L-function of flows using Manhattan distance

code.py: The extracted aggregation by existing Euclidean flow L-function may be overestimated and causes dispersion. To solve the drawback, we propose a clustering method based on L-function of Manhattan space. This code reveals our method. flowdata-test.csv: We selected taxi OD data from the Shunyi district, which is a subcenter of Beijing and near the inside of the eastern Sixth Ring Road. The road network in this area is roughly in grid pattern, it can well demonstrate the effectiveness and robustness of our method. For this area, we chose the taxi OD flows for October 20–24, 2014. We examined flows that exceeded 2 km and we identified 1081 taxi OD flows in the district for our experiment.