Big Data for Big Cities: A Solution to Auckland’s Traffic Congestion

If people across the world were to agree on one thing, they would agree on that traffic congestion is bad for everybody. Traffic congestion is a major drawback of big cities. It negatively affects the environment through increasing CO2 emissions, slows down the economy through lowering productivity and mobility, stimulates social issues such as road rage, wastes a considerable amount of time, and the list goes on.

Increases in urbanisation and migration have exacerbated the problem of traffic congestion worldwide. This seems to be the case in Auckland, the largest city in New Zealand. Auckland population increased by 21% from 2001 to 2013. This significant growth in population resulted in more people needing to commute. Unlike other fast-growing cities, however, Auckland’s transport system does not seem to accommodate its growth.

Traffic congestion in Auckland has become an unbearable nightmare for many Aucklanders. The morning bus going from Huapai to the CBD leaves at 6.30am and arrives at 8.33am (NZ Harald). A couple of years ago, car drivers travelling in the morning from Albany to the CBD used to spend less than 30 minutes in the motor way whereas now it takes up to an hour to get to Harbour Bridge. Aucklanders are increasingly frustrated and dissatisfied with the performance of the city’s transport system (Wilson, NZ Harald).

In attempts to improve the transport system, Auckland Transport (AT) has been exploring long and short term solutions. These included upgrading existing facilities, “bus lanes, and different rail options including … a diesel shuttle, electric trains with and without double tracking” (Wade, NZ Harald). Unfortunately, these solutions were found to be costly, are faced with bureaucratic barriers, and/or require political motives (Wade, NZ Harald). Although Auckland Council and AT might have long-term plans to improve the system, a solution to reduce the current traffic congestion is very much needed.

There seems to be three approaches to reducing and controlling traffic congestion. The first approach is the use of physics to control the flow and movement of vehicles on the road. A classic example is the elimination of intersection that resulted in the idea of highways. Traffic engineers also influence the flow of traffic through using traffic lights, ramp stop lights/signs, merging lanes …etc. Most of these are already implemented in Auckland’s motorways. The cost of adding more of these traffic-management tools will probably be high and not exceed the benefit.

Another approach to reducing traffic congestion is introducing new policies. Road choices can be influenced by introducing an economic externality (e.g. taxing certain roads/lanes during peak hours) and/or incentives (e.g. tax exemption for jobs in rural areas and less populated towns). For example, the capital of Sweden, Stockholm, started charging about ₤2 on the use of bridges during rush hours, resulting in a 20% decrease in traffic congestion. In Auckland’s case, however, this approach does not seem suitable because of the bureaucratic and political barriers as pointed out by Ms Rose (Wade, NZ Harald).

The third approach is the use of high-tech solutions. These are devices built on IT systems and GIS technology to predict traffic flows and provide feedback on current traffic congestions. These devices might be good when there is plenty of alternative roads to use. The problem is that the device uses similar algorithms and same satellite, leading all device users to receive similar information and escape to the same direction which ends up creating traffic congestion.

Given the limited options AT has, an efficient solution would require looking into the current resources that are available and abundant. What is available and abundant is data… Big Data!

This article proposes a solution to the increasing traffic congestion in Auckland using marketing theory and Big Data analytics. The objectives include ensuring that the solution:

  • reduces the current traffic congestion in Auckland significantly;
  • is feasible; and
  • results in an efficient allocation of resources (minimise costs while increasing returns).

 

The reminder of this article is structured as follows. The next section introduces the solution and explains the theoretical background on which the solution is built. This is followed by an empirical illustration of the solution using traffic data collected in North Shore, Auckland. Then, the article points out the limitation, provides recommendations and concludes.

 

The Solution: Traffic Basket Analysis!

Theoretical background

Driving and consumption behaviours are similar in that choices and decisions are made to meet needs. Most of these choices and decisions are common and somewhat repetitive. One neglected cause of traffic congestion is the shared needs of road users. Humans live in structured and organised societies. Hence, geographic areas in big cities tend to be meaningfully clustered. There are commercial areas, residential areas, tourist areas, sport areas, and so on. Commuters move between these areas to achieve common goals (i.e. shopping, visit family, get to work…etc.). Thus, one would find a group of people making similar road choices. Examples include choosing to use the motorway to get to work, pick up children from school using the South-North side of the road, and use Road #10 to stop at the supermarket on the way back home. While these similarities between road users create traffic congestion, these also form patterns of co-decisions. The task is to identify associations between congested road choices and link the associated ones.

Market Basket Analysis is one powerful tool that marketers use to determine whether associations between choices/decisions are significant/strong, true, and co-occurring. In marketing, it is common to find associations between purchases of two, or more, complementary products. Examples include BBQ sauces in the meat section, laundry baskets with washing powders/liquids, bread and jams …etc.

Similarly, it is not unusual to find associations between two, or more, congested roads. A significant number of drivers using the Northern Motorway (SH1) also use Queen Street. Roads are complementary in nature; Figure 2 explains this. Travelling through Points B, C, and D complements the journey to Point E. But, what if we know that the association between Point A and Point E is strong and that, say, 80% of commuters departing from Point A go all the way to Point E?  Linking Points A and E directly makes the roads to Points B, C, and D substitute while it makes the road linking Points A and E complementary. The outcome should then be the removal of unnecessary traffic congestions at Points B, C, and D.

Empirical Illustration

AT has kindly made Auckland’s traffic data from July 2012 to February 2017 available on their website. The dataset contains traffic count data and statistics of over 7000 roads across Auckland. North Shore area is selected for this illustration. The focus here is on morning peak hours. There are three important steps required to perform the analysis and apply the solution to Auckland’s case.

Procedure

First, the top 20 most-congested roads are identified and selected for analysis. This is to ensure that the effect is large enough to take actions; in other words, ensuring that the solution helps a large portion of road users. Figure 3 ranks the top 20-most congested roads in North Shore, Auckland by volume of traffic. During the morning peak hour, Beach Road is the most congested road in North Shore while Apollo Drive is #20 most congested road in that area.

Second, the association rules are applied to the 20 roads to ensure that association between roads are true so that the same associations will be found the next morning, next week, and next month …etc. One major limitation of this illustration is that the data are recorded at road-level instead of commuter-level. Thus, it is impossible to know where exactly the traffic goes from these congested roads. For the sake of this illustration, the following assumptions are made:

  1. not all road users leave the congested area (stay to work, live, or any other reason);
  2. traffic flows randomly to the nearest roads;
  3. most drivers travel from North to South (i.e. Albany to Davenport/CBD).

 

Random proportions of road users are assumed to travel through other roads. The closer the travel distance is, the more likely it is to be the commuter’s next destination.

Third, Zhang’s (2000) Measure of Association and Dissociation is used to examine the level of co-occurrence between associations. In other words, the measure indicates whether Road Y is used when Road X is used. There is a co-occurring association when Zhang’s Measure is positive and a dissociation when it is negative. The higher the outcome of Zhang’s Measure, the stronger the co-occurrence. Figure 4 presents the result of the analysis; the map shows only the co-occurring associations between congested roads, and illustrates how traffic congestion can be reduced.

Refer to Figure 4, each coloured line on the map represents an association/s with one congested road/s. For example, drivers who used Beach Road also used East Coast Road (Taka/Glenfield/E.C), East Coast Road (Silverdale/Redvale), and Apollo Drive. Some of these associations are more obvious than others. An example is the co-use (0.66) of Lake Road-Devonport and Hurstmere Road.

Application

Now that we know the co-use of congested roads, our next task is to link these associated roads/areas. There are several ways to link two, or more, roads including cycling paths, bus routes, and special/smart lanes. For this to work properly, links between two congested areas should not be interrupted by intersections, bus stops …etc.. Linking the associated roads using the modern smart lanes/roads, such as the award-winning project SMART Tunnel in Malaysia, may be good for reducing traffic jams but are very expensive. An efficient solution is to offer a non-stop bus service. The non-stop bus goes back and forth using less congested roads to pick up and drop off commuters from two co-used congested roads/areas. For instance, the bus travels from Leak Road-Devonport to Sunnybrae Road and returns without stopping on the way.

Limitation

One major limitation of the illustration in this article is that the data are at road-level. This makes it harder to know which roads were exactly used by commuters. In case AT has no access to such data (driver-level data), a survey may need to be conducted to obtain these data.

Recommendations

  • The non-stop bus should not be interrupted by stops (dropping off and picking up commuters);
  • The non-stop bus service should be frequent during peak hours (mornings & evenings);
  • The waiting time for the next non-stop bus should not be long;
  • The price should not exceed price tickets for the normal bus service;
  • The bus route should be selected carefully so that it takes commuters less time to get to their destination than using their cars;
  • The non-stop bus departs from a point where car parks are available (other than bus stations).

 

Conclusion

Auckland’s traffic congestion is a serious issue causing a lot of the frustration in the city. Although AT has a number of solutions to reduce the stress in the city, these solutions are either costly, have bureaucratic barriers, and/or lack political motives. Because traffic congestion is increasing by the day, a solution to control it and reduce it is very much needed. This article proposed the solution of using big data and marketing theory to reduce traffic congestion in Auckland. Applying the logic of the well-known market basket analysis to traffic congestion in North Shore, it has been found that congested roads are co-used.

The proposed solution meets three important objectives (stated in the beginning of this article). First, one can be assured that the impact of using this solution is significant. This is because the statistical analysis used in this solution includes the most congested roads (top 20) only. Hence, the outcome of the analysis will influence the largest portion of road users. Second, the solution is feasible as all that is required is 1) data analysis; and 2) offering non-stop bus service. Third, this solution results in a more efficient allocation of resources. The costs of traffic congestion are minimized through encouraging road users to use the non-stop bus service; Thus, less pollution, less frustration, and less time wasted on the roads. The returns are maximized through improving mobility on the roads, improving productivity (workers travel time is reduced), and charging for the non-stop bus service.

The article, therefore, concludes that linking co-used roads using non-stop bus service has the potential to remove significant amounts of unnecessary traffic jams, resulting in a more efficient transport system and, ultimately, large smiles on the faces of morning commuters.

 

We, at NeoRetailing, are happy to offer AT and/or Auckland Council our help in testing and applying this solution.

 

Share/like this article if you:

  • support testing and applying this solution; and/or
  • would like to let AT know about this idea/solution.

 

PS: Data presented in this article are for illustration purposes only.