Have you ever needed to write a quick program or calculator? Have you ever had a problem that would be trivial to write a program for, but you still didn’t do it because you didn’t want the overhead of planning the interface, setting up version control, creating a GUI, distributing it to different systems, and figuring out how to share the code?

Utility Mill will handle all of that overhead for you, you just provide the core Python code to transform input to output.

(A storm is coming … a utilty storm)

I have the first prototype up and running at www.utilitymill.com

Please try it out. I really need alpha testers, and if anyone wants to work with me on this project (think Utility Mill co-founder!), please drop me a line. And of course, if nothing else, feel free to post ideas for utilities to make in the comments here.

If you need some ideas for utilities you could try making:

1. Alpha Phrase To Phone Number Calculator a la this
2. mysqldump simple GUI (to generate a command to paste into your console)
3. ls simple GUI
4. rsync simple GUI
5. .htaccess generator
6. Do something interesting using Python sets
7. Given 2 lists, list all duplicate entries
8. Convert number to binary, vice versa

Stay tuned, I’ve got lots of exciting features planned soon: web services to access utilities, chain utilities together, download a utility as a desktop app, and more.

I remember hearing in an Economics class long ago that the Federal Reserve has to give hints to the public about future interest rate changes since surprising financial markets is usually a bad thing.

I got to wondering recently if there were certain key words the Fed might use in its statements to warn the markets about upcoming changes in rates.

Coincidentally I also recently came across a very easy to use Python based Bayesian classifier called Reverend. (Here and here are some good introductions to the concept of Bayesian classification.)

So I came up with a plan put all of the FOMC statements I could locate into a Bayesian classifier and see if I could use words in past statements to predict future rate changes.

Update:
Latest (and better) code with more recent FOMC statements here.

Here is the data I compiled:
FOMC Statements with Rates (Includes two files, a Gnumeric format and a wierd CSV format)

Next, I put together a little Python script to read in the data I compiled, and train two Bayesian models.
Here’s everything you need to run the program for yourself.

One model will try to predict the exact rate change e.g., -.5, -.25, 0, .25, or .5 percent.
The other model tries to predict only the direction (or no change) of an interest rate change.

To test these models, I used 10 fold cross validation on a randomly shuffled data set and then I did 10 runs of that.

From these runs:
The direction-only model succesfully predicted the next rate change direction an average of 72.17% of the time with a standard deviation of 18.8 across all of the folds for all of the runs.
The model to predict the exact rate change was correct an average of 69.5 % of the time with a standard deviation of 17.28.
I don’t have any other methods to compare these results to so I’m not sure if they are good or not. But it looks like the predictions are correct a fair amount of the time, so perhaps this method is at least plausible.

What words in FOMC statements signal rate changes?
I’ll pull out a few of the high probability words for somes of the classes (rate change amounts) so you can have a look.

Lowered rate by .5% at next meeting:
soften
segments
damped
eroding
decrease
weaken
(mostly words associated with a poor economy, pretty neat!)

Raised rate by .5% at next meeting:
met
essentially
considerations
record
exceed
credit
(Maybe words associated with an overheating economy?)

Left rate unchanged:
limited
stabilizing
none
choice
Ongoing
prolonged
enhance
trend

And finally, my predictions for interest rates for July, 2007:

Both models predict a 60.99% chance of no change to the Fed funds rate.

Here are all of the two models’ predictions:

Exact Rate Change model predicts:
‘0′, 0.60992435592948235
‘0.25′, 0.2681162694505409
‘-0.5′, 0.022171742716859288
‘-0.25′, 0.004682228470558325
‘0.5′, 0.00040881787835661543

Direction only model predicts:
‘No Change’, 0.60992435592948235
‘Increase’, 0.28084524515677251
‘Decrease’, 0.13120882568327658

As you probably know, my website Just One Club Card lets you combine all of your club cards into one card. The final product looks like this:

The Problem:
Sometimes the barcode scanners at stores get confused and don’t know which barcode on the card to read. People have tried all kinds of solutions from bending or folding the card to covering the unneeded barcodes with your hand. These solutions do work but are awkward.

The New and Improved Solution:
Make a sliding, windowed, paper sleeve like this:

Here is a PDF with a template for the sleeve which you can print, cut out, fold and tape. (If someone wants to document this process in an instructable with clearer, step by step instructions, that would be most excellent!)

Source:
I created the template using Inkscape, a free vector drawing tool. Here is the SVG file in case you want to improve upon the template. I’ll admit the current one was hastily made. Please post improvements in the comments here.

–

So I posted Grocist on the Meta Filter projects page. (Feel free to vote for it if you’re a member )
And a user commented that he was sad it didn’t support CueCats. Previously I had thought this wouldn’t be a popular feature.

So I’m just taking a poll (comment in this post) to see how many users would be interested in CueCat support? Secondly has anyone tried a CueCat to see if it already worked. How about a “neutered” CueCat?

If a lot users or potential users have CueCats I’m happy to add support for them (assuming they don’t work right now).

People get annoyed when I assign them tasks in Basecamp and check that “notify by email” checkbox. So I added this even more annoying feature:

No, no, it’s not a real Basecamp feature. I used this method to make the fake screenshot. But it would be a fun feature to have, no?