notes.txt

Tue 150118 1540
# on inspection of the html - the locations are not spread out 
# rather they are pinned at specific locations - for example, the
# most common entry in 100 randomly chosen addresses is Peshawar itself
# on further inspection - when using geocode to find address, sometimes, 
# it gives a suggestion which is usually Peshawar - need to monitor these
# suggestions somehow
# Results:
#  some results are given as partial matches, often leading to a city result
#  need to monitor the output on the partials - usually revert back to a result
# of just peshawar - which is an issue 
# Actions:
#  save the geocode object DONE
#  also save parital result in a spearate column too DONE
#  create new dataframe with above so we dont need to keep using api DONE
#  test the result against the input to see how close the match is** - called
#address tester

Tue 150118 1640
# the new heat/scatter map could be improved - maybe a pcolormesh
# the accuracy of the results from google maps is 48% which increases to 60%
when we include lev distances of 0 or 1

Wed 160118 1029
# focus on visualisation now - getting a nice image with the 60% accurate data
#this is possible with folium - very nice (finished)
# could also run the 15,000 addresses - will start