Length L‐function for network‐constrained point data

Abstract

AbstractNetwork‐constrained points are constrained by and distributed on road networks, for example, taxi pick‐up and drop‐off locations. The aggregation pattern (clustering) of network‐constrained points (significantly denser than randomly distributed) along roads may indicate spatial anomalies. For example, detecting and quantifying the aggregation with the highest intensity (i.e., the number of taxi pick‐up points per network length) can reveal high taxi demand. The network K‐function and its derivative (incremental network K‐function) have been utilized to identify point aggregations and measure aggregation scale, yet can only identify radius‐based planar‐scale results, thereby mis‐estimating aggregation patterns owing to the network topology configuration heterogeneity. Specifically, complex road networks (e.g., intersections) may incur aggregations despite their low intensity. This study constructs the length L‐function for network‐constrained points, using its first derivative to quantify the true‐to‐life aggregation scale and the local function to extract aggregations. Synthetic and practical data experiments show innovative detection of aggregations at the length‐based scale and with high intensity, providing a new approach to point pattern analysis of networks unaffected by topological complexity.