analytics. My problem is that there are only 40 or so months worth of data. I am wondering if this is enough to even make anything relevant and/or accurate.
So far I've made a time series mining structure, that just uses the booking number as input and predict. I am having a hard time justifying this as they view it as something that could have just been done in excel.
Also I have made a neural network. The company has several (continuous) metrics that they track along with bookings. So I made a neural network to use these inputs to predict the bookings.
This made another problem. The prediction that the neural network makes is different from the next month that that time series predicts and is causing distrust by the potential end users.
I've thought about combining the two models. I could use the output of the neural net and do a natural prediction join of that result into the time series.
Seeking advice, and something that I can do with Analysis Services that could be useful to a company with only a small number of months of data to work with.
View Complete Post