in reply to using a stochastic matrix to simulate weather conditions
Aldebaran, here is something to get you started on collecting the numbers,
#!/usr/bin/env perl use strict; use warnings; use DateTime::Format::Strptime; #use DateTime; use REST::Client; use Data::Roundtrip qw/:all/; my $lat = 31; my $long = 80; # this is our fetcher, similar to LWP::UserAgent # but better suited for this kind of web service: REST my $rest = REST::Client->new() or die "failed to construct client"; # see examples in # https://www.weather.gov/documentation/services-web-api # set the host $rest->setHost('https://api.weather.gov'); # and this is our query with lat,long specified above my $query = "/gridpoints/TOP/$lat,$long/forecast"; # make the request and check the response code, 200 is good my $response = $rest->GET($query) or die "failed to GET($query)"; if( $rest->responseCode() != 200 ){ die "failed to GET(".$rest->getHos +t()."/$query) with ".$rest->responseCode() } # we get back JSON my $jsonstr = $response->responseContent(); # convert JSON string to a perl variable my $pv = json2perl($jsonstr); if( ! defined $pv ){ die "something wrong with this alleged json : '$j +sonstr'" } # go to the interesting part my $forecasts = $pv->{'properties'}->{'periods'}; # or print it all and examine it # print perl2dump($pv); # we have some dates in the data in ISO8601 format # this is a parser to convert that date to a DateTime # object which we can query about things (like seconds-unix-epoch) my $dateparser = DateTime::Format::Strptime->new( # parses 2020-04-26T06:00:00-05:00, # %F then literal T then %T then timezone offset pattern => '%FT%T%z' ) or die "failed to DateTime::Format::Strptime->new()"; # we store each prediction in this hash, keyed on start-end dates # see below $key my %parsed; for my $aforecast (@$forecasts){ print "go: ".$aforecast->{'name'}."\n"; # extract various things from 1 prediction record # examine the record like : print perl2dump($aforecast); # like "This Afternoon" my $period_str = $aforecast->{'name'}; # start and end times of the prediction, convert them to objec +ts my $start_time_str = $aforecast->{'startTime'}; my $start_time_dateobj = $dateparser->parse_datetime($start_time_s +tr) or die "parsing date '$start_time_str'."; my $end_time_str = $aforecast->{'endTime'}; my $end_time_dateobj = $dateparser->parse_datetime($end_time_str) +or die "parsing date '$start_time_str'."; # sunny? my $forecast_str = $aforecast->{'shortForecast'}; # temperature as a number, see later for the units my $temp = $aforecast->{'temperature'}; # store this record/prediction in our %parsed hash # keyed on this: my $key = $start_time_str." to ".$end_time_str; $parsed{$key} = { 'date-human-str' => $period_str, # edit: added this 'date-from' => $start_time_str, 'date-to' => $end_time_str, 'date-span-hours' => ($end_time_dateobj->epoch()-$start_time_d +ateobj->epoch())/3600, 'date-from-epoch' => $start_time_dateobj->epoch(), 'date-to-epoch' => $end_time_dateobj->epoch(), 'forecast-string' => $forecast_str, # we append temp unit to the key, e.g. 'F' 'forecast-temp-'.$aforecast->{'temperatureUnit'} => $temp, } } print perl2dump(\%parsed);
Edit: added comments to above
You still need to figure out how to convert the received data to machine-readable format, re: "sunny", and, most importantly, figure out how to store received data for easy access over time (SQLite database?, csv files?, json files? perl variable files which you can later eval? - in order of my preference) - what search keys, what identifies a unique record? etc.
Then you need to figure out how to create the sequence of "sunny"->"cloudy"->... given the time periods of the forecast etc.
But it looks ;ole I missed something because you should be training your model with actual data and not forecasts. But that's trivial to correct hopefully, just change your urls.
those who can do and those who can't blahblah, so Aldebaran++
bw, bliako (who is not a medical doctor and has no opinion about the efficacy of said dissinfectation methods).
|
|---|
| Replies are listed 'Best First'. | |
|---|---|
|
Re^2: using a stochastic matrix to simulate weather conditions
by Aldebaran (Curate) on May 01, 2020 at 22:57 UTC | |
by hippo (Archbishop) on May 02, 2020 at 08:11 UTC |