The Twitterer model has the following attributes and metrics. Model Functions are available in the documentation
Filtering Attributes
These attributes are indexed to make intelligent queries. All geo-relevant queries should only look at users where path_affected == true
. Queries to determine evacuation behavior should further filter for shelter_in_place == false
Basic Filtering Attributes
Variable |
Type |
Description |
unclassifiable |
Boolean |
At some point, the method has deemed this user as unclassifiable due to a lack of information Update: Users with no affected path are also considered unclassifiable currently |
path_affected |
Boolean |
Whether or not this user’s path intersects with the bounding box |
shelter_in_place |
Boolean |
If the method has determined at any point the user sheltered in place, then this value is set to true |
unclassified_percentage |
Integer |
A rounded integer [0,100], that describes the percentage of the user’s tweets that did not land in a cluster during the DBScan clustering. |
Advanced Filtering Attributes
hazard_level_before
:
A ranking variable for a user’s calculated pre-storm home/shelter location that decreases as their level of potential danger / storm affectedness increases.
Value |
Description |
10 |
The user’s before home/shelter location exists in NYC Evacuation Zone A |
20 |
The user’s before home/shelter location exists in NYC Evacuation Zone B |
30 |
The user’s before home/shelter location exists in NYC Evacuation Zone C |
|
|
50 |
The user’s before home/shelter location exists within the Bounding Box |
|
|
100 |
The user’s before home/shelter location does not fall within the bounding box, however, their full movement path between tweets does intersect with the bounding box. (Could be a user traveling through NYC) |
Movement Variables
Variable |
Type |
Description |
shelter_in_place_location |
Array |
If a user sheltered in place, then this is the location that was calculated for that |
during_storm_movement |
2D Array |
The x,y points of the user’s during storm movement |
cluster_movement_pattern |
Array |
A list of cluster IDs, including the before and after the storm. Example: ["1","1","2","2","1"] implies that a user’s before and after storm location are calculated as cluster ID “1” and during the storm they moved from cluster 1 to cluster 2, but did not return to cluster 1 in the during storm window. |
Stored Location Attributes
Variable |
Type |
Description |
cluster_locations |
Hash |
Key : Value pair of Cluster ID : [x,y] . If available, :before_home and :after_home are keys |
Basic Attributes
Variable |
Type |
Description |
handle |
String |
User’s screen name. If they had more than one screen name, then it appears as a comma separated list, but still a string. |
tweet_count |
Integer |
The number of tweets embedded in the user document |
id_str |
String |
User’s id_str attribute from Twitter |
Sample User Document:
"_id" : ObjectId("53f61f802ddc1832ce001021"),
"id_str" : "256002542",
"handle" : "Alexandroh82",
"tweet_count" : 146,
"issue" : 40,
"during_storm_movement" : [
[
-74.005592,
40.78649
],
[
-74.00619055,
40.7886864
]
],
"cluster_locations" : {
"before_home" : [
-74.00619055,
40.7886864
],
"after_home" : [
-74.00619055,
40.7886864
],
"8" : [
-74.00619055,
40.7886864
],
"10" : [
-74.005592,
40.78649
]
},
"cluster_movement_pattern" : [
"8",
"10",
"8",
"8"
],
"hazard_level_before" : 50,
"shelter_in_place" : false,
"shelter_in_place_location" : [ ],
"confidence" : 40,
"unclassified_percentage" : 60,
"path_affected" : true,
"before" : [ ],
"during" : [ ],
"after" : [ ]