This project is the implementation of the paper Topic Model-based Road Network Inference from Massive Trajectories (MDM 2017)

Data Structure: (tid, x, y, lon, lat, time)

• Trajectories
  • utm_axis = (442551, 447326, 4634347, 4637377)
  • gps_axis = (41.858952 -87.69215 41.886565 -87.634896)
  • TrajMap
    • /Data/Chicago/chicago.pickle
      • id, x (utm), y (utm), t, tid
  • Biagioni
    • /Data/Chicago/all_trips
• Real map
  • osm
    • /Data/Chicago/chicago_edges_osm.txt
    • /Data/Chicago/chicago_vertices_osm.txt
  • dataframe
    • /Data/Chicago/chicago_map_df.csv

• Trajectories
  • Shanghai small data
    • utm_axis = (347500, 352500, 3447500, 3452500)
    • gps_axis = (121.4, 121.452, 31.1515, 31.197)
    • Data:
      • /Data/Shanghai/minsh_1000.pickle
      • /Data/Shanghai/minsh_1000_biagioni
      • /Data/Shanghai/minsh_5000.pickle
      • /Data/Shanghai/minsh_10000.pickle
  • Shanghai big data:
    • utm_axis = (345000, 365000, 3445000, 3465000)
• Map:
  • /Data/Shanghai/sh_map_df.csv

chicago_biagioni LineSegment 798 3.7801833333333335

• data need to change
  • ./bounding_boxes
• data need store
  • ./skeleton_maps/*
• data after runing
  • ./*.png
  • ./trips/*
  • ./skeleton_images/*
  • ./skeleton_maps/*

python script.py

data.index = range(len(data))
data[['tLon','tLat','pLon','pLat']]

python setup.py build_ext --inplace --force

• $u_{ij}$: topic matrix with $j$ th topic and $i$ th cell
• $k$: number of topics
• $w$: grid width
• $p$: padding