The goal of our project is to provide flexible mobility solution such as an electric scooter within the school campus. With the electric scooter, students can avoid sweating it out rushing off to classes under the hot sun, avoid waiting 10minutes for a missed bus and avoid getting stuck at a crowded bus stop during peak times. Flexible mobility also allows students to cut across narrow paths to avoid long detours that are defined by the NUS shuttle bus route.
1 Problem Space
Compared to a bulky bicycle, the electric scooter is also lightweight, foldable and versatile, allowing navigation along narrow sheltered paths and hills, and easy lift while escalating staircases.
In this project, we want to find an optimal location for designating our electric scooter stand, so that students can rent them at a central location that is at the center of the most visited spots in campus.
We followed an input-process-output framework for our project. The input phase consists of data collection, data cleaning, finding new data layers that complement our objective and solution.
The process consists of exploring the distribution of our data, understanding our data, exploring the patterns and generating some initial hypothesis of the activity of our students. It also involves generating heatmaps, finding hotspots and generating relevant statistics that can be used to support our objective of locating the right site for an electric scooter stand.
The output consists of identifying the optimal location for our rental stands, and generating relevant insights about distribution of our data points.
2 Data cleaning
We collected coordinates of public bus stops near NUS campus, as well as coordinates for NUS internal shuttle bus stops.
We split the date and time into two separate columns for the data from moves.
Since Openpaths time is GMT+8, we converted all times to Singapore timestamp by adding 8 hours. Missing values was deleted. Data was not imputed.
After compiling moves and Openpaths data, we removed repeated IDs or overlapping data.
Last, we converted students names to student email ID as index.
3 Data Exploration Process
Our database consists of data of 52 students from Moves and Openpath. During data exploration process, we use Tableau to explore the data. The number of records for each ID shows below.
We also used the ‘Find Identical’ tool to filter out the 3601 different locations out of 7650 rows of data. This indicates that more than 53% of data are repeated checked-in points.
Exploring values of longitude and latitude:
- To study how change of longitude and latitude changes with real life distance.we plotted the latitude and longitude values on google map to get a feel.
- We discovered that 0.0001, 4th decimal place indicates a change of 10-20metres.In other words, 10 to 20 metres means the person is simply moving to different parts within the same facility.
- For a change in 0.001, 100 – 200metres. So we can easily detect when a person is moving when we see a change in long/lat values by 0.001
- Using this heuristic, it is easy for us to zoom in on time periods when change in movement is occurring during analysis.
4 Model Building
Use of ArcCatalog to build a default Geodatabase, for ease of managing and storing data.
To see the distribution of our data points for all 52 IDs across two weeks, we plotted all our data points on a Singapore basemap.
Next, we generated a heatmap by counting the number of points. From the heat map we can see where are the popular check-in places, or the places with high level of activity for past 2 weeks. As seen, the heatmap shows that the area with highest level of check-ins by ISS students for the two weeks is at NUS Clementi area. We can see numerous scattered spots around Singapore. These spots could be attributable to the workplaces of part-time students or residences of students staying outside campus. The high level of activity at the NUS-Clementi area aligns well with our objective to provide a flexible mobility solution for the students. We narrow our scope to the NUS campus where highest level of activity occur. Next we generated a word-cloud to have an overview of frequently checked-in places, both within and outside NUS. The word-cloud is generated from a frequency table we prepared by filtering and selecting the long/lat values that occurred frequently in our database.
The word cloud is simply a rough guide of which are the places with high level of activity, and is no way a part of our statistical analysis yet. The size of the word indicates the frequency of check-ins. We still need to validate these places have high level of activity by running proper statistical analysis methods on ArcGISOnline.
The reason behind wanting to know the location of the most visited places is such that we can find the mean center of these popular spots, in order to designate our electric scooter rental stands.
Using Statistical Analysis to find popular check-in places
Step 1: we begin by plotting the data points that are within the NUS campus area. We generated a heatmap on ArcgisOnline to display the hot spots. The heatmap is shown below:
Step 2: We did the hotspot analysis and generate the hexagons. Each hexagon is 23metre wide. And we have a total of 299 hexagons. There is a circle/bubble within each hexagon. The size of the bubble corresponds to its value. The smallest value or bubble has a value of 1, and the largest value is 225. Average value across all bubbles is 4. The purpose of this step is not to generate insight, but rather it is a statistical method that would be followed up in step 3.
Step 3: Based on the result of hotspot analysis, we filter the Statistical Significance and Number of points. By setting high statistical significance setting the number of points to be >= 6, we get 17 popular spots.
Next we ranked the 17 popular spots in descending order.
On the NUS map, we placed green markers at these 17 popular spots. The map below shows a few of the markers.
We split our area of analysis into two separate regions for the campus. The reason for this is that there is a huge separation or distance between the area at PGP/NUH and the region at central library. Using “Mean Center” tool to do spatial analysis to find out the mean center of the popular locations, we designated a scooter station at a place near to this point represented the mean center. The purple five-pointed stars represent the mean center we get.
After generating our mean center, we realised that the mean center can be located in the middle of a road or contoured hill. To address this issue. We did a buffer zone of radius of around 10 metres that contains the neighbouring area, such that we can recommend an electric scooter spot at the perimeter of our buffer zone.
For our project we did not designate at stand at NUS U-town, which is pretty far out from the rest of the campus. The reason being that ofo services are already set up in NUS U-Town.
We calculate the distance between two stations to confirm that these two scooter stations are quite far apart. Measuring the distance apart with the blue line drawn on the map, the two stations are approximately 1km apart.
We believe the little bits of time saved through flexible mobility will reduce procrastination and improve productivity for time-packed students.
Travel to the school gym you want at YIH or U-town without hesitation of time wastage or carrying loads around.
Travel to your favourite food vendors across different faculties with the increased mobility.
Travel across faculties for electives without being late for 20 minutes waiting for or switching buses in campus.
Arrive at your destination without needing to cool down your body during hot seasons.
Frees up your mind to carry your laptop around in school on a electric scooter, instead of having the burden of carrying it on your back while walking.
The public online map: https://iss-arcgis.maps.arcgis.com/home/item.html?id=0ff8dfd0386b4262b552f2bb29b23d15
Ng Wei Ping[A0056380H]
Wang Ruoshi [A0163338A]
Zheng Weiyu [A0163212R]