Objective:
To give indicators about choosing a good place of residence for students in NUS ISS by visualizing household surroundings.

Background&Tools:
Most of students in NUS ISS are foreigners, so the first thing after they came to Singapore was to find an ideal place to live in. However, they won’t know whether it is a perfect match for them until they have moved in for a while, and some houses have weaknesses like higher price, long distance or bad environment.
It is very difficult to find a better place for yourself, because no one can have a trial for changing the accommodation, so the students need some indicators to help them to define a good place of residence.Here we use ArcGIS to do all visualization and analysis.

Data Sources:
The basic data we use is from the app “Moves”. It contains 1,845 records from 26 students, and we have putted marked all records on the map——each point represents one record and the size of it represents the length of duration the person stayed. Then we add some other data like bus stop location, highway network, pharmacy and playfield to do the analysis.

Data Processing:
Step 1: Visualize students movement track data from MOVES by using ANN statistical methods.
Figure 1 Visualization of data from MOVES

Note: Each point represents one record and the size of it represents the length of duration the person spent on that place.
Insight: This figure shows where ISS students spent majority of their time and that the most popular area is around the NUS campus.

Step 2: Apply clustering and outlier detection.
Figure 2 Cluster and outlier

Note: “Duration” is the input.The points of high cluster, like the purple ones, indicate the places we often go and stay for long, like campus and home. And the points of high outlier are places we stayed for long but only visit once or twice, like hotel or house of friends and relatives.
Insight: For example, the bright green point, located in somewhere in the campus but not ISS, may be another place like Utown or central library where the student spent a whole day there for self-study or group discussion, but it is not the place he or she usually goes, so defined as an outlier. The points of low cluster are places that we often go but only stayed for a while, such as food courts and bus stops near our houses, and the points of low outlier are places that we also spent a little time there and only visit once or twice, like a shopping mall far away from home.
Purpose: To identify and define points represent homes.(as Figure 3 shows)

Figure 3 Points represent homes

Step 3: Add highway and main road data to explore the influence of noise to students’ homes by using buffer.

Note: Different kinds of road will have different average decibels of influence to the surroundings.

Purpose: To find homes which are seriously influenced by noise.

Step 4: Add bus stop, pharmacy and playfield location data to explore the convenience of daily life.
Figure 6 Bus Stop of Singapore

Insight and process: We put the location of all the bus stops in Singapore on the map.They are very crowded, implying that the foundation facilities are well-constructed in Singapore.We use the ‘Select by location’ tool and set the search distance as 500m, which means that if one’s home is less than 500m from the nearest bus stop, it will be selected.The result is shown in green ticks as shown in Figure 7.
Figure 7 Selected Homes by bus stop station

Then, we put the location of all playfields in Singapore. With a distance of 1000m, we select some of the homes and the result is shown in blue ticks.
Next we put the location of all the registered pharmacies (shown as red crosses in Figure 8)in Singapore on the map, too. Some of the homes are selected with a distance no more than 1000m, which are shown in purple ticks.

Figure 8 Selected homes by playfields and pharmacies

Step 5: Use the result we conclude before we finally marked those homes with the least noise and best convenience in the map.
Figure 9 Final result of good homes

Conclusion:
We have applied clustering, outlier detection, buffer, ANN methods and selection functions to do this geospatial analysis about household surroundings of students in NUS ISS.

We just used the students data for sampling, actually we can add more data about life and from different kinds of people to improve this analysis and make some recommendations.

Submitted by Team HAHAHA EBAC04 Full-time
GAO RUOFEI     A0163436E
HE SHANGUI    A0163348B
LIU YIZHE         A0163191J
ZOU YUMENG  A0163394Y