Bobby, I'm still not 100% on what method(s) I'm going to need to use. It might end up being the case that a lot of tuning is needed to make predictions usable. At the very least it should be able to note regularities and flag them.
Right now this is just something that I need, and thought it would be useful to others if it works. Thanks, Brian On Tue, Nov 18, 2008 at 1:21 PM, Metz, Bobby <[EMAIL PROTECTED]> wrote: > Brian, > How are you planning to tackle the prediction piece? Have any > particular algorithm in mind? I started speculating after reading your > e-mail, but nothing concrete jumped out at me. I thought that a purely > numerical date representation would work...using some type of standard > regression for prediction...with conversion from/to DateTime objects > simply for display. Of course that now seems impractical to me since > timezones and daylight savings irregularities would screw that up in my > mind. > > B > >> -----Original Message----- >> From: Brian Hann [mailto:[EMAIL PROTECTED] >> Sent: Monday, November 17, 2008 5:40 PM >> To: Steffen Mueller >> Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]; datetime@perl.org >> Subject: Re: Module submission DateTime::Event::Predict >> >> I thought maybe since I would be working with DateTime objects that >> would be the best fit. >> >> If that's not the case then I'm completely open to suggestions. >> >> Thank you, >> >> Brian >> >> On Mon, Nov 17, 2008 at 11:41 AM, Steffen Mueller >> <[EMAIL PROTECTED]> wrote: >> > Hi Brian, hi DateTime people, >> > >> > Perl Authors Upload Server wrote: >> >> The following module was proposed for inclusion in the Module List: >> >> >> >> modid: DateTime::Event::Predict >> >> description: Predict next date from a set of dates >> > [...] >> >> This module would take a set of dates (probably DateTime > objects) >> >> and find trends in recurrence ('every Thursday', 'every weekend >> >> day', 'every other week', 'once a month', etc) and also provide >> >> predictions for the next date most likely to follow the set. >> >> >> >> If workable I would think this would have any number of >> >> applications. >> > [...] >> > >> > I asked Dave Rolsky about this. He requested that for namespaces in > the >> > DateTime::* family, you should please clear your request with the >> > DateTime folks via their mailing list which I'm CCing. >> > >> > Best regards, >> > Steffen, for the PAUSE admins >> > >
