Konsortiumstreffen meinGrün
meinGrün Konsortiumstreffen, 17. Oktober 2019
meinGrün: ergänzende Social Media Analyse
Dr.-Ing Alexander Dunkel
Technische Universität Dresden
Geowissenschaften, Institut für Kartographie
Struktur:
- Stand: Auswertung Beliebtheit/ Frequentierung (Alexander)
- Stand: Thematische Analyse (Madalina)
- Backend, Service-Dienst, Datenschutz (HyperLogLog)
Web Präsentation
URL: http://ad.vgiscience.org/meingruen-konsortiumstreffen
➡ Pfeil rechts: Hauptfolien
⬇ Pfeil unten: Unterfolien
s-Taste
für Präsentationsmodus und Stichpunkte zu Folien
Frequentierung/Beliebtheit
External Links:
- Code Processing (Jupyter NB - Flickr UPL HTML)
- Download the Jupyter Notebook itself + the target shapes here
Summary Results
- Interactive maps Dresden (Flickr+Twitter+Instagram)
(in case of viewing problems, check Antivirus or Adblock)
- L0: UP Normalized
- L1: UP Normalized
- L2: UP Normalized
- L3: UP Normalized (careful: can be slow on slow hardware!)
- Interactive maps Heidelberg:
- L0: UP Normalized
- L1: UP Normalized
- L2: UP Normalized
- L3: UP Normalized
Abbreviations:
- L0-L3: These are the meinGrün target shape levels
- L0 - main targets (targets.json)
- L1 - main target parts (targetParts_Wege.json)
- L2 - mein target LC parts (targetParts_Wege_LC.json)
- L3 - main target LULC parts (targetParts_Wege_LULC.json)
- UP, UPL, UD: These are measurements for Social Media posts
- UP: Number of Photos
- UD: User days (each user is only counted once per day)
- UPL: User Post Locations (each user is only counted once per distinct location)
- A0, A1, A2, A3: These are short references to areas selected as example regions below for Heidelberg and Dresden
Data Structure Results:
Filename conventions for results: filenames follow the following pattern:
filename = {TODAY}_meingruenshapes_{TARGETSHAPE_VERSION}_{CITY_NAME.lower()}_{LEVEL}_weighted_{INTERSECT_VERSION}
where:
{TODAY}
- the date of processing (yyyy-mm-dd){TARGETSHAPE_VERSION}
- version of target shapes (e.g. as of 2019-09-12, v4.1 is the most recent version){CITY_NAME.lower()}
- either Dresden or Heidelberg{LEVEL}
- the target shape level (L0-L3, see above){INTERSECT_VERSION}
- the version of the intersection code used, as of 2019-09-12, v1.1 is the most recent version
Columns:
Both shapefile and CSV include the following columns
- TARGET_ID - reference ID to input target shapes
- UP - Number of total user posts per shape
- UP_Flickr - Number of total Flickr user posts per shape
- UP_Twitter - Number of total Twitter user posts per shape
- UP_Instagram - Number of total Instagram user posts per shape
- UP_Norm - Normalized/Weighted total User posts based on Area, interpolated to 1-1000 range
- UP_Norm_Origin - Normalized/Weighted total User posts based on Area and Source of Origin, interpolated to 1-1000 range
- popularity - Natural breaks applied to UP_Norm_Origin for generating 6 bins of classes.
These classes are suggestions for the overall city scale (but not meaningful when zoomed in):
- 5 : 'very high',
- 4 : 'high',
- 3 : 'average',
- 2 : 'low',
- 1 : 'very low',
- 0 : 'no data'
Column UP_Norm_Origin is the primary meinGrün Indicator for Frequentation/Popularity
1. Data processing
1.1 Intersection Examples (L0)
1.2 Normalizing and Weighting data
There are different measurements that provide different insights into social media patterns. What is counted:
- User Posts (UP)
- User Days (UD)
- User Post Location (UPL)
TARGET_ID | UP | UD | UPL | UP_Instagram | UP_Flickr | UP_Twitter | |
---|---|---|---|---|---|---|---|
24384 | dd_EBK-TSP-OSM_30690 | 35621 | 28946 | 24810 | 35054 | 567 | 0 |
1758 | dd_EBK-TSP-OSM_33708 | 8011 | 6375 | 5965 | 6983 | 1028 | 0 |
22702 | dd_EBK-TSP-OSM_20688 | 37789 | 31835 | 28202 | 34809 | 2967 | 13 |
8449 | dd_EBK-TSP-OSM_26112 | 3457 | 2375 | 1895 | 3025 | 432 | 0 |
16013 | dd_EBK-TSP-OSM_24972 | 3368 | 2368 | 1879 | 3025 | 343 | 0 |
Examples Normalization
The following graphics shows the top target shape for Dresden, based on Post Count (UP)
- yellow: top target polygon without area normalization
- red: top target polygon with area normalization
L2:
2. Results
2.1 Dresden
2.1.1 A1
- L0
- L1
- L2
- L3
2.1.2 A2
- L0
- L1
- L2
- L3
2.1.3 A3
- L0
- L1
- L2
- L3
ArcMap Natural Breaks (L3):
2.2 Heidelberg
2.2.1 A0
- L0
- L1
- L2
- L3
2.2.2 A1
- L0
- L1
- L2
- L3
2.3 Selected areas
- Pillnitz
- Dresden Altstadt
- Schloss Heidelberg
- Schloss und Schlossgarten Schwetzingen
3. Datasets
Exported datasets include:
- a CSV with column TARGET_ID (meinGrün Target-geometry ref) and LBSM columns
- UP,UP_Flickr,UP_Twitter,UP_Instagram,UD,UPL,UP_Norm,UP_Norm_Origin,popularity
- a Shapefile with the same columns
The two recommended measurement:
- popularity: labeled based on natural breaks and UP_Norm_Origin - recommended classes for overall city level
- UP_Norm_Origin: 1 to 1000 interpolated range weights for Social Media - apply individual classes
Backend, Service-Dienst, Datenschutz (HyperLogLog)
HLL Link (Marc Löchner)
Backend & Service Dienst
Vielen Dank