SAS-SCHULZ-MC1

VAST Challenge 2022

Challenge 1

Team Information

Team members:

• Falko Schulz, SAS Institute, Falko.Schulz@sas.com (PRIMARY)
• Don Chapman, SAS Institute, Don.Chapman@sas.com
• Steven Harenberg, SAS Institute, Steven.Harenberg@sas.com
• Stu Sztukowski, SAS Institute, Stu.Sztukowski@sas.com
• Cheryl LeSaint, SAS Institute, Cheryl.LeSaint@sas.com

Student team?

No

Analytics tools used:

• SAS Visual Analytics
• NETWORK Procedure from SAS Visual Data Mining and Machine Learning

Approximately how many hours were spent working on this submission in total?

100 hours

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2022 is complete?

Yes

Video

Watch Video on YouTube

Questions

Question 1 | Question 2 | Question 3 | Question 4

Demographics and Relationships involves understanding the city's demographics. Given social networks and other information about the city, you will analyze the available data to prepare a one-page fact sheet about the city's demographics, its neighborhoods, and its business base.

1Assuming the volunteers are representative of the city's population, characterize what you can about the demographics of the town. Provide your rationale and supporting data.

Limit your response to 10 images and 500 words.

Age

Age ranges from 18 to 60. Average Age is 39. Most cases (704 of 880) have an age between 22 and 56.

Apartment Location

Central is the most common value representing 45.68% (402 of 880) of Apartment Location. Other common values are Northwest (at 29.43%) and East (at 14.09%).

Apartment Cost

"$$" is the most common value representing 31.97% (485 of 1.5K) of Apartment Cost. Other common values are "$$$" (at 27.88%) and "$" (at 24.65%).

Households Size

Household Size ranges from 1 to 3. Average Household Size is 1.9.

Children (yes/no)

"no" is the most common value representing 71.93% (633 of 880) of Children (yes/no). Another common value is "yes" (at 28.07%).

Education Level

"HighSchoolOrCollege" is the most common value representing 47.95% (422 of 880) of Education Level. Other common values are "Bachelors" (at 26.36%) and "Graduate" (at 19.32%).

Schools

Enrollment (max) ranges from 242 to 418. Average Enrollment (max) is 360. Most cases (3 of 4) have an Enrollment (max) between 242 and 418.

Restaurants

Food Cost ranges from $4.1 to $5.9. Average Food Cost is $5. Most cases (16 of 20) have a Food Cost between $4.2 and $5.9.

Pubs

Hourly Cost ranges from $6.4 to $15. Average Hourly Cost is $11. Most cases (9 of 12) have an Hour Cost between $7.5 and $14.

Employer Job Count

Employer Job Count ranges from 2 to 9. Average Employer Job Count is 5.2. Most cases (202 of 253) have an Employer Job Count between 2 and 8.

Employer Education Requirements

"HighSchoolOrCollege" is the most common value representing 72.56% (2.2k of 3K) of Education Level (min req). other common values are "Bachelors" (at 12.64%) and "Low" (at 8.93%).

Wage Range

$10 - $20 is the most common value representing 67.24% (893 of 1.3K) of Wage Range. Other common values are $20 - $50 (at 29.59%) and $50 - $100 (at 3.16%).

Hourly Rate

Hourly Rate ranges from $10 to $100. Average Hourly Rate is $19. Most cases (1.1K of 1.3K) have an Hourly Rate between $10 and $33.

Job Start Time

8:00 to 8:29 is the most common value representing 26.20% (348 of 1.3K) of Job Start Time.

Job End Time

16:00 to 16:29 is the most common value representing 26.20% (348 of 1.3K) of Job End Time.

Job in Study

"In Study" is the most common value representing 89.6% (1.2K of 1.3K).

2Consider the social activities in the community. What patterns do you see in the social networks in the town? Describe up to ten significant patterns you observe, with evidence and rationale.

Limit your response to 10 images and 500 words.

There are several anomalies in the first few days of the study, such as people having multiple jobs with some lasting only 5 minutes. The study started with 1,011 participants with 32 dropping out after 4 days and another 99 dropping out after 5 days. None of the participants who dropped out of the study lived in an apartment and they all traveled to School. Our conclusion is these participants stayed in the dorms at college/university and traveled home at the end of the school year. The remaining 880 participants completed the study, which ran 450 days.

The analysis in this report is only on those that participated in the entire study.

On weekends, the longest average travel time was between Apartments and Pubs. On weekdays, the longest average travel time was between Apartments and Employers. Participants also had a longer average time staying at their Employer on weekdays versus weekends.

Using SAS Viya's network analytics procedure, we searched for various travel patterns to better understand the daily social activities of the participants. When analyzing patterns, we took into account the travel end location, purpose and check-in time. Using pattern match algorithm, we are able to determine the frequency of given patterns in the data. These were the primary patterns included in our search:

workDays

Number of days where the participant goes to work.

pubEvents

Number of times where the participant goes to a pub.

eatEvents

Number of times where the participant goes to a restaurant.

goingToEatAtWork

The participant goes to work, eats in the middle of work, then goes home.

twoJobs

The participant goes to work at one job then goes to work at a different job.

goingToPubAtWork

The participant goes to work, goes to a pub in the middle of work, then goes home.

goingToPubAfterWork

The participant goes to work, comes home, then goes to a pub.

goingToEatAfterWork

The participant goes to work, comes home, then goes to a restaurant.

multiplePubsOneDay

The participant goes out to at least two pubs.

These patterns can be grouped into Working, Pub, and Restaurant visits to better describe their daily social activities.

Eating out is a common activity after work and in general. The median participant ate out 469 times in total and ate out 130 times after a work day (40% of the work days).

Visiting pubs were a common activity even though 13 participants refrained from going to pubs altogether. Thankfully no participant went to a pub during a work shift. However going to pubs after work and on non-work days was common, often multiple times in one day. The median participant went to a pub 364 times in total, 71 times after work (22% of the work days), and had 100 days that contained multiple pub trips.

3Identify the predominant business base of the town, and describe patterns you observe.

Limit your response to 10 images and 500 words.

The citizens in the study primarily went to two locations during their free time: pubs and restaurants. Pubs were the dominant place to engage in social gatherings. There are five primary locations of pubs and restaurants:

• Central
• North Central
• Northwest
• East
• South

Throughout the study, two pubs dominated all sales in the central part of town: 1342 and 1344. These two pubs made up over 20% of all sales alone. Additionally when performing a network analysis of the travel network involving business pubs 1342 and 1344 had nearly twice the centrality score of any other business, implying that they are a common gathering location.

Nearly 78% of all pub and restaurant sales occurred between the Northwest, Southwest Central, and Central parts of town.

Restaurants had the lowest number of sales within the study. Sales tended to be distributed throughout the city relatively evenly, but the central part of the city had the lowest number of sales. The best performing restaurant, 1801, is located in the northwest part of the city. The poorest performing restaurant, 1349, is located in the central part of the city.

Throughout the course of the study, Restaurants and Pubs were shown to have very different behavior in terms of the times that participants spent money. Regardless of the location of the pub, participants tended to spend the most money on weekends between 9AM and 12:00AM, and weekdays from 5PM to 12:00AM. Pubs tended to have fewer sales on weekdays between 6AM and 4PM.

Restaurants told another story. Restaurants, however, generate most of their revenue between 7AM and 8AM on weekends and 11AM and 3PM on weekdays.

Sales for pubs across all regions exhibit identical seasonality, with the most sales occurring on Saturday and the least sales occurring on Wednesdays. There is no growth in spending for either restaurants or pubs based on these participants' habits. Sales declined between March and May before stabilizing to an overall average of approximately $340,000/month.

Restaurants, however, tell a different story. Seasonality is rarely exhibited within the identified districts, but individual restaurants often exhibit their own seasonality. When grouped together in districts, restaurants tend to have erratic seasonality. This is due to each individual restaurant's patterns not matching up with other restaurants. For example, Restaurant 1804 has very clear seasonality that peaks on Sundays, but this seasonality is not identical to other restaurants in the Southern district.

We discovered 6 clusters of weekly sale patterns by restaurants and pubs. There is no clear relationship between weekly sales patterns and location within the city. The y axis represents normalized total sales. The higher the value, the greater the sales at that time. The clusters below are zoomed in to a representative date range and is designed to show the general shape of sales patterns over time.

Cluster 1

Consists almost entirely of pubs with the exception of a single restaurant (895) that has similar demand patterns to pubs. Saturday and Sunday both have sharp increases in sales. Weekdays tend to have much fewer sales.

Cluster 2

Restaurants whose sales tend to be greatest during weekdays. Weekends have fewer sales.

Cluster 3

Restaurants whose sales tend to be greatest on Saturday followed by Sunday. Weekdays have fewer sales.

Cluster 4

Two restaurants whose sales are relatively evenly distributed across weekdays with few peaks in-between.

Cluster 5

Four restaurants whose sales are generally higher on weekdays but tend to have slightly higher sales on weekends compared to clusters 2, 3, and 4. Sales tend to be more erratic throughout the week compared to other clusters.

Cluster 6

Four restaurants whose sales tend to peak on Sundays, with Saturday close behind.

4From your answers to questions 1-3, assemble a one-page summary that provides the key information to share with residents about the town.