gistfile2.txt

plot_ushahidi_streams.py

import csv
from networkx import Graph
from pylab import *

class MsgData:
    def __init__(self, fname='haiti_sms.csv',
                 start_date='JAN 12',
                 end_date='FEB 11'):

        self.start_date = start_date
        self.end_date = end_date
        
        self.categories = {}
        self.data = Graph()

        f = open(fname)
        f.readline()
        for d in csv.DictReader(f):
            d_id = int(d.pop('#'))
            for c in d['CATEGORY'].split(','):
                if c in [' ', '']:
                    continue

                c_code, c_str = c.split('.')

                c_code = c_code.strip()
                c_str = c_str.strip()
                
                self.data.add_node(d_id, d)
                self.data.add_edge(c_code, d_id)

                self.categories[c_code] = c_str

    def streams(self):
        streams = []
        for c in sorted(self.categories):
            times_c = []
            for n in self.data[c]:
                times_c.append(datestr2num(self.data.node[n]['INCIDENT DATE']))
            stream_c, bins_c = histogram(times_c, arange(datestr2num('%s 2010' % self.start_date),
                                                         datestr2num('%s 2010' % self.end_date), .01))
            streams.append(stream_c)
        return bins_c, streams

def txt_col(col):
    return (.5*col[0], .5*col[1], .5*col[2])

def plot_report_streams(D, cmap=cm.spectral):
    t, streams = D.streams()
    n = float(len(streams))
    x = t.copy()
    
    k = normpdf(arange(-10,10,.1), 0, 4)
    streams = [np.convolve(s_c, k) for s_c in streams]

    for i in range(len(streams)):
        plot_date(x, i+streams[i][99:-99], '-',
                  linewidth=3,
                  alpha=1, color=cmap(i/n), zorder=.5-i)
        fill(hstack((x, x[::-1])),
             hstack((i+streams[i][99:-99], (i+zeros(len(x)))[::-1])),
             linewidth=2,
             color=cmap(i/n), alpha=.25, zorder=-i)
        text(x[0], i+.4,
             sorted(D.categories.items())[i][1].split('| ')[-1],
             color=txt_col(cmap(i/n)), va='top', ha='right', fontsize=8, rotation=30)
        text(x[-1]+.01, i,
             sorted(D.categories.items())[i][1].split('| ')[-1],
             color=txt_col(cmap(i/n)), va='bottom', ha='left', fontsize=8, rotation=30)
        x -= .1
        
    yticks([])
    xticks(.5+arange(datestr2num(D.start_date), datestr2num(D.end_date)),
            ['%d-%d' % (num2date(i).month, num2date(i).day)
             for i in arange(datestr2num(D.start_date), datestr2num(D.end_date))],
            fontsize=8, rotation=75, ha='center')
    l,r,b,t = axis()
    axis([l-4.5,r+4.5,b-4,t])
    
def hist_reports(D):
    """ Generate histogram of report type frequency"""

    hist = [[len(D.data[c]), c] for c in D.categories]
    hist = sorted(hist, reverse=True)

    left = arange(len(hist))
    bar(left, [h for h,c in hist])
    xticks(left+.5,
           [D.categories[c].split('| ')[-1] for h,c in hist],
           fontsize=8, rotation=45, ha='right')
    ylabel('Number of Reports')
    title('Number of Reports by Type')
    l,r,b,t = axis()
    axis([l-.5, len(hist)+.5, b, t])

if __name__ == 'main':
    D = MsgData()
    plot_report_streams(D)
    savefig('reports.png')