How to test an urban theory - Data Science Lecture

Slide 1

Slide 1 text

Testing an urban theory Marco De Nadai (http://www.marcodena.it)

Slide 2

Slide 2 text

2 Cities have always been studied IDEAL CITY (XV century)

Slide 3

Slide 3 text

3 Cities have always been studied IDEAL CITY (XV century) SYSTEM (XIX century)

Slide 4

Slide 4 text

4 Cities have always been studied IDEAL CITY (XV century) SYSTEM (XIX century) LIVING ORGANISM (XX century)

Slide 5

Slide 5 text

5 Understand cities

Slide 6

Slide 6 text

6 Understand cities New data

Slide 7

Slide 7 text

7 Understand cities New methods New data

Slide 8

Slide 8 text

How can we test an urban theory? 8 1 Take the theory “Operationalize” the theory Does it work? Is it still valid? Why does it matter? 2 3 4

Slide 9

Slide 9 text

9 Follow and ask please! How to test an urban theory 1 There are no stupid questions 2

Slide 10

Slide 10 text

Take the theory STEP 1

Slide 11

Slide 11 text

The theory: Jane Jacobs One of the most influential books in city planning • planning models that dominated mid-century planning • American housing policy (HOPE VI) • Melbourne, Toronto etc. 11 1 2 3 4 THEORY TEST Klemek, C. (2011) ‘Dead or Alive at Fifty? Reading Jane Jacobs on her Golden Anniversary’ Dissent, Vol. 58, No. 2, 75–79.

Slide 12

Slide 12 text

The theory: not tested! • Not empirically tested until 2015 • Tested in Seoul, from costly surveys collected in years • Theory from 1961! 12 1 2 3 4 THEORY TEST Sung, Hyungun, Sugie Lee, and SangHyun Cheon. "Operationalizing Jane Jacobs’s Urban Design Theory Empirical Verification from the Great City of Seoul, Korea." Journal of Planning Education and Research (2015.

Slide 13

Slide 13 text

The theory: Jane Jacobs One of the most influential books in city planning • Death: caused by the elimination of pedestrian activity • Life: created by a vital urban fabric at all times of the day 13 1 2 3 4 THEORY TEST

Slide 14

Slide 14 text

Jacobs’ diversity conditions Diversity => Urban vitality There are 4 diversity conditions To be ensured in each city’s district (10,000+ inhabitants) 14 1 2 3 4 THEORY TEST SMALL BLOCKS LAND USE AGED BUILDINGS DENSITY

Slide 15

Slide 15 text

Land Use Mix 2+ primary uses (contemporarily) JACOBS’ VIEW: People come for different purposes, continuously EFFECT: “sidewalk ballet” and “eyes on the street” 15 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY 1 2 3 4 THEORY TEST

Slide 16

Slide 16 text

Small blocks City blocks should be small/short LAND USE SMALL BLOCKS 16 1 2 3 4 THEORY TEST AGED BUILDINGS DENSITY JACOBS’ VIEW: improves walkability EFFECT: Increase face-to-face interactions

Slide 17

Slide 17 text

Aged buildings Buildings mixed (age and types) 17 AGED BUILDINGS 1 2 3 4 THEORY TEST LAND USE DENSITY SMALL BLOCKS JACOBS’ VIEW: To ensure economic diversity EFFECT: high-/low-income residents new/small enterprises

Slide 18

Slide 18 text

Density Concentration of people and enterprises JACOBS’ VIEW: People have a reason to live in a district EFFECT: Attract people 18 SMALL BLOCKS DENSITY 1 2 3 4 THEORY TEST LAND USE AGED BUILDINGS

Slide 19

Slide 19 text

Necessary, diversity conditions All four factors are necessary 19 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY 1 2 3 4 THEORY TEST

Slide 20

Slide 20 text

Border Vacuums • Patches of land dedicated to one single use • They could be either bad and good: • Parks are good for pedestrian activity • But they are exposed to criminality and deprivation if not well managed (e.g. night) 20 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 2 3 4 THEORY TEST

Slide 21

Slide 21 text

“Operationalize” the theory STEP 2

Slide 22

Slide 22 text

The data: 22 1 3 4 THEORY TEST 2

Slide 23

Slide 23 text

The data: 23 1 3 4 THEORY TEST 2 Get the data at: https://developer.foursquare.com/

Slide 24

Slide 24 text

The data: 24 1 3 4 THEORY TEST 2

Slide 25

Slide 25 text

The data: 25 1 3 4 THEORY TEST 2

Slide 26

Slide 26 text

The data: 26 1 3 4 THEORY TEST 2 Get the data at: https://overpass-turbo.eu/

Slide 27

Slide 27 text

The data: 27 1 3 4 THEORY TEST 2

Slide 28

Slide 28 text

The data: 28 1 3 4 THEORY TEST 2 • Urban Atlas: https://land.copernicus.eu/local/urban- atlas/urban-atlas-2012

Slide 29

Slide 29 text

The data: 29 1 3 4 THEORY TEST 2 • ISTAT (Census): https://www.istat.it/it/archivio/104317

Slide 30

Slide 30 text

30 2 1 3 4 URBAN DESCRIPTION GIS data “Operationalize” the theory

Slide 31

Slide 31 text

31 2 1 3 4 URBAN DESCRIPTION GIS data “Operationalize” the theory

Slide 32

Slide 32 text

32 2 1 3 4 URBAN DESCRIPTION GIS data “Operationalize” the theory

Slide 33

Slide 33 text

33 2 1 3 4 URBAN DESCRIPTION GIS data “Operationalize” the theory

Slide 34

Slide 34 text

34 2 1 3 4 URBAN DESCRIPTION GIS data Does it work? Is it still valid? 3 “Operationalize” the theory

Slide 35

Slide 35 text

“Operationalize” Land Use Mix For district : ! = − ' "∈$ !," log(!," ) log || !,#: % square footage of land use : {residential, commercial, recreation} 35 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2 Ref: R. Cervero. Land-use mixing and suburban mobility. University of California Transportation Center, 1989 EFFECT: The higher, the better. 1 0

Slide 36

Slide 36 text

36 1 3 4 THEORY TEST 2

Slide 37

Slide 37 text

37 1 3 4 THEORY TEST 2

Slide 38

Slide 38 text

“Operationalize” Small blocks Block size is a proxy for an high number of peoples’ interactions For district : 1 | | * 0∈023456(7) () EFFECT: The lower, the better 38 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2

Slide 39

Slide 39 text

39 1 3 4 THEORY TEST 2

Slide 40

Slide 40 text

Aged buildings Aged buildings are supposed to be a proxy for new, small enterprises. For district : 1 |7 | * 8∈9! 7 : set of companies EFFECT: The higher, the worse 40 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2

Slide 41

Slide 41 text

Aged buildings For district the weighted standard deviation of buildings age. EFFECT: The higher, the better 41 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2 !"# = ∑!"# $ %!('!( ̅ ')% + ('(#) ' ∑!"# $ %!

Slide 42

Slide 42 text

42 1 3 4 THEORY TEST 2

Slide 43

Slide 43 text

43 1 3 4 THEORY TEST 2

Slide 44

Slide 44 text

“Operationalize” Density For district : Employment density: |:;<23=>? <>3<2>!| @A>@! Population density: |B3@! EFFECT: The higher, the better 44 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2

Slide 45

Slide 45 text

45 1 3 4 THEORY TEST 2

Slide 46

Slide 46 text

“Operationalize” Vacuums Distance to huge parks for district : 1 7 * 8∈F! ( , , ) 7 : the set of the blocks : the set of parks EFFECT: The higher, the better 46 LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS 1 3 4 THEORY TEST 2

Slide 47

Slide 47 text

47 1 3 4 THEORY TEST 2

Slide 48

Slide 48 text

Diversity Vitality

Slide 49

Slide 49 text

Call Detail Records Data collected by mobile operators for billing reasons • Unique userID • Gender and age • Geographical location (Antenna) • Datetime 49 1 2 3 4 THEORY TEST

Slide 50

Slide 50 text

Call Detail Records 50 1 2 3 4 THEORY TEST

Slide 51

Slide 51 text

“Operationalize” Vitality • Mobile phone Internet activity as a proxy for urban vitality • We calculate the activity density in each district 1 || % "∈$ | % | : set of hours (180 days x 24h) • Six Italian cities with 100,000+ inhabitants (e.g. Rome, Milan…) • 6 months time span (in 2014) 51 0.2 0.0 0.2 0.4 0.6 0.8 1.0 Land Use Mix 0.8 1.2 1.8 2.7 4.1 6.1 9.3 14.0 21.1 31.9 Activity density ⇥ 103 ROME MILAN 0.8 1.2 1.8 2.7 4.1 6.1 9.3 14.0 21.1 31.9 Activity density 103 1 2 3 4 THEORY TEST MILAN

Slide 52

Slide 52 text

52 2 1 3 4 URBAN DESCRIPTION Mobile data GIS data “Operationalize” the theory

Slide 53

Slide 53 text

53 2 1 3 4 URBAN DESCRIPTION Mobile data GIS data “Operationalize” the theory

Slide 54

Slide 54 text

54 2 1 3 4 URBAN DESCRIPTION Mobile data GIS data “Operationalize” the theory

Slide 55

Slide 55 text

55 2 1 3 4 URBAN DESCRIPTION Mobile data GIS data “Operationalize” the theory

Slide 56

Slide 56 text

56 2 1 3 4 URBAN DESCRIPTION Mobile data GIS data Does it work? Is it still valid? 3 “Operationalize” the theory

Slide 57

Slide 57 text

Demo 1 https://github.com/denadai2/test-the- theory- lesson/blob/master/step1_compute_fe atures.ipynb 57

Slide 58

Slide 58 text

Is the theory still valid? STEP 3

Slide 59

Slide 59 text

−3 −2 −1 0 1 2 3 −9 −8 −7 −6 −5 −4 −3 −2 . Is the theory still valid? 59 Intersections density (log + Z-score) Activity density (log) 1 2 3 4 THEORY TEST

Slide 60

Slide 60 text

−3 −2 −1 0 1 2 3 −9 −8 −7 −6 −5 −4 −3 −2 R2 : 0.63 Is the theory still valid? 60 Intersections density (log + Z-score) Activity density (log) 1 2 3 4 THEORY TEST

Slide 61

Slide 61 text

Demo 2 https://github.com/denadai2/test-the- theory- lesson/blob/master/step2_exploration. ipynb 61

Slide 62

Slide 62 text

The log Linear Regression 62 = ! ! + " " + ⋯ + "# "# + 1 2 3 4 THEORY TEST

Slide 63

Slide 63 text

The log Linear Regression 63 Activity density 1 2 3 4 THEORY TEST = ! ! + " " + ⋯ + "# "# +

Slide 64

Slide 64 text

The log Linear Regression 64 Activity density Land Use Mix Employment density 1 2 3 4 THEORY TEST = ! ! + " " + ⋯ + "# "# +

Slide 65

Slide 65 text

Demo 3 http://setosa.io/ev/ordinary-least- squares-regression/ and https://github.com/denadai2/test-the- theory- lesson/blob/master/step3_Regression.i pynb 65

Slide 66

Slide 66 text

Jacobs’ theory holds and is still valid 66 1 2 3 4 THEORY TEST Urban metric Beta coefficient Employment density 0.434*** Intersections density 0.191*** Housing types 0.1854*** Closeness highways -0.102*** 3rd places x closeness highways 0.07** Closeness parks x closeness highways -0.07*** − 0.77 *** p-value < 0.001; ** p-value < 0.01; 4-fold Cross-validation: 75% training – 25% testing, 1000 interactions

Slide 67

Slide 67 text

Jacobs’ theory holds and is still valid 67 URBAN VITALITY 1 2 3 4 THEORY TEST URBAN METRICS Predict $: 0.77

Slide 68

Slide 68 text

…But something is different 68 1 2 3 4 THEORY TEST LAND USE SMALL BLOCKS AGED BUILDINGS DENSITY VACUUMS

Slide 69

Slide 69 text

Why does it matter? STEP 4

Slide 70

Slide 70 text

Web data and mobile phone records offer insights on how most urban dwellers experience entire cities 1 2 3 4 THEORY TEST

Slide 71

Slide 71 text

Why does it matter? • Evaluate the districts vitality • Know in advance the best places for retails • Quantifying regulatory interventions • We created the recipe for city that works 71 1 2 3 4 THEORY TEST

Slide 72

Slide 72 text

Let’s test an urban theory 72 1 The Jacobs’ theory We created the metrics We tested the theory Framework for urban vitality 2 3 4

Slide 73

Slide 73 text

Test the un-tested And now?

Slide 74

Slide 74 text

Broken windows theory • City mismanagement • Dirty places • Poor infrastructure Lead to misbehavior => Crime Q: Are people avoiding places where they feel unsafe? 74 Wilson, James Q., and George L. Kelling. "Broken windows." Critical issues in policing: Contemporary readings (1982): 395- 407.

Slide 75

Slide 75 text

75 Urban perception from Place Pulse Salesses, P., Schechtner, K., & Hidalgo, C. A. (2013). The collaborative image of the city: mapping the inequality of urban perception. PloS one

Slide 76

Slide 76 text

76 … 1 5 Place Pulse • New York • Boston • Linz • Salzburg Place Pulse 2 • Rome • Milan URBAN PERCEPTION Safety perception: MIT Place Pulse

Slide 77

Slide 77 text

77 Security perception prediction * B. Zhou, A. Lapedriza, J. Xiao, A. Torralba, and A. Oliva. “Learning Deep Features for Scene Recognition using Places Database.” NIPS, 2014. • Learning human security perception PERCEPTION SCORE [0-10]

Slide 78

Slide 78 text

Demo 4 https://github.com/denadai2/test-the- theory- lesson/blob/master/step_extra_vision.i pynb 78

Slide 79

Slide 79 text

79 Urban metric Standardized Beta coefficient Population density 0.155** Employees density 0.328** Deprivation -0.022 Distance from the center -0.257** Security perception 0.105** adj − R$ 0.91 ** p-value < 0.001; * p-value < 0.01; Security perception -> presence of people