Traditional methods of counting bicyclists are resource-intensive and generate data with sparse spatial and temporal detail. Previous research suggests big data from crowdsourced fitness apps offer a new source of bicycling data with high spatial and temporal resolution. However, crowdsourced bicycling data are biased as they oversample recreational riders. Our goals are to quantify geographical variables, which can help in correcting bias in crowdsourced, data and to develop a generalized method to correct bias in big crowdsourced data on bicycle ridership in different settings in order to generate maps for cities representative of all bicyclists at a street-level spatial resolution. We used street-level ridership data for 2016 from a crowdsourced fitness app (Strava), geographical covariate data, and official counts from 44 locations across Maricopa County, Arizona, USA (training data); and 60 locations from the city of Tempe, within Maricopa (test data). First, we quantified the relationship between Strava and official ridership data volumes. Second, we used a multi-step approach with variable selection using LASSO followed by Poisson regression to integrate geographical covariates, Strava, and training data to correct bias. Finally, we predicted bias-corrected average annual daily bicyclist counts for Tempe and evaluated the model’s accuracy using the test data. We found a correlation between the annual ridership data from Strava and official counts (R2 = 0.76) in Maricopa County for 2016. The significant variables for correcting bias were: The proportion of white population, median household income, traffic speed, distance to residential areas, and distance to green spaces. The model could correct bias in crowdsourced data from Strava in Tempe with 86% of road segments being predicted within a margin of ±100 average annual bicyclists. Our results indicate that it is possible to map ridership for cities at the street-level by correcting bias in crowdsourced bicycle ridership data, with access to adequate data from official count programs and geographical covariates at a comparable spatial and temporal resolution.
The ongoing novel coronavirus (COVID-19) pandemic has highlighted the need for individuals to have easy access to healthcare facilities for treatment as well as vaccinations. The surge in COVID-19 hospitalizations during 2020 also underscored the fact that accessibility to nearby hospitals for testing, treatment and vaccination sites is crucial for patients with fever or respiratory symptoms. Although necessary, quantifying healthcare access is challenging as it depends on a complex interaction between underlying socioeconomic and physical factors. In this case study, we deployed a Multi Criteria Decision Analysis (MCDA) approach to uncover the barriers and their effect on healthcare access. Using a least cost path (LCP) analysis we quantified the costs associated with healthcare access from each census block group in the Los Angeles metropolitan area (LA Metro) to the nearest hospital. Social vulnerability reported by the Centers for Disease Control and Prevention (CDC), the daily number of COVID-19 cases from the Los Angeles open data portal and built environment characteristics (slope of the street, car ownership, population density distribution, walkability, traffic collision density, and speed limit) were used to quantify overall accessibility index for the entire study area. Our results showed that the census block groups with a social vulnerability index above 0.75 (high vulnerability) had low accessibility owing to the higher cost of access to nearby hospitals. These areas were also coincident with the hotspots for COVID-19 cases and deaths which highlighted the inequitable exposure of socially disadvantaged populations to COVID-19 infections and how the pandemic impacts were exacerbated by the synergistic effect of socioeconomic status and built environment characteristics of the locations where the disadvantaged populations resided. The framework proposed herein could be adapted to geo-target testing/vaccination sites and improve accessibility to healthcare facilities in general and more specifically among the socially vulnerable populations residing in urban areas to reduce their overall health risks during future pandemic outbreaks.
With the rapid proliferation of smartphones, human beings act as social sensors by means of carrying GPS-enabled devices that share location data. This has resulted in an abundance of sensor data gathered over long periods of time. Gaining meaningful insights from such massive amounts of spatio-temporal data accumulated by several disparate sources is often a challenge for organizations. Identifying demographics of mobile phone users by telecommunication providers is one such example. Demographic information plays a very significant role in targeting online advertisements to focused user groups by gaining insights about userfis mobility patterns. However, in practice, demographic information such as age and gender are mostly unavailable to app developers for open access due to privacy concerns. In this paper, we try to address the gap of how to enrich location data with demographics, which could be valuable for app developers. In our study, we use a machine learning approach to predict the gender and age of mobile phone users from a set of 3,252,950 anonymised GPS trajectories with 60,865 unique devices using a predictive model which is based upon the concept of Geohashes. We study to what extent usersfi demographics could be inferred from their frequently visited locations by encoding by formulating a multi-level classification algorithm to find the most frequently visited Geohashes and associating them with nearest points of interests which would enable predicting age-group and gender of the users who prefer to visit a specific location in a sequential manner. Experiments are conducted on a real dataset of mobile phone users collected and shared by a telecommunication provider. Th The experimental results show that the proposed algorithm can achieve mean prediction accuracy scores of 71.62% and 96.75% for predicting gender and age groups of the users respectively.
Smartphones and wearable devices are driving a boom in mobility data. We use data-driven tools for classifying movement data into five different travel modes (bicycle, walk, bus, motor vehicle and SkyTrain) in Vancouver and St. John’s, Canada. Using data from a GPS-enabled smartphone app (Itinerum) combined with a wrist-worn accelerometer (GENEActiv) collected over a period of 67 days, we classified modes using Support Vector Machines from 4071 trips. Pre-labelled data were used to classify modes with 90.9% accuracy when data from both devices were combined in comparison to a single data source with accuracy ranging between 55.5% and 79.4%.