Wednesday, July 18, 2012 9:50 AM
we are currently working on a data-mining project for a restaurant. We are in the conceptual phase right now, so we haven’t started to put in real work yet, but we thought about how the future solution could look like. The aim of this project is an as precise as possible sales forecast for the upcoming week. Additionally if we know that next Saturday will be a very sunny day we would like to use that information for the sales forecast for that particular day.
We want to include in our analysis/data-mining:
-store master data (size, outdoor seats etc.)
-weather: both historical weather data and a future weather forecast
-historical sales patterns (weekly, monthly and seasonal patterns)
-data about the location for instance if it is close to a shopping center etc.
We are working with sql server ssas, which has some built in data-mining algorithms.
We thought about the decision tree algorithm in order to see if there is a relation between weather and sales or between holidays and sales. In order to do that we could discretize the sales figures in buckets and the weather data too. After that we could run prediction queries to see in which sales bucket the upcoming day would fall into.
For the historical sales patterns requirement we thought about thetime series algorithm. As input we could use past sales figures.
The problem we have right now is that we cannot think about a solution for combining all the information we got from multiple algorithms in one simple predicted forecast figure.
One the one hand the decision tree cannot work with historical sales patterns and on the other hand the time series algorithm cannot use the weather forecast. I mean we could do a cross prediction, but as far as I understand it, there is no way to say: “look time series algorithm it’s going to be a Friday and it is going to be cloudy give me a sales number”!
It would be really nice if you have some advice for that particular project. It is our first real data-mining project and none of us have worked with the algorithms before, so it would be really nice if you guys could share your experience with us and how you would tackle that problem.
Saturday, July 21, 2012 3:17 AMAnswererThis thread discusses exactly the same situation http://social.msdn.microsoft.com/Forums/en-US/sqldatamining/thread/ebd022d5-bdde-4b48-be26-878ef2c900d4
Tatyana Yakushev [PredixionSoftware.com]
- Marked As Answer by Tatyana YakushevEditor Saturday, July 21, 2012 3:17 AM