Developing a Traffic Congestion Model based on Google Traffic

Data: A Case Study in Ecuador

Yasmany García-Ramírez

Department of Civil Engineering, Universidad Técnica Particular de Loja, San Cayetano Street, Loja, Ecuador

Keywords: Urban Streets, Developing Countries, Traffic Congestion Model, Fundamental Parameters, Google.

Abstract: Congestion on urban streets has negative impacts on the urban economy, environment, and lifestyle.

Congestion, in developing countries, will increase despite knowing its cons. One way to control or reduce

congestion is by sharing traffic information through traffic model congestion. This model includes the

estimation of the travel time from the desired place of origin-destination. Speed-flow-density parameters

help to calculate travel time. These fundamental parameters could be estimated using Floating Car Data

from Google. Therefore, the objective of this research is to calibrate equations for the fundamental

parameters with traffic state indicators by Google, relating them to ground truth data. Six density-flow

equations and six speed-density equations were calibrated using power and linear curve, and some of them

were validated. Other cities can use these equations to build their traffic congestion model. With this model,

road users can plan the journey and choice the best route or travel in times of low congestion or uptake of

public transport, decongesting the city and saving traffic costs related. This comprehensive research extends

the knowledge of how Google traffic information can employ in developing cities.

1 INTRODUCTION

Congestion on roads, especially in urban areas, has a

large negative social and economic impact on the

community as well as on the environment (Bacon et

al., 2011). Congestion may cause delay and noise

that frustrates motorists and commuters, which also

would have health implications. It may also lead to

road traffic crashes and the degradation of the road

infrastructure (Ackaah, 2019). In developed nations,

traffic congestion is taken seriously, applying

several measures to reduce it or control it.

Unfortunately, most cities in developing countries

are experimented and will be doing, hard times with

traffic congestions (Yokota, 2004).

In developing economies, the main problem is

that congestion keeps on increasing because the

number of people owning cars keeps increasing

(Ackaah, 2019; Mfenjou, Abba Ari, Abdou, Spies,

& Kolyang, 2018). The scenario is complicated

when many people live in these cities, intensifying

transportation of good and passengers (Jain, Jain, &

Jain, 2017). Also, when public transportation does

https://orcid.org/0000-0002-0250-5155

not offer enough quality for drivers to leave their

cars at home, or road infrastructure does not

encourage drivers to change the mode of transport.

Although the problems of vehicular congestion are

known, very little has been done, due to the lack of

personnel and technology, and especially to financial

constraints (Yokota, 2004; Singh, Bansal, & Sofat,

2014).

Congestion can be tackled either by increasing

street capacity or through demand management.

Increasing capacity is very difficult in urban

environments and very expensive that developing

countries cannot afford (Baratian-Ghorghi & Zhou,

2015). A more practical means of handling the

existing infrastructure to optimize its use has

become necessary (Ackaah, 2019). Some demand

must be reduced, displaced to other routes, or move

to other days if users have access to timely, accurate,

and reliable traffic information (Bagloee, Ceder, &

Bozic, 2014). This information could influence

travel behaviour (Reza & Kermanshah, 2005;

Andersson, Hiselius, & Adell, 2018) and could

reduce journey time; and traffic congestion along

with reduced vehicle emissions and fuel

consumption (Hall, 1996). To build a successful

García-Ramírez, Y.

Developing a Trafﬁc Congestion Model based on Google Trafﬁc Data: A Case Study in Ecuador.

DOI: 10.5220/0009594501370144

In Proceedings of the 6th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2020), pages 137-144

ISBN: 978-989-758-419-0

137

traffic congestion model is necessary to collect

some.

Data collection could be performed using

traditional on-road sensors such as inductive loops

or road tube counters. These sensors require

specialized equipment for their installation in the

outdoor. Moreover, maintenance requires that

personnel visit the locations and repairs disrupt

traffic. Those sensors are limited in terms of their

coverage because it is prohibitively expensive to

instrument representative road sections in the city.

Floating Car Data (FCD) is an alternative data

resource, that has high coverage (Altintasi, Tuydes-

Yaman, & Tuncay, 2017; van den Haak et al., 2018).

FCD is possible due to the rise in the number of

mobile phones (Gunawan & Chandra, 2014) and the

increase in Internet use (Jiang, 2019). An indirect

measure of the number of mobile phones is the

mobile cellular telephone subscriptions (for every

100 people) that employ cellular technology. In

2018, in Ecuador, the subscriptions were 92, while

in the whole world was 104.94 (World Bank,

2019b). Other countries that, according to the World

Bank (World Bank, 2019a), have similar income had

98 (Serbia), 106 (Tonga), 132 (Argentina), 153

(South Africa), and 180 (Thailand). Internet use was

also increased worldwide, and in 2017, around 50%

of the population used it via a computer, mobile

phone or digital TV (World Bank, 2019b). In the

same year, Ecuador had 57%, 73% for Serbia, 41%

for Tonga, 74% for Argentina, 56% for South

Africa, and 57% for Thailand. Regarding

smartphones, in 2017, the percentage of people who

use them was 63% in Serbia, 73% in Argentina, 60%

in South Africa, and 71% in Thailand (Google,

2018). Considering these values and its growing

trend, FCD has a very good opportunity to be used

in developing countries.

FCD collects real-time traffic data by locating

the vehicle via mobile phones or GPS over the road

network (Altintasi et al., 2017). This data is then

processed, to calculate travel time or average speeds

in every road segment. This information is sharing to

users through an online map or mobile phone

applications. For example, Google’s application

combines location data taken from participants’

GPS-equipped mobile phones with a traditional

sensor (Google, 2009). Car location is map-matched,

and speed and direction of travel are sent

anonymously to a central processing centre. Their

aggregated results are shown overlaying road maps

with congestion information through four colour

codes. Traffic information provided by Google is

85% accurate for cars and 71% for motorbikes

(Ahmed, Mehdi, Ngoduy, & Abbas, 2019). This

traffic congestion information is valuable by road

users and road system managers.

Road users can plan the journey and choice the

route, while road system managers view travel time

as an essential network performance indicator (Rose,

2006). Travel time is calculated using the segment

length, the number of intersections in the route,

traffic flow, speed, and traffic density. The last three

ones are the fundamental parameters in traffic

engineering (Garber & Hoel, 2014). This

information helps to identify different traffic states

(congested, free-flow, etc.) and events (i.e., entering

or exiting from a queue/bottleneck, shockwave

propagation, etc.) (Altintasi et al., 2017). Despite the

importance of these fundamental relationships, some

cities in developing nations have not invested in

getting them. One option for those cities is to use

Google traffic information to calculate speed-flow-

density parameters. Google shares aggregate data,

after applying some “noise” (Knoop, Van Erp,

Leclercq, & Hoogendoorn, 2018), and only shares

that information with few institutions in the world

(university, institutes or transportation centre’s) in

their program Better Cities (Eland, 2015). These

institutions belong to developed countries, so it is

difficult to obtain this numerical information for

cities in developing nations.

However, Google codes the numerical

information using four colours and gives it for free

through its platforms (web and mobile app). This

colour-coded traffic (live and typical) is available in

several cities worldwide. Google traffic information

is a result of shared data from more than 2 billion

monthly active users (Matney, 2017). By default, the

user shares their location data by Google's location

service and sent to the Google database for further

processing. It may be stored on the device until it

has an Internet connection. For traffic, users’

information is classified based on speed. It is worth

mentioning that the user can turn it off this option to

avoid sharing his/her location data. In spite of the

growth of smartphone use, Internet access, and the

number of Google active users, it is not known if the

colour-coded traffic indicator is accurate in

developing countries.

The aim of this research is to calibrate equations

for the colour-coded traffic indicators provided by

Google using ground truth data. Data were collected

in urban streets from a medium city (Loja-Ecuador).

As a result, several equations were calibrated y

validated. In order to show this research, the rest of

this paper is structured as follows. Section 2 gives an

overview of the materials and methods. It describes

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

138

the sample size, data collection variables, and

procedure, data processing. Also, it analyses the

relationship between speed data and the colour-

coded condition. Section 3 shows the model

calibration process and validation. Lastly, the final

part presents the principal conclusions.

2 MATERIALS AND METHODS

2.1 Sample Size

Loja, a medium city from Ecuador, was selected for

this study. Ecuador is a developing country located

in South America. Loja has about 215,000

inhabitants (INEC, 2010) and around 50,000

registered vehicles (INEC, 2014). In its urban area,

collection data process included two stages: the

calibration and validation. For calibration it

collected data in 16 urban streets, while, for

validation it collected in 3 urban streets from the

same city (see figure 1). Streets from both data

collection had two lanes and one direction of traffic

circulation, and less than 5% of the longitudinal

slope. Also, streets had a speed limit of 50 km/h.

2.2 Data Collection Variables

Two groups of variables were collected: Google

traffic information and ground truth data. First, from

Google applications (web or mobile app), the

colour-coded was recorded. Four colours are

available: green = no traffic delay, orange = medium

amount of traffic, red = traffic delays and darker red

= the slower speed of traffic on the road (Google,

2019). Also, it was collected when the street was

closure and when the application did not show any

colour. In the ground truth data were collected the

traffic flow and vehicles speeds.

2.3 Data Collection Procedure

Data were recorded from 5 January 2019 to 18

January 2019 in the 19 selected streets. Colour-

coded was collected manually during the daytime

(06h00 to 22h00). It was selected this time range due

to the typical traffic information in Google for this

city is between those hours. This range also avoids

the noise that occurs in low flows that are in the

night (Knoop et al., 2018). Traffic flow and vehicle

speeds were collected manually in situ in the middle

of the street.

Figure 1: Map of the downtown of Loja city (Ecuador)

with the studied streets.

Traffic flow was estimated with the collected

vehicles in a time interval (mostly 10 minutes). The

vehicle speed was estimated from two marks on the

pavement (usually 2 meters) and with the time that

the vehicle spent passing that marks. All data

collection was performed under good weather

conditions.

2.4 Data Processing

Speeds of every vehicle were estimated using the

ground truth data. It calculated the average speed

and traffic flow every 10 minutes. It selected this

period time due the Google typical traffic

information is given in that range. Density was

estimated using calculated speed and flow. The

colour-coded was related to those parameters, and

their results are shown in a section later.

2.5 Speed Analysis

Google only presents traffic conditions as colour-

coded. Exactly it cannot be said what parameters

considered in their calculation or what are their

thresholds. Speed behaviour patterns were explored

using several boxplots, plotting the ground truth

speed and the colour-coded traffic indicator, as can

be seen in figure 2.

Figure 2: Boxplots of average ground truth speed clustered

by several traffic indicators from Google. RC: road

closure, NTI: no traffic information, NTD: no traffic

delay, MAT: medium amount of traffic, TD: traffic delays,

TSSTR: the slower speed of traffic on the road.

Developing a Trafﬁc Congestion Model based on Google Trafﬁc Data: A Case Study in Ecuador

139

Google does not provide any colour when it does

not have enough information, or the street is closure;

however, in-situ vehicles were circulating in those

indicators. So, boxplots also included the indicator

of no traffic information and road closure sign. This

leaves doubts about the reliability of Google traffic

information in this city. However, another study

found that Google data can be used for analysing

traffic management scenarios and for informing and

signalling users on the road, after comparing in-situ

speed and Google information in The Netherlands

(van den Haak et al., 2018).

Consequently, figure 2 shows that the speed

decreases from "no traffic delay" (green colour) to

"the slower speed of traffic" (darker red). However,

boxplots have a wide range of speeds and the same

speed is found in another boxplot. For example, 40

km/h is found in green colour (no traffic delay),

orange colour (medium amount of traffic), and red

colour (traffic delays). This particularity makes it

difficult to calibrate equations, as there are no

unique data for each condition. Thus, it analysed the

speed-flow-density relationships with the colour-

coded indicators. Then, in the next section, its results

are shown.

A cluster analysis was performed to get a better

understand of Google traffic indicators. Ground truth

speed and colour-coded conditions were used in this

analysis employing statistical Software R (R Core

Team, 2013). Google offers four colours, so in the

first analysis was assumed four clusters (>83.5% of

similarity) with average linkage and Euclidean

distance. It shows its results in table 1.

Table 1: Cluster analysis results between ground truth

speeds and colour-coded traffic indicators.

Four clusters analysis

Cluster

Number

of obs.

Similarity

(%)

Average

speed

(km/h)

Speed

thresholds

(km/h)*

1 158 83.5 44.33 >38

2 330 87.5 31.06 26-38

3 929 89.1 20.70 16-26

4 189 89.8 11.38 <16

Six clusters analysis

1 21 91.9 51.65 >47

2 137 90.8 43.21 39-47

3 100 93.9 35.37 32-39

4 230 92.4 29.19 25-32

5 929 89.1 20.70 16-25

6 189 89.8 11.38 <16

*Adding or resting half of the average speed difference

among clusters.

An alternative cluster analysis was added to the

table 1, considering six clusters (>89.1% of

similarity) in analogy to the six levels of service

(LOS) from the Highway Capacity Manual for urban

streets (TRB, 2010) (see table 2). Also, it used the

average linkage and Euclidean distance as

parameters for the analysis.

In table 2, clusters from 1 (green = no traffic

delay) to 4 or 6 (darker red=the slower speed of

traffic on the road). Speed thresholds are calculated

adding or resting half or the speed difference among

the clusters. For example, the speed difference

between cluster 4 and 5 is 8.49 km/h, so half of this

is 4.25 km/h, and then lower thresholds will be

29.19-4.25 = 24.9 ≈ 25 km/h. The upper threshold

will be calculated using 3 and 4 cluster speeds.

According to table 2, streets in this study should be

classified as class III or IV, because their speed limit

is 50 km/h. According to the four cluster analysis

from table 1, the four average speeds do not fit in III

or IV class. If the speed thresholds from Table 3 are

rearranged, data could fit in class III, for example, A

and B (green), C and D (orange), E (red), F (darker

red). However, with six clusters, almost every value

matches with the thresholds of urban street class III.

In this way, Google could offer traffic information in

terms of the level of service. Also, it would solve the

problem that it found one speed on several levels or

colour codes, so it can be used in the practice.

3 RESULTS

Although some traffic indicators from Google have a

trend with speed, and it has some relationship with

the level of service, there is not clear how this can be

used to build a traffic congestion mode from Google.

Table 2: Level of service (LOS) of urban streets.

Urban street

class

I II III IV

Speed*

(km/h)

90-70 70-55 55-50 55-40

FFS**

(km/h)

80 65 50 45

LOS Average travel speed (km/h)

A > 72 > 59 > 50 > 41

B > 56-72 > 46-59 > 39-50 > 32-41

C > 40-56 > 33-46 > 28-39 > 23-32

D > 32-40 > 26-33 > 22-28 > 18-23

E > 26-32 > 21-26 > 17-22 > 14-18

F ≤ 26 ≤ 21 ≤ 17 ≤ 14

* Range of free-flow speed, ** Typical FFS.

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

140

Therefore, other analyses were conducted to

calibrate equations from speed-density-flow

variables. After calibrated them, a validation process

was performed to evaluate the quality of the

developed equations. Those equations will help to

build the traffic congestion model.

3.1 Calibration of Equations

Figure 3 plotted density and flow data clustering by

Google traffic indicators. Also, a power trend line

was also plotted in that figure; in order to compare

R-squared with other trend lines. The power curve

was chosen due to its consistency when density is

zero flow is also zero. Table 3 shows the power

curve equations.

The regular shape of the curve of density-flow

relationship is an inverted U. In these cases, the data

only covers the first part of the curve. When the flow

gets higher, also, density gets higher until an

inflection point, where the flow starts decreasing

when density continues growing. The trends in

figure 3 are not close to the inflection point. It is

interesting that data slope is getting flattered when is

more congested (NDT, MAT, TD, and TSSTR).

This trend is consistent with the theory of

fundamental diagrams because when it is more

congested, adding more vehicles will increase the

density more slowly than traffic without delay.

NDT, MAT and TD have similar flow data until the

density of 20 veh/km. It is also interesting that the

highest density in every figure (NDT, MAT, TD,

and TSSTR) increases approximately from 20 in 20:

green colour is up to 40 veh/km, orange colour is up

to 60 veh/km, red colour is up to 80 veh/km and

darker red is up to 100 veh/km. The curves in RC

and NTI have similar shape than the others even

when there is no traffic information or it has the road

closure sign.

Secondly, figure 4 plotted density and speed data

clustering using the same Google traffic indicators.

Furthermore, a linear trend line was plotted

according to the fundamental diagram theory. Table

3 also shows these linear equations. The data trends

from figure 4 are consistent with the fundamental

diagram. Most conditions (NTI, NTD, MAT, TD,

and TSSTR) have higher R-squared with

exponential or power trend line. However, there is a

straight line used in the flow-density relationship, so

it selected that regression. The trend line in each

indicator has a different slope than the others,

similar in figure 3.

Models from table 3 are applicable in the showed

range. Equations 8 and 9 have similar parameters, so

only one equation can be calibrated. However, in

this investigation, the models have been left in their

original version, to see the traffic indicators

separately. In general, R-squared from density-flow

equations is bigger than density-speed models.

3.2 Validation of Calibrated Equations

A validation process was performed to evaluate the

quality of the calibrated models from table 3. For

this validation, it was collected data from three

streets in the same city. These streets had similar

characteristics to the ones in the calibration process.

Collecting data and data processing was the same

than in the calibration process.

Figure 3: Density and flow data clustered by several traffic

indicators from Google. RC: road closure, NTI: no traffic

information, NTD: no traffic delay, MAT: medium

amount of traffic, TD: traffic delays, TSSTR: the slower

speed of traffic on the road.

Figure 4: Density and speed data clustered by several

traffic conditions from Google. RC: road closure, NTI: no

traffic information, NTD: no traffic delay, MAT: medium

amount of traffic, TD: traffic delays, TSSTR: the slower

speed of traffic on the road.

Developing a Trafﬁc Congestion Model based on Google Trafﬁc Data: A Case Study in Ecuador

141

Table 3: Calibrated equations for density/flow and

speed/density for several Google traffic indicators.

Traffic

indicator

Colour-

coded

Calibrated

equation

RC None q = 27,25k

0.87

0,83 (1)

NTI None q = 39,46k

0.85

0,94 (2)

NTD Green q = 42,80k

0.77

0,88 (3)

MAT Orange q = 70,65k

0.58

0,64 (4)

TD Red q = 65,36k

0.57

0,71 (5)

TSSTR Darker red q = 184,14k

0.27

0,29 (6)

RC None s = -0,33k + 24,77 0,20 (7)

NTI None s = -0.82k + 38,20 0,45 (8)

NTD Green s = -0,87k + 36,36 0,41 (9)

MAT Orange s = -0,57k + 32,91 0,42 (10)

TD Red s = -0,33k + 26,98 0,57 (11)

TSSTR Darker red s = -0,19k + 21,71 0,65 (12)

q = traffic flow (veh/h), k = traffic density (veh/km), s

= average speed (km/h), RC: road closure, NTI: no

traffic information, NTD: no traffic delay, MAT:

medium amount of traffic, TD: traffic delays, TSSTR:

the slower speed of traffic on the road.

The prediction errors were calculated to validate

the previous calibrated speed models. Those errors

were: mean absolute error (MAE) and mean absolute

percentage error (MAPE) (see table 4). An analysis

of variance (ANOVA) was carried out to validate

the models, determining whether the difference

between predicted values (equations) and collected

data from validation means are statistically

significant. Those values should not differ in a 95%

level of confidence. It shows in table 4 the predicted

errors and ANOVA results.

Table 4: Calibrated equations for density/flow and

speed/density for several Google traffic indicators.

# MAE*

MAPE

(%)

ANOVA

95% CI P value

(1) - - - -

(2) - - - -

(3) 2.05 21.17 (8.67; 13.34) 0.138

(4) 4.19 27.01 (14.65; 18.39) 0.057

(5) 7.19 31.68 (21.56; 24.85) 0.050

(6) - - - -

(7) - - - -

(8) - - - -

(9) 5.61 27.68 (20.69; 24.01) 0.051

(10) 4.57 22.41 (20.18; 23.05) 0.011

(11) 2.92 17.11 (18.07; 19.83) 0.409

(12) - - - -

- Not enough data to validate models, MAE = mean

absolute error, MAPE = mean absolute percentage

error, 95% CI= confidence interval, * In equations

3-5 MAE is in veh/h and in equations 9-11 is in

km/h.

Table 4 does not have prediction errors or

ANOVA analysis for equations 1, 2, 6-8 and 12;

because the collected data from the validation

process were not enough to do it. The highest

density error was 7.19 veh/h, while the highest speed

error was 5.61 km/h. Predicted error was away

31.68% and 27.68% from the calibrated values.

Despite these high values, the p-value exceeds from

the assumed level of significance (α=0.05) in almost

all equations. This means that the average predicted

values do not differ from the collected validation

ones; in consequence, those models are valid.

However, caution is suggested in equations 5 and 9,

because they are close to that level of significance.

4 CONCLUSIONS

The aim of this article was to calibrate equations for

the colour-coded traffic indicators provided by

Google using ground truth data. After analysing the

results, it presents the following conclusions:

Colour-coded from Google have a reasonable

trend with the ground truth speeds. However, their

data dispersion makes it difficult to calibrate

equations. Therefore, a new analysis was conducted

with variables from fundamental diagrams (speed-

flow-density) and the colour-code traffic indicators.

In the density-flow analysis, data were consistent

with the theory of traffic engineering and equations

were calibrated using the power curve. Data were

also consistent in the density-speed analysis, and

calibrated some linear models. The density-speed

models have lower R-squared values than the

density-flow models, so it recommended taking

caution when using. Those models were validated

using prediction errors and ANOVA analysis.

After the cluster analysis of average speed

ground truth and traffic indicators from Google,

their relationships are unclear. The HCM has 6 LOS,

and Google offers 4 levels (four colours). However,

if it is divided the speed data into six levels, Google

could offer the information in terms of the level of

service, considering the average speed thresholds

approximately fits in the urban street LOS. An

advantage of this arrangement is that a speed range

will belong to a particular LOS and therefore to a

single colour. In contrast, in Google traffic

information, the same speed range belongs to several

colour-coded indicators. This information will be

helpful for cities that want to build a low-cost traffic

congestion model.

This study has a number of limitations. First, it

performed collection data in just one city, which

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

142

probably will not have the same urban environments

than in others. Also, the urban streets have a speed

limit of 50 km/h, have two lanes, one direction, and

are flat. Additionally, this study starts from the

assumption that the data in the middle of the tangent

belong to the whole street, while other elements

should consider when approaching or exiting from

the intersection. Furthermore, the calibrated

equations are valid in a specific range, so they

should not use out of those ranges.

Despite these limitations, the present study helps

to understand the use of Google traffic indicators in

urban streets, offering useful information for urban

planners and street designers. It showed the

relationship between LOS and the average speed

ground truth. It showed that when Google does not

provide colour or in a road closure sign, real traffic

was circulating through those streets. Also, based on

the growth of smartphone use, Internet access, and

the number of Google active users, the calibrated

equations can be used by other cities to create their

traffic model. This methodology could employ in

other places or help to develop ITS.

ACKNOWLEDGEMENTS

The author acknowledges the support of

SENESCYT and Universidad Técnica Particular de

Loja, and some students that helped collecting data.

REFERENCES

Ackaah, W. (2019). Exploring the use of advanced traffic

information system to manage traffic congestion in

developing countries. Scientific African, 4(e00079).

https://doi.org/10.1016/j.sciaf.2019.e00079

Ahmed, A., Mehdi, M. R., Ngoduy, D., & Abbas, M.

(2019). Evaluation of accuracy of advanced traveller

information and commuter behaviour in a developing

country. Travel Behaviour and Society, 15, 63–73.

https://doi.org/10.1016/j.tbs.2018.12.003

Altintasi, O., Tuydes-Yaman, H., & Tuncay, K. (2017).

Detection of urban traffic patterns from Floating Car

Data (FCD). In 19th EURO Working Group on

Transportation Meeting (pp. 382–391). Istanbul,

Turkey: Transportation Research Procedia.

https://doi.org/10.1016/j.trpro.2017.03.057

Andersson, A., Hiselius, L. W., & Adell, E. (2018).

Promoting sustainable travel behaviour through the

use of smartphone applications: A review and

development of a conceptual model. Travel Behaviour

and Society, 11, 52–61. https://doi.org/

10.1016/j.tbs.2017.12.008

Bacon, J., Bejan, A. I., Beresford, A. R., Evans, D.,

Gibbens, R. J., & Moody, K. (2011). Using real-time

road traffic data to evaluate congestion. In Dependable

and Historic Computing. Lecture Notes in Computer

Science (Vol. 6875, pp. 93–117). Berlin, Heidelberg:

Springer. https://doi.org/10.1007/978-3-642-24541-

1_9

Bagloee, S. A., Ceder, A., & Bozic, C. (2014).

Effectiveness of route traffic information in

developing countries using conventional discrete

choice and neural-network models. Journal of

Advanced Transportation, 48(6), 486–506.

https://doi.org/10.1002/atr.1198

Baratian-Ghorghi, F., & Zhou, H. (2015). Investigating

Women’s and Men’s Propensity to Use Traffic

Information in a Developing Country. Transportation

in Developing Economies, 1(1), 11–19.

https://doi.org/10.1007/s40890-015-0002-5

Eland, A. (2015). Tackling Urban Mobility with

Technology. Retrieved June 6, 2019, from

https://europe.googleblog.com/2015/11/tackling-

urban-mobility-with-technology.html

Garber, N. J., & Hoel, L. A. (2014). Traffic and Highway

Engineering. Stamford (USA): Cengage Learning.

Google. (2009). Arterial traffic available on Google.

Retrieved September 20, 2019, from http://google-

latlong.blogspot.com/2009/08/arterial-traffic-

available-on-google.html

Google. (2018). Consumer Barometer. Retrieved

November 12, 2019, from

https://www.consumerbarometer.com/en/

Google. (2019). View places, traffic, terrain, biking, and

transit - Computer - Google Maps Help. Retrieved

June 20, 2019, from https://support.google.com/maps/

answer/3092439?hl=en&visit_id=1-

636415200196506079-2363811973&rd=2

Gunawan, F. E., & Chandra, F. Y. (2014). Optimal

Averaging Time for Predicting Traffic Velocity Using

Floating Car Data Technique for Advanced Traveler

Information System. In The 9th International

Conference on Traffic & Transportation Studies (Vol.

138, pp. 566–575). Procedia Social and Behavioral

Sciences. https://doi.org/10.1016/j.sbspro.2014.07.240

Hall, R. W. (1996). Route choice and advanced traveler

information systems on a capacitated and dynamic

network. Transportation Research Part C: Emerging

Technologies, 4(5), 289–306. https://doi.org/10.1016/

S0968-090X(97)82902-6

INEC. (2010). Censo de Población y Vivienda en el

Ecuador, Fascículo provincial Loja.

INEC. (2014). Vehículos Matriculados – Serie Histórica

2008-2014. Retrieved June 25, 2019, from

http://www.ecuadorencifras.gob.ec/vehiculos-

matriculados-serie-historica-2008-2014/

Jain, S., Jain, S. S., & Jain, G. (2017). Traffic Congestion

Modelling Based on Origin and Destination. In 10th

International Scientific Conference Transbaltica 2017:

Transportation Science and Technology (Vol. 187, pp.

442–450). Elsevier. https://doi.org/10.1016/

j.proeng.2017.04.398

Developing a Trafﬁc Congestion Model based on Google Trafﬁc Data: A Case Study in Ecuador

143

Jiang, D. (2019). The construction of smart city

information system based on the internet of things and

cloud computing. Computer Communications, 147,

21. https://doi.org/10.1016/j.comcom.2019.10.035

Knoop, V. L., Van Erp, P. B. C., Leclercq, L., &

Hoogendoorn, S. P. (2018). Empirical MFDs using

Google Traffic Data. In IEEE Conference on

Intelligent Transportation Systems, Proceedings, ITSC

(pp. 3832–3839). IEEE. https://doi.org/10.1109/

ITSC.2018.8570005

Matney, L. (2017). Google has 2 billion users on Android,

500M on Google Photos | TechCrunch. Retrieved

November 12, 2019, from https://techcrunch.com/

2017/05/17/google-has-2-billion-users-on-android-

500m-on-google-photos/

Mfenjou, M. L., Abba Ari, A. A., Abdou, W., Spies, F., &

Kolyang. (2018). Methodology and trends for an

intelligent transport system in developing countries.

Sustainable Computing: Informatics and Systems, 19,

96–111. https://doi.org/10.1016/j.suscom.2018.08.002

R Core Team. (2013). R: A language and environment for

statistical computing. R Foundation for Statistical

Computing. Vienna, Austria. Retrieved from

http://www.r-project.org/

Reza, A., & Kermanshah, M. (2005). Traffic information

use modeling in the context of a developing country.

Periodica Polytechnica Transportation Engineering,

33(1–2), 125–137.

Rose, G. (2006). Mobile Phones as Traffic Probes:

Practices, Prospects and Issues. Transport Reviews,

26(3), 275–291. https://doi.org/10.1080/

01441640500361108

Singh, G., Bansal, D., & Sofat, S. (2014). Intelligent

Transportation System for Developing Countries-A

Survey. International Journal of Computer

Applications, 85(3), 34–38. https://doi.org/10.5120/

14824-3058

TRB. (2010). Highway Capacity Manual (HCM2010).

Washington D.C (USA): Transportation Research

Board of the National Academies.

van den Haak, P., Bakri, T., Van Katwijk, R., Emde, M.,

Agricola, N., & Snelder, M. (2018). Validation of

Google Floating Car Data for Applications in Traffic

Management. In Transportation Research Board 97th

Annual Meeting. Washington DC, United States.

World Bank. (2019a). World Bank Country and Lending

Groups – World Bank Data Help Desk. Retrieved

October 11, 2019, from https://datahelpdesk.

worldbank.org/knowledgebase/articles/906519-world-

bank-country-and-lending-groups

World Bank. (2019b). World Bank Open Data | Data.

Retrieved November 11, 2019, from https://datos.

bancomundial.org/

Yokota, T. (2004). ITS for Developing Countries ITS for

Developing Countries. Retrieved from

http://documents.worldbank.org/curated/en/90253146

8313538337/pdf/356770ITS0deve1ntries0Note101PU

BLC1.pdf

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

144