Spatial Models and Neural Network for Identifying Sustainable Transportation Projects with Study Case in Querétaro, an Intermediate Mexican City

Abstract

The construction of urban and transport indicators aims for a better diagnosis that enables technical and precise decision-making for the public administration or private investment. Therefore, it is common to make comparisons and observe which has better diagnosis results in a diversity of indexes and models. The present study made a comparative analysis of spatial models using artificial intelligence to estimate transport demand. To achieve this goal, the audit field was recollected in specific urban corridors to measure the indicators. A study case in Querétaro, an emergent city in the Mexican region known as El Bajío, is conducted. Two similar urban avenues in width and length and close to each other were selected to apply a group of spatial models, evaluating the avenues by segments and predicting the public transport demand. The resulting database was analyzed using Artificial Neural Networks. It displays specific indicators that have around 80% of correlations. The results facilitate the localization of the avenue segments with the most volume of activity, supporting interventions in urban renewal and sustainable transportation projects.

1. Introduction

Cities are complex systemic phenomena [1]. They contain a significant amount of information that must be analyzed from different perspectives, tools, and processes [2]. This increasing volume of data needs to consider the sustainable development standard tripartite system, such as the social, economic, and environmental variables, and its evolution into the Sustainable Development Goals (SDG) [3,4]. Specifically, Borowy brings to attention the word need, in the famous phrase from the Brundtland Commission: development that meets the need of the present without compromising the ability of future generations to meet their own needs. In reality, the word need is oriented to the needs of the poor people worldwide. Transport is a critical need of the population at lower and medium lower socio-economic levels.

The developing countries face multiple ongoing problems, adding to this amount of work. Specifically, Mexican cities have, extensively and historically, negative results in their construction: low densities, low or null connectivity, and dispersion [5,6]. This kind of development generates a negative cycle in which the cities grow until it is complicated to be sustainable. All of this adds to changes in the land use in vast territories, among other problems. The majority of these changes are from agricultural to urban activities, given the attraction for these investments in the last few decades [7].

Therefore, the need for robust urban analysis is evident. This analysis must be interdisciplinary and include several diagnosis processes to establish technical and practical conclusions, as well as support for decision-making in public administration and private investment [8]. There is an actual need for technological and innovative processes such as deep learning, machine learning, or, in this case, Artificial Neural Networks (ANN). This article compares spatial models for road infrastructure and the public transport demand as an urban phenom, trying to resolve:

• Which spatial model better correlates to the actual public transport demand?

It is essential to mention that the models selected are not constructed to estimate public transport demand but for the general qualification of an urban corridor. That is why we decided to compare them with a natural activity of an urban corridor, such as the public transport demand. The response to the question helps find the prioritization and potential zones for intervention and as a solution to public transport demand. It is important to remember that public transport demand has some recent solutions such as park and ride [9,10], or bike-sharing systems [11], because it promotes a gradual interchange to a more sustainable way of transport.

To obtain urban renewal project resources, the projects need to benefit the most people possible. Unfortunately, not enough data about public transport, or pedestrian counting, are available because of the costs of doing that on many streets. Many studies are not available from official administrations, among other problems [12,13]. Therefore, an indicator that predicts better helps to locate the avenue segments with the most people using them, as [14] confirms for car ownership prediction, or [15] uses for predicting demand of mobility in e-scooters, or [16] use for traffic forecasting. This article confirms the forecast for the public transport demand.

As a first exercise process, Section 2, there is a review of the selected urban indicators and information about the city and the chosen avenues for the analysis. Then, in Section 3, the utilization of the Public Space Accessibility Tool (PSAT) considers social, economic, and urban variables. There are two more indicators from Salvador Rueda’s planning concept Urban Complexity defined for an urban city plan [17]. The three spatial models also have sub-indicators that will be measured against each other with the public transport demand. A passenger count in peak hours is carried out to display the current demand.

Afterward, a second process is conducted, with the results of the three indicators and the public transport count, which are compared and valued using an Artificial Neural Network (ANN), in Section 4. The comparison results are displayed in Section 5, where the PSAT is one of the indicators with the most correlation with the public transport demand. Then, Section 6, there is a discussion about the prioritization of urban renewal projects in the Querétaro Metropolitan Area (MA) and the application. Finally, Section 7, a conclusion with a summary and future work.

2. Materials: Selected Spatial Models and Area of Analysis

Developing indicators to measure access is a common practice, favoring specific sectors [18]. The PSAT is a spatial model based on recommendations of variables by some authors, applied in Mexican cities with official data [19].

In 2010, Salvador Rueda (2010) established more than 130 variables in 50 urban indicators and implemented them in the Vitória-Gasteiz Plan. Rueda uses a reticular mesh to store the information, although he prefers to use various formulations and techniques applied to the street segments, with existing statistical information from Vitoria-Gasteiz [17]. Similar works confirm that the list of variables is valid [20,21], and it is used by international consultancy agencies such as the Institute for Transportation and Development Policy [22]. Most of the variables are also used for urban and transport planning and design, as [23].

A process diagram was made for better comprehension of the exercise in Figure 1.

2.1. Selected City of Analysis

Querétaro, MA, is part of the emergent metropolitan areas of the country and is the eighth in size. It is part of the vibrant and well-located cities within El Bajío, a centric area of Mexico, with potential and conditions for great attraction of investments, calling for urban growth. Querétaro municipality alone, grew approximately ten times its size, in forty years, from 1985 to 2015 [24], adding to the urban phenom occurring in the majority of cities around the world [25].

Similar to the Mexican metropolitan areas, most of the renewal projects made by public administrations of the Querétaro, MA, are oriented toward motorized private mobility [26]. This diminishes other sectors such as public transport or walking, which are more significant in population. The last massive poll for the city mobility study has evidence: 60% of the people are moving sustainably: active mobility (walking 22%, cycling 1%) or collective mobility (public transport 33%, private transport 3%). When mixed with the socio-economic levels, the results are more revealing. People who move sustainably, are between 60% and 70% of the middle to lower levels [27].

One of the problems that result from extensive and dispersed city construction, is how to manage the prioritization of urban renewal projects, given the extension of the street network. In the case of Querétaro municipality, the road network has grown immensely during the last decade. The official data from Instituto Nacional de Estadística y Geografía (INEGI) provides an approximate extension of 3000 km of roads.

Given that one public administration of three years has the resources to build up to 300 km of roads [28], it needs ten administrations or the equivalent of 30 years, to do urban renewal projects in all the areas of the city. This is one of the recurrent problems in polls about the city and its competitiveness. Society wants better streets, with less incidence, more illumination, transit reduction, and others [29].

2.2. Selected Area and Road Axes

The selected roads correspond to a basic premise: to have similar width and length characteristics and show a differentiation concerning the infrastructure for public transport. Since there are already some rapid transit (BRT) bus lines in the Querétaro, MA, it is preferable to start from an already built axis with an acceptable adaptation time. Avenida de la Luz (AL) was chosen because it has been in operation for three years. It has one of the first structural axes in the Querétaro, MA: established stops, incorporation of confined lanes, and even a track cycle lane.

Figure 2 displays a section of Av. De la Luz, made using the streetmix web application. It can be seen that even if the space dedicated to the pedestrian is significant in both cases, most of it is located in the median strip, away from the activities. The street should be reconfigured to get closer to the standard of the indicator.

The other urban axis selected is Av. Pirineos (AP), which does not have this specialization for sustainable mobility, appearing to be an avenue with classic engineering design and outdated regulations for sustainable mobility. The section of Figure 3 exposes that, but with little more space for the pedestrian. Nevertheless, there is already an urban renewal project on this avenue [30]. Prior to that, there were reports of the Town Planning Thematic Advice, consulting about priority projects, and the AP was one [31]. This serves as an example of analysis and comparison between both avenues with different levels of needs.

The AL roadway is classified as a zone boulevard, while the AP roadway is referred to as zone avenue. This is according to the latest mobility study conducted in the municipal government [27], which uses a classification methodology similar to the federal government, represented in the Manual de Calles, Diseño Vial para las Ciudades Mexicanas [32].

3. Method: Phase A. Implementation of Selected Indicators

The implementation consists of assigning a value to each street segment in both urban avenues. Each indicator has a value. A data-sheet was constructed to compare the indicators and the public transport demand.

3.1. Public Space Accessibility Tool

The first indicator, the PSAT, is implemented using 14 variables that put qualifications in every street segment. The majority of the variables are related to the walkability index by [33], as road incidence, infrastructure for the disabled, and also the habitability variables proposed in the neighborhood design research proposed by [34], as the diversity of transport modes, block length, street network grid, and mixed land use. Other variables were considered as proximity to housing, jobs, schools, commerce, presence of trees, urban lights, road signals, and sidewalks. The data are available by INEGI for all the variables [35,36]. This indicator is run for the entire road network in Querétaro, and the analysis went down to the two selected avenues. Recent results by [37], provide an implementation of the PSAT on a bigger scale, with an extensive amount of the public transport data for the correlation. This paper represents an approach of a shorter scale in order to probe the effectiveness with less taken data. This is because using public transport data results in a high cost in time and resources.

It is important to also refer to exercises such as the one from [38], about the influences of the built environment and the volume of pedestrian activity. In addition, one of the objectives of this group is the support of intervention projects.

Figure 4 locates the area of the selected urban axes, with the results of the indicator by street segment. The streets in AL have a better rating compared to the streets of AP. It also displays the characteristics of the average street on both roads, including the used variables. Lastly, the street segments that have the highest values are located.

3.2. Indicator of Spatial and Functional Continuity of Street Corridor

Rueda defines this indicator as “The degree of interaction of the spatial sequences through the density of activities per stretch of the street”. Therefore, the objective of this indicator is to confirm two characteristics: The diversity of opportunities that a city should offer along its streets and the distribution of public space, which is favorable for pedestrian mobility. We define these two characteristics as sub-indicators. They are reviewed separately to unite the conceptions and define an overall result.

The first sub-indicator, named range of activities, measures the number of non-residential activities for every 100 m of the street. Here, Rueda establishes some ranges:

• more than ten activities per 100 m is considered optimal;
• between five and ten activities is considered intermediate;
• between two and five activities is considered low;
• less than two activities is considered very low;
• without activities is considered null.

Every kind of land use, except housing, is selected for the activities. Therefore, it includes all economic units data provided officially by the Directorio Estadístico Nacional de Unidades Económicas (DENUE) used in its latest version (2020). It is anticipated that the AP axis has very few segments with significant activity density, while the AL axis has much more activity along the avenue. The process of dividing activities is performed every 100 m.

The AP axis shows null activity. Along the axis, a zone with little activity is observed. This range is low, and this is explained by the fact that the axis has many enclosed housing complexes, and most of them turn their backs to the avenue. It is also noted that there are vacant lots along much of the axis, so there is an opportunity to reverse the situation. On the other hand, the AL axis has more segments with better ranges, some at the intermediate level and others at the optimal level, while there are few street segments where the ranges are low, very low, or zero.

The second sub-indicator, is named Public Space Distribution, concerning the percentage of area distribution favorable to the pedestrian. Rueda proposes only two ranges:

• 75% or more of pedestrian-oriented public space is considered optimal;
• 74% or less is considered flawed.

Unlike the first characteristic, where five ranges were shown, the distribution of public space is a basic and unjustifiable categorization. In addition, it is proposed to establish more ranges in order to obtain a greater diversity of categorizations to improve the analysis. So then:

• 50% to 74% is considered intermediate;
• 40% to 50% is considered low;
• 30% to 40% is considered very low;
• less than 30% is considered null.

Both avenues are divided into segmented polygons to verify the space distribution along the section width and length, adding a dimension to the analysis. The polygons are surveyed using maps from Google Earth from 2020 and then processed with the Quantum Geographic Information System to obtain areas of each polygon, and thus the spatial distribution, for pedestrians and other modes of mobility: by bicycle, by public transport, and by private transport.

Table 1. Summary of public space distribution dedicated to mobility in Av. De la Luz.

Land Use	Total Area	Percentage
Sidewalks	11,839	8.12
Median strips	37,960	26.04
Cycle lanes	10,193	6.99
Road lanes	59,596	40.88
Public Transport lanes	26,179	17.95
Total	146,319	100

shows the summary per area of the AL axis. The space for pedestrians is the sum of the sidewalks and the median strips, representing 34.16%. On the other hand,

Table 2. Summary of public space distribution dedicated to mobility in Av. Pirineos.

Land Use	Total Area	Percentage
Sidewalks	11,115	10.69
Median strips	34,185	32.88
Cycle lanes	0	0
Road lanes	58,103	55.89
Public Transport lanes	0	0
Waterway channel	553	0.53
Total	103,956	100

displays the AP axis, and it can be seen that there is more space for pedestrians with 43.57%. The range has a difference of almost ten percentage points, which is considered very low for the avenue. This also locates where some segments on both axes stand out. They achieve a greater pedestrian area, adding percentages more significant than 50% to be graded in an intermediate range.

Finally, both characteristics are combined to obtain the final result for the spatial and functional Continuity of Corridor (Cco) street indicator.

Figure 5 displays that, despite the best rating of the distribution of public space favorable to pedestrians on the AP axis, it is on the AL axis that there is a much more significant presence of activities along the study boundary. On the other hand, the AP axis has only a tiny group of segments, where the activity level is similar to that prevailing on the AL axis. There are some plot lands in the western part of AP, which opens up the opportunity to improve the rank on the indicator.

The results of the indicator are harmful to both roads, although with particular areas of opportunity:

• on the AL axis, there is already a good density of activities in much of the corridor;
• the street should be reconfigured to raise the rating and thus obtain the optimal range of the indicator.

3.3. Indicator of Balance between Residence and Activities

This indicator aims to locate the city areas, where the necessity of mobility is reduced, because of the mix of uses, including the living. This is similar to the recent proposal of urban renewal policies as the city of the fifteen minutes [39]. This also enables the Jane Jacobs objective of eyes on the street [40]. To achieve the latter, it is required to do extensive research about typologies of the diversity of uses and activities in a determined neighborhood or region. This is in order to integrate the mixing better.

This indicator is about the m² of tertiary construction areas, such as commercial, retail, offices, light industry, and even public buildings. The recommendation of Rueda is about an area. Given that the comparative up to this point is about the axis, the area summary is made on the buildings in front of the avenue. This area is then divided between the number of houses on the avenue.

The construction footprint of each building was drawn on the corridor to obtain the total area of activities. Then, the buildings for residential use are recognized and marked, and then the dealing with commerce, services, and other activities, is also recognized and added to the total area of the building. Both housing and tertiary activities data were obtained by INEGI [35,36].

There is much more housing than on the AL road, although much of it is not accessible on the road, but represents an enclosed system, which is why its use is discussed.

3.4. Public Transport Demand and Pedestrian Flow

As seen on the ascents and descents, the public transport demand is part of the public space activity and is used as an independent variable for this process. These data are obtained with field works in both avenues. Therefore, a series of information counting was carried out in the field, with capacity at peak times at public transport stops along the two roads. There are two hours of maximum demand, 7:00 a.m. to 8:00 a.m. and 6:00 p.m. to 7:00 p.m., during which capacity is used to quantify ascents and descents at each stop established and/or found during previous tours.

Figure 6 displays at least two results:

• there is a big difference on both avenues;
• there is a central segment with the most demand in AL.

Visually, there is a similarity between the indicators and the public transport demand. A technical comparative for confirming this is next in the process.

4. Method: Phase B Comparative Analysis Using Neural Network

In this section, an ANN is used to identify and estimate the demand for public transport based on the indicators mentioned in this work. A backpropagation (BP) ANN was used because it is the most extensive network with well-known advantages and disadvantages [41,42]. The training and development of the neural networks were carried out in Matlab. Urban indicators are calculated using algebraic equations, so the computation time is negligible. In the case of ANNs, the execution time is also insignificant. However, the training time depends on the number of data and the selected architecture. Due to the amount of data, the training and processing time ranges from 1 to 30 s. The urban indicators were used as inputs in the neural network, and the output was the demand for public transport. This allows inferring the relationship of each indicator with the demand for transport through the ANN.

The original data are normalized and filtered to obtain a helpful database for training the network. This process makes it possible to sensitize the algorithm to the maximum and minimum values in each variable. In addition to comparing the different accessibility indices, the function of the ANN is to estimate the demand for public transport in segments where there are no public transport stops. For this purpose, the segments that do not contain these stops were filtered to determine the demand for public transport in each segment. The first step is the selection of the network architecture. The approach used in this research is Supervised Machine Learning as it requires less data than a deep approach. The choice was made by selecting the inputs and outputs and making multiple performance tests that mainly vary in three main parameters: the transfer functions, the number of hidden layers, and the number of neurons in each layer. A neuron generally contains an input p, a final output a, as shown in Figure 7. On the other hand, w represents the weights, and b is a gain that reinforces the output of the adder n. Therefore, n is the transfer function input that adjusts the value of the adder to the final output and is chosen depending on the problem specifications that the neuron has to solve.

To obtain an objective comparison, different indicators should be compared under the same network architecture. Therefore, for the architecture selection process, all the indicators are selected as input and the variable of demand for public transport as output. On the other hand, the transfer functions used are the three most common functions in BP ANNs. The functions are listed below and are described by Equations (1)–(3).

• Tangential Sigmoid Function ($TSF$) (see Equation (1));
• Logistic Sigmoid Function ($LSF$) (see Equation (2));
• Linear Transfer Function ($LTF$) (see Equation (3)).

$$TSF=\frac{1-e^{-2t}}{1+e^{-2n}}$$

(1)

$$LSF=\frac{1}{1+e^{-2n}}$$

(2)

$$LTF=n$$

(3)

The $LSF$ is widely used in many complex system processes and shows a low temporal progression from initial stages to maximum levels after a time. The $TSF$ behaves similarly to the $LSF$ but allows exploring negative values. Finally, the $LTF$ allows obtaining a linear and proportional output. Although other transfer functions are less used since their behavior is similar, for example, the Saturating Linear Transfer Function ($SLTF$) case, which is similar to $LSF$. Additionally, between one and two hidden layers were tested, and finally, the number of neurons was varied according to the performance shown by the network. The network performance was measured by root–mean-square error ($RMSE$) which is described by Equation (4) and is a typical quality index for characterization of the performance [43,44]. The starting point was a minimum number determined by the geometric pyramid rule, and the number of neurons was gradually increased. The most relevant results of the architecture selection are summarized in

Table 3. Summary of the tests for architecture selection.

BP-ANN	Transfer Functions	Neurons	RMSE(%)
1	TSF-LTF	3	30.52
2	TSF-LTF	5	19.52
3	TSF-LTF	8	17.70
4	TSF-LTF	12	31.27
5	TSF-TSF-LTF	[5 3]	14.18
6	TSF-TSF-LTF	[8 4]	29.18
7	LSF-LSF-LTF	[8 4]	25.32
8	TSF-LSF-LTF	[8 4]	23.44
9	TSF-TSF-TTF	[5 3]	15.49

.

$$RMSE=\sqrt{\frac{{\sum }_{i=1}^N{(Predicte{d}_i-Actua{l}_i)}^2}{N}}$$

(4)

The best performing architecture is architecture 5, consisting of two hidden layers with 5 and 3 neurons, respectively. Therefore, subsequent tests were carried out through this network architecture. The selected network architecture is executed using each indicator as input to compare the performance of the indicators. Each network was executed ten times to validate the results of these runs. Subsequently, the average RMSE was calculated for all indicators. However, the accessibility indicator contains data related to public transport. Therefore, the indicator was recalculated, and the original indicator was distinguished with the letter A. On the other hand, the indicator without transport was named B. The results of these tests are exhibited in

Table 4. Summary of the tests for each indicator.

Indicator	BP-ANN	RMSE (%)	R2 (%)	MBE (%)
1	Accessibility indicator A	18.68	66.78	−2.77
2	Accessibility indicator B	20.72	65.83	−1.69
3	Spatial and functional continuity of the corridor street	21.66	56.06	1.76
4	Range of activities	20.60	63.51	−2.96
5	Housing by levels	31.67	1.17	−4.86
6	m2 of tertiary use	27.91	55.25	4.48

.

5. Results

The results of the comparison display positive conclusions in the prediction of the public transport demand. Table 4 is made for better comprehension.

The results show that indicator 1 has a better performance than indicator 2, but the Accessibility indicator B does not have any intrinsic public transport data. Additionally, it was observed that indicators 2, 3, and 4 have a similar performance in estimating the demand for public transport. The last two indicators have a significant increment in their RMSE. Additionally, the R² indicator is added to show the degree of prediction of the models and the mean bias error (MBE) to observe the bias in each model.

It could not be concluded that indicators 2, 3, and 4 presented a better performance due to the closeness of their results. However, these indicators show similar performance for this task. It also observed that the network benefits by using all the indicators as inputs, reducing up to six percentage points of RMSE.

It is worth mentioning that the model with less RMSE is one of the sub-indicators, as the 3: Range of activities is part of the 4: Spatial and functional continuity of the corridor street. Since this indicator reflects the number of activities in the street, it will need the testing on a bigger scale and with a more diversified number of avenues or in an entire road network.

Nevertheless, the PSAT model is the only indicator that was implemented in all the Querétaro street network. Since the results do not heavily influence the intrinsic data of the public transport variable, Indicator A is implemented.

Figure 8 displays the same study area, consisting of the two urban corridors, AL and AP, with all the street segments around, qualified by the PSAT tool. For better visualization, the figure displays only the segments with the better qualification ranges in blue color.

AL has an important priority above AP, compared to the reduced area of AP that merits intervention. This supports a change of strategy for urban renewal projects related to the public transport demand instead of the actual public administration projects commented on in the introduction [30].

Given the results, the PSAT is used to compare the other avenues that have a Bus Rapid corridor. Figure 9 shows a segment of the Avenida Constituyentes (AC) and Figure 10 displays Calzada Belen (CB). It is important to mention that AC was the first Bus Rapid Transit system corridor in the city, and it extends for more than 12 kms. The analysis was made in a segment with the same 3 kms distance that is composed AL, and with one more road lane than AL. For its part, CB was the third BRT corridor in the city, and it functions as a complementary route to AL. The extension is similar in 3 kms but CB has one less lane than AL.

Nevertheless, the results are similar to what is happening to AL, both figures displays the similarity in the value segments along the corridor, but also, and very importantly, the necessity of intervention in local streets around the avenues.

Further research imply the necessity of comparing with the public transport demand of both avenues. It is important to mention that, this PSAT indicator has been evaluated in bigger metropolitan area with better results in the RMSE prediction [37].

6. Discussion

It is important to describe the disparity between the models. For instance, models with few variables and simpler processes, such as the Range of Activities, have similar results to more complex models such as the PSAT tool, which uses 14 variables. This also directs the need to test the indicator Range of Activities on a big scale, as a metropolitan area.

The complexity of having 14 variables means that the PSAT indicator also functions as a classifier. This means the indicator is capable of specifying which kind of activities have the street, creating a context with the variables. Apart from locating the most activities, as the Range of Activities does.

To the research question: Which spatial model better correlates to the actual public transport demand?

After implementing the spatial models, the public transport demand, and the ANN process, it was concluded that the PSAT model has a better correlation. The result enables to keep using this tool in other zones, cities, and metropolitan areas of the country since the data are official and public.

The results also shows that it is more of a priority to intervene in street segments around the AL corridor. This kind of project alludes to the last-mile concept. One big contribution is the location of urban renewal projects for an entire metropolitan area. The data available and the cost of the process made possible the results instead of using extensive and expensive field work.

In addition, these projects bring synergy to the public transport system, as it benefits all the people that ascend or descend until the next 400–500 m. In the same manner, these interventions support the analysis and location of other sustainable transport solutions such as the park and ride or the bike-share systems.

7. Conclusions

This paper compares six spatial models, giving diverse results, as seen in the BP ANN table. The following is a summary of the main results.

• Indicator 1 is the most assertive prediction, with at least a 1.92% difference from the next best indicators.
• The results in the extended area analysis support a change of strategy for urban renewal projects.
• There is little difference between the PSAT alternatives, and indicators 1 and 2, implying lousy influence on the public transport data. This variable consists of all the bus lines, pondered by their amount by street segment.
• Indicators 3 and 4 predict the public transport demand with around 80% precision. These two indicators are correlated, meaning that density and diversity of activities have much influence as volume predictors.
• There are two spatial models, indicators 5 and 6, with an RMSE superior to 25%, which prompts to eliminate them. It means that neither housing density nor specific activities have enough influence as volume predictors.

It also implies a prioritization in the intervention between both urban avenues. The AL corridor has the most value in all three principal indicators, so the need for intervention in most of the corridors is justified. On the contrary, AP has only specific segments with enough value, supporting an exact location for an urban renewal project, such as a public transport station, plus some segments of complete streets projects. Another trained network application estimates the non-stop public transport segment with the highest transport demand.

The indicator with better results is used in other urban corridors of the same city. The objective is to compare with corridors that have bus rapid transit lanes, in order to comprehend better the necessity of urban renewal and sustainable transportation interventions where cities develop faster. The results are very similar, and the need is also around the corridor, as well as AL also displays.

These studies are needed in developing countries where the technology is obsolete, and it is difficult to locate the volume of people ascending or descending from a public transportation system.

The latter is a base to predict the exact location where people are moving sustainably. The prediction will have an 18.68% margin for error and can be scaled for a metropolitan area. It should be noted that this confirmation is for the public transport demand, translated in the ascents and descents of people.

Part of the limitations of this paper is the r2 results. It implies further research detailing the performance of every one of the PSAT indicator variables. It potentially adjusts the indicator performance.

The next step is to confirm the tendency of the results, doing the exercise in more cities, and on a bigger scale, as a city or a metropolitan area, a more extensive scale will provide more diversified and complete data.

Another step will further the primary objective: to create an indicator that enhances both urban and transport strategies and policies of intervention in a road network. Such policies as prioritization of urban renewal projects where they are needed the most due to the volume of people already using them.