a good start at linking gtfs nodes to osm nodes

link_nodes.py

from sklearn.neighbors import KDTree
import pandas as pd
from geopy.distance import vincenty

WALKING_SPEED = 3 # walking assumed at 3 miles per hour
TRANSFER_TIME = 10 # transfer time in minutes

# assume df1 and df2 each have 2 float columns specifying x and y
# in the same order and coordinate system and no nans.  returns the indexes
# from df1 that are closest to each row in df2
def nearest_neighbor(df1, df2):
    kdt = KDTree(df1.as_matrix())
    indexes = kdt.query(df2.as_matrix(), k=1, return_distance=False)
    return df1.index.values[indexes]

df1 = pd.HDFStore("2015_06_01_osm_bayarea4326.h5").nodes
df2 = pd.read_csv("nodes_v4.csv", index_col="node_id")

df2["nearest_osm_node"] = nearest_neighbor(df1[['x', 'y']], 
	                                       df2[['x', 'y']])

new_edges = []

for transit_node_id, row in df2.iterrows():
	
	# need to make a new edge between the node in df2 (the transit node)
	# and the node in openstreetmap (the street node)

	osm_node_id = int(row.nearest_osm_node)
	osm_row = df1.loc[osm_node_id]
	
	distance = vincenty((row.y, row.x),
		                (osm_row.y, osm_row.x)).miles
	time = distance / WALKING_SPEED * 60 + TRANSFER_TIME

	# save the edge
	new_edges.append((transit_node_id, osm_node_id, time))
	# make the edge bi-directional as it's a walking edge
	new_edges.append((osm_node_id, transit_node_id, time))

new_edges = pd.DataFrame(new_edges, columns=["from", "to", "weight"])
print new_edges.describe()