HN Explorer

Ask HN: Would you use Excel to release your new algorithm?

2 points by ninjamayo ↗ HN
I have been working for a while now, with a good friend of mine, on a data mining algorithm (decision tree) that has some pretty cool features. It is fast, robust, very accurate and parameterless. We didn't know where to start using it so we thought an Excel addin might do the job. We haven't launched yet but we are going to do it soon. Excel is for us a tool that we use on a daily basis among other data mining tools and seemed to require a lot less investment in time and money to begin with. So I know we are probably not following the optimal path for making this algorithm available to a wide range of users but do you think Excel is a dead end? Any suggestions are more than welcome.

4 comments

[ 3.4 ms ] story [ 14.8 ms ] thread
dalke
All of the people I know who do data mining use R or Python.

Since it's a good idea to compare your methods to comparable existing methods, and there are plenty of existing decision tree implementations (eg, C4.5), you'll likely have to validate/compare against off-the-shelf systems before you can convince people to invest in your tool.

There are also a lot of data mining algorithm, most of which are free, so you're going to have stiff competition. An Excel system might work, if you have a specific target audience and can reach them easily. The details will depend on your problem domain. It wouldn't work in mine.

ninjamayo
A lot of the people I know work in R or Python too however I also know a lot of people who dont have the coding or stats skills and need to do this kind of work. We thought it would be useful to build something simple in a familiar environment such as that of Excel. The tool doesn't require any kind of coding skills and its pretty simple to use.

Thanks for the feedback though.

dalke
Which means it depends very much on your target market, so is a question that cannot be answered on HN, at least not in the generic form given here.

Best of hopes for your product.

jslove
Excel = Small Data big $$

ETL = Bank people

Java = reel developers

Python, R = Scientists

c++ = old skool