kleem · November 12, 2015 14:53
diff --git a/README.md b/README.md
diff --git a/bluelight_post_ts.csv b/bluelight_post_ts.csv
diff --git a/index.css b/index.css
diff --git a/index.php b/index.php
 <?php
  $LOG = json_encode(shell_exec('Rscript main.R'));
 ?>
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <title>R template</title>
    <script src="http://d3js.org/d3.v3.min.js"></script>
    <style>
      html, body { padding: 0; margin: 0; }
    </style>
    <link type="text/css" href="index.css" rel="stylesheet"/>
  </head>
  <body>
    <script>
      var __log = <?= $LOG; ?>;
      if (__log)
        console.log(__log);
        
      d3.xml('output.svg', function(error, documentFragment) {
        if (error) { throw error; }
        
        svg_dom = documentFragment
          .getElementsByTagName('svg')[0];
          
        d3.select('body').node().appendChild(svg_dom);
        svg_d3 = d3.select('svg');
        svg_d3
          .attr('width', '960px')
          .attr('height', '500px');
      });
    </script>
  </body>
 </html>
diff --git a/main.R b/main.R
 svg('output.svg', width=9.6, height=5)

 # Lettura del CSV
 post_ts = read.csv("bluelight_post_ts.csv", header=FALSE, sep=",")
 dates <- as.Date(post_ts$V2, "%Y-%m-%d")
 plot(dates, post_ts$V1, type="l", xlab="Date", ylab="Posts volumes", main="Posts in Drugs Forum")

 #Integrazione dei valori mancanti con 0
 all_days = seq(as.Date("1999-01-01"), as.Date("2015-07-12"), "days")
 all_values = c(length(all_days))
 for (d in 1:length(all_days)) {
    pos = match(all_days[d], dates, nomatch=-1)
    if (pos ==-1){
        all_values[d] <- 0
    }
    else{
        all_values[d] <- post_ts$V1[pos]
    }
 }
 plot(all_days, all_values, type="l", xlab="Date", ylab="Published posts", main="Posts volumes")

 # Creazione della TS e decomposizione secondo Loess
 my_ts=ts(all_values, start=c(1999,1), end=c(2015,1), frequency=365)
 plot(my_ts)
 loess=stl(my_ts, 365)
 plot(loess)

 # Prendo i remainders e cerco le anomalie
 loess_remainders = loess$time.series[,3]
 plot(loess_remainders)
 mu = mean(loess_remainders)
 sigma = sd(loess_remainders)
 abline(h=mu, col="red")
 abline(h=mu+3*sigma, col="green")
 abline(h=mu-3*sigma, col="green")

 #Locate outliers
 plot(all_days, all_values, type="l", xlab="Date", ylab="Published posts", main="Posts volumes")
 outliers = c(length(loess_remainders))
 for (i in 1:length(loess_remainders)){
    outliers[i]<-ifelse(loess_remainders[i] > mu+3*sigma || loess_remainders[i] < mu-3*sigma, TRUE, FALSE)
    if(outliers[i]>0){
        points(all_days[i], all_values[i], type='p', col='red')
    }
 }
diff --git a/output.svg b/output.svg
diff --git a/Rplots.pdf b/Rplots.pdf
diff --git a/thumbnail.png b/thumbnail.png
	<?php
	$LOG = json_encode(shell_exec('Rscript main.R'));
	?>
	<!doctype html>
	<html lang="en">
	<head>
	<meta charset="utf-8">
	<title>R template</title>
	<script src="http://d3js.org/d3.v3.min.js"></script>
	<style>
	html, body { padding: 0; margin: 0; }
	</style>
	<link type="text/css" href="index.css" rel="stylesheet"/>
	</head>
	<body>
	<script>
	var __log = <?= $LOG; ?>;
	if (__log)
	console.log(__log);

	d3.xml('output.svg', function(error, documentFragment) {
	if (error) { throw error; }

	svg_dom = documentFragment
	.getElementsByTagName('svg')[0];

	d3.select('body').node().appendChild(svg_dom);
	svg_d3 = d3.select('svg');
	svg_d3
	.attr('width', '960px')
	.attr('height', '500px');
	});
	</script>
	</body>
	</html>
	svg('output.svg', width=9.6, height=5)

	# Lettura del CSV
	post_ts = read.csv("bluelight_post_ts.csv", header=FALSE, sep=",")
	dates <- as.Date(post_ts$V2, "%Y-%m-%d")
	plot(dates, post_ts$V1, type="l", xlab="Date", ylab="Posts volumes", main="Posts in Drugs Forum")

	#Integrazione dei valori mancanti con 0
	all_days = seq(as.Date("1999-01-01"), as.Date("2015-07-12"), "days")
	all_values = c(length(all_days))
	for (d in 1:length(all_days)) {
	pos = match(all_days[d], dates, nomatch=-1)
	if (pos ==-1){
	all_values[d] <- 0
	}
	else{
	all_values[d] <- post_ts$V1[pos]
	}
	}
	plot(all_days, all_values, type="l", xlab="Date", ylab="Published posts", main="Posts volumes")

	# Creazione della TS e decomposizione secondo Loess
	my_ts=ts(all_values, start=c(1999,1), end=c(2015,1), frequency=365)
	plot(my_ts)
	loess=stl(my_ts, 365)
	plot(loess)

	# Prendo i remainders e cerco le anomalie
	loess_remainders = loess$time.series[,3]
	plot(loess_remainders)
	mu = mean(loess_remainders)
	sigma = sd(loess_remainders)
	abline(h=mu, col="red")
	abline(h=mu+3*sigma, col="green")
	abline(h=mu-3*sigma, col="green")

	#Locate outliers
	plot(all_days, all_values, type="l", xlab="Date", ylab="Published posts", main="Posts volumes")
	outliers = c(length(loess_remainders))
	for (i in 1:length(loess_remainders)){
	outliers[i]<-ifelse(loess_remainders[i] > mu+3sigma \|\| loess_remainders[i] < mu-3sigma, TRUE, FALSE)
	if(outliers[i]>0){
	points(all_days[i], all_values[i], type='p', col='red')
	}
	}