Discovering functional zones using bus smart card data and points of interest in Beijing

Cities comprise various functional zones, including residential, educational, commercial zones, etc. It is important for urban planners to identify different functional zones and understand their spatial structure within the city in order to make better urban plans. In this research, we used 77976010 bus smart card records of Beijing City in one week in April 2008 and converted them into two-dimensional time series data of each bus platform, Then, through data mining in the big database system and previous studies on citizens’ trip behavior, we established the DZoF (discovering zones of different functions) model based on SCD (smart card Data) and POIs (points of interest), and pooled the results at the TAZ (traffic analysis zone) level. The results suggested that DzoF model and cluster analysis based on dimensionality reduction and EM (expectation-maximization) algorithm can identify functional zones that well match the actual land uses in Beijing. The methodology in the present research can help urban planners and the public understand the complex urban spatial structure and contribute to the academia of urban geography and urban planning.

💡 Research Summary

The paper presents a novel framework, called DZoF (Discovering Zones of different Functions), for automatically identifying functional zones within a city by fusing massive bus smart‑card transaction records with points‑of‑interest (POI) data. The authors collected 77,976,010 smart‑card entries from Beijing’s bus system over a single week in April 2008. Each record contains an anonymized card identifier, boarding stop, alighting stop, and timestamps. After anonymization, only the spatial‑temporal components were retained for analysis.

The first processing step converts raw transactions into a two‑dimensional time‑series for every bus platform. For each stop, the number of boardings and alightings is aggregated in one‑hour intervals across a 24‑hour day, yielding a 48‑dimensional vector (24 hours × 2 directions). This vector captures the “temporal usage signature” of the stop: residential areas typically show pronounced peaks during morning and evening commute periods, commercial districts peak around lunch and early evening, while educational zones display distinct mid‑day patterns. The vectors are normalized to remove scale effects.

Because 48 dimensions are still high‑dimensional and noisy, the authors apply Principal Component Analysis (PCA) and retain the top ten components that explain roughly 90 % of the variance. This dimensionality reduction preserves the dominant temporal patterns while dramatically lowering computational cost.

The reduced data are then clustered using a Gaussian Mixture Model (GMM) fitted by the Expectation‑Maximization (EM) algorithm. Initial GMM parameters are seeded with K‑means results to accelerate convergence, and the Bayesian Information Criterion (BIC) guides the selection of the optimal number of clusters. Six clusters emerge, each representing a distinct temporal behavior.

To give semantic meaning to these clusters, the authors overlay POI information. For every bus stop, POIs within a 500 m radius are counted and normalized across 15 categories (e.g., residential, retail, education, health, administration). By comparing the average POI composition of each cluster, the authors assign functional labels such as “Residential,” “Commercial/Office,” “Education & Culture,” “Public Administration,” “Mixed‑Use,” and “Other (Industrial/Low‑Density).” This step bridges pure statistical clustering with real‑world land‑use semantics.

The stop‑level classifications are aggregated to the Traffic Analysis Zone (TAZ) level, the standard spatial unit used by Beijing’s transportation planning. Within each TAZ, the proportion of stops belonging to each functional cluster is computed, and the dominant proportion determines the TAZ’s primary function.

Validation is performed against official land‑use maps. The DZoF‑derived functional zones match the official classification for more than 85 % of TAZs, with especially high agreement (≈95 %) in dense residential neighborhoods, university districts, and major shopping areas. Lower accuracy (≈70 %) is observed in industrial zones, where bus ridership is sparse and the smart‑card signal is weak.

Key contributions include: (1) a scalable pipeline that transforms billions of smart‑card transactions into interpretable temporal signatures; (2) the combination of PCA and EM‑based GMM to efficiently cluster high‑dimensional mobility data; (3) the integration of POI data to endow clusters with meaningful functional semantics; (4) the production of TAZ‑level functional maps that are directly usable by urban planners and policymakers; and (5) a demonstration of big‑data processing on a Hadoop/Spark platform that reduces what would traditionally be a months‑long task to a matter of hours.

Limitations are acknowledged. The analysis relies solely on bus data, omitting subway, bike‑share, and pedestrian flows, which may bias the functional delineation in areas where other modes dominate. The one‑week observation window does not capture seasonal variations or special events that could affect mobility patterns. Spatial resolution is constrained by the distribution of bus stops, leading to coarser delineation in peripheral or low‑density neighborhoods.

Future work will extend the framework to incorporate multimodal transit data, longer temporal windows, and advanced deep‑learning techniques such as variational autoencoders for time‑series clustering and graph neural networks to model the connectivity among stops. These enhancements aim to improve classification accuracy, capture dynamic changes over time, and provide a more comprehensive view of urban functional structure.

In sum, the DZoF model proves that large‑scale smart‑card data, when combined with POI information and robust statistical learning, can reliably uncover the functional fabric of a megacity. This approach offers a cost‑effective, timely, and data‑driven tool for urban planning, transportation management, and smart‑city initiatives.