Script.

Geoffrey Challen
1 parent 1cbbad00
Showing 6 changed files with 101 additions and 72 deletions
figures/.gitignore
figures/scripts/survey.py
figures/scripts/FormResponses.csv → figures/survey.csv
figures/survey.pdf
figures/survey.py
results.tex
+/.virtualenv
 *.pyc
-#!/usr/bin/env python
-
-import os
-import numpy as np
-import matplotlib.cm as cm
-import json
-import operator
-from datetime import *
-from pylab import *
-import matplotlib.pyplot as plt
-import csv
-from matplotlib import rc
-
-#rc('font',**{'family':'serif','serif':['Times'], 'size': 8})
-#rc('text', usetex=True)
-
-all_ = []
-eff = []
-usg = []
-filename = '/home/anudipa/logcat/data/FormResponses.csv'
-with open(filename, 'rb') as csvfile:
- for line in csv.reader(csvfile, dialect="excel"):
- uCount = 0
-	eCount = 0
-	if 'Timestamp' in line[0]:
-	continue
-	for i in range(2,8):
-	if 'No' in line[i]:
-	continue
-	elif 'Maybe' in line[i]:
-	if i < 5:
-	uCount += 0.5
-	else:
-	eCount += 0.5
-	else:
-	if i < 5:
- uCount += 1
- else:
- eCount += 1
-#	print line[i]
-	eff.append(eCount)
-	usg.append(uCount)
-#	print line[2:8], eff[-1], usg[-1]
-print len(eff),eff
-print len(usg),usg
-fig, ax = plt.subplots()
-#ax = fig.add_subplot(111)
-#fig.set_canvas(plt.gcf().canvas)
-left = 0.0
-for i in range(len(eff)): 
- ax.bar(left, eff[i], width=1, color = 'b', linewidth = 0.25)
- ax.bar(left+1, usg[i], width=1, color ='r', linewidth = 0.25)
- left += 2.5
-#ax.legend([r'{\textbf {Efficiency Metric}}', r'{\textbf {Usage Metric}}'], loc='upper center')
-legend = ax.legend(['Efficiency Metric','Usage Metric'], loc='upper center')
-# Set the fontsize
-for label in legend.get_texts():
- label.set_fontsize('small')
-ax.set_yticks(ax.get_yticks()[1:])
-ax.tick_params(axis='both', which='both', top = 'off', right = 'off', bottom = 'off', labelleft='off', labelbottom='off')
-ax.set_xlim([0,(len(eff))])
-#ax.set_xlabel(r'{\textbf {Users}}')
-#ax.set_ylabel(r'{\textbf {Value}}')
-ax.set_xlabel('Users')
-ax.set_ylabel('Value')
-fig.set_size_inches(6.5,2.5)
-fig.subplots_adjust(left=0.06, right=0.99, top=0.95, bottom=0.2)
-fig.savefig('survey.pdf', dpi=300)
+#!/usr/bin/env python
+
+import argparse,numpy,os,sys,csv,re
+
+from matplotlib import rc
+
+rc('font',**{'family':'serif','serif':['Times'], 'size': 9})
+rc('text', usetex=True)
+
+import matplotlib.pyplot as plt
+
+parser = argparse.ArgumentParser()
+parser.add_argument('output', type=str, help='Output file.')
+args = parser.parse_args()
+
+class Score(object):
+ def __init__(self, usage_score, efficiency_score):
+ self.usage_score, self.efficiency_score = usage_score, efficiency_score
+
+scores = []
+
+for line in csv.DictReader(open('survey.csv', 'rb'), dialect='excel'):
+ usage_keys = [key for key in line.keys() if re.search(r"""U\d""", key) != None]
+ efficiency_keys = [key for key in line.keys() if re.search(r"""E\d""", key) != None]
+ 
+ usage_score = 1. * len([1 for key in usage_keys if line[key] == 'Yes.']) + \
+ 0.5 * len([1 for key in usage_keys if line[key] == 'Maybe.'])
+ 
+ efficiency_score = 1. * len([1 for key in efficiency_keys if line[key] == 'Yes.']) + \
+ 0.5 * len([1 for key in efficiency_keys if line[key] == 'Maybe.'])
+
+ scores.append(Score(usage_score, efficiency_score))
+
+scores.sort(key=lambda s: s.efficiency_score - s.usage_score)
+
+efficiency_wins = len([1 for s in scores if s.efficiency_score > s.usage_score])
+usage_wins = len([1 for s in scores if s.usage_score > s.efficiency_score])
+
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.bar(numpy.arange(len(scores)) - 0.3,
+ [s.usage_score for s in scores],
+ width=0.3, color='b', linewidth=0.,
+ label='{\\small \\textbf{Usage-Based Measure}: %d Wins}' % (usage_wins,))
+ax.bar(numpy.arange(len(scores)),
+ [s.efficiency_score for s in scores],
+ width=0.3, color='r', linewidth=0.,
+ label='{\small \\textbf{Efficiency-Based Measure}: %d Wins}' % (efficiency_wins,))
+ax.legend(loc='upper center', fontsize=9)
+
+ax.set_yticks(ax.get_yticks()[1:])
+ax.xaxis.set_ticks(numpy.arange(len(scores)))
+ax.xaxis.set_tick_params(which='both', direction='out')
+ax.tick_params(axis='both', which='both', top='off', right='off', labelbottom='off')
+ax.axis(xmin=-0.7,xmax=len(scores) + 0.7,
+ ymin=0,ymax=max([max(s.efficiency_score,s.usage_score) for s in scores]) + 0.5)
+
+for tick_location in ax.yaxis.get_majorticklocs():
+ ax.axhline(tick_location, color='black', ls=':', linewidth=0.1, zorder=-1)
+
+ax.set_xlabel('\\textbf{%d Responses}' % (len(scores)))
+ax.set_ylabel('\\textbf{Score}')
+fig.subplots_adjust(right=0.99,top=0.98,left=0.08,bottom=0.10)
+
+fig.set_size_inches(6.5,2.5)
+
+fig.savefig('survey.pdf')
@@ -2,8 +2,37 @@
 \label{sec-results}
  
 To examine the potential components of a value measure further, we utilize a
-large dataset of energy consumption measurements collected on the \PhoneLab{}
-testbed.
+large dataset of energy consumption measurements collected by an IRB-approved
+experiment run on the \PhoneLab{} testbed. \PhoneLab{} is a public smartphone
+platform testbed located at the University at
+Buffalo~\cite{phonelab-sensemine13}. 220~students, faculty, and staff carry
+instrumented Android Nexus~5 smartphones and receiv subsidized service in
+return for willingness to participate in experiments. \PhoneLab{} provides
+access to a representative group of participants balanced between genders and
+across a wide variety of age brackets, making our results more
+representative.
+
+Understanding fine-grained energy consumption dynamics such as what apps ran
+for how long and how much energy each interactive session consumed while
+running in the background required more information than Android normally
+exposes to apps. In addition, to explore our content deliver metric we also
+wanted to capture information about app usage---including foreground and
+background time and use of the display and audio interface---that was not
+possible to measure on unmodified Android devices.
+
+So to collect our dataset we took advantage of \PhoneLab{}'s ability to
+modify the Android platform itself. Our modification augmented the platform
+to collect the fine-grained energy consumption and app behavior information
+required to understand smartphone energy consumption. We instrumented the
+\texttt{SurfaceFlinger} and \texttt{AudioFlinger} Android platform components
+to record usage of the screen and audio, and altered the Activity Services
+package to record energy consumption at each app transition, allowing energy
+consumption by components such as the screen to be accurately attributed to
+the foreground app, a feature that Android's internal battery monitoring
+component (the Fuel Gauge) lacks. The dataset of 67~GB of compressed log
+files represents \num{6806} user days during which \num{1328}~apps were
+started \num{277785} times and used for a total of \num{15224} hours of
+active use.
  
 At \PhoneLab{} based on the analysis of data collected about foreground and
 background energy consumption by applications running on the participants'
@@ -69,7 +98,7 @@ users spend actively interacting with them.
 \input{./figures/tables/tableCONTENT.tex}
  
  
-\begin{figure}[t]
+\begin{figure*}[t]
 \centering
 \includegraphics[width=\textwidth]{./figures/survey.pdf}
  
@@ -78,7 +107,7 @@ users spend actively interacting with them.
  
 \label{fig-survey}
  
-\end{figure}
+\end{figure*}
  
 To evaluate our efficiency metric against usage based metric, we sent out a
 survey to our participants asking to answer if they would remove the 3 top