# reading training data
    zf_test = zipfile.ZipFile('C:/Users/Umar/Downloads/test.csv.zip')
    df_test = pd.read_csv(zf_test.open('test.csv'), converters={'POLYLINE': lambda x: json.loads(x)[-1:]})
    latlong_test = np.array([[p[0][1], p[0][0]] for p in df_test['POLYLINE'] if len(p)>0])

    zf_train = zipfile.ZipFile('C:/Users/Umar/Downloads/train.csv.zip')
    df_train = pd.read_csv(zf_train.open('train.csv'), converters={'POLYLINE': lambda x: json.loads(x)[-1:]})
    latlong_train = np.array([[p[0][1], p[0][0]] for p in df_train['POLYLINE'] if len(p)>0])
# cut off long distance trips
lat_low, lat_hgh = np.percentile(latlong_train[:,0], [2, 98])
lon_low, lon_hgh = np.percentile(latlong_train[:,1], [2, 98])



     # create image
        bins = 513
        lat_bins = np.linspace(lat_low, lat_hgh, bins)
        lon_bins = np.linspace(lon_low, lon_hgh, bins)
        H2, _, _ = np.histogram2d(latlong_train[:,0], latlong_train[:,1], bins=(lat_bins, lon_bins))

prior = H2 + np.ones(shape=prior.shape)
prior = prior/np.sum(prior,dtype=np.float64)

    lat_centroids = [(lat_bins[i]+lat_bins[i+1])/2.0 for i in xrange(len(lat_bins)-1)]
    lon_centroids = [(lon_bins[i]+lon_bins[i+1])/2.0 for i in xrange(len(lon_bins)-1)]

name 'prior' is not defined