Estimating pedestrian volume (AADPT) for each crossing given the total AADPT and site characteristics

The goal of this project is to develop a method to calculate the average annual daily pedestrian traffic (AADPT) for each crossing in a given signalized intersection with a given total AADPT.

Dataset

The data set for this project include pedestrian/cyclist counts for signalized intersections every 15 minutes interval from around 2018 to 2022 from three regions:

• Milton, Ontario, Canada: 3,600,929 counts for 50 intersections
• Toronto, Ontario, Canada: 6,618,624 counts for 65 intersections
• Pima County Arizona, U.S.: 9,119,014 counts for 108 intersections The following plot showes the time series plot for the aggregated daily counts for four appraoches of an intersection for 2019 for Milton.

Phase 1: Use one/multiple short-term counts to estimate AADPT ratio

The simplest method is to take the ratio from one or multiple available short-term count(s). To evaluate this method, the mentioned data sets are used to:

1. Find the true AADPT and true ratios for each crossing
2. Use the valid days and hours to estimate the ratios from one or multiple short-term counts
   • The times of day to collect short-term count data includes 7am to 9am, 11am to 2pm and 3pm to 6pm occurring on Tuesdays, Wednesdays, and Thursdays from Sep to Nov.
   • Exlcuding the holidays (e.g. new year`s day)
3. Evaluate the error related to this estimate

The developed code can be used on any data set with similar structure.

How to Use

First, you need to prepare three csv files and save in data/ directory as follows:

• datasetName.csv: the timee series pedstrian counts for each crossing with column neams as (name, ped_N, ped_S, ped_E, ped_W)
• datasetName_holidays.csv: holiday dates in the region with (YYYY-MM-DD) format
• datasetName_intersections.csv: intersection names to be considered in the analysis

Then you can follow these steps to run analysis.

# 1. First, clone the repo
$ git clone https://github.com/mz-zarei/ped-vol-disagg-project.git
$ cd ped-vol-disagg-project/

# 2. Create a vitural Python environment
$ python3 -m venv .venv
$ source .venv/bin/activate

# 3. Install all requirements
$ pip install -r requirements.txt

# 4. Run!
$ python stc_estimation/main.py --dataset DatasetName

You can customize step four adding following features to the command:

• --dataset: the prefix of the input and output datasets (default: milton)
• --data-path: the directory where inputs are located (default: /data/)
• --out-path: the directory to save outputs (default: ./outs)
• --start-date: start date for the analysis period (default: 2021-10-01)
• --end-date: end date for the analysis period (default: 2022-09-30)
• --Max15min: Max valid value for 15min counts (default: 100)
• --Max24h: Max valid value for 24h counts (default: 500)
• --Min24h: Min valid value for 24h counts (default: 72)
• --stc-num: Number of short-term counts to be used for ratio estimation (default: 1)
• --repeat: number of samples to be taken with given stc_num for ratio estimation (default: 100)
• --percentile: nth percentile to be reported in the output (default: 85)
• --verbos: enables showing result in each step (default: False)

The outputs will be saved at stc_estimation/outs/ including two files:

• datasetName_error_df.csv: all ratio estimation errors for each intersection using each STC count(s)
• datasetName_res_df.csv: aggregated error results in terms of 95% confidence intervals and and Nth percentile for each intersection

To Do:

• Analysis the results for given intersections
• Use statitical methods to estimate AADPT ratio and compare it with the ones from STC estimation method